Surgical decision-making for long bone metastases Stein J. Janssen. Surgical decision-making for long bone metastases Stein J.

Transcription

1 Surgical decision-making for long bone metastases Stein J. Janssen Surgical decision-making for long bone metastases Stein J. Janssen

2 Stellingen behorend bij het proefschrift getiteld Surgical decision-making for long bone metastases Stein J. Janssen 1. Intramedullary nailing and endoprosthetic reconstruction are preferred treatment methods for patients with proximal femoral metastasis (this thesis) 2. Implant-specific reasons for reoperation and their timing should be considered when deciding between endoprosthetic reconstruction and intramedullary nailing for proximal femoral metastasis (this thesis) 3. Clinical CT scans can be useful for predicting occurrence of a pathological fracture through a femoral metastasis (this thesis) 4. Life expectancy is a key factor in surgical decision making for patients with long-bone metastases (this thesis) 5. Metastasectomy with negative margins is recommended for renal cell bone metastasis in patients with a relatively good life expectancy (this thesis) 6. A nomogram is a stable and moderately accurate method for predicting survival probabilities in patients with long-bone metastasis (this thesis) 7. Without data you re just another person with an opinion (W. Edwards Deming) 8. Prediction is very difficult, especially about the future (Niels Bohr) 9. The devil is in the detail 10. Chance favours only the prepared mind (Louis Pasteur) 11. It is not the critic who counts; not the man who points out how the strong man stumbles, or where the doer of deeds could have done them better. The credit belongs to the man who is actually in the arena, whose face is marred by dust and sweat and blood (Theodore Roosevelt)

3 Surgical decision-making for long bone metastases Stein J. Janssen

4 Copyright Stein J. Janssen, Amsterdam, The Netherlands. No part of this thesis may be reproduced, stored or transmitted in any form or by any means, without the prior permission of the author and the original copyright holder. Work performed at Academic Medical Center, University of Amsterdam and Massachusetts General Hospital, Harvard Medical School. The research fellowship was supported by grants from: KWF Kankerbestrijding De Drie Lichten Anna Fonds Prof. Michaël-van Vloten Fonds Printing of this thesis was financially supported by: Academic Medical Center Amsterdam, Department of Orthopaedic Surgery Nederlandse Orthopaedische Vereniging Implantcast Van Campen Consulting Chipsoft ISBN: Layout and printing: Optima Grafische Communicatie, Rotterdam, The Netherlands

5 SURGICAL DECISION-MAKING FOR LONG BONE METASTASES ACADEMISCH PROEFSCHRIFT ter verkrijging van de graad van doctor aan de Universiteit van Amsterdam op gezag van de Rector Magnificus prof. dr. ir. K.I.J. Maex ten overstaan van een door het College voor Promoties ingestelde commissie, in het openbaar te verdedigen in de Agnietenkapel op dinsdag 13 november 2018, te 10:00 uur. door Stein Jasper Janssen geboren te Utrecht

6 Promotiecommissie: Promotor: prof. dr. C.N. van Dijk AMC-UvA Copromotor: dr. J.A.M. Bramer AMC-UvA Overige leden: prof. dr. P.D.S. Dijkstra Universiteit Leiden prof. dr. J.C. Goslings AMC-UvA prof. dr. M. Maas AMC-UvA dr. G.R. Schaap AMC-UvA prof. dr. H.W.B. Schreuder Radboud Universiteit Nijmegen prof. dr. E.M.A. Smets AMC-UvA dr. A.M. Westermann AMC-UvA Faculteit der Geneeskunde

7 Table Of Contents Chapter 1: General Introduction 7 Part I: Metastatic Femoral Lesions Chapter 2: Complications After Surgery For Proximal Femoral Metastasis: A Retrospective Study Of 417 Patients Chapter 3: Outcome After Fixation Of Metastatic Proximal Femoral Fractures: A Systematic Review Of 40 Studies Chapter 4: A Comparison Of Questionnaires For Assessing Physical Function In Patients With Lower Extremity Bone Metastases Chapter 5: Predicting Pathological Fracture In Femoral Metastases Using A Clinical CT Scan Based Algorithm: A Case-Control Study Part II: Metastatic Humeral Lesions Chapter 6: Outcome Of Operative Treatment Of Metastatic Fractures Of The 109 Humerus: A Systematic Review Of 23 Clinical Studies Chapter 7: Complications After Surgery For Metastatic Humeral Lesions 129 Chapter 8: Management Of Metastatic Humeral Fractures: Variations According 149 To Orthopedic Subspecialty, Tumor Characteristics Part III: Survival Chapter 9: Metastasectomy, Intralesional Resection, Or Stabilization Only In The Treatment Of Bone Metastases From Renal Cell Carcinoma Chapter 10: Prognostication In Patients With Long Bone Metastases: Does A Boosting Algorithm Improve Survival Estimates? Chapter 11: Are Allogeneic Blood Transfusions Associated With Decreased Survival After Surgery For Long Bone Metastatic Fractures? Chapter 12: Summary 229 Chapter 13: General Discussion 239 Chapter 14: Summary In Dutch (Samenvatting In Het Nederlands) 255 Abbreviations 263 Acknowledgments 265 Portfolio 269 Report Of Scholarship 273 About The Author 283

8

9 CHAPTER 1 1 General Introduction

10 Chapter 1 Bone metastasis occurs when cancer cells spread from their original site via the bloodstream to bone tissue. 1 Skeletal metastases (secondary bone tumors) account for 70% of all malignant bone tumors; the remainder being primary bone tumors such as osteosarcoma and chondrosarcoma. 2 Management of skeletal metastases is almost always palliative and therefore different from management of primary bone tumors. HISTORY The earliest archaeological examples of metastatic carcinoma of the skeleton were found in Egypt and date from 3000 to 500BC. 3-6 Historically, bone metastases were less common as a result of shorter life expectancy most people did not live long enough to develop cancer and differences in life-style (smoking, dietary constituents, pollution). 4,7,8 Advances in medical care, sanitation, and nutrition in the 19 th and 20 th century considerably improved life expectancy and markedly changed disease patterns: cancer became more prevalent and is currently the second leading cause of death in developed countries (after cardiovascular disease). 9,10 In the United States, 1.7 million people were diagnosed with cancer in 2015 and about 600,000 died as a result of cancer (United States population in 2015: 321 million). 11 In the Netherlands, 100,000 people were diagnosed with cancer in 2015 and about 40,000 died as a result of cancer (Netherlands population in 2015: 17 million). 12 The number of people with a history of cancer continues to grow due to: (1) a growing number of cases in the aging population, and (2) longer survival secondary to early cancer detection and improved medical treatment. 13,14 In the United States on January 1, 2014, there were 14.5 million living people with a history of cancer and in ten years, this number is estimated to increase to 19 million. 14 INCIDENCE Breast, lung, and prostate cancer are the most common primary tumors in developed countries and are also tumors that are most likely to metastasize to bone. 11,12,15,16 After the lungs and liver, bone is the most common site of metastases and these metastases are more common in the axial skeleton than in the appendicular skeleton. 1,17-19 The femur and humerus are the most frequently affected long bones. 1,18 In autopsy studies, the incidence of bone metastases among cadavers varies substantially based on the primary tumor type: among people who died of breast or prostate carcinoma about 70% had bone metastases, while among people who died of lung, kidney, or thyroid carcinoma about 30% had bone metastases. 19,20 It is estimated that approximately 280,000 adults were living with bone metastatic disease in the United States in The economic burden 8

11 Introduction of patients with metastatic bone disease was estimated at $12.6 billion, which is 17% of the total direct medical costs related to cancer ($74 billion) in the United States (in 2004). 13 Bone metastases often occur in the final stages of life and can result in skeletal related events such as bone pain, nerve root compression, spinal cord compression, hypercalcemia, and pathological fracture. 22 Three population based cohort studies from Denmark present the rates of bone metastases and skeletal related events among patients with newly diagnosed breast, lung, or prostate cancer between 1999 and Jensen et al. demonstrated that among 35,912 breast cancer patients, 4.2% (n=1,494) had bone metastases either at the time of the primary cancer diagnosis (0.6%) or during followup (3.5%), and 48% (712 of 1,494) developed a skeletal related event (median followup: 3.5 years). 23 Cetin et al. demonstrated that among 29,720 lung cancer patients, 6.8% (n=2,032) had bone metastases either at the time of the primary cancer diagnosis (1.1%) or during followup (5.7%), and 56% (1,146 of 2,032) developed a skeletal related event (median followup: 7.3 months). 24 Nørgaard et al. demonstrated that among 23,087 men with prostate cancer, 14% (n=3,261) had bone metastases either at the time of the primary cancer diagnosis (3.0%) or during followup (11%), and 52% (1,691 of 3,261) developed a skeletal related event (median followup 2.2 years). 25 Hence, about half of the patients with bone metastases develop skeletal related events. 26 Pathological fracture is a common skeletal related event with rates varying from 16% to 42% among patients with bone metastases. 1,27-31 The increasing number of patients with a history of cancer has been linked to an absolute increase in the number of patients living with bone metastatic disease, which probably results in a larger number of patients developing a pathological fracture. Skeletal related events pathological fractures in particular cause decreased quality of life, decline in physical function, loss of independence, and decreases survival. 20,22,32 1 WORK-UP Metastatic carcinoma of bone should be considered in patients older than 40 years with a suspect bone lesion. The diagnostic workup for a patient with a lesion suspect for metastatic carcinoma and unknown primary tumor starts with a clinical history and physical examination of the chest, abdomen, and thyroid, and the breast in female patients and prostate in male patients. 1,33-35 This is followed by laboratory analysis for multiple myeloma and prostate cancer. Chest radiographs and radiographs of every painful bone should be obtained. Bone metastasis typically appears as osteolytic permeative (i.e. moth-eaten appearance) lesions of the diaphysis or metaphysis. 35,36 However, depending on the primary tumor, lesions can also be osteoblastic (e.g. prostate) or mixed with osteolytic and osteoblastic components (e.g. breast). Bone CT and MRI are preferred methods to further characterize the bone lesion in case a primary bone tumor is suspected. A CT-scan of 9

12 Chapter 1 the chest, abdomen, and pelvis can be made if the primary tumor has not been identified yet. 33,35 A technetium 99m or FDG-PET scan can be made to assess for other skeletal lesions and search for the primary tumor. Finally, a diagnostic biopsy of the bone lesion preferably one that is most easily accessible in case of multiple bone lesions can be obtained for histological confirmation. A biopsy is recommended if a patient has no previous histological diagnosis of metastatic bone disease, the patient has been disease-free for a prolonged period of time, or if the lesion is not characteristic of the known primary cancer. 1,37-39 Multiple myeloma and lymphoma are considered primary bone tumors as they originate from hematopoietic and lymphoid tissue; however, they are often grouped with bone metastases as surgical management is comparable. 40 SURGICAL MANAGEMENT In general, the aim of surgical treatment for a patient with a metastatic lesion is to optimize quality of life and physical function for the remaining life span while minimizing the risk of complications, mortality, and secondary surgical interventions. Patients with a pathological fracture typically present to orthopaedic oncology surgeons and trauma surgeons as it often mandates surgical intervention. As a rule of thumb, surgical treatment for a pathological fracture through a bone metastasis is indicated if the estimated life expectancy of the patient at least exceeds the anticipated recovery time from surgery. 1,41,42 This translates into a minimum of six to twelve weeks of expected survival for surgical procedures such as intramedullary nailing. 37,38 However, accurately estimating life expectancy in patients with bone metastases is difficult and therefore a potential barrier to providing optimal care. 43,44 Nonoperative management (i.e. closed reduction and immobilization) of longbone pathological fractures is ineffective as these fractures demonstrate poor fracture healing potential due to the tumor characteristics and as a result of radiation therapy. 37,38,45 The indication for surgery is less clear for non-fractured metastatic lesions. The two most commonly cited indications for surgical management of non-fractured metastatic lesion are an impending fracture (i.e. a metastatic lesion at risk of pathological fracture) and a solitary metastasis. 1,37,38 Impending fractures are easier to treat, with less morbidity, less costs, and faster recovery as compared to complete pathological fractures. 38,46-50 However, the downside is overtreatment of patients who would not develop a pathological fracture. Therefore, many studies aimed to establish predictors of pathological fracture occurrence to better define lesions at risk of fracture. Although no single clinical or radiographic predictor or combination is sufficiently accurate at predicting occurrence of a pathological fracture, several risk factors have been identified: pain on weight bearing, defect size >30 millimeters, lytic appearance on radiographs, location of the lesion, and more than 50% circumferential cortical destruction Solitary metastasis especially in 10

13 Introduction renal cell carcinoma and thyroid carcinoma is often mentioned as an indication for enbloc surgical resection as some studies demonstrate improved survival after complete resection; however, this is refuted by others. 1,54-57 Common methods of fixation for long bone metastatic lesions can be categorized into: open reduction internal fixation, intramedullary nailing, and endoprosthetic reconstruction. Many different implants exist within these categories and can be combined with intraoperative adjuvants such as polymethyl methacrylate or bone grafts, creating numerous treatment strategies. However, there is no consensus about which surgical strategy is most adequate for femoral and humeral metastatic bone lesions. 58 In addition, postoperative external beam radiation therapy is often used as it might reduce the risk of tumor progression and improve function, although evidence is scarce. 59,60 As bone metastases can originate from many primary tumors, occur everywhere in the skeleton, and have multiple morphological appearances, numerous factors need to be considered in surgical decision making, including primary tumor type, life expectancy, location of the bone metastasis, presence of a pathological fracture, and presence of visceral or other bone metastases In conclusion, patients with bone metastatic lesions have on average a poor prognosis and surgical management predominantly aims to optimize quality of life by providing a stable construct that outlives the patient. The purpose of this PhD thesis is to: develop tools for better patient selection for surgery, improve implant selection based on patient- and tumor characteristics, identify risk factors for adverse outcomes, and evaluate outcome after treatment for patients with long bone metastases. OUTLINE OF CHAPTERS Part I: Metastatic Femoral Lesions There is debate about which implant is most appropriate for treatment of metastatic proximal femoral lesions and only a few studies compare surgical strategies. Chapter 2 compares surgical outcomes after commonly used implants for proximal femoral metastases in a large multi-institutional retrospective cohort study. Chapter 3 puts these findings into perspective by pooling surgical outcomes over a large number of studies using a systematic review of the literature. This review also demonstrates that although considered important functional outcome is only scarcely and inconsistently reported. Chapter 4 therefore compares different questionnaires that measure physical function in patients with lower extremity bone metastases in a cross-sectional survey study. Measuring functional outcome is an important step towards establishing which implant is optimal. Chapter 5 describes and tests an algorithm that can be used to predict occurrence of a pathological fracture through a non-fractured femoral metastatic lesion using a CT scan. 11

14 Chapter 1 This helps to better assess which lesion is at risk of fracture and potentially benefits from prophylactic fixation. Part II: Metastatic Humeral Lesions Chapter 6 is a systematic review of the literature and provides an overview of outcomes after commonly used surgical techniques for metastatic humeral lesions. There is substantial variation in reported outcomes and surgical strategies used; it is unclear which implant is most appropriate for a specific situation. In Chapter 7, we established complication and reoperation rates in a multi-institutional retrospective cohort study with a large number of patients that underwent surgery for metastatic humeral lesions. In addition, this study determined risk factors for these outcomes to anticipate postoperative problems. These findings could help inform our patients and risk stratify them. Chapter 8 is a case-vignette study assessing how orthopaedic oncologists and trauma surgeons approach metastatic humeral lesions. We determined which factors influence the decision for surgical treatment and the choice for a specific implant. This study also sheds light on areas with relative consensus among surgeons regarding treatment options, and areas where surgeons disagree and more evidence is needed. Part III: Survival Chapter 9 specifically addresses bone metastases from renal cell carcinoma and how different methods of resection influence tumor recurrence, reoperation, and survival. Chapter 10 describes the development of different algorithms for survival prognostication in patients with long bone metastases and compares their accuracy. These algorithms help estimate life expectancy which is an important factor in surgical decision making. Chapter 11 assesses if perioperative allogeneic blood transfusions influence survival in patients who undergo surgery for long bone metastastases. Finally, a summary followed by a general discussion including conclusions and future perspectives are provided in Chapter 12 &

15 Introduction PRIMARY STUDY QUESTIONS 1 Part I: Metastatic Femoral Lesions Is there a difference in outcome physical function, reoperations, and complications between endoprosthetic reconstruction, intramedullary nailing, and open reduction internal fixation for proximal femoral metastasis? What questionnaire is most useful for measurement of physical function in patients with lower extremity bone metastasis? Can a CT-scan based algorithm predict occurrence of a pathological fracture through a metastatic femoral lesion? Part II: Metastatic Humeral Lesions What outcome physical function, reoperations, and complications can be expected after surgical treatment of humeral metastasis? What factors are associated with reoperations and systemic complications after surgical treatment of humeral metastasis? Is there a difference in surgical decision making for humeral metastasis based on physician, patient, or tumor characteristics? Part III: Survival Is there a difference in local tumor recurrence, reoperation, and survival between metastasectomy, intralesional resection, and stabilization only for renal cell metastasis? What factors are associated with worse survival among patients who underwent surgery for long bone metastases? What type of algorithm is most accurate for predicting survival probability after surgery for long bone metastases? Are allogeneic blood transfusions associated with worse survival after surgery for long bone metastases? 13

16 Chapter 1 REFERENCES 1. Randall RL. Metastatic Bone Disease: An Integrated Approach to Patient Care. New York: Springer-Verlag New York; Greenspan A, Jundt G, Remagen W. Differential Diagnosis in Orthopaedic Oncology. Second ed Binder M, Roberts C, Spencer N, Antoine D, Cartwright C. On the antiquity of cancer: evidence for metastatic carcinoma in a young man from ancient Nubia (c BC). PLoS One. 2014; 9(3): e Nerlich AG, Rohrbach H, Bachmeier B, Zink A. Malignant tumors in two ancient populations: An approach to historical tumor epidemiology. Oncol Rep. Jul 2006; 16(1): Strouhal E. Ancient Egyptian case of carcinoma. Bull N Y Acad Med. Mar 1978; 54(3): Strouhal E, Kritscher H. Neolithic Case of a Multiple Myeloma from Mauer (Vienna, Austria). Anthropologie (Brno). 1990; 28(1): Doll R, Peto R. The causes of cancer: quantitative estimates of avoidable risks of cancer in the United States today. J Natl Cancer Inst. Jun 1981; 66(6): Purdue MP, Hutchings SJ, Rushton L, Silverman DT. The proportion of cancer attributable to occupational exposures. Ann Epidemiol. Mar 2015; 25(3): Eurostat. Causes of death statistics. 2012; Accessed 31/ 03/ 2016, National Center for Health Statistics. Health, United States, Hyattsville, American Cancer Society. Cancer Facts & Figures Atlanta integraal kankercentrum nederland (iknl). Incidentie en sterfte van kanker. 2016; cijfersoverkanker.nl/. Accessed 30/03/2016, Schulman KL, Kohles J. Economic burden of metastatic bone disease in the U.S. Cancer. Jun ; 109(11): DeSantis CE, Lin CC, Mariotto AB, et al. Cancer treatment and survivorship statistics, CA Cancer J Clin. Jul-Aug 2014; 64(4): Tofe AJ, Francis MD, Harvey WJ. Correlation of neoplasms with incidence and localization of skeletal metastases: An analysis of 1,355 diphosphonate bone scans. J Nucl Med. Nov 1975; 16(11): Scher HI, Yagoda A. Bone metastases: pathogenesis, treatment, and rationale for use of resorption inhibitors. Am J Med. Feb ; 82(2A): Swanson DA, Orovan WL, Johnson DE, Giacco G. Osseous metastases secondary to renal cell carcinoma. Urology. Dec 1981; 18(6): Krishnamurthy GT, Tubis M, Hiss J, Blahd WH. Distribution pattern of metastatic bone disease. A need for total body skeletal image. JAMA. Jun ; 237(23): Abrams HL, Spiro R, Goldstein N. Metastases in carcinoma; analysis of 1000 autopsied cases. Cancer. Jan 1950; 3(1): Coleman RE. Clinical features of metastatic bone disease and risk of skeletal morbidity. Clin Cancer Res. Oct ; 12(20 Pt 2): 6243s-6249s. 21. Li S, Peng Y, Weinhandl ED, et al. Estimated number of prevalent cases of metastatic bone disease in the US adult population. Clin Epidemiol. 2012; 4: Wilkinson AN, Viola R, Brundage MD. Managing skeletal related events resulting from bone metastases. BMJ. 2008; 337: a

17 Introduction 23. Jensen AO, Jacobsen JB, Norgaard M, Yong M, Fryzek JP, Sorensen HT. Incidence of bone metastases and skeletal-related events in breast cancer patients: a population-based cohort study in Denmark. BMC Cancer. 2011; 11: Cetin K, Christiansen CF, Jacobsen JB, Norgaard M, Sorensen HT. Bone metastasis, skeletalrelated events, and mortality in lung cancer patients: a Danish population-based cohort study. Lung Cancer. Nov 2014; 86(2): Norgaard M, Jensen AO, Jacobsen JB, Cetin K, Fryzek JP, Sorensen HT. Skeletal related events, bone metastasis and survival of prostate cancer: a population based cohort study in Denmark (1999 to 2007). J Urol. Jul 2010; 184(1): Oster G, Lamerato L, Glass AG, et al. Natural history of skeletal-related events in patients with breast, lung, or prostate cancer and metastases to bone: a 15-year study in two large US health systems. Support Care Cancer. Dec 2013; 21(12): Daniele S, Sandro B, Salvatore I, et al. Natural History of Non-Small-Cell Lung Cancer with Bone Metastases. Sci Rep. 2015; 5: Coleman RE, Rubens RD. The clinical course of bone metastases from breast cancer. Br J Cancer. Jan 1987; 55(1): Saad F, Gleason DM, Murray R, et al. A randomized, placebo-controlled trial of zoledronic acid in patients with hormone-refractory metastatic prostate carcinoma. J Natl Cancer Inst. Oct ; 94(19): Lipton A, Colombo-Berra A, Bukowski RM, Rosen L, Zheng M, Urbanowitz G. Skeletal complications in patients with bone metastases from renal cell carcinoma and therapeutic benefits of zoledronic acid. Clin Cancer Res. Sep ; 10(18 Pt 2): 6397S-6403S. 31. Zekri J, Ahmed N, Coleman RE, Hancock BW. The skeletal metastatic complications of renal cell carcinoma. Int J Oncol. Aug 2001; 19(2): van der Vliet QM, Paulino Pereira NR, Janssen SJ, et al. What Factors are Associated With Quality Of Life, Pain Interference, Anxiety, and Depression in Patients With Metastatic Bone Disease? Clin Orthop Relat Res. Oct Rougraff BT, Kneisl JS, Simon MA. Skeletal metastases of unknown origin. A prospective study of a diagnostic strategy. J Bone Joint Surg Am. Sep 1993; 75(9): Piccioli A, Maccauro G, Spinelli MS, Biagini R, Rossi B. Bone metastases of unknown origin: epidemiology and principles of management. J Orthop Traumatol. Jun 2015; 16(2): Dutch Orthopedic Tumor Society. Botmetastasen. Oncoline integraal kankercentrum nederland (iknl) Rosenthal DI. Radiologic diagnosis of bone metastases. Cancer. Oct ; 80(8 Suppl): Bickels J, Dadia S, Lidar Z. Surgical management of metastatic bone disease. J Bone Joint Surg Am. Jun 2009; 91(6): Quinn RH, Randall RL, Benevenia J, Berven SH, Raskin KA. Contemporary management of metastatic bone disease: tips and tools of the trade for general practitioners. J Bone Joint Surg Am. Oct ; 95(20): Clayer M, Duncan W. Importance of biopsy of new bone lesions in patients with previous carcinoma. Clin Orthop Relat Res. Oct 2006; 451: Alvi HM, Damron TA. Prophylactic stabilization for bone metastases, myeloma, or lymphoma: Do we need to protect the entire bone? Tumor. Clin Orthop Relat Res. 2013; 471: Damron TA, Sim FH. Surgical treatment for metastatic disease of the pelvis and the proximal end of the femur. Instr Course Lect. 2000; 49:

18 Chapter Bryson DJ, Wicks L, Ashford RU. The investigation and management of suspected malignant pathological fractures: a review for the general orthopaedic surgeon. Injury. Oct 2015; 46(10): Bauer HC, Wedin R. Survival after surgery for spinal and extremity metastases. Prognostication in 241 patients. Acta Orthop Scand. Apr 1995; 66(2): Nathan SS, Healey JH, Mellano D, et al. Survival in patients operated on for pathologic fracture: implications for end-of-life orthopedic care. J Clin Oncol. Sep ; 23(25): Gainor BJ, Buchert P. Fracture healing in metastatic bone disease. Clin Orthop Relat Res. Sep 1983(178): Arvinius C, Parra JLC, Mateo LS, Maroto RG, Borrego AF, Stern LLD. Benefits of early intramedullary nailing in femoral metastases. Int Orthop. 2014; 38(1): Dijkstra S, Wiggers T, van Geel BN, Boxma H, Dijstra S. Impending and actual pathological fractures in patients with bone metastases of the long bones. A retrospective study of 233 surgically treated fractures. Eur J Surg. 1994; 160: Blank AT, Lerman DM, Patel NM, Rapp TB. Is Prophylactic Intervention More Cost-effective Than the Treatment of Pathologic Fractures in Metastatic Bone Disease? Clin Orthop Relat Res. Mar Gitelis S, Sheinkop MB, Hammerberg K, Brugliera P. The role of prophylactic surgery in the management of metastatic hip disease. Orthopedics. Aug 1982; 5(8): Ward WG, Holsenbeck S, Dorey FJ, Spang J, Howe D. Metastatic disease of the femur: surgical treatment. Clin Orthop Relat Res. 2003: S Van der Linden YM, Dijkstra PD, Kroon HM, et al. Comparative analysis of risk factors for pathological fracture with femoral metastases. J Bone Joint Surg Br. May 2004; 86(4): Mirels H. Metastatic disease in long bones. A proposed scoring system for diagnosing impending pathologic fractures. Clin Orthop Relat Res. Dec 1989(249): Dijkstra PD. Pathological fractures of long bones due to bone metastases: Orthopaedic Surgery, Erasmus Universiteit Rotterdam; Fuchs B, Trousdale RT, Rock MG. Solitary bony metastasis from renal cell carcinoma: significance of surgical treatment. Clin Orthop Relat Res. Feb 2005(431): Hoshi M, Takada J, Ieguchi M, Takahashi S, Nakamura H. Prognostic factors for patients with solitary bone metastasis. Int J Clin Oncol. Feb 2013; 18(1): Les KA, Nicholas RW, Rougraff B, et al. Local progression after operative treatment of metastatic kidney cancer. Clin Orthop Relat Res. Sep 2001(390): Lin PP, Mirza AN, Lewis VO, et al. Patient survival after surgery for osseous metastases from renal cell carcinoma. J Bone Joint Surg Am. Aug 2007; 89(8): Steensma M, Healey JH. Trends in the surgical treatment of pathologic proximal femur fractures among Musculoskeletal Tumor Society members. Clin Orthop Relat Res. Jun 2013; 471(6): Townsend PW, Rosenthal HG, Smalley SR, Cozad SC, Hassanein RE. Impact of postoperative radiation therapy and other perioperative factors on outcome after orthopedic stabilization of impending or pathologic fractures due to metastatic disease. J Clin Oncol. Nov 1994; 12(11): Willeumier JJ, van der Linden YM, Dijkstra PD. Lack of clinical evidence for postoperative radiotherapy after surgical fixation of impending or actual pathologic fractures in the long bones in patients with cancer; a systematic review. Radiother Oncol. Oct 2016; 121(1):

19

20

21 PART I Metastatic Femoral Lesions

22

23 CHAPTER 2 Complications After Surgery For Proximal Femoral Metastasis: A Retrospective Study Of 417 Patients 2 S.J. Janssen J.T.P. Kortlever J.E. Ready K.A. Raskin M.L. Ferrone F.J. Hornicek S.A. Lozano-Calderon J.H. Schwab Journal of the American Academy of Orthopaedic Surgeons Jul;24(7): Presented at: Musculoskeletal Tumor Society Annual Meeting 2014, Houston, Texas, USA. American Academy of Orthopaedic Surgeons Annual Meeting 2015, Las Vegas, Nevada, USA.

24 Chapter 2 ABSTRACT Objectives To compare outcomes among surgical strategies for proximal femoral metastases. Design Retrospective cohort study. Setting Two tertiary care referral centers for orthopaedic oncology. Participants 417 consecutive patients with proximal femoral metastasis who underwent surgery between 1999 and Interventions Intramedullary nailing (n = 302), endoprosthetic reconstruction (n = 70), and open reduction and internal fixation (n = 45). Outcome Measures Primary outcome measures were reoperations and 30-day systemic complications. Secondary outcome measures were total estimated blood loss, anesthesia time, duration of hospital admission, and 30-day survival. Results Reoperation rates did not differ among surgical strategies (5.3% after intramedullary nailing, 11% after endoprosthetic reconstruction, and 13% after open reduction and internal fixation; p = 0.134). When reasons for reoperation were assessed separately, fixation failure was most common after open reduction and internal fixation (13% versus 3.0% after intramedullary nailing and none after endoprosthetic reconstruction; p < 0.001), whereas deep infection was most common after endoprosthetic reconstruction (8.6% versus 2.0% after intramedullary nailing and none after open reduction and internal fixation; p = 0.010). Overall 30-day systemic complication rates did not differ among surgical strategies (8.3% after intramedullary nailing, 14% after endoprosthetic reconstruction, and 11% after open reduction and internal fixation; p = 0.268). Conclusions Implant-specific complications and their timing should be considered in the choice of surgical strategy. Analysis of secondary outcomes and risk factors for systemic complications could aid in surgical decision making. 22

25 Complications After Surgery For Proximal Femoral Metastasis INTRODUCTION The proximal femur is the long bone most commonly affected by metastatic disease. 1,2 Bone metastases weaken the bone and reduce load bearing capabilities. These changes can result in pain and eventually pathological fracture. Surgical treatment is often indicated in patients with pathological fracture, whereas indications are less clear in patients with non-fractured lesions. 3,4 Factors considered in the decision to pursue surgical treatment and the selection of surgical technique include the location of the lesion, presence of a fracture, tumor type, cortical destruction, the patient s life expectancy, patient preferences, and the expected outcome. 3-7 Only a few studies have rigorously compared outcomes among implant types. 3,4 Although most studies are limited by retrospective design and inherent bias, they provide useful information about implant durability and complications. 3-5,8 However, studies comparing surgical strategies need large numbers of patients because outcomes such as reoperation are relatively rare (6.4 to 10.3%). 3,4 A survey study of 98 orthopaedic oncologists demonstrated large variation in the physicians preferred surgical strategies, emphasizing the need for further study to improve understanding of surgical outcomes. 9 The aim of this study was to assess reoperation rates and systemic complications after surgical treatment of patients with metastases and multiple myeloma of the proximal femur. Multiple myeloma was included because the surgical approach is comparable to metastatic lesions resulting from solid tumors. 5 Specifically, we compared reoperation rate and 30-day systemic complication rate among surgical strategies. Secondarily, we compared blood loss, anesthesia time, duration of hospital admission, and 30-day survival among surgical strategies. Additionally, we assessed risk factors for 30-day systemic complications. 2 METHODS Study Design This retrospective study was approved by our institutional review board. All electronic medical records of patients who had an International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis code for any pathological fracture (ICD-9-CM code 733.1) or a Current Procedural Terminology (CPT) code for prophylactic fixation of any bone (CPT codes 24498, 23491, 23490, 27495, 27187, 27745, 25490, 25492, 25491, or 24498) between January 1999 and January 2014 at two tertiary care referral centers for orthopaedic oncology were flagged. All flagged electronic medical records were manually reviewed to establish whether the patient fulfilled our predefined eligibility criteria. We included a consecutive series of 417 eligible patients who underwent 23

26 Chapter 2 I II III Figure 1: Illustration of the proximal femur demonstrating the anatomic areas included in the study. Area I consists of the isthmus to the base of the femoral neck, area II is the trochanteric region, and area III is the subtrochanteric region to 5 cm below the lesser trochanter. If the lesion spanned multiple areas, we included it in the area most affected by the lesion. surgical treatment of a proximal femoral metastasis or myeloma. In the 21 patients who had bilateral lesions, we included only the first surgical procedure so as to not violate the assumption of independence. 10 The proximal femur is defined as the region extending from the isthmus of the femoral neck to 5 cm below the lesser trochanter (Figure 1). 11 We excluded patients younger than 18 years, patients undergoing revision procedures, patients with substantial tumor involvement of the acetabulum, and patients with a lesion proximal to the isthmus of the femoral neck. The latter were excluded because these proximal lesions are almost exclusively managed with endoprostheses, hampering comparison of techniques. Patients were included regardless of followup duration because we considered both short-term and long-term outcomes to be relevant. Surgical Strategies Although preferences varied among surgeons, in general, a pathological fracture was treated surgically when the patient was expected to live longer than 30 days. An impending fracture was treated surgically when the patient was expected to live longer than 30 days and had substantial bone destruction or pain on load bearing. In terms of surgical strategy, proximal femur resection and endoprosthetic reconstruction was generally preferred in patients with extensive bone destruction (and therefore more often in patients with pathological fractures), in patients with tumors resistant to radiation therapy (e.g. renal cell carcinoma), and in patients with more proximal metastatic lesions. Intramedullary nailing was primarily used in patients with limited bone loss and trochanteric or subtrochanteric lesions. Open reduction and internal fixation (ORIF) was typically used in patients with small focal lesions around the trochanteric area

27 Complications After Surgery For Proximal Femoral Metastasis Ten orthopaedic oncologists performed 372 of 417 procedures (89%). These surgeons had a median of 10 years (range, 1 to 40 years) of experience at the time of the procedure. The remaining 45 procedures were performed by 34 trauma or arthroplasty surgeons who were not supervised by orthopaedic oncologists. No difference was found in reoperation rates between orthopaedic oncologists (7.3% [27 of 372]) and other surgeons (6.7% [3 of 45]) by the Fisher exact test (p = 0.999). The procedure was performed at hospital 1 for 237 patients (57%) and at hospital 2 for 180 patients (43%). No difference was found in the number of pathological fractures between institutions (40% [94 of 237] at hospital 1 and 44% [79 of 180] at hospital 2) by the Fisher exact test (p = 0.422). 2 Table 1: Surgical strategy by anatomic location Anatomic location Operation type Implant type n (%) Neck area (I) IMN (n = 28) Intramedullary nailing 28 (47) (n = 59) EPR (n = 28) Unipolar hemiarthroplasty 10 (17) Bipolar hemiarthroplasty 10 (17) Total Hip Arthroplasty 4 (7) Long-stem bipolar hemiarthroplasty 2 (3) MTP - Total Hip Arthroplasty 1 (2) MTP Bipolar hemiarthroplasty 1 (2) ORIF (n = 3) Dynamic Hip Screw 3 (5) Trochanteric area (II) IMN (n = 146) Intramedullary nailing 146 (72) (n = 203) EPR (n = 30) MTP Bipolar hemiarthroplasty 15 (7) Long-stem bipolar hemiarthroplasty 6 (3) MTP - Total Hip Arthroplasty 3 (1) Total Hip Arthroplasty 2 (1) Bipolar hemiarthroplasty 2 (1) Long-stem Total Hip Arthroplasty 1 (0) Unipolar hemiarthroplasty 1 (0) ORIF (n = 27) 1 Plate with screws in head/neck area 9 (4) Dynamic Hip Screw 18 (9) Subtrochanteric area (III) IMN (n = 128) Intramedullary nailing 128 (83) (n = 155) EPR (n = 12) MTP Bipolar hemiarthroplasty 11 (7) Long-stem bipolar hemiarthroplasty 1 (1) ORIF (n = 15) 1 Plate with screws in head/neck area 8 (5) 2 Plates with screws in head/neck area 1 (1) 1 Plate 1 (1) Long-plate with Dynamic Hip Screw 5 (3) n = 417 IMN = intramedullary nailing, EPR = endoprosthetic reconstruction, ORIF = Open Reduction and Internal Fixation, MTP = Modular Tumor Prosthesis 25

28 Chapter 2 Most patients with femoral neck lesions (area I) were treated with endoprosthetic reconstruction (47% [28 of 59]) or non-cemented intramedullary nailing (47% [28 of 59]). The remaining three patients (5.1%) with femoral neck lesions were treated with ORIF, with cement packing after curettage of the lesion in two of those patients (Figure 1 and Table 1). Twenty-six of the 28 prostheses (93%) were cemented. Most patients with trochanteric lesions (area II) were treated with intramedullary nailing (72% [146 of 203]; 145 non-cemented, 1 cemented). The remaining patients with trochanteric lesions were treated with endoprosthetic reconstruction (15% [30 of 203], all cemented) or ORIF (13% [27 of 203]). In 16 of the 27 ORIF procedures (59%), cement was used after curettage of the lesion. Most patients with subtrochanteric lesions (area III) were treated with intramedullary nailing (83% [128 of 155]; 122 non-cemented, 6 cemented). The remaining patients with subtrochanteric lesions were treated with ORIF (10% [15 of 155]) or endoprosthetic reconstruction (8% [12 of 155], all cemented). In 9 of the 15 ORIF procedures (60%), cement was used after curettage of the lesion. Postoperative care and rehabilitation varied among patients depending on the severity of the underlying disease. However, immediate and unrestricted postoperative weight bearing was allowed in most patients. Outcome Measures The primary outcome measures were reoperation and 30-day systemic complications. We included any reoperation described in the medical record, but only the first reoperation was accounted for in the analyses. Two research fellows (S.J.J. and J.T.P.K.) independently reviewed all reports of subsequent surgeries to capture reoperations. We included the following systemic complications that occurred 30 days postoperatively: pneumonia, pulmonary embolism, fat/cement embolism, myocardial infarction, sepsis, and intraoperative death. All medical records of patients with an ICD-9-CM code for any of these complications (Appendix 1) were flagged and subsequently reviewed independently by two research fellows (S.J.J. and J.T.P.K.) to assess whether the infection fulfilled the predefined criteria: a diagnosis of pneumonia is based on symptoms consistent with pneumonia, chest radiographs, and a positive sputum culture or empiric start of antibiotics; a pulmonary embolism is based on a CT or ventilation/perfusion scan plus symptoms; a diagnosis of myocardial infarction is based on electrocardiography or echocardiography plus symptoms; and sepsis is defined as systemic inflammatory response syndrome requiring intensive care admission with a positive culture. Secondary outcome measures were total estimated blood loss, anesthesia time, duration of hospital admission, and 30-day survival. These outcome measures were derived from medical records. In addition, we used the Social Security Death Index to establish date of death. 13 Data on age, body mass index (BMI), comorbidity status, tumor type, preoperative white blood cell count and hematocrit level, sex, fracture type and area, visceral and bone 26

29 Complications After Surgery For Proximal Femoral Metastasis metastases, previous local radiation therapy, and previous systemic therapy were derived from medical records. We reviewed radiographs to assess the location of the lesion when surgical or radiologic reports were unclear or when the lesion was located in the femoral neck, subtrochanteric area, or proximal shaft. We used an algorithm based on the ICD-9- CM code to assess the modified Charlson Comorbidity Index, an index ranging from 0 to 24 based on 12 weighted comorbidities (Appendix 2). 14,15 The modified Bauer score was used as a surrogate for cancer status. 16,17 The Bauer score is commonly used to estimate life expectancy in patients with bone metastases and is a composite of four prognostic factors: (1) no visceral metastases, (2) no lung cancer, (3) the presence of breast cancer, kidney cancer, multiple myeloma, or lymphoma, and (4) the presence of one solitary bone metastasis. The score ranges from 0 to 4, with a higher score indicating relatively better prognosis. 2 Followup Three hundred forty-seven (83%) patients were followed for 3 months or until death at 3 months. At 3 months, 241 patients (58%) were alive, 106 patients (25%) were deceased, and 70 patients (17%) were lost to followup. Three hundred forty-five patients (83%) were followed for 1 year or until death at 1 year. At 1 year postoperatively, 105 patients (25%) were alive, 240 patients (58%) were deceased, and 72 patients (17%) were lost to followup. The median followup was 4 months (range, 0 to 144 months), primarily because of the poor survival rate. Statistical Analysis Categorical variables are presented with frequencies and percentages. Continuous variables are presented as median with interquartile range because inspection of histograms suggested non-normality. Baseline characteristics were compared among implants using the Fisher exact test for categorical variables and the Kruskal-Wallis test for continuous variables. The Levene test indicated homogeneity of variances. Patients with missing values for one of the variables (BMI, 21% [87 of 417]; white blood cell count, 5% [19 of 417]; hematocrit level, 3% [11 of 417]) were not included in the respective baseline analyses. Log-rank analysis was used to compare the reoperation rate among surgical strategies. Reasons for reoperation were subdivided into fixation failure (i.e. nonunion, implant fracture, implant loosening, tumor progression), deep infection, and tumor progression, and log-rank analyses were used to compare these reasons among surgical strategies. Two patients died intraoperatively and were not included in the log-rank analyses. No left censoring occurred, and right censoring was assumed to be non-informative. Visual inspection of log-log plots suggested no serious violation of the proportional hazards assumption

30 Chapter 2 The Fisher exact test was used to compare 30-day systemic complication rates among surgical strategies. Secondary outcome measures were compared among surgical strategies with the Kruskal-Wallis test for estimated blood loss, anesthesia time, and duration of hospital stay and with the Fisher exact test for 30-day survival. Patients with missing values for estimated blood loss (8% [32 of 417]) and anesthesia time (16% [66 of 417]) were excluded from the analyses. The Levene test indicated a violation of the assumption of equality of variances for blood loss but not for anesthesia time or duration of hospital stay. We therefore reported the p value from a Kruskal-Wallis test on the log of blood loss, which resolved the inequality of variance. Additionally, we assessed risk factors for systemic complications. Variables were selected on the basis of previous studies or theoretic association with systemic complications. 19 Exploratory bivariate logistic regression analysis was used to test associations of the variables with systemic complications. Multivariable logistic regression analysis was used to further explore the cause-effect relationship and adjust for sex and age. All statistical analyses were performed using Stata 14.0 (StataCorp LP, College Station, TX, USA). A two-tailed p value less than 0.05 was considered statistically significant. RESULTS Baseline Characteristics Among the 417 patients, 163 (39%) were men. The median age was 62 years (range, 23 to 94 years). There were 173 pathological fractures (41%) and 244 impending fractures (59%). Breast (30% [125 of 417]) and lung (23% [95 of 417]) tumors were most common (Appendix 3). The median survival was 8 months. In comparing baseline characteristics among surgical strategies, we found that pathological fractures were more common in the endoprosthetic reconstruction group (79%) than in the intramedullary nailing and ORIF groups (33% and 40% respectively; p < 0.001); the presence of multiple bone metastases was more common in the intramedullary nailing group (85%) than in the endoprosthetic reconstruction and ORIF groups (74% and 73% respectively; p = 0.023); and previous systemic therapy was more common in the intramedullary nailing group (73%) than in the endoprosthetic reconstruction and ORIF groups (51% and 47% respectively; p < 0.001; Table 2). Anatomic location (p < 0.001; Table 2) and tumor type (p = 0.014; Figure 2) also varied among surgical strategies. Age, sex, BMI, comorbidity and cancer status, white blood cell count, hematocrit level, presence of visceral metastases, and previous radiation therapy did not differ among surgical strategies (Table 2). 28

31 Complications After Surgery For Proximal Femoral Metastasis Table 2: Baseline characteristics of patients by surgical strategy Intramedullary nailing (n = 302) Median (Interquartile range) Endoprosthetic reconstruction (n = 70) Median (Interquartile range) ORIF (n = 45) Median (Interquartile range) n = 417 p value Age in years 62 (52-70) 63 (55-72) 65 (54-75) 0.44 Body mass index* 26 (23-30) 24 (27-29) 25 (23-30) 0.90 Modified CCI 6 (6-8) 6 (6-8) 7 (6-8) 0.94 Modified Bauer score 2 (1-3) 2 (2-3) 2 (1-3) 0.17 WBC count (in K/uL)* 8.6 ( ) 9.2 ( ) 10.8 ( ) 0.15 Hematocrit (in %)* 33 (30-36) 33 (30-34) 34 (31-37) n (%) n (%) n (%) Men 114 (38) 32 (46) 17 (38) 0.46 Pathological fracture 100 (33) 55 (79) 18 (40) <0.001 Anatomical location Neck area (below isthmus) 28 (9) 28 (40) 3 (7) Trochanteric area 146 (48) 30 (43) 27 (60) Subtrochanteric area 128 (42) 12 (17) 15 (33) <0.001 Visceral metastases 147 (49) 32 (46) 23 (51) 0.84 Multiple bone metastases 258 (85) 52 (74) 33 (73) Previous local radiotherapy 63 (21) 15 (21) 7 (16) 0.72 Previous systemic therapy 220 (73) 36 (51) 21 (47) <0.001 ORIF = Open Reduction and Internal Fixation, CCI = Charlson Comorbidity Index, WBC = White blood cell. bold font indicates a significant difference (two-tailed p value below 0.05) *Body mass index was available in 330 (79%) of the patients: 259 (86%) in the intramedullary nailing group, 45 (64%) in the endoprosthetic reconstruction group, and 26 (58%) in the ORIF group. White blood cell count was available in 398 (95%) of the patients: 286 (95%) in the intramedullary nailing group, 69 (99%) in the endoprosthetic reconstruction group, 43 (96%) in the ORIF group. Hematocrit was available in 406 (97%) of the patients: 294 (97%) in the intramedullary nailing group, 69 (99%) in the endoprosthetic reconstruction group, 43 (96%) in the ORIF group. Reoperations The overall reoperation rate was 7.2% (30 of 417 patients) and did not differ among surgical strategies (5.3% [16 of 302] after intramedullary nailing, 11% [8 of 70] after endoprosthetic reconstruction, and 13% [6 of 45] after ORIF; p = 0.134; Figure 3). Reoperation for failure of fixation was highest after ORIF (13% [6 of 45] versus 3.0% [9 of 302] after intramedullary nailing and zero [0 of 70] after endoprosthetic reconstruction; p < 0.001; Figure 4). The rate of reoperation for deep infection was highest after endoprosthetic reconstruction (8.6% [6 of 70] versus 2.0% [6 of 302] after intramedullary nailing and zero [0 of 45] after ORIF; 29

32 Chapter 2 IMN EPR ORIF Percent Breast Lung Myeloma Prostate RCC Other Figure 2: Bar graph depicting tumor distribution among surgical strategies (p = 0.014, by chi-square test). Figure 3: Kaplan-Meier failure plot demonstrating the probability of reoperation for any reason among surgical strategies (log-rank analysis, p = 0.134). The probability of reoperation was 1.6% at 3 months (95% confidence interval [CI], 0.60% to 4.2%) and 4.4% at 12 months (95% CI, 2.1% to 8.9%) for intramedullary nailing; 6.8% at 3 months (95% CI, 2.6% to 17%) and 12% at 12 months (95% CI, 5.4% to 25%) for endoprosthetic reconstruction; and 5.4% at 3 months (95% CI, 1.4% to 20%) and 12% at 12 months (95% CI, 3.5% to 35%) for open reduction and internal fixation. 30

33 Complications After Surgery For Proximal Femoral Metastasis 2 Figure 4: Kaplan-Meier failure plot demonstrating the probability of reoperation for failure of fixation (nonunion, implant fracture, implant loosening, and tumor progression) among surgical strategies (log-rank analysis, p < 0.001). The probability of reoperation for failure of fixation was 0.47% at 3 months (95% confidence interval [CI], 0.07% to 3.3%) and 2.6% at 12 months (95% CI, 0.95% to 7.0%) for intramedullary nailing; 0% at 3 and 12 months for endoprosthetic reconstruction; and 5.4% at 3 months (95% CI, 1.4% to 20%) and 12% at 12 months (95% CI, 3.5% to 35%) for open reduction and internal fixation. p = 0.010; Figure 5). The rate of reoperation for tumor progression did not differ among surgical strategies (0.66% [2 of 302] after intramedullary nailing, 2.2% [1 of 45] after ORIF, and zero [0 of 70] after endoprosthetic reconstruction; p = 0.366; Figure 6). The number of patients who underwent reoperation increased steadily over time, with the exception of reoperations for deep infections in the endoprosthetic reconstruction group, all of which occurred 4 months postoperatively (Appendix 4). 30-Day Systemic Complications The study identified 40 patients (9.6%) with 46 instances of complications: pneumonia (5.3% [22 of 417]), pulmonary embolism (2.2% [9 of 417]), sepsis (1.7% [7 of 417]), myocardial infarction (1.2% [5 of 417]), fat embolism (0.24% [1 of 417] during cementation of a modular prosthesis), and intraoperative death (0.48% [2 of 417]). Both intraoperative deaths occurred during endoprosthetic reconstructions with cemented femoral implants (a unipolar hemiarthroplasty and a modular bipolar tumor hemiarthroplasty) and were attributed to pulmonary fat/cement embolisms, although the cause of death was not confirmed. Overall 30-day systemic complication rates did not differ among surgical strategies (8.3% [25 of 302] after intramedullary nailing, 14% [10 of 70] after endoprosthetic reconstruction, and 11% [5 of 45] after ORIF; p = 0.268). Looking specifically at all pulmonary 31

34 Chapter 2 Figure 5: Kaplan-Meier failure plot demonstrating the probability of reoperation for deep infection among surgical strategies (log-rank analysis, p = 0.010). The probability of reoperation for deep infection was 1.1% at 3 months (95% confidence interval [CI], 0.37% to 3.5%) and 2.1% at 12 months (95% CI, 0.71% to 6.0%) for intramedullary nailing; 6.8% at 3 months (95% CI, 2.6% to 17%) and 12% at 12 months (95% CI, 5.4% to 25%) for endoprosthetic reconstruction; and 0% at 3 and 12 months for open reduction and internal fixation. Figure 6: Kaplan-Meier failure plot demonstrating the probability of reoperation for tumor progression among surgical strategies (log-rank analysis, p = 0.366). The probability of reoperation for tumor progression was 0% at 3 months and 1.3% at 12 months (95% confidence interval [CI], 0.32% to 5.1%) for intramedullary nailing; 0% at 3 and 12 months for endoprosthetic reconstruction; and 0% at 3 months and 6.7% at 12 months (95% CI, 0.97% to 39%) for open reduction and internal fixation. 32

35 Complications After Surgery For Proximal Femoral Metastasis embolic events, we did not find a difference among surgical strategies (2.3% [7 of 302] after intramedullary nailing, 5.7% [4 of 70] after endoprosthetic reconstruction, and 2.2% [1 of 45] after ORIF; p = 0.296). Secondary Outcome Measures Blood loss (p < 0.001) and anesthesia time (p = 0.003) were both highest for endoprosthetic reconstruction rather than for intramedullary nailing or ORIF, whereas duration of hospital admission (p = 0.017) was longest for ORIF (Table 3). No difference was found in 30-day survival among the three surgical strategies (p = 0.99). Median 30-day survival for each treatment strategy was 93%. 2 Table 3: Secondary outcome measures by surgical strategy Intramedullary nailing (n = 302) Median (Interquartile range) Endoprosthetic reconstruction (n = 70) Median (Interquartile range) ORIF (n = 45) Median (Interquartile range) n = 417 p value Estimated blood loss* (in ml) 200 ( ) 400 ( ) 275 ( ) <0.001 Anesthesia time* (in minutes) 181 ( ) 222 ( ) 178 ( ) <0.001 Duration of hospital admission (in days) 6 (4-10) 7 (5-10) 8 (5-11) ORIF = Open Reduction Internal Fixation, bold font indicates a significant difference (two-tailed p value below 0.05) *Estimated blood loss was available in 385 (92%) of the patients: 290 (96%) in the intramedullary nailing group, 57 (81%) in the endoprosthetic reconstruction group, and 38 (84%) in the ORIF group. Anesthesia time was available in 351 (84%) of the patients: 258 (85%) in the intramedullary nailing group, 60 (86%) in the endoprosthetic reconstruction group, 33 (73%) in the ORIF group Risk Factors For Systemic Complications Exploratory bivariate analysis revealed an association of age, modified Charlson Comorbidity Index, and modified Bauer score with systemic complications (Table 4). We included only the modified Bauer score in the multivariable model accounting for age and sex because we considered the comorbidity index to be an intermediary variable rather than a true confounder. This model, which accounted for male sex (odds ratio [OR], 1.36; 95% CI, 0.69 to 2.68; p = 0.372) and age (OR, 1.03; 95% CI, 0.99 to 1.05; p = 0.067), demonstrated an effect of the modified Bauer score on postoperative systemic complications (OR, 0.67; 95% CI, 0.49 to 0.92; p = 0.014; Hosmer-Lemeshow goodness-of-fit test, p = 0.56; c-statistic, 0.67). This finding suggests that an increase in modified Bauer score indicating better prognosis is associated with a decreased chance of systemic complication. 33

36 Chapter 2 Table 4: Bivariate analysis of factors associated with 30-day systemic complications Unadjusted Odds Ratio (95% confidence interval) p value Age 1.03 ( ) Men 1.64 ( ) Body mass index* 0.97 ( ) Modified CCI 1.22 ( ) Modified Bauer score 0.68 ( ) WBC count (in K/uL)* 1.04 ( ) Hematocrit (in %)* 1.03 ( ) Pathological fracture 1.18 ( ) Previous systemic therapy 0.83 ( ) n = 417 CCI = Charlson Comorbidity Index, WBC = White blood cell. bold font indicates a significant difference (two-tailed p value below 0.05). *Body mass index was available in 330 (79%) patients, white blood cell count in 398 (95%), and hematocrit in 406 (97%). DISCUSSION In the surgical management of metastatic femoral lesions, the goal is a single procedure that allows for immediate weight bearing and remains stable throughout the patient s lifetime while minimizing complications. 12 We assessed differences in reoperation rate, 30-day systemic complication rate, blood loss, anesthesia time, duration of hospital admission, and 30-day survival among surgical strategies in the largest (to the best of our knowledge) cohort of patients surgically treated for proximal femoral metastases. The relatively large sample size also allowed us to compare reasons for reoperation among surgical strategies. We found that the overall reoperation rate did not differ among intramedullary nailing, endoprosthetic reconstruction, and ORIF. However, when patients were subdivided by reasons for reoperation, endoprosthetic reconstruction was found to be the most durable over time (no failures) but most likely to require (early) reoperation for infection. This finding emphasizes the importance of estimating life expectancy when weighing treatment strategies in these terminally ill patients. 16,20,21 No difference was found in 30-day systemic complication rates among surgical strategies, but cancer status was identified as a risk factor for development of a systemic complication. This study has several limitations. First, no uniform criteria were used in the decision to perform surgical treatment, and many surgeons were involved. However, most of the procedures (89%) were performed by 10 orthopaedic oncologists, and we found no difference in reoperation rate between orthopaedic oncologists and other subspecialty surgeons. Selection bias could not have been eliminated from the study and is an important 34

37 Complications After Surgery For Proximal Femoral Metastasis limitation. We explored differences in baseline characteristics among surgical strategies to understand possible confounding and found that, in patients with pathological fractures and more proximal lesions, endoprosthetic reconstruction was used more often than intramedullary nailing or ORIF. These factors could have compromised the comparison of surgical strategies. Unfortunately, the low number of reoperations and complications did not allow correction for potential confounders in our analyses. 22 Second, although we found no difference in overall reoperation rate or 30-day complication rate among implant types, a larger sample size might have revealed a significant difference. A post-hoc power analysis for reoperation rate comparing intramedullary nailing with ORIF (chosen because the effect size [i.e. difference] between these groups was largest) demonstrated that, to achieve a power of 0.80, we would have needed slightly more patients (i.e. 334 in the intramedullary nailing group and 50 in the ORIF group) to demonstrate a significantly higher overall reoperation rate in the ORIF group. The current sample size achieved a power of A post-hoc power analysis for the 30-day systemic complication rate comparing intramedullary nailing with endoprosthetic reconstruction (chosen because the effect size [i.e. difference in proportion: 6.0%] between these groups was largest) demonstrated that, to achieve a power of 0.80, we would have needed substantially more patients (i.e. 1,166 in the intramedullary nailing group and 271 in the endoprosthetic reconstruction group) to demonstrate a significantly higher complication rate in the endoprosthetic reconstruction group. The current sample size achieved a power of Third, we did not account for postoperative radiation therapy because it was not always provided at the included institutions, nor did we account for preoperative ambulatory status because the Eastern Cooperative Oncology Group performance status was available for only a limited number of patients. Fourth, diagnostic and billing codes were used to flag potentially eligible patients, flag systemic complications, and determine comorbidity status. Because inaccuracies in coding could have occurred, we reviewed all electronic medical records to assess whether the flagged patients fulfilled eligibility criteria and whether the flagged systemic complications met definitions. This methodology resolved the issue of overcoding but did not address the possibility of undercoding. However, we think that undercoding is relatively uncommon and is independent of the surgical strategy and, therefore, that it did not influence the results. Fifth, poor patient survival affected followup duration in our study, as in other studies of patients with bone metastases. Our median followup was short (4 months), but 83% of the patients were available for followup or were deceased at 3 and 12 months postoperatively. Implant survival in this patient population should be considered relative to the patients short life expectancy. We accounted for variation in followup duration among surgical strategies by using log-rank analysis for reoperations and using a relatively short time frame for systemic complications. Sixth, we assessed nine risk factors for systemic complications in an exploratory fashion, which increases the chance of a type I error. Future research should assess how the identified risk factor modified 2 35

38 Chapter 2 Bauer score relates to systemic complications before it is used for risk stratification. Seventh, we were unable to account for the possible scenario in which a patient s illness precluded a necessary reoperation, nor were we able to account for treatment performed at a different institution. The overall reoperation rate in our study (7.2%) is consistent with that reported in previous studies (6.4 to 10.3%). 3,4 Our results are also consistent with those of previous studies that demonstrated that endoprosthetic reconstruction is more durable (i.e. resulted in fewer failures) than intramedullary nailing or ORIF. Steensma et al 3 found that 0.5% of endoprosthetic reconstructions (1 of 197) required implant exchange after failure compared with 6.1% of intramedullary nailing procedures (5 of 82) and 42% of ORIF procedures (8 of 19) (p < 0.01). Wedin and Bauer 4 demonstrated an 8.3% rate of failure requiring reoperation after endoprosthetic reconstruction (9 of 109) compared with 13% (3 of 24) after intramedullary nailing and 25% (3 of 12) after the use of a dynamic hip screw. However, no formal statistical analyses were performed in the latter study. Prostheses are associated with additional complications, such as higher rates of infection (8.6% in our study) and dislocation. Wedin and Bauer 4 reported deep infections after endoprosthetic reconstruction (3.7% [4 of 109]) but none after intramedullary nailing or ORIF; however, no statistical analysis was conducted. No infections requiring reoperation were reported by Steensma et al. 3 We found one instance of dislocation after endoprosthetic reconstruction that required open reduction (1.4%). Steensma et al reported open reduction for dislocation in 5 of 197 endoprosthetic reconstructions (2.5%), and Wedin and Bauer 4 reported that 3 of 109 endoprosthetic reconstructions (2.8%) required reoperation for dislocation. Our results and these studies demonstrate that infection is more common after endoprosthetic reconstruction than after other treatment strategies. The overall systemic complication rate in our study (9.6%) is in line with that reported in previous studies (1.4% to 12.5%), and we found no difference among surgical strategies. 4,5,23 We demonstrated that poor cancer status independent of age and sex is a risk factor for the development of systemic complications. A study of 5,716 patients undergoing musculoskeletal tumor surgery demonstrated that older age, male sex, blood transfusion, longer duration of anesthesia, and higher Charlson Comorbidity Index score were associated with postoperative complications. 19 The modified Bauer score is commonly used to estimate life expectancy in patients with bone metastases, but its association with systemic complications has not been shown. 16,17 This risk factor could be used to anticipate postoperative complications, stratify patients by risk, and, if possible, optimize a patient s health before surgical treatment. Postoperative complications, such as infection, can hamper a patient s ability to resume or begin medical treatment (e.g. chemotherapy) for the underlying malignancy, potentially affecting the patient s survival. We found that blood loss was almost twice as high and anesthesia time was approximately 40 minutes longer in patients treated with endoprosthetic reconstruction than in 36

39 Complications After Surgery For Proximal Femoral Metastasis patients treated with intramedullary nailing or ORIF. These findings are comparable to those of previous studies demonstrating longer duration of surgery and increased blood loss after hip arthroplasty than after internal fixation for non-pathological femoral neck fractures. 24,25 However, the difference in blood loss in our study might be explained in part by the difference in vascularity of the tumors or by the proportion of fractures because prostheses were more commonly used in patients with renal cell carcinoma and pathological fractures than in patients with other conditions. 26 These factors should be considered when planning surgical treatment because patients might be too ill to withstand lengthy procedures or substantial blood loss. Acknowledging the limitations of this retrospective study, we found that ORIF and intramedullary nail fixation fail more often than does endoprosthetic reconstruction, whereas deep infections are most common following endoprosthetic reconstruction and often occur within the first months after surgical treatment. These implant-specific complications and their timing should be considered in the choice of surgical strategy. The findings also emphasize the importance of incorporating estimated life expectancy into surgical decision making. The risk factor of poor cancer status could be used to predict systemic complications. 2 REFERENCES 1. Bickels J, Dadia S, Lidar Z. Surgical management of metastatic bone disease. J Bone Joint Surg Am. Jun 2009; 91(6): Ratasvuori M, Wedin R, Keller J, et al. Insight opinion to surgically treated metastatic bone disease: Scandinavian Sarcoma Group Skeletal Metastasis Registry report of 1195 operated skeletal metastasis. Surg Oncol. 2013; 22: Steensma M, Boland PJ, Morris CD, Athanasian E, Healey JH. Endoprosthetic treatment is more durable for pathologic proximal femur fractures. Clin Orthop Relat Res. 2012; 470: Wedin R, Bauer HC. Surgical treatment of skeletal metastatic lesions of the proximal femur: endoprosthesis or reconstruction nail? J Bone Joint Surg Br. 2005; 87: Alvi HM, Damron TA. Prophylactic stabilization for bone metastases, myeloma, or lymphoma: Do we need to protect the entire bone? Clin Orthop Relat Res. 2013; 471: Talbot M, Turcotte RE, Isler M, Normandin D, Iannuzzi D, Downer P. Function and health status in surgically treated bone metastases. Clin Orthop Relat Res. 2005; 438: Ward WG, Holsenbeck S, Dorey FJ, Spang J, Howe D. Metastatic disease of the femur: surgical treatment. Clin Orthop Relat Res. 2003: S Forsberg JA, Wedin R, Bauer H. Which implant is best after failed treatment for pathologic femur fractures? Clin Orthop Relat Res. Mar 2013; 471(3): Steensma M, Healey JH. Trends in the surgical treatment of pathologic proximal femur fractures among Musculoskeletal Tumor Society members. Clin Orthop Relat Res. Jun 2013; 471(6):

40 Chapter Bryant D, Havey TC, Roberts R, Guyatt G. How many patients? How many limbs? Analysis of patients or limbs in the orthopaedic literature: a systematic review. J Bone Joint Surg Am. Jan 2006; 88(1): Weiss RJ, Ekstrom W, Hansen BH, et al. Pathological subtrochanteric fractures in 194 patients: a comparison of outcome after surgical treatment of pathological and non-pathological fractures. J Surg Oncol. 2013; 107: Quinn RH, Randall RL, Benevenia J, Berven SH, Raskin KA. Contemporary management of metastatic bone disease: tips and tools of the trade for general practitioners. J Bone Joint Surg Am. Oct ; 95(20): Huntington JT, Butterfield M, Fisher J, Torrent D, Bloomston M. The Social Security Death Index (SSDI) most accurately reflects true survival for older oncology patients. Am J Cancer Res. 2013; 3(5): Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol. Jun 1992; 45(6): Quan H, Li B, Couris CM, et al. Updating and validating the Charlson comorbidity index and score for risk adjustment in hospital discharge abstracts using data from 6 countries. Am J Epidemiol. Mar ; 173(6): Bauer HC, Wedin R. Survival after surgery for spinal and extremity metastases. Prognostication in 241 patients. Acta Orthop Scand. Apr 1995; 66(2): Leithner A, Radl R, Gruber G, et al. Predictive value of seven preoperative prognostic scoring systems for spinal metastases. Eur Spine J. Nov 2008; 17(11): Allison PD. Survival Analysis Using SAS: A Practical Guide. 2 ed Ogura K, Yasunaga H, Horiguchi H, Fushimi K, Kawano H. What is the effect of advanced age and comorbidity on postoperative morbidity and mortality after musculoskeletal tumor surgery? Clin Orthop Relat Res. Dec 2014; 472(12): Nathan SS, Healey JH, Mellano D, et al. Survival in patients operated on for pathologic fracture: implications for end-of-life orthopedic care. J Clin Oncol. Sep ; 23(25): Janssen SJ, van der Heijden AS, van Dijke M, et al Marshall Urist Young Investigator Award: Prognostication in Patients With Long Bone Metastases: Does a Boosting Algorithm Improve Survival Estimates? Clin Orthop Relat Res. Oct 2015; 473(10): Katz MH. Multivariable analysis: a primer for readers of medical research. Ann Intern Med. Apr ; 138(8): Sarahrudi K, Greitbauer M, Platzer P, Hausmann JT, Heinz T, Vecsei V. Surgical treatment of metastatic fractures of the femur: a retrospective analysis of 142 patients. J Trauma. 2009; 66: Chammout GK, Mukka SS, Carlsson T, Neander GF, Stark AW, Skoldenberg OG. Total hip replacement versus open reduction and internal fixation of displaced femoral neck fractures: a randomized long-term follow-up study. J Bone Joint Surg Am. Nov ; 94(21): Tidermark J, Ponzer S, Svensson O, Soderqvist A, Tornkvist H. Internal fixation compared with total hip replacement for displaced femoral neck fractures in the elderly. A randomised, controlled trial. J Bone Joint Surg Br. Apr 2003; 85(3): Pazionis TJ, Papanastassiou ID, Maybody M, Healey JH. Embolization of hypervascular bone metastases reduces intraoperative blood loss: a case-control study. Clin Orthop Relat Res. Oct 2014; 472(10):

41 Complications After Surgery For Proximal Femoral Metastasis Appendix 1: International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) Codes Used to Flag Systemic Complications Complication Codes Flagged Pulmonary embolism , , 451.2, , 451.9, , , 453.8, 453.9, 453, 453.4, 415.1, , , 453.9, , , , , 416.2, 444.1, , , , , 444.9, , , , , 453.2, 453.3, , , , , , , 453.6, , , , , , , , , , , , , , , , , Pneumonia 481, 482.0, 482.1, 482.2, , , , , , , , , , , , , 482.9, 483.8, 485, 486, 495.7, 507 Sepsis 038.0, , , , 038.3, , , , , , , 038.8, 038.9, Myocardial infarction 427.5, , , , , , , , , , , , , , , , , , , , Appendix 2: Modified Charlson Comorbidity Index Algorithm Based on International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) Codes Comorbidity Weight* Codes AIDS/HIV Any malignancy, including leukemia and lymphoma* , , , , , , , 189.9, , , , , , , , , , , 230.8, 231.2, 231.9, , 233.0, 233.1, , , 233.4, 233.7, , 235.7, 235.8, 236.2, 236.4, 236.5, , , 237.6, , , , 239.6, 239.7, , Chronic pulmonary disease , 416.9, , , , 506.4, 508.1, Congestive heart failure , , , , , , , , , , , Dementia 2 290, 290.0, 290.3, , 294.1, , Diabetes with chronic complications , Hemiplegia or paraplegia , , Metastatic solid tumor* , , , Mild liver disease* , , , , , , 070.6, 070.9, 570, 570.1, 573.3, 573.4, 573.8, 753.9, V42.7 Moderate or severe liver disease* , Renal disease , , , , , , , , , , , 588.0, V42.0, V45.1, V56-V56.8 Rheumatologic disease , , , 714.8, 725 *The following comorbidities were mutually exclusive: mild liver disease and moderate or severe liver disease, and any malignancy and metastatic solid tumor. For example, a patient with a metastatic solid tumor received 6 points total (not 6 points for metastatic solid tumor and 2 points for any malignancy). 39

42 Chapter 2 Appendix 3: Tumor Distribution Tumor Type Bone metastasis n (%) Breast 125 (30) Lung 95 (23) Prostate 29 (7) Kidney 26 (6) Melanoma 13 (3) Unknown 13 (3) Esophageal 12 (3) Adenocarcinoma of unknown origin 10 (2) Thyroid 9 (2) Other* 28 (7) Primary bone tumor (myeloma) 57 (14) N=417 *Other primary tumor types were colorectal (n = 7), bladder (n = 4), hepatocellular (n = 5), vulvar (n = 3), neuroendocrine (n = 3), cutaneous squamous cell carcinoma (n = 2), uterus (n = 2), pancreas (n = 1), and ovarian (n = 1). Appendix 4: Reasons for Reoperation and Surgical Strategy Patient Sex (Age in Yr) Fracture Type Location Primary Tumor Surgical Strategy Reason for Reoperation Months to Reoperation (Procedure) F (62) Impending Subtrochanteric Breast IMN Deep infection 1 (I & D) F (61) Pathological Trochanteric Breast IMN Deep infection 1 (I & D) F (56) Impending Neck Breast IMN Deep infection 1 (I & D) M (63) Impending Trochanteric RCC IMN Peri-implant fracture after a minor fall M (68) Impending Trochanteric Bladder IMN Painful recurrence F (65) Pathological Trochanteric Myeloma IMN Nonunion, bone collapse, blade prominence M (61) Impending Neck Lung IMN Soft-tissue painful recurrence F (76) Pathological Subtrochanteric Breast IMN Deep infection and helical blade prominence F (30) Impending Subtrochanteric Breast IMN Implant fracture (distal screw), nonunion 2 (distal screw replacement) 4 (MTP THA) 4 (long-stem THA) 4 (resection of recurrence) 7 (I & D and screw and blade replacement) 14 (new IMN) 40

43 Complications After Surgery For Proximal Femoral Metastasis Appendix 4: Reasons for Reoperation and Surgical Strategy (continued) Patient Sex (Age in Yr) Fracture Type Location Primary Tumor Surgical Strategy Reason for Reoperation F (58) Pathological Subtrochanteric Breast IMN Implant fracture, nonunion M (58) Pathological Subtrochanteric Myeloma IMN Implant fracture (distal screw), nonunion Months to Reoperation (Procedure) 25 (blade plate with intramedullary plate) 36 (New IMN and bone graft) M (79) Pathological Trochanteric Myeloma IMN Deep infection 42 (I & D) F (44) Pathological Subtrochanteric Breast IMN Painful bursitis 43 (trimming of intramedullary nail) M (78) Pathological Neck Prostate IMN Nonunion, bone collapse, blade prominence F (37) Impending Subtrochanteric Breast IMN Implant fracture, nonunion F (52) Impending Trochanteric Breast IMN Deep infection, secondary to arm cellulitis F (69) Pathological Subtrochanteric Breast EPR (MTP bipolar hemiarthroplasty) F (57) Impending Neck Adenocarcinoma of unknown origin EPR (long-stem bipolar hemiarthroplasty) M (70) Pathological Trochanteric RCC EPR (MTP bipolar hemiarthroplasty) F (53) Pathological Trochanteric Breast EPR (MTP bipolar hemiarthroplasty) M (56) Pathological Neck RCC EPR (unipolar hemiarthroplasty) M (58) Pathological Neck Lung EPR (MTP bipolar hemiarthroplasty) F (74) Impending Trochanteric Thyroid EPR (MTP bipolar hemiarthroplasty) F (48) Impending Trochanteric Breast EPR (MTP bipolar hemiarthroplasty) Hematoma, deep infection, nerve compression 52 (THA) 55 (new IMN) 58 (removal of intramedullary nail) 1 (I & D) Deep infection 1 (I & D) Deep infection 1 (I & D) Hematoma, deep infection Deep infection Deep infection Dislocation Bursitis and painful exostoses 1 (I & D) 3 (I & D and bipolar head exchange) 4 (I & D and bipolar head exchange) 12 (open reduction) 37 (trimming of exostoses) 2 41

44 Chapter 2 Appendix 4: Reasons for Reoperation and Surgical Strategy (continued) Patient Sex (Age in Yr) Fracture Type Location Primary Tumor Surgical Strategy Reason for Reoperation F (76) Pathological Subtrochanteric Thyroid ORIF (1 plate) Peri-implant fracture Months to Reoperation (Procedure) 1 (MTP bipolar hemiarthroplasty) M (75) Pathological Subtrochanteric Lung ORIF (1 plate) Implant fracture 2 (MTP bipolar hemiarthroplasty) M (64) Impending Trochanteric Esophageal ORIF (DHS) Local recurrence M (46) Pathological Subtrochanteric Myeloma ORIF (DHS) Implant fracture (screws), nonunion F (52) Impending Trochanteric Breast ORIF (DHS) Fixation failure (loosening) F (56) Impending Subtrochanteric Thyroid ORIF (1 plate) Implant fracture, nonunion 8 (MTP bipolar hemiarthroplasty) 20 (new intramedullary nail) 39 (MTP bipolar hemiarthroplasty) 49 (MTP bipolar hemiarthroplasty) F = female, DHS = dynamic hip screw, EPR = endoprosthetic reconstruction, I & D = irrigation and débridement, IMN = intramedullary nailing, M = male, MTP = modular tumor prosthesis, ORIF = open reduction and internal fixation, RCC = renal cell carcinoma, THA = total hip arthroplasty 42

45 CHAPTER 3 Outcome After Fixation Of Metastatic Proximal Femoral Fractures: A Systematic Review Of 40 Studies 3 S.J. Janssen T. Teunis F.J. Hornicek C.N. van Dijk J.A.M. Bramer J.H. Schwab Journal of Surgical Oncology 2016 Sep;114(4): Poster at: European Musculoskeletal Oncology Society Annual Meeting 2016, La Baule, France.

46 Chapter 3 ABSTRACT Objectives To assess: (1) functional outcome, (2) local complication rate, and (3) systemic complication rate after surgery for proximal femoral metastases. These outcomes were compared between three commonly used surgical strategies: endoprosthetic reconstruction, intramedullary nailing, and open reduction and internal fixation (ORIF). Design Systematic review and meta-analysis. Data Sources On September 24 th, 2015, we searched the Pubmed, Embase, and Cochrane libraries for studies (published after 1980) using the keywords: pathologic and fracture and surgery, including synonyms, in title and abstract, without any limits, yielding 7,670 publications. Eligibility Criteria For Selecting Studies Studies reporting on functional outcome or complications after surgery for impending or actual pathological proximal femoral fractures caused by metastatic disease. Exclusion criteria were: case-reports, studies with less than 10 patients within a treatment arm, reviews, letters to the editor, meeting abstracts, technique papers, revision procedures, and indiscernible treatment arms for proximal femoral metastasis. Results All three surgical strategies result in reasonable function on average; however, wide ranges indicate that both poor and good functional levels are obtained. We found that the overall reoperation rate was comparable for endoprosthesis and intramedullary nailing, but was higher after ORIF. Definitions and reporting of systemic complications varied widely among studies. Conclusions This study provides an overview of functional outcome and complications after common surgical strategies for impending and pathological fractures in proximal femoral metastasis. The results reported here can aid patients and their surgeons in the decision for surgical treatment. 44

47 Outcome After Fixation Of Metastatic Proximal Femoral Fractures INTRODUCTION The proximal femur is commonly affected by bone metastases. 1,2 Bone metastases weaken the bone and reduce load bearing capabilities which can lead to pain and eventually a pathological fracture. Surgical treatment is often indicated in case of a pathological or impending fracture and aims to restore function in a single procedure while minimizing the risk of complications. 1-3 Several surgical strategies are practiced for the treatment of proximal femoral fractures caused by bone metastasis. 4 The three mostly used concepts of treatment are: endoprosthetic reconstruction, intramedullary nailing, and open reduction and internal fixation with plate and screws. 1,2,4 The decision about whether to undergo surgery and the choice of surgical strategy is ideally made by the patient together with their doctor. This decision depends on many factors including estimated life expectancy, location and size of the lesion, and the expected functional outcome and complications. A survey among orthopaedic oncologists demonstrated large variation in preference for treatment and previous retrospective studies support the use of all three surgical techniques. 4 Outcomes such as reoperations and complications are relatively rare (<10%) and therefore require large cohorts to adequately compare surgical strategies. This review aims to summarize current literature to help inform patients about expected outcome and compare the outcome between surgical strategies for proximal femoral fractures caused by bone metastasis. Specifically, we asked the following questions: (1) What is the functional outcome after surgery?, and (2) What is the local and systemic complication rate? We compared these outcomes between the three aforementioned surgical strategies. 3 METHODS Article Selection We report our results according to the PRISMA Statement and our review protocol was registered on PROSPERO prior to study selection (#2014:CRD ). 5 On 24 September 2015, we searched the Pubmed, Embase, and Cochrane libraries for studies, published after 1980 using the keywords: ( pathologic* OR impending ) AND ( fracture* ) AND ( surgery OR surgeries OR operation OR operations OR operativ* OR surgical* OR intramedull* OR fixation* OR resection* OR osteosynth* OR endoprosth* OR prosth* OR arthroplas* ) in title and abstract, without any limits, yielding 7,670 publications (Figure 1). Two reviewers (S.J., T.T.) independently screened titles and abstracts and subsequently full texts using predefined criteria. We included studies reporting on functional outcome 45

48 Chapter 3 Figure 1: Flowchart demonstrating articles identified, excluded, and included in the systematic review. Thirty-eight studies were identified by our search, one was found through bibliography screening, and we included our recent retrospective study reviewing the cases at our institution. or complications after surgery for impending or actual pathological proximal femoral fractures caused by metastatic disease. Myeloma and lymphoma were also included as these are commonly grouped with metastases and have similar mechanical implications. 6 We excluded: case-reports, studies with less than 10 patients within a treatment arm, reviews, letters to the editor, meeting abstracts, technique papers, revision procedures, and indiscernible treatment arms for proximal femoral metastasis (e.g. studies that also include primary tumors, other anatomical areas, or revision procedures without reporting results separately). We contacted authors if studies published after the year 2000 included indiscernible treatment arms but were otherwise eligible (20 studies) Only one of these could eventually be included. 8 In case of overlapping cohorts (14 studies), we included the most comprehensive or largest study Authors were contacted if studies 46

49 Outcome After Fixation Of Metastatic Proximal Femoral Fractures published after the year 2000 insufficiently reported outcomes (14 studies). 33,35-38,40-47 Eight authors 36-38,41,42,44,45,47 responded and additional information on outcome was obtained for 4 studies which could then be included. 36,41,42,47 The bibliographies of included studies were checked for publications missed by our search; 1 additional study was found and included (Figure 1). 48 Two reviewers (S.J., T.T.) independently appraised the quality of the studies using predetermined criteria study design, disclosure, direct comparison of surgical techniques, eligibility criteria for patient selection, baseline reporting, outcome reporting, attrition bias and extracted data using a standardized sheet. Discordant judgments were resolved by consensus discussion. 3 Outcome Measures The following outcome measures were collected from the selected studies per treatment: physical function measured by an internationally accepted standardized instrument, local complications, and systemic complications. Seven studies reported on functional outcome using a standardized measure; five used the Musculoskeletal Tumor Society (MSTS) score, one used the Toronto Extremity Salvage Score (TESS), and one used both. The MSTS score is a clinician completed assessment rating six domains, resulting in a score of 0 to 100% with a higher score indicating better function. 49 The TESS is a patient completed questionnaire containing 30 items resulting in a score of 0 to 100, with a higher score indicating better function. 50,51 We only included those local complications that required reoperation and grouped these into: deep infection (e.g. incision and debridement with or without implant removal), fixation failure (including: implant or peri-implant fracture, nonunion, implant loosening, acetabular protrusion of hemiarthroplasty), dislocation (e.g. open reduction with or without component revision), tumor recurrence or progression, and total number of reoperations. We included all reported systemic complications, regardless of treatment consequences. The following data was extracted for all patients with femoral metastases: number of patients and affected femurs, number of reported actual and impending fractures, age, sex, minimum followup in months, percentage oneyear survival, and tumor distribution. We also extracted number of affected femurs, number of pathological and impending fractures, use of cement, and anatomic location of the lesion (head/neck, trochanteric, subtrochanteric, proximal shaft area) per treatment concept. We considered the following surgical techniques (i.e. treatment arms): endoprosthetic reconstruction (including: total hip arthroplasty, hemiarthroplasty, and modular tumor prosthesis), intramedullary nailing, and ORIF (including: plate-screw fixation and dynamic hip screw). 47

50 Chapter 3 Statistical Analysis We reported the average functional outcome score with range and followup per surgical technique. No meta-analysis could be performed on these data because of the variation in outcome measures used, small number of studies, and varying followup. Complication rates are reported as frequencies with percentages per surgical technique. We used a random-effects meta-analysis to generate pooled effect estimates for: (1) the total number of reoperations, (2) deep infections, and (3) fixation failures per surgical technique. The pooled effect estimate is a weighted average of included studies and presented as a percentage with 95% confidence interval (95% CI). Studies that did not report on these specific outcomes were not included in the calculations. We used the Freeman-Tukey double arcsine transformation to obtain 95% confidence intervals that are admissible when the complication rates are close to 0 or 100%. 52,53 The random-effects meta-analysis accounts for heterogeneity in effect estimates across studies (i.e. studies are permitted to have different effects and characteristics) 52. We reported the I-squared statistic; an estimate of the percentage of the variation in the effect estimate due to the heterogeneity across studies. No meta-analysis could be performed for systemic complications because of the low number of studies reporting on this outcome per treatment and substantial variation in definition of systemic complications among studies. All statistical analyses were performed using Stata 14.0 (StataCorp LP, College Station, TX, USA). RESULTS Study Characteristics Forty studies were included (Figure 1); all had a retrospective design. Critical appraisal demonstrated that 32 (80%) studies had clear eligibility criteria and clear methods of selecting patients, leaving the remainder subject to selection bias. Only 11 (28%) studies described a clear definition of the outcomes to be reported, for the remaining studies outcome bias cannot be excluded. Overall, 17 (43%) of the studies were assessed as being at high or unclear risk of attrition bias due to possibly high loss to followup (Figure 2, Appendix 1). Patient Characteristics The 40 studies reported on 3,211 metastatic lesions in the complete femur; average age ranged from 54 to 78 years and 39% (range among studies: 15 to 55%) were men (Table 1). The oneyear survival percentage reported in 23 studies ranged from 0 to 62%. Breast (35%), Lung (15%), Prostate (10%), and Kidney (8.2%) were the most common 48

51 Outcome After Fixation Of Metastatic Proximal Femoral Fractures Prospective Design Disclosure Comparison of Techniques Patient Selection 3 Baseline Outcome No Attrition Bias Percentage Properly reported Figure 2: Overall quality of the included studies. All studies had a retrospective study design. (Appendix 1 includes the critical appraisal per study). primary tumors (Figure 3, Appendix 2). Myeloma and lymphoma accounted for 9.8% of the tumors. The primary tumor type was unknown in 2.4% of the cases. The 40 studies describe 58 treatments including 2,748 proximal femoral metastases: 23 studies reporting on 1,461 endoprostheses (664 [45%] pathological fractures, 347 [24%] impending fractures, and not specified in 450 [31%]); 24 studies reporting on 1,054 intramedullary nails (468 [44%] pathological fractures, 389 [37%] impending fractures, and not specified in 197 [19%] cases); and 11 studies reporting on 233 ORIFs (108 [46%] pathological fractures, 80 [34%] impending fractures, and not specified in 45 [19%] cases). Fracture location was head/neck in 17%, trochanteric/subtrochanteric in 34%, and not specified in 48% of the endoprosthetic reconstructions; head/neck in 3%, trochanteric/ subtrochanteric in 75%, and not specified in 22% of the intramedullary nails; and head/ neck in 3%, trochanteric/subtrochanteric in 55%, and not specified in 43% of the ORIFs. Cement was used in 87% of the endoprosthetic reconstructions (864/998, not specified in 463 cases), in 13% of the intramedullary nails (84/649, not specified in 405 cases), and in 71% of the ORIFs (113/160, not specified in 73 cases) (Appendix 3). Functional Outcome Seven studies reported on functional outcome following endoprosthetic reconstruction: five studies reported the MSTS score in 95 patients; The average MSTS score ranged 49

52 Chapter 3 Table 1: Characteristics of all patients with femoral metastases per study (%) Author, Year Implant Patients Femurs Pathological fractures Median Age Men Minimum followup# Oneyear survival in % Choy et al IMN (100) 66* 8 (42) - 26 Hettwer et al EPR (63) 65* 45 (41) 0 42 Janssen et al EPR, IMN, ORIF (41) (39) 0 58 Arvinius et al IMN (68) 68* 28 (43) 3 - Piccioli et al IMN (100) 61* 42 (53) Shemesh et al IMN (52) 63 8 (38) 4 28 Fakler et al IMN (100) (55) Sorensen et al EPR Weiss et al IMN, EPR (77) (52) 0 33 Asavamongkolgul et al. ORIF (33) (37) 2 42 Harvey et al IMN, EPR (57) 60* 72 (45) C 51 Steensma et al EPR, IMN, ORIF (38) Hattori et al EPR (81) 65 8 (50) 3 62 Parker et al EPR, IMN, ORIF (100) 72* 57 (39) C - Zacherl et al EPR, IMN, ORIF (100) 64* 24 (38) - 31 Potter et al EPR (41) 62* Sarahrudi et al EPR, ORIF (100) (29) 0 17 Chandrasekar et al EPR (69) Selek et al EPR (62) 55* 21 (47) 2 27 Park et al EPR Rethnam et al IMN (82) 72 5 (45) - - Wedin et al EPR, IMN, ORIF (86) (47) C 30 50

53 Outcome After Fixation Of Metastatic Proximal Femoral Fractures Table 1: Characteristics of all patients with femoral metastases per study (%) (continued) Men Minimum followup# Oneyear survival in % Author, Year Implant Patients Femurs Pathological fractures Median Age Datir et al IMN (76) Moholkar et al IMN (54) 66* 12 (25) - - Ramakrishnan et al IMN (18) (43) 7 - Piatek et al IMN (93) Ward et al EPR, ORIF (47) Assal et al IMN (50) 71 3 (25) C 10 Giannoudis et al IMN (77) 68* 9 (30) 2 32 Chien et al EPR (81) 61* 14 (44) 1 20 Algan et al IMN Nargol et al IMN (100) 78 2 (20) 4 0 Rompe et al EPR, ORIF (34) 56* 11 (22) - 50 Broos et al EPR, ORIF (92) 65* 12 (16) - - Korkala et al EPR (100) Borel Rinkes et al EPR (79) 61* 5 (15) 12 - Yazawa et al EPR, IMN Fasano et al IMN (82) 70 5 (45) 7 22 Behr et al IMN, ORIF (46) 70* 14 (29) 7 - Lane et al EPR (75) 54* 38 (23) = not available, * = Mean age, # = in months, C = complete followup, EPR = Endoprosthetic reconstruction, IMN = Intramedullary nailing, ORIF = Open reduction and internal fixation 3 51

54 Chapter 3 2.4% 1.8% 1.9% 7.9% 10% 17% 35% Breast Lung Kidney Prostate Multiple Myeloma Lymphoma Colon/Rectum Unknown Other 8.2% 15% Figure 3: Primary tumor distribution (Appendix 2 includes the numbers). from 51 to 74%, but followup varied from 6 to 27 months (Table 2). The score in individual patients ranged from 28 to 100%. Two studies reported TESS scores: one found a score of 67 (range 52 to 82) in 5 patients after 18 months, and another found a score of 71 (46 to 84) in 11 patients after 27 months. Table 2: Functional outcome per implant type for proximal femur metastatic fractures (%) Author, Year Implant Followup Patients Outcome measure Median score (Range) Harvey et al EPR 9 months 21 MSTS 70* (40-90) Hattori et al EPR 6 months 12 MSTS 58 (28-100) Potter et al EPR 24 months 39 MSTS 67* (43-90) Chandrasekar et al EPR 18 months 5 TESS 67* (52-82) Selek et al EPR 2 months 29 MSTS 51* 12 months 12 MSTS 66* Park et al EPR 27 months 11 MSTS 74 (57-83) 27 months 11 TESS 71 (46-84) Harvey et al IMN 9 months 24 MSTS 80* (27-100) Asavamongkolgul et al ORIF 3 months 17 MSTS 80 (25-96) * = Mean score, MSTS = Musculoskeletal Tumor Society Score, TESS = Toronto Extremity Salvage Score, EPR = Endoprosthetic reconstruction, IMN = Intramedullary nailing, ORIF = Open reduction and internal fixation. Range was not available for the patients in the study by Selek et al. 52

55 Outcome After Fixation Of Metastatic Proximal Femoral Fractures One study reported the MSTS score after intramedullary nailing; the average score was 80% in 24 patients with a followup of 9 months (range 27 to 100%). One study reported the MSTS score after ORIF; the average score was 80% in 17 patients with a followup of 3 months (range 28 to 96%). Local Complications The overall reoperation rate varied from 0 to 31% after endoprosthetic reconstruction, 0 to 26% after intramedullary nailing, and 0 to 42% after ORIF (Table 3). Meta-analysis demonstrates pooled reoperation rates of 5.2% (95% CI 2.9 to 8.1%) for endoprostheses, 4.2% (95% CI 2.0 to 6.8%) for intramedullary nails, and 14% (95% CI 7.3 to 22%) for ORIF (Figure 4, left column). 3 Table 3: Local and systemic complication rates per implant type for proximal femur metastatic fractures (%) Author, Year Femurs Implant Pathological fractures Deep infection Fixation Failure Disloc. Recurr. Reop. Sys. Comp. Hettwer et al EPR (63) 4 (4) 1 (1) 6 (5) 1 (1) 13 (12) - Janssen et al EPR (79) 6 (9) 0 (0) 1 (1) 0 (0) 8 (11) - Sorensen et al EPR (1) 0 (0) 3 (3) 0 (0) 4 (4) - Weiss et al EPR 82-0 (0) 4 (5) 1 (1) 0 (0) 5 (6) 1 (1) Harvey et al EPR (62) 10 (9) 0 (0) 8 (7) 0 (0) 16 (14) - Steensma et al EPR (41) 0 (0) 0 (0) 5 (3) 1 (1) 6 (3) - Hattori et al EPR (81) 0 (0) 0 (0) - 0 (0) - - Parker et al EPR (100) - 3 (6) 2 (4) 0 (0) 5 (9) - Zacherl et al EPR (100) 2 (15) 0 (0) 1 (8) 0 (0) 4 (31) - Potter et al EPR (41) Sarahrudi et al EPR (100) 0 (0) 2 (9) 1 (4) 1 (4) 5 (22) 1 (4) Chandrasekar et al EPR (69) - 0 (0) 0 (0) 1 (1) 1 (1) - Selek et al EPR (62) 1 (2) 2 (4) 0 (0) 0 (0) 3 (7) 1 (2) Park et al EPR 31-0 (0) 0 (0) 1 (3) 0 (0) 1 (3) - Wedin et al EPR (0) 5 (5) 3 (3) 0 (0) 9 (8) 6 (6) Ward et al EPR 46-0 (0) 0 (0) 0 (0) 0 (0) 0 (0) - Chien et al EPR (83) - 0 (0) 0 (0) Rompe et al EPR 25 9 (36) 0 (0) 1 (4) 1 (4) 0 (0) 3 (12) 2 (8) Broos et al EPR 36-0 (0) 0 (0) 1 (3) 0 (0) 1 (3) - Korkala et al EPR (100) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) - Borel Rinkes et al EPR (79) 0 (0) 1 (3) 0 (0) 1 (3) 1 (3) - Yazawa et al EPR 41-0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 1 (2) Lane et al EPR (77) 1 (1) 0 (0) 0 (0) 0 (0) 1 (1) 2 (1) Choy et al IMN (100) 0 (0) 1 (10) 0 (0) 0 (0) 1 (10) - 53

56 Chapter 3 Table 3: Local and systemic complication rates per implant type for proximal femur metastatic fractures (%) (continued) Author, Year Femurs Implant Pathological fractures Deep infection Fixation Failure Disloc. Recurr. Reop. Sys. Comp. Janssen et al IMN (33) 6 (2) 9 (3) 0 (0) 2 (1) 16 (5) - Arvinius et al IMN (66) - 2 (4) 0 (0) 2 (4) 2 (4) 6 (11) Piccioli et al IMN (100) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 13 (16) Shemesh et al IMN (52) 1 (5) 0 (0) 0 (0) 0 (0) 1 (5) 0 (0) Fakler et al IMN (100) 0 (0) 2 (17) 0 (0) 0 (0) 2 (17) - Weiss et al IMN (0) 10 (9) 0 (0) 1 (1) 11 (10) 2 (2) Harvey et al IMN (46) 1 (2) 10 (22) 0 (0) 0 (0) 12 (26) - Steensma et al IMN (33) 0 (0) 2 (2) 0 (0) 3 (4) 5 (6) - Parker et al IMN (100) - 1 (3) 0 (0) 0 (0) 1 (3) - Zacherl et al IMN (100) 0 (0) 2 (5) 0 (0) 0 (0) 3 (8) - Rethnam et al IMN 11 9 (82) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 3 (27) Wedin et al IMN 24-0 (0) 3 (13) 0 (0) 0 (0) 3 (13) 0 (0) Datir et al IMN (76) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) Moholkar et al IMN (54) 1 (2) 1 (2) 0 (0) 0 (0) 2 (4) 6 (13) Ramakrishnan et al IMN 28 5 (18) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 2 (7) Piatek et al IMN 18-1 (6) 2 (11) 0 (0) 0 (0) 3 (17) - Assal et al IMN 12 6 (50) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 1 (8) Giannoudis et al IMN 17-0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) Algan et al IMN 12-0 (0) 0 (0) 0 (0) 0 (0) 0 (0) - Nargol et al IMN (100) - 0 (0) 0 (0) 0 (0) - - Yazawa et al IMN 18-0 (0) 1 (6) 0 (0) 0 (0) 1 (6) 0 (0) Fasano et al IMN 11 9 (82) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 4 (36) Behr et al IMN (44) 0 (0) 1 (3) 0 (0) 0 (0) 1 (3) - Janssen et al ORIF (40) 0 (0) 6 (13) 0 (0) 1 (2) 6 (13) - Asavamongkolgul et al. ORIF 27 9 (33) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) - Steensma et al ORIF 19 6 (32) 0 (0) 5 (26) 0 (0) 3 (16) 8 (42) - Parker et al ORIF (100) - 6 (17) 0 (0) 0 (0) 6 (17) - Zacherl et al ORIF (100) 0 (0) 2 (17) 0 (0) 0 (0) 3 (25) - Sarahrudi et al ORIF (100) 0 (0) 3 (20) 0 (0) - 3 (20) 0 (0) Wedin et al ORIF 13-0 (0) 3 (23) 0 (0) 0 (0) 3 (23) 0 (0) Ward et al ORIF 13-0 (0) 1 (8) 0 (0) 0 (0) 1 (8) - Rompe et al ORIF 25 8 (32) 0 (0) 0 (0) 0 (0) 3 (12) 4 (16) 2 (8) Broos et al ORIF 19-0 (0) 1 (5) 0 (0) 0 (0) 1 (5) - Behr et al ORIF 10 5 (50) 0 (0) 1 (10) 0 (0) 0 (0) 1 (10) - - = not available, Disloc. = dislocation, Recurr. = Recurrence, Reop. = total reoperations, Sys. Compl. = Systemic Complications. EPR = Endoprosthetic reconstruction, IMN = Intramedullary nailing, ORIF = Open reduction and internal fixation. 54

57 Outcome After Fixation Of Metastatic Proximal Femoral Fractures Total Reoperations Reoperation for Deep Infection Reoperation for Fixation Failure Author, Year ES (95% CI) % Weight ES (95% CI) % Weight ES (95% CI) % Weight Endoprosthesis Hettwer et al Janssen et al Sorensen et al Weiss et al Harvey et al Steensma et al Hattori et al Parker et al Zacherl et al Potter et al Sarahrudi et al Chandrasekar et al Selek et al Park et al Wedin et al Ward et al Chien et al Rompe et al Broos et al Korkala et al Borel Rinkes et al Yazawa et al Lane et al Subtotal: Intramedullary nail Choy et al Janssen et al Arvinius et al Piccioli et al Shemesh et al Fakler et al Weiss et al Harvey et al Steensma et al Parker et al Zacherl et al Rethnam et al Wedin et al Datir et al Moholkar et al Ramakrishnan et al Piatek et al Assal et al Giannoudis et al Algan et al Nargol et al Yazawa et al Fasano et al Behr et al Subtotal: ORIF Janssen et al Asavamongkolgul et al. Steensma et al Parker et al Zacherl et al Sarahrudi et al Wedin et al Ward et al Rompe et al Broos et al Behr et al Subtotal: Heterogeneity: Overall: 12 (6.4, 19) 11 (5.1, 21) (1.0, 9.0) 8.6 (3.2, 18) (0.0, 4.9) 0.0 (0.0, 5.1) (1.0, 9.5) (0.0, 5.2) (0.0, 3.5) (2.0, 14) (0.0, 4.4) (1.3, 12) (8.3, 22) (4.3, 16) (0.0, 3.2) (1.1, 6.5) (0.0, 1.9) (0.0, 1.9) (0.0, 21) (0.0, 21) (3.1, 20) (1.2, 15) (9.1, 61) (1.9, 46) (0.0, 25) (7.5, 44) (0.0, 15) (1.1, 28) (0.0, 6.7) (0.0, 4.5) (1.4, 18) (0.1, 12) (0.5, 15) (0.1, 17) (0.0, 11) (0.0, 11) (3.8, 15) (0.0, 3.3) (1.5, 10) (0.0, 7.7) (0.0, 7.7) (0.0, 7.7) (0.0, 27) (2.5, 31) (0.0, 14) (0.1, 21) (0.1, 15) (0.0, 9.7) (0.0, 9.7) (0.0, 22) (0.0, 22) (0.0, 22) (0.1, 15) (0.0, 11) (0.1, 15) (0.0, 8.6) (0.0, 8.6) (0.0, 8.6) (0.0, 3.4) (0.0, 3.4) (0.0, 2.2) (2.9, 8.1) (0.0, 2.1) (0.0, 1.3) 100 (I^2 = 73%, p = 0.00) (I^2 = 61%, p = 0.00) (I^2 = 46%, p = 0.01) 10 (0.3, 45) (3.1, 8.5) (0.4, 12) (0.0, 4.5) (0.1, 24) (2.1, 48) (5.2, 18) (14, 41) (2.0, 14) (0.1, 13) (1.7, 22) (0.0, 29) (2.7, 32.4) (0.0, 20) (0.5, 14) (0.0, 12) (3.6, 41) (0.0, 26.5) (0.0, 20) (0.0, 27) (0.1, 27) (0.0, 29) (0.1, 15) (2.0, 6.8) 100 (I^2 = 54%, p = 0.00) 0.0 (0.0, 31) (0.7, 4.3) (0.0, 4.5) (0.1, 24) (0.0, 27) (0.0, 3.4) (0.1, 12) (0.0, 4.4) (0.0, 9.5) (0.0, 29) (0.0, 14) (0.0, 20) (0.1, 11) (0.0, 12) (0.1, 27) (0.0, 27) (0.0, 20) (0.0, 27) (0.0, 19) (0.0, 29) (0.0, 10) (0.0, 0.5) 100 (I^2 = 0.0%, p = 0.94) 10 (0.3, 45) (1.4, 5.6) (0.4, 12) (0.0, 4.5) (0.0, 16) (2.1, 48) (4.5, 16) (11, 36) (0.3, 8.5) (0.1, 13) (0.7, 18) (0.0, 29) (2.7, 32) (0.0, 20) (0.1, 11) (0.0, 12) (1.4, 35) (0.0, 27) (0.0, 20) (0.0, 27) (0.0, 31) (0.1, 27) (0.0, 29) (0.1, 15) (1.1, 5.0) 100 (I^2 = 47%, p = 0.01) 13 (5.1, 27) (0.0, 7.9) (5.1, 27) (0.0, 13) (0.0, 13) (0.0, 13) (20, 67) (0.0, 18) (9.1, 51) (6.6, 34) (6.6, 34) (5.5, 57) (0.0, 27) (2.1, 48) (4.3, 48) (0.0, 22) (4.3, 48) (5.0, 54) (0.0, 25) (5.0, 54) (0.2, 36) (0.0, 25) (0.2, 36) (4.5, 36) (0.0, 14) (0.0, 14) (0.1, 26) (0.0, 18) (0.1, 26) (0.3, 45) (0.0, 31) (0.3, 45) (7.3, 22) (0.0, 0.9) (4.3, 17) 100 (I^2 = 54%, p = 0.02) (I^2 = 0.0%, p = 0.99) (I^2 = 54%, p = 0.02) p = 0.01 p = 0.82 p = (4.0, 7.9) 0.1 (0.0, 0.6) 2.2 (1.0, 3.7) (I^2 = 67%, p = 0.00) (I^2 = 15%, p = 0.18) (I^2 = 66%, p = 0.00) Percentage Percentage Percentage Figure 4: Random-effects meta-analysis with reoperation rates (solid black diamonds) per study per treatment arm including the 95% confidence interval [95% CI] (black horizontal lines crossing the solid diamonds). The green open diamonds are the pooled reoperation rates based on all studies per treatment arm (subtotal). The vertical red dashed lines indicate the reoperation rate for all treatment arms (overall). The left column includes all reoperations, the middle column includes reoperations for deep infections, and the right column includes reoperations for fixation failure. ES = effect estimate, ORIF = Open reduction and internal fixation. The I-squared statistic demonstrates the estimate of the percentage of the variation in the effect estimate due to the heterogeneity across studies. Note the varying percentage-scale on the x-axis. 55

58 Chapter 3 The reoperation rate for deep infection varied from 0 to 15% after endoprosthetic reconstruction, 0 to 6% after intramedullary nailing, and was 0% in all studies after ORIF (Table 3). Meta-analysis demonstrates pooled deep infection rates of 0.68% (95% CI 0.0 to 2.05%) for endoprostheses, 0.04% (95%CI 0.00 to 0.54%) for intramedullary nails, and 0.00% (95%CI 0.00 to 0.92%) for ORIF (Figure 4, middle column). The reoperation rate for fixation failure varied from 0 to 9% after endoprosthetic reconstruction, 0 to 22% after intramedullary nailing, and 0 to 26% after ORIF (Table 3). Meta-analysis demonstrates pooled fixation failure rates of 0.4% (95% CI 0.0 to 1.3%) for endoprostheses, 2.8% (95% CI 1.1 to 5.0%) for intramedullary nails, and 10% (95% CI 4.3 to 17%) for ORIF (Figure 4, right column). Dislocations requiring reoperation in the endoprosthesis group ranged from 0 to 8% (Table 3). Tumor recurrence or progression requiring reoperation was reported in 20 patients; distribution of the recurring tumors was: 2 kidney, 2 breast, 1 bladder, 1 lung, 1 esophageal, 1 multiple myeloma, and not specified in the remaining 12 cases. Other reasons for reoperation (included in the overall reoperation rates) were: an exostosis (1), massive hematoma (1), wound healing problems (2), and a loose piece of cement in the hip joint after endoprosthetic reconstruction; 37,38,74,76 bursitis (1), wound healing problems (1), and painful hardware, after intramedullary nailing; 48,76 and wound healing (2) after ORIF. 37,76 Systemic Complications Only fourteen studies reported on systemic complications after 16 treatments. The systemic complication rate varied from 1 to 8% after endoprosthetic reconstruction, 0 to 27% after intramedullary nailing, and 0 to 8% after ORIF. There were 54 systemic complications reported in 53 patients: deep venous thrombosis (17), pneumonia (9), pulmonary embolism (6), respiratory failure (5), intraoperative mortality (3), fat-embolism (2), cardiac failure (2), urinary tract infection (2), perforated colon (1), cerebrovascular accident (1), multi-organ failure (1), ileus (1), respiratory arrest (1), cardiovascular embolism (1), and not specified (2). Fat-embolism and intraoperative mortality occurred in the same case. 55 DISCUSSION Endoprosthetic reconstruction, intramedullary nailing, and ORIF are common treatments for pathological proximal femoral fracture caused by bone metastasis; all aim to preserve the patient s independence and quality of the remaining life. Most studies are relatively small hampering comparison of outcomes among these surgical techniques. We aimed to establish functional outcome and complications both local and systemic and compared these among common surgical strategies. We found that only seven studies reported functional outcome and that on average all three techniques result in good function. 56

59 Outcome After Fixation Of Metastatic Proximal Femoral Fractures In terms of local complications, we found that the pooled overall reoperation rate was comparable for endoprosthesis and intramedullary nailing, but was substantially higher for ORIF. Deep infection seems to occur more commonly after endoprosthetic reconstruction, while fixation failure more commonly occurred after intramedullary nailing and ORIF. Deep venous thrombosis, pneumonia, and pulmonary embolism were the most commonly reported systemic complications. These findings could aid in surgical decision making and help inform the patient. Our study has limitations. First, although three large literature databases have been used, we might have missed possible relevant publications that are not listed in these libraries. Additionally, our search missed one relevant paper identified by bibliography screening of included studies ; this means that other relevant studies might have been missed as well. 48 However, we see these as a minor limitations because this was the only study identified after extensive screening of the bibliographies of the included studies. Second, the type of implants used within the three treatment concepts vary; the surgical technique, experience of the surgeons, and postoperative rehabilitation probably varied as well. We see this as an important limitation as we were not able to stratify for this; however, we feel that this affects all three treatment arms and reflects clinical practice. Third, indications fracture type, location of the fracture, extend of bone destruction for the specific strategies vary and selection bias can therefore not be ignored. We tried to minimize this by focusing on proximal femoral metastases. Furthermore, we described fracture types and location per surgical technique to help understand the variation in indications. Fourth, adjuvant treatment radiation therapy and chemotherapy can influence complication rates; These are poorly reported in the included studies and we were therefore not able to stratify for this, we consider this as an important limitation. Fifth, definitions of outcome, such as complications, might have varied between studies and we therefore only included local complications that required reoperation. Sixth, followup varied between studies and was not always clearly reported. We see this as an important limitation as short followup or loss to followup attrition bias might preclude identification of a complication. This might result in underestimating the complication rates; however, all three treatment arms are affected by this. Despite these limitations, we feel that meta-analysis is justified for estimation of pooled reoperation rates per surgical strategy as this outcome is often the primary outcome of interest and therefore generally well reported. The limited number of studies assessing functional outcome and variation in use of outcome measures and duration of followup did not allow us to directly compare surgical strategies. All three techniques seem to result in reasonable function on average; however, the range is wide indicating that both poor and good functional levels are obtained. The MSTS score is most commonly used; 49 however, it is a clinician completed assessment of the patients functional level and results might therefore be biased as clinicians tend to overestimate function and underestimate pain. 51,77-79 Future studies assessing function 3 57

60 Chapter 3 should therefore focus on patient reported outcome measures, such as the TESS score or a more general instrument like the PROMIS Physical Function Cancer questionnaire. 51,80,81 Furthermore, studies should aim to measure both preoperative and postoperative functional levels in order to establish the efficacy of surgical fixation. Lastly, including measures of quality of life (e.g. the EQ-5D or SF-36 instruments) would help us to understand how treatment impacts the overall wellbeing of the patient. 77 Acknowledging the limitations of directly comparing local complications among surgical strategies, we found that ORIF results in the highest overall reoperation rate. These reoperations after ORIF were predominantly performed for failure of fixation. We found comparable overall reoperation rates when comparing endoprosthetic reconstruction with intramedullary nailing. However, reoperation for deep infection seems to be more common after endoprosthetic reconstruction, while reoperation for failure of fixation is more common after intramedullary nailing. Dislocations can technically only occur after endoprosthetic reconstruction and required reoperation in 0 to 8% of the cases. The finding of comparable overall reoperation rates but different reasons for reoperation are confirmed when focusing on the seven studies that describe both endoprosthetic reconstructions and intramedullary nailing. 44,46-48,62,74-76 Four out of these seven studies describe timing of the reoperations; deep infections seem to occur early (within the first months), while fixation failures seem to occur late. 47,48,62,74 Based on these findings, we feel that intramedullary nailing is an acceptable option for treatment of proximal femoral metastasis in the trochanteric and subtrochanteric area with limited bone loss in patients with poor life expectancy. Endoprosthetic reconstruction should be considered in patients with femoral head/neck lesions, lesions with substantial bone loss in the trochanteric or subtrochanteric area, radioresistant lesions (e.g. renal cell carcinoma), and in patients with reasonable life expectancy. Several algorithms exist to help the surgeon estimate life expectancy Deep venous thrombosis, pneumonia, and pulmonary embolism were the most commonly reported systemic complications. Intraoperative mortality was rare (3 cases described). Measures should be taken to prevent systemic complications from occurring and to limit their consequences in these patients. Unfortunately, poor quality of reporting systemic complications did not allow direct comparison of this outcome between surgical strategies using a meta-analysis approach. Future studies should be more clear about: which systemic complications are to be reported, what the treatment consequences are, and the timing and followup of the systemic complications. 86 Lastly, studies should only include 1 limb per 1 patient as to avoid violation of the statistical rule of independence; or authors should use adequate statistical method to account for this. 87 In conclusion, this study provides an overview of functional outcome and complications after common surgical strategies for impending and pathological fractures in proximal femoral metastasis. All three surgical strategies result in reasonable function on average; however, functional levels vary substantially. ORIF results in a high reoperation rate, while 58

61 Outcome After Fixation Of Metastatic Proximal Femoral Fractures reoperation rates for endoprosthesis and intramedullary nailing are comparable. The results reported here can aid patients and their surgeons in the decision for surgical treatment. REFERENCES 1. Issack PS, Barker J, Baker M, Kotwal SY, Lane JM. Surgical management of metastatic disease of the proximal part of the femur. J Bone Joint Surg Am ; 96-A: Quinn RH, Randall RL, Benevenia J, Berven SH, Raskin KA. Contemporary management of metastatic bone disease: tips and tools of the trade for general practitioners. J Bone Joint Surg Am. Oct ; 95(20): Forsberg JA, Wedin R, Bauer H. Which implant is best after failed treatment for pathologic femur fractures? Clin Orthop Relat Res. Mar 2013; 471(3): Steensma M, Healey JH. Trends in the surgical treatment of pathologic proximal femur fractures among Musculoskeletal Tumor Society members. Clin Orthop Relat Res. Jun 2013; 471(6): Liberati A, Altman DG, Tetzlaff J, et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate healthcare interventions: explanation and elaboration. BMJ ; 339(2700). 6. Alvi HM, Damron TA. Prophylactic stabilization for bone metastases, myeloma, or lymphoma: Do we need to protect the entire bone? Tumor. Clin Orthop Relat Res. 2013; 471: Ampil FL, Sadasivan KK. Prophylactic and therapeutic fixation of weight-bearing long bones with metastatic cancer. South Med J. 2001; 94: Arvinius C, Parra JLC, Mateo LS, Maroto RG, Borrego AF, Stern LLD. Benefits of early intramedullary nailing in femoral metastases. Int Orthop. 2014; 38(1): Ashford RU, Hanna SA, Park DH, et al. Proximal femoral replacements for metastatic bone disease: financial implications for sarcoma units. Int Orthop. 2010; 34: Bohm P, Huber J. The surgical treatment of bony metastases of the spine and limbs. J Bone Joint Surg Br. 2002; 84: Camnasio F, Scotti C, Peretti GM, Fontana F, Fraschini G. Prosthetic joint replacement for long bone metastases: analysis of 154 cases. Arch Orthop Trauma Surg. 2008; 128: Cole AS, Hill GA, Theologis TN, Gibbons CLMH, Willett K. Femoral nailing for metastatic disease of the femur: a comparison of reamed and unreamed femoral nailing. Injury. 2000; 31: Fischer T, Felenda M, Dittel KK. Pathological fractures of the long bones due to bone metastases. Aktuelle Traumatol. 2004; 34: Gibbons CER, Pope SJ, Murphy JP, Hall AJ. Femoral metastatic fractures treated with intramedullary nailing. Int Orthop. 2000; 24: Gruber G, Zacherl M, Leithner A, et al. [Surgical treatment of pathologic fractures of the humerus and femur]. Orthopade. 2009; 38: 324, , Katzer A, Meenen NM, Grabbe F, Rueger JM. Surgery of skeletal metastases. Arch Orthop Trauma Surg. 2002; 122: Kinkel S, Stecher J, Gotterbarm T, Bruckner T, Holz U. Compound osteosynthesis for osteolyses and pathological fractures of the proximal femur. Orthopedics. 2009; 32:

62 Chapter Kollender Y, Bickels J, Price WM, et al. Metastatic renal cell carcinoma of bone: indications and technique of surgical intervention. J Urol. 2000; 164: Krastman P, Welvaart WN, Breugem SJM, Van Vugt AB. The Holland nail(registered trademark): A universal implant for fractures of the proximal femur and the femoral shaft. Injury. 2004; 35: Ratasvuori M, Wedin R, Keller J, et al. Insight opinion to surgically treated metastatic bone disease: Scandinavian Sarcoma Group Skeletal Metastasis Registry report of 1195 operated skeletal metastasis. Surg Oncol. 2013; 22: Sehat K, Baker RP, Pattison G, Price R, Harries WJ, Chesser TJS. The use of the long gamma nail in proximal femoral fractures. Injury. 2005; 36: Sokolovski VA, Voloshin VP, Aliev MD, et al. Total hip replacement for proximal femoral tumours: our midterm results. Int Orthop. 2006; 30: Talbot M, Turcotte RE, Isler M, Normandin D, Iannuzzi D, Downer P. Function and health status in surgically treated bone metastases. Clin Orthop Relat Res. 2005; 438: Toliusis V, Kalesinskas RJ, Kiudelis M, Maleckas A, Griksas M. Surgical treatment of metastatic tumors of the femur. Medicina (Kaunas). 2010; 46: Zore Z, Filipovic Zore I, Matejcic A, Kamal M, Arslani N, Knezovic Zlataric D. Surgical treatment of pathologic fractures in patients with metastatic tumors. Coll Antropol. 2009; 33: Ward WG, Spang J, Howe D, Gordan S. Femoral recon nails for metastatic disease: indications, technique, and results. Am J Orthop (Belle Mead NJ). 2000; 29: Rompe JD, Eysel P, Hopf C, Heine J, Schaub T. Application of tumor protheses in secondary neoplastic destruction of the proximal femur. Z Orthop Ihre Grenzgeb. 1993; 131: Sarahrudi K, Hora K, Heinz T, Millington S, Vecsei V. Treatment results of pathological fractures of the long bones: a retrospective analysis of 88 patients. Int Orthop. 2006; 30: Dobozi WR, Dvonch VM, Saltzman ML. Treatment of impending pathological fractures of the femur with flexible intramedullary nails. Orthopedics. 1984; 7: Dijkstra S, Wiggers T, van Geel BN, Boxma H, Dijstra S. Impending and actual pathological fractures in patients with bone metastases of the long bones. A retrospective study of 233 surgically treated fractures. Eur J Surg. 1994; 160: Chandrasekar CR, Grimer RJ, Carter SR, Tillman RM, Abudu A, Buckley L. Modular endoprosthetic replacement for tumours of the proximal femur. J Bone Joint Surg Br. Jan 2009; 91(1): Fourneau I, Broos P. Pathologic fractures due to metastatic disease. A retrospective study of 160 surgically treated fractures. Acta Chir Belg. Dec 1998; 98(6): Behr JT, Dobozi WR, Badrinath K. The treatment of pathologic and impending pathologic fractures of the proximal femur in the elderly. Clin Orthop Relat Res. 1985: Borel Rinkes IH, Wiggers T, Bouma WH, van Geel AN, Boxma H. Treatment of manifest and impending pathologic fractures of the femoral neck by cemented hemiarthroplasty. Clin Orthop Relat Res. 1990: Broos P, Reynders P, van den Bogert W, Vanderschot P. Surgical treatment of metastatic fracture of the femur improvement of quality of life. Acta Orthop Belg. 1993; 59 Suppl 1: Chandrasekar CR, Grimer RJ, Carter SR, Tillman RM, Abudu AT. Modular endoprosthetic replacement for metastatic tumours of the proximal femur. J Orthop Surg Res. 2008; 3: Rompe JD, Eysel P, Hopf C, Heine J. Metastatic instability at the proximal end of the femur. Comparison of endoprosthetic replacement and plate osteosynthesis. Arch Orthop Trauma Surg. 1994; 113:

63 Outcome After Fixation Of Metastatic Proximal Femoral Fractures 38. Sarahrudi K, Greitbauer M, Platzer P, Hausmann JT, Heinz T, Vecsei V. Surgical treatment of metastatic fractures of the femur: a retrospective analysis of 142 patients. J Trauma. 2009; 66: Ward WG, Holsenbeck S, Dorey FJ, Spang J, Howe D. Metastatic disease of the femur: surgical treatment. Clin Orthop Relat Res. 2003: S Algan SM, Horowitz SM. Surgical treatment of pathologic hip lesions in patients with metastatic disease. Clin Orthop Relat Res. 1996: Hattori H, Mibe J, Yamamoto K. Modular megaprosthesis in metastatic bone disease of the femur. Orthopedics. 2011; 34: e Hettwer WH, Horstmann PF, Hovgaard TB, Grum-Scwensen TA, Petersen MM. Low infection rate after tumor hip arthroplasty for metastatic bone disease in a cohort treated with extended antibiotic prophylaxis. Adv Orthop. 2015; 2015: Lane JM, Sculco TP, Zolan S. Treatment of pathological fractures of the hip by endoprosthetic replacement. J Bone Joint Surg Am. 1980; 62: Parker MJ, Khan AZ, Rowlands TK. Survival after pathological fractures of the proximal femur. Hip Int. 2011; 21: Potter BK, Chow VE, Adams SC, Letson GD, Temple HT. Endoprosthetic proximal femur replacement: metastatic versus primary tumors. Surg Oncol. 2009; 18: Steensma M, Boland PJ, Morris CD, Athanasian E, Healey JH. Endoprosthetic treatment is more durable for pathologic proximal femur fractures. Clin Orthop Relat Res. 2012; 470: Weiss RJ, Ekstrom W, Hansen BH, et al. Pathological subtrochanteric fractures in 194 patients: a comparison of outcome after surgical treatment of pathological and non-pathological fractures. J Surg Oncol. 2013; 107: Harvey N, Ahlmann ER, Allison DC, Wang L, Menendez LR. Endoprostheses last longer than intramedullary devices in proximal femur metastases. Clin Orthop Relat Res ; 470: Enneking WF, Dunham W, Gebhardt MC, Malawar M, Pritchard DJ. A system for the functional evaluation of reconstructive procedures after surgical treatment of tumors of the musculoskeletal system. Clin Orthop Relat Res. Jan 1993(286): Clayer M, Davis A. Can the Toronto Extremity Salvage Score produce reliable results when used online? Clin Orthop Relat Res. Jun 2011; 469(6): Davis AM, Wright JG, Williams JI, Bombardier C, Griffin A, Bell RS. Development of a measure of physical function for patients with bone and soft tissue sarcoma. Qual Life Res. Oct 1996; 5(5): Normand SL. Meta-analysis: formulating, evaluating, combining, and reporting. Stat Med ; 18: Nyaga VN, Arbyn M, Aerts M. Metaprop: a Stata command to perform meta-analysis of binomial data. Arch Public Health ; Asavamongkolgul A, Thanakarasombat S. Treatment of bone tumors in the femoral trochanteric area. J Med Assoc Thai. 2012; 95 Suppl 9: S Assal M, Zanone X, Peter RE. Osteosynthesis of metastatic lesions of the proximal femur with a solid femoral nail and interlocking spiral blade inserted without reaming. J Orthop Trauma. 2000; 14: Chien SH, Hung SH, Cheng YM, et al. Surgical treatment of pathologic fracture of the femur. Kaohsiung J Med Sci. 1997; 13:

64 Chapter Choy WS, Kim KJ, Lee SK, et al. Surgical treatment of pathological fractures occurring at the proximal femur. Yonsei Med J. 2015; 56(2): Datir SP, Bedi GS, Curwen CH. Unreamed femoral nail with spiral blade in subtrochanteric fractures: experience of 55 cases. Injury. 2004; 35: Fakler JKM, Hase F, Bohme J, Josten C, Böhme J, Josten C. Safety aspects in surgical treatment of pathological fractures of the proximal femur - modular endoprosthetic replacement vs. intramedullary nailing. Patient Saf Surg. 2013; 7(1): Fasano Jr. FJ, Olysav DJ, Stauffer ES. Intramedullary stabilization of neoplastic destructive disease involving the subtrochanteric region of the femur. Orthopedics. 1988; 11: Giannoudis PV, Bastawrous SS, Bunola JA, Macdonald DA, Smith RM. Unreamed intramedullary nailing for pathological femoral fractures. Good results in 30 cases. Acta Orthop Scand. 1999; 70: Janssen SJ, Kortlever JTP, Ready JE, et al. Complications After Surgical Management of Proximal Femoral Metastasis: A Retrospective Study of 417 Patients. Accepted for publication in the J Am Acad Orthop Surg Korkala OL, Karaharju EO. Metastatic fractures of long bones. Int Orthop. 1991; 15: Moholkar K, Mohan R, Grigoris P. The Long Gamma Nail for stabilisation of existing and impending pathological fractures of the femur: an analysis of 48 cases. Acta Orthop Belg. 2004; 70: Nargol AV, Port A, Seif-Asaad SS. Pathological fractures of the proximal femur treated with the Variwall reconstruction nail. Injury. 1996; 27: Park DH, Jaiswal PK, Al-Hakim W, et al. The use of massive endoprostheses for the treatment of bone metastases. Sarcoma. 2007; 2007: Piatek S, Westphal T, Bischoff J, Schubert S, Holmenschlager F, Winckler S. [Intramedullary stabilisation of metastatic fractures of long bones]. Zentralbl Chir. 2003; 128: Piccioli A, Rossi B, Scaramuzzo L, Spinelli MS, Yang Z, Maccauro G. Intramedullary nailing for treatment of pathologic femoral fractures due to metastases. Injury. 2014; 45(2): Ramakrishnan M, Prasad SS, Parkinson RW, Kaye JC. Management of subtrochanteric femoral fractures and metastases using long proximal femoral nail. Injury. 2004; 35: Rethnam U, Cordell-Smith J, Kumar TM, Sinha A. Complex proximal femoral fractures in the elderly managed by reconstruction nailing - complications & outcomes: a retrospective analysis. J Trauma Manag Outcomes. 2007; 1: Selek H, Basarir K, Yildiz Y, Saglik Y. Cemented endoprosthetic replacement for metastatic bone disease in the proximal femur. J Arthroplasty. 2008; 23: Shemesh S, Kosashvili Y, Sidon E, Yaari L, Cohen N, Velkes S. Intramedullary nailing without curettage and cement augmentation for the treatment of impending and complete pathological fractures of the proximal or midshaft femur. Acta Orthop Belg. 2014; 80(1): Sorensen MS, Gregersen KG, Grum-Schwensen T, Hovgaard D, Petersen MM. Patient and implant survival following joint replacement because of metastatic bone disease: A crosssectional study of 130 patients with 140 joint replacements. Acta Orthop. 2013; 84(3): Wedin R, Bauer HC. Surgical treatment of skeletal metastatic lesions of the proximal femur: endoprosthesis or reconstruction nail? J Bone Joint Surg Br. 2005; 87: Yazawa Y, Frassica FJ, Chao EY, Pritchard DJ, Sim FH, Shives TC. Metastatic bone disease. A study of the surgical treatment of 166 pathologic humeral and femoral fractures. Clin Orthop Relat Res. 1990:

65 Outcome After Fixation Of Metastatic Proximal Femoral Fractures 76. Zacherl M, Gruber G, Glehr M, et al. Surgery for pathological proximal femoral fractures, excluding femoral head and neck fractures: resection vs. stabilisation. Int Orthop. 2011; 35: Cheng EY. Prospective quality of life research in bony metastatic disease. Clin Orthop Relat Res. Oct 2003(415 Suppl): S Nelson E, Conger B, Douglass R, et al. Functional health status levels of primary care patients. JAMA. Jun ; 249(24): Rosenberger PH, Jokl P, Cameron A, Ickovics JR. Shared decision making, preoperative expectations, and postoperative reality: differences in physician and patient predictions and ratings of knee surgery outcomes. Arthroscopy. May 2005; 21(5): Garcia SF, Cella D, Clauser SB, et al. Standardizing patient-reported outcomes assessment in cancer clinical trials: a patient-reported outcomes measurement information system initiative. J Clin Oncol. Nov ; 25(32): Jensen RE, Potosky AL, Reeve BB, et al. Validation of the PROMIS physical function measures in a diverse US population-based cohort of cancer patients. Qual Life Res. May Bauer HC, Wedin R. Survival after surgery for spinal and extremity metastases. Prognostication in 241 patients. Acta Orthop Scand. Apr 1995; 66(2): Janssen SJ, van der Heijden AS, van Dijke M, et al Marshall Urist Young Investigator Award: Prognostication in Patients With Long Bone Metastases: Does a Boosting Algorithm Improve Survival Estimates? Clin Orthop Relat Res. Jul Katagiri H, Takahashi M, Wakai K, Sugiura H, Kataoka T, Nakanishi K. Prognostic factors and a scoring system for patients with skeletal metastasis. J Bone Joint Surg Br. May 2005; 87(5): Nathan SS, Healey JH, Mellano D, et al. Survival in patients operated on for pathologic fracture: implications for end-of-life orthopedic care. J Clin Oncol. Sep ; 23(25): Clavien PA, Barkun J, de Oliveira ML, et al. The Clavien-Dindo classification of surgical complications: five-year experience. Ann Surg. Aug 2009; 250(2): Bryant D, Havey TC, Roberts R, Guyatt G. How many patients? How many limbs? Analysis of patients or limbs in the orthopaedic literature: a systematic review. J Bone Joint Surg Am ; 88-A(1):

66 Chapter 3 Appendix 1: Critical appraisal of included studies (Figure 2) Author, Year Prospective Design Disclosure Comparison of Techniques Patient Selection Baseline Outcome No Attrition Bias Choy et al Hettwer et al Janssen et al Arvinius et al Piccioli et al Shemesh et al Fakler et al Sorensen et al Weiss et al Asavamongkolgul et al Harvey et al Steensma et al Hattori et al Parker et al Zacherl et al Potter et al Sarahrudi et al Chandrasekar et al Selek et al Park et al Rethnam et al Wedin et al Datir et al Moholkar et al Ramakrishnan et al Piatek et al Ward et al Assal et al Giannoudis et al Chien et al Algan et al Nargol et al Rompe et al Broos et al Korkala et al Borel Rinkes et al

67 Outcome After Fixation Of Metastatic Proximal Femoral Fractures Appendix 1: Critical appraisal of included studies (Figure 2) (continued) Author, Year Prospective Design Disclosure Comparison of Techniques Patient Selection Baseline Outcome No Attrition Bias Yazawa et al Fasano et al Behr et al Lane et al Prospective design 1, Prospective design 0, Retrospective design Disclosure 1, Disclosure is reported 0, Disclosure is not reported Comparison of Techniques 1, Direct comparison of surgical techniques 0, Single treatment arm or not a comparison Patient selection 1, Clear eligibility criteria 0, Potential selection bias or eligibility criteria unclear Baseline 1, Detailed baseline characteristics per treatment arm 0, Mixed or unspecified baseline characteristics Outcome 1, Clear definition of outcomes to be reported 0, Outcomes not specified or unclear Completeness of outcome data (no attrition bias) 1, <20% Lost to followup within a year 0, >20% Lost to followup within a year, or unclear 65

68 Chapter 3 Appendix 2: Primary tumor characteristics of all patients with femoral metastases per study (Figure 3) (%) Author, Year Patients Femurs Breast Lung Kidney Prostate MM Lymph. Colon/ Rectum Unknown Other Choy et al (32) 3 (16) 3 (16) 0 (0) 0 (0) 0 (0) 1 (5) 0 (0) 6 (32) Hettwer et al (33) 19 (18) 13 (12) 10 (10) 8 (8) 4 (4) 4 (4) 3 (3) 9 (9) Janssen et al (30) 95 (23) 26 (6) 29 (7) 57 (14) 0 (0) 7 (2) 13 (3) 65 (16) Piccioli et al (29) 8 (10) 2 (3) 12 (15) 16 (20) 4 (5) 6 (8) 0 (0) 9 (11) Shemesh et al (47) 5 (26) 0 (0) 1 (5) 3 (16) 0 (0) 1 (5) 0 (0) 0 (0) Fakler et al (40) 1 (5) 0 (0) 2 (10) 2 (10) 0 (0) 1 (5) 2 (10) 4 (20) Weiss et al (36) 19 (10) 27 (14) 34 (18) 0 (0) 0 (0) 0 (0) 16 (8) 28 (14) Asavamongkolgul et al (37) 3 (11) 0 (0) 2 (7) 1 (4) 0 (0) 2 (7) 1 (4) 8 (30) Harvey et al (26) 19 (12) 19 (12) 10 (6) 14 (9) 2 (1) 7 (4) 3 (2) 43 (27) Steensma et al (25) 50 (17) 39 (13) 12 (4) 26 (9) 12 (4) 12 (4) 0 (0) 72 (24) Hattori et al (50) 1 (6) 2 (13) 3 (19) 0 (0) 0 (0) 0 (0) 1 (6) 1 (6) Parker et al (36)* 17 (12)* 4 (3)* 33 (23)* 9 (6)* 13 (9)* 2 (1)* 11 (8)* 4 (3)* Zacherl et al (41) 10 (17) 5 (8) 8 (14) 0 (0) 0 (0) 2 (3) 0 (0) 10 (17) Sarahrudi et al (46) 14 (10) 7 (5) 11 (8) 6 (4) 0 (0) 0 (0) 0 (0) 38 (27) Selek et al (39) 13 (30) 2 (5) 3 (7) 0 (0) 0 (0) 0 (0) 0 (0) 9 (20) Rethnam et al (45) 2 (18) 0 (0) 3 (27) 1 (9) 0 (0) 0 (0) 0 (0) 0 (0) Wedin et al (32) 16 (11) 8 (6) 36 (25) 0 (0) 0 (0) 0 (0) 0 (0) 37 (26) Moholkar et al (50)* 7 (15)* 2 (4)* 5 (10)* 6 (13)* 1 (2)* 1 (2)* 0 (0)* 2 (4)* Ramakrishnan et al (29)* 4 (14)* 0 (0)* 9 (32)* 5 (18)* 0 (0)* 1 (4)* 0 (0)* 1 (4)* Ward et al (21) 41 (23) 16 (9) 17 (9) 28 (15) 9 (5) 0 (0) 3 (2) 29 (16) Assal et al (50) 3 (30) 1 (10) 1 (10) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) Giannoudis et al (56) 8 (30) 2 (7) 2 (7) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 66

69 Outcome After Fixation Of Metastatic Proximal Femoral Fractures Appendix 2: Primary tumor characteristics of all patients with femoral metastases per study (Figure 3) (%) (continued) Unknown Other Author, Year Patients Femurs Breast Lung Kidney Prostate MM Lymph. Colon/ Rectum Chien et al (44) 4 (13) 2 (6) 1 (3) 0 (0) 0 (0) 0 (0) 6 (19) 5 (16) Algan et al (59) 1 (4) 5 (19) 0 (0) 3 (11) 0 (0) 0 (0) 1 (4) 1 (4) Nargol et al (50) 1 (10) 0 (0) 1 (10) 2 (20) 0 (0) 0 (0) 1 (10) 0 (0) Rompe et al (72) 0 (0) 14 (28) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) Broos et al (75) 3 (5) 4 (6) 4 (6) 0 (0) 0 (0) 0 (0) 1 (2) 4 (6) Borel Rinkes et al (79) 2 (6) 0 (0) 3 (9) 2 (6) 0 (0) 0 (0) 0 (0) 0 (0) Fasano et al (40) 2 (20) 1 (10) 2 (20) 1 (10) 0 (0) 0 (0) 0 (0) 0 (0) Behr et al (45) 4 (11) 2 (5) 1 (3) 9 (24) 0 (0) 1 (3) 1 (3) 3 (8) Lane et al (61) 12 (7) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 51 (31) MM = Multiple Myeloma, Lymph. = Lymphoma. Primary tumor type for femur metastases not available for the following studies: Arvinius et al. 2014, Sorensen et al. 2013, Potter et al. 2009, Chandrasekar et al. 2008, Park et al. 2007, Datir et al. 2004, Piatek et al. 2003, Korkala et al. 1991, Yazawa et al

70 Chapter 3 Appendix 3: Affected anatomical area and use of cement per implant type for proximal femur metastatic fractures (%) Author, Year Implant Femurs Pathological fractures Cement Head /Neck Troch. Area Subtroch. Area Prox. Shaft Unspec. Hettwer et al EPR (63) - 0 (0) 0 (0) 0 (0) 0 (0) 111 (100) Janssen et al EPR (79) 68 (97) 28 (40) 30 (43) 12 (17) 0 (0) 0 (0) Sorensen et al EPR (0) 0 (0) 0 (0) 0 (0) 105 (100) Weiss et al EPR (0) 0 (0) 82 (100) 0 (0) 0 (0) Harvey et al EPR (62) 113 (100) 0 (0) 0 (0) 0 (0) 0 (0) 113 (100) Steensma et al EPR (41) 197 (100) 0 (0) 115 (58) 82 (42) 0 (0) 0 (0) Hattori et al EPR (81) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 16 (100) Parker et al EPR (100) 29 (54) 0 (0) 0 (0) 0 (0) 0 (0) 54 (100) Zacherl et al EPR (100) 13 (100) 0 (0) 0 (0) 0 (0) 0 (0) 13 (100) Potter et al EPR (41) 39 (100) 0 (0) 0 (0) 0 (0) 0 (0) 39 (100) Sarahrudi et al EPR (100) 23 (100) 0 (0) 0 (0) 0 (0) 0 (0) 23 (100) Chandrasekar et al EPR (69) 81 (100) 0 (0) 0 (0) 0 (0) 0 (0) 81 (100) Selek et al EPR (62) 45 (100) 0 (0) 0 (0) 0 (0) 0 (0) 45 (100) Park et al EPR (0) 0 (0) 0 (0) 0 (0) 31 (100) Wedin et al EPR (97) 54 (50) 28 (26) 27 (25) 0 (0) 0 (0) Ward et al EPR (0) 0 (0) 0 (0) 0 (0) 46 (100) Chien et al EPR (83) - 11 (92) 1 (8) 0 (0) 0 (0) 0 (0) Rompe et al EPR 25 9 (36) - 0 (0) 0 (0) 0 (0) 0 (0) 25 (100) Broos et al EPR (42) 10 (28) 11 (31) 0 (0) 0 (0) Korkala et al EPR (100) - 8 (53) 7 (47) 0 (0) 0 (0) 0 (0) Borel Rinkes et al EPR (79) 34 (100) 27 (79) 0 (0) 0 (0) 0 (0) 0 (0) Yazawa et al EPR (100) 27 (66) 14 (34) 0 (0) 0 (0) 0 (0) 68

71 Outcome After Fixation Of Metastatic Proximal Femoral Fractures Appendix 3: Affected anatomical area and use of cement per implant type for proximal femur metastatic fractures (%) (continued) Unspec. Prox. Shaft Subtroch. Area Troch. Area Head /Neck Cement Pathological fractures Implant Femurs Author, Year Lane et al EPR (77) 75 (46) 78 (48) 55 (34) 30 (18) 0 (0) 0 (0) Choy et al IMN (100) 10 (100) 0 (0) 2 (20) 8 (80) 0 (0) 0 (0) Janssen et al IMN (33) 7 (2) 28 (9) 146 (48) 128 (42) 0 (0) 0 (0) Arvinius et al IMN (66) - 0 (0) 0 (0) 0 (0) 0 (0) 56 (100) Piccioli et al IMN (100) 8 (10) 0 (0) 30 (38) 22 (28) 28 (35) 0 (0) Shemesh et al IMN (52) 0 (0) 0 (0) 11 (52) 5 (24) 5 (24) 0 (0) Fakler et al IMN (100) 3 (25) 0 (0) 0 (0) 0 (0) 0 (0) 12 (100) Weiss et al IMN (0) 0 (0) 108 (100) 0 (0) 0 (0) Harvey et al IMN (46) - 0 (0) 0 (0) 0 (0) 0 (0) 46 (100) Steensma et al IMN (33) 31 (38) 0 (0) 7 (9) 75 (91) 0 (0) 0 (0) Parker et al IMN (100) - 0 (0) 0 (0) 0 (0) 0 (0) 40 (100) Zacherl et al IMN (100) 4 (11) 0 (0) 8 (22) 14 (38) 10 (27) 5 (14) Rethnam et al IMN 11 9 (82) - 0 (0) 0 (0) 0 (0) 0 (0) 11 (100) Wedin et al IMN 24-6 (25) 0 (0) 4 (17) 20 (83) 0 (0) 0 (0) Datir et al IMN (76) - 0 (0) 0 (0) 17 (100) 0 (0) 0 (0) Moholkar et al IMN (54) - 0 (0) 0 (0) 0 (0) 0 (0) 48 (100) Ramakrishnan et al IMN 28 5 (18) - 0 (0) 0 (0) 28 (100) 0 (0) 0 (0) Piatek et al IMN 18-0 (0) 0 (0) 0 (0) 0 (0) 6 (33) 12 (67) Assal et al IMN 12 6 (50) 0 (0) 1 (8) 1 (8) 7 (58) 3 (25) 0 (0) Giannoudis et al IMN (0) 0 (0) 0 (0) 17 (100) 0 (0) Algan et al IMN 12-0 (0) 1 (8) 4 (33) 5 (42) 0 (0) 2 (17) Nargol et al IMN (100) 0 (0) 0 (0) 5 (50) 5 (50) 0 (0) 0 (0) 3 69

72 Chapter 3 Appendix 3: Affected anatomical area and use of cement per implant type for proximal femur metastatic fractures (%) (continued) Author, Year Implant Femurs Pathological fractures Cement Head /Neck Troch. Area Subtroch. Area Prox. Shaft Unspec. Yazawa et al IMN (67) 0 (0) 18 (100) 0 (0) 0 (0) 0 (0) Fasano et al IMN 11 9 (82) 3 (27) 0 (0) 0 (0) 11 (100) 0 (0) 0 (0) Behr et al IMN (44) - 0 (0) 8 (24) 7 (21) 19 (56) 0 (0) Janssen et al ORIF (40) 27 (60) 3 (7) 27 (60) 15 (33) 0 (0) 0 (0) Asavamongkolgul et al ORIF 27 9 (33) 19 (70) 0 (0) 27 (100) 0 (0) 0 (0) 0 (0) Steensma et al ORIF 19 6 (32) 8 (42) 0 (0) 16 (84) 3 (16) 0 (0) 0 (0) Parker et al ORIF (100) - 0 (0) 0 (0) 0 (0) 0 (0) 35 (100) Zacherl et al ORIF (100) 12 (100) 0 (0) 0 (0) 0 (0) 0 (0) 12 (100) Sarahrudi et al ORIF (100) - 0 (0) 0 (0) 0 (0) 0 (0) 15 (100) Wedin et al ORIF 13-3 (23) 1 (8) 8 (62) 4 (31) 0 (0) 0 (0) Ward et al ORIF (0) 0 (0) 0 (0) 0 (0) 13 (100) Rompe et al ORIF 25 8 (32) 25 (100) 0 (0) 0 (0) 0 (0) 0 (0) 25 (100) Broos et al ORIF (100) 0 (0) 5 (26) 14 (74) 0 (0) 0 (0) Behr et al ORIF 10 5 (50) - 2 (20) 5 (50) 3 (30) 0 (0) 0 (0) - = not available, Troch. = Trochanteric, Subtroch = Subtrochanteric, Prox. Shaft = Proximal shaft, Unspec. = Unspecified, EPR = Endoprosthesis, IMN = Intramedullary nail, ORIF = Open reduction internal fixation 70

73 CHAPTER 4 A Comparison Of Questionnaires For Assessing Physical Function In Patients With Lower Extremity Bone Metastases 4 S.J. Janssen N.R. Paulino Pereira K.A. Raskin M.L. Ferrone F.J. Hornicek C.N. van Dijk S.A. Lozano-Calderon J.H. Schwab Journal of Surgical Oncology 2016 Nov;114(6): Presented at: Musculoskeletal Tumor Society Annual Meeting 2015, Orlando, Florida, USA. Poster at: Massachusetts General Hospital Clinical Research Day 2015, Boston, Massachusetts, USA. American Academy of Orthopaedic Surgeons Annual Meeting 2016, Orlando, Florida, USA.

74 Chapter 4 ABSTRACT Objectives To assess: (1) the degree to which five different questionnaires measure physical function in patients with lower extremity bone metastases, (2) differences in coverage and reliability, and (3) difference in completion time. Design Prospective survey study. Setting Two tertiary care referral centers for orthopaedic oncology. Participants 100 of 115 (87%) consecutively invited patients with lower extremity metastases participated in this study between June 2014 and September Outcome Measures The PROMIS Physical Function Cancer, PROMIS Neuro-QoL Mobility, Toronto Extremity Salvage Score (TESS), Lower Extremity Function Score (LEFS), and Musculoskeletal Tumor Society score (MSTS) questionnaires. Results All questionnaires measured the same concept; demonstrated by high correlations (>0.7). Floor effect was absent, while ceiling effect was present in all, but highest for the PROMIS Neuro-QoL Mobility (7%). The standard error of measurement was below the threshold indicating reliability over a wide range of ability levels for the PROMIS-Physical Function, TESS, and LEFS. Completion time differed between questionnaires (p<0.001) and was shortest for the PROMIS questionnaires. Conclusions The PROMIS Physical Function Cancer questionnaire is the most useful for measurement of physical function in patients with lower extremity bone metastases. This is due to its reliability over a wide range of ability levels, validity, brevity, and good coverage. 72

75 Comparison Of Questionnaires For Assessing Physical Function INTRODUCTION Metastatic disease to bone is common in the lower extremity and is a major contributor to deteriorating function and quality of life. 1-4 Physical function is an important outcome measure to determine success of treatment. 3-5 The Musculoskeletal Tumor Society (MSTS) score and Toronto Extremity Salvage Score (TESS) are most commonly used to evaluate function in patients with bone metastases. 4,6,7 The MSTS score is developed to be completed by the clinician who evaluates function through six items. 7,8 The TESS is the first patient reported outcome for patients with musculoskeletal tumors and assesses function through 30 items. 6,9 Other non-tumor-specific patient reported outcome measures exist to assess physical function; however, questionnaires often lack responsiveness (i.e. cannot detect subtle differences in function) for the condition under study, suffer from floor and ceiling effect (i.e. situations in which a subject scores at the lower or upper limit of the questionnaire), or are burdensome due to the number of questions. 10 Therefore, the National Institute of Health funded the Patient-Reported Outcomes Measurement Information System (PROMIS) to develop standardized item banks to measure physical, mental, and social health. 10,11 PROMIS questionnaires allow for Computer Adaptive Testing (CAT) a dynamic selection of items wherein the response to each item guides the system s choice of the next item resulting in a tailored series of questions, reducing floor and ceiling effect and questionnaire burden, while maintaining precision. 11,12 Because function can vary substantially among patients with bone metastases, finding a reliable and valid instrument covering the full range of lower extremity function, while minimizing burden, is beneficial for both clinical and research practice. We therefore sought to determine the most effective, reliable, and efficient questionnaire by comparing the MSTS score, TESS, two PROMIS CAT questionnaires PROMIS Physical function Cancer and Neuro-QoL Mobility, and the generic Lower Extremity Function Scale (LEFS). We assessed which questionnaire is most useful for assessment of physical function in patients with lower extremity bone metastases. Specifically, our study questions evaluated: (1) the degree to which questionnaires measure the same underlying trait (physical function), (2) differences in coverage and reliability, and (3) difference in completion time. 4 METHODS Study Design Our institutional review board approved this prospective study and verbal informed consent was obtained. One hundred fifteen consecutive new or followup patients who presented to two tertiary care referral centers for orthopaedic oncology between June 2014 and September 2015 were invited. We approached English-speaking patients aged 18 years 73

76 Chapter 4 or older with a metastatic bone lesion, myeloma, or lymphoma of the lower extremity. Patients were enrolled either before or immediately after their visit with the surgeon. Twelve (10%) patients declined participation and 3 (3%) did not finish all questionnaires, leaving 100 patients for analysis. Patients were included once to avoid an effect of learning curve. We collected data using Assessment Center (Northwestern University Feinberg School of Medicine, Chicago, IL, USA, 2007) on a tablet computer. Assessment Center is a free online data collection tool allowing researchers to create surveys and includes the PROMIS CAT item banks. We calculated that a minimum sample size of 100 participants would provide 90% statistical power (beta 0.10; alpha 0.05) to detect an R-squared of 0.10 (moderate effect size: 0.32) on a bivariate correlation model between questionnaires. Outcome Measures Date of birth, race, sex, and tumor type were collected from medical records. Location of the lesion, education, marital status, presence of other disabling conditions, prior surgery or radiation therapy for the lesion, and presence of other bone or visceral metastases were collected from the patient. This was followed by the PROMIS Pain Intensity questionnaire. Subsequently, the PROMIS Physical Function Cancer (CAT), PROMIS Neuro-QoL Mobility (CAT), TESS lower extremity, LEFS, and MSTS lower extremity questionnaires were completed by the patient in random order (Assessment Center has a build in option to randomize the order of questionnaires). The survey consisted of 77 to 93 questions depending on the number of questions in the CAT questionnaires and the median completion time was 12 minutes (interquartile range: 9 to 15 minutes). The PROMIS Pain Intensity is generic and assesses how much pain a patient had over the past seven days and the pain intensity right now. The PROMIS CAT questionnaires consist of a dynamic selection of items. Computer Adaptive Testing administers a minimum of 4 questions and the response to each item guides the system s choice of the next most appropriate item until either the standard error of measurement (a measure of preciseness) drops below a specified level, or the participant answered the maximum number of questions (set at 12) Every PROMIS item comprises 5 response options on a Likert scale. The final score for PROMIS measures is represented by the t-score, a standardized score with a mean of 50 (United States general population) and a standard deviation of 10. A higher t-score represents more of the concept (i.e. better physical function) being measured. The PROMIS-Physical Function Cancer CAT is an item bank that evaluates general disability in cancer patients The PROMIS Neuro-QoL Mobility CAT is an item bank that measures lower extremity disability in patients with chronic neurological disease

77 Comparison Of Questionnaires For Assessing Physical Function The Toronto Extremity Salvage Score (TESS) assesses disability in patients with musculoskeletal tumors. 6,14 The questionnaire consists of 30 items rated on a 5-point Likert scale, but has an additional option this task is not applicable for me. The raw score is converted to a score ranging from 0 to 100, with a higher score indicating better function. The Lower Extremity Functional Scale (LEFS) measures lower extremity function in patients with lower extremity conditions (e.g. osteoarthritis). 15 The questionnaire consists of 20 items, and every item is rated on a 5-point Likert scale. The raw score is converted to a score from 0 to 100, with a higher score indicating better function. The Musculoskeletal Tumor Society (MSTS) score is developed to be completed by the clinician, and consists of six items rated on a 6-point Likert scale. 7,8 The raw score is converted to a score from 0 to 100, with a higher score indicating better function. Although the score is developed to be completed by the clinician, we administered the lower extremity version to patients. 4 Statistical Analysis Categorical variables were presented as frequencies and continuous variables were presented as median with interquartile range as histograms suggested non-normal distributions. We used exploratory factor analysis to assess if included questionnaires measure the same concept (physical function). Exploratory factor analysis is a method that looks for a common underlying mathematically derived trait measured by all questionnaires and subsequently correlates the included questionnaires with this trait. 16,17 A factor score of 1 indicates perfect correlation of the questionnaire with the underlying trait and 0 indicates no correlation. Subsequently, we assessed pairwise correlations between questionnaires using the Spearman rank correlation coefficient to further explore if questionnaires measure the same. A coefficient of 1 indicates perfect correlation, while a correlation of 0 indicates no correlation. 18 We considered a correlation coefficient above 0.7 as strong, indicating that questionnaires measure the same. 18,19 Coverage of the questionnaires was demonstrated by the floor and ceiling effect; floor effect is a term used to describe a situation in which a subject has scores at the lowest possible limit (floor), ceiling effect reflects scores at the upper limit (ceiling). 19,20 Reliability precision of measurement was assessed by calculating the standard error of measurement (SEM) as a function of ability per questionnaire using Item Response Theory. 21 The ability score per questionnaire is set to a standard scale with a mean of 0 and a standard deviation of 1 (Figure 1). The SEM quantifies the degree to which a measurement contains error; i.e. a patient will, by definition, have a true ability score and the SEM describes the likely range of actual scores such a patient might have with a specific questionnaire as a result of the unreliability of the questionnaire. 21,22 A SEM below 0.32 is accepted as precise, and corresponds to a Cronbach alpha internal consistency 75

78 Chapter 4 of ,22 The SEM for the range of ability scores per questionnaire is calculated through Graded Response Models, which fit ordered polytomous items. Difference in completion time between questionnaires was tested using the Friedman test. Subsequently, pairwise comparisons were made using the Wilcoxon signed-rank test; the alpha level was adjusted using Bonferroni correction for multiple comparisons (Bonferroni alpha level: 0.005, [0.05/10]). All statistical analyses were performed using Stata 14.0 (StataCorp LP, College Station, TX, USA). A two-tailed p value less than 0.05 was considered statistically significant unless stated otherwise. RESULTS Among the 100 included patients, 59 were women and the median age was 63 years (Table 1). Sixty-six patients had previous surgery and 56 had previous radiation therapy. All five questionnaires measure the same construct as demonstrated by substantial correlation (factor scores > 0.7) of every questionnaire with the underlying trait (Table 2). This Table 1: Baseline characteristics of patients with lower extremity bone metastases (n = 100) Median (Interquartile Range) Age (in years) 63 (54-70) Sex n Men 41 Women 59 Tumor location Femur 71 Acetabulum 14 Pelvis 12 Tibia 2 Fibula 1 Race White 93 Black or African American 7 Highest grade completed High school or below 32 Some college or Bachelor s degree 41 Professional degree 27 76

79 Comparison Of Questionnaires For Assessing Physical Function Table 1: Baseline characteristics of patients with lower extremity bone metastases (n = 100) (continued) Marital status Median (Interquartile Range) Married 66 Widowed 12 Single 11 Separated/divorced 7 Living with partner 4 Presence of other disabling conditions* Yes 26 No 73 Previous surgery for lesion Yes 66 No 34 Previous radiotherapy for lesion Yes 56 No 41 Unknown 3 Tumor distribution Bone metastases Breast 29 Renal cell 11 Prostate 9 Lung 8 Melanoma 4 Bladder 3 Leiomyosarcoma 3 Colorectal 2 Thyroid 2 Stomach 1 Hepatocellular 1 Oesophageal 1 Neuroendocrine 1 Sarcoma 1 Primary bone tumor Lymphoma 12 Myeloma 12 4 *One patient did not complete this question. 77

80 Chapter 4 Standard Error of Measurement Musculoskeletal Tumor Society Score PROMIS Neuro QoL Mobility CAT PROMIS Physical Function Cancer CAT Toronto Extremity Salvage Score Lower Extremity Function Score Ability score Figure 1: The standard error of measurement a measure of precision is demonstrated as a function of the ability scores per questionnaire. Each dot represents a patients ability and the corresponding standard error of measurement for the specific questionnaire. A higher standard error of measurement means less measurement precision. The ability score is standardized with a mean of 0 and a standard deviation of 1 for all questionnaires to allow for comparison between questionnaires. CAT = Computer Adaptive Testing. QoL = Quality of Life. finding is supported by high (> 0.7) interquestionnaire correlations, indicating substantial shared variance (Table 3). Floor effect was not seen in any of the questionnaires, meaning that all questionnaires differentiate well among disabled patients. Ceiling effect was seen in all questionnaires, but was highest (7%) for the PROMIS Neuro-QoL Mobility questionnaire (Table 4), meaning that this questionnaire performed worst when differentiating higher functioning patients. The standard error of measurement was below the threshold (0.32) of excellent measurement precision indicating reliability for all questionnaires in the midrange ability Table 2: Factor loadings (correlations) from exploratory factor analysis measuring the underlying trait Factor1 (underlying trait) PROMIS Physical Function CAT* PROMIS Neuro-QOL Mobility CAT* Toronto Extremity Salvage Score Lower Extremity Function Scale Musculoskeletal Tumor Society Score CAT = Computer Adaptive Testing, QOL = Quality of life, PROMIS = Patient Reported Outcomes Measurement Information System 78

81 Comparison Of Questionnaires For Assessing Physical Function Table 3: Spearman rank correlation coefficient between questionnaires Promis Physical Function CAT PROMIS Physical Function CAT Promis Neuro-QOL Mobility CAT PROMIS Neuro-QOL Mobility CAT Toronto Extremity Salvage Score Toronto Extremity Salvage Score Lower Extremity Function Scale Lower Extremity Function Scale Musculoskeletal Tumor Society Score Musculoskeletal Tumor Society Score CAT = Computer Adaptive Testing, QOL = Quality of life, PROMIS = Patient Reported Outcomes Measurement Information System; p values for all interquestionnaire correlations was below scores (theta between -1 and 1), except for the MSTS score (Figure 1). The MSTS score was least reliable as indicated by a high standard error of measurement for the complete range of ability scores. The PROMIS Neuro-QoL Mobility becomes less reliable at ability levels above 1, this means that it loses precision when assessing function in relatively well functioning patients (in line with the ceiling effect). The PROMIS-Physical Function, TESS, and LEFS questionnaires have excellent precision (low standard error of measurement) over a broad range of ability levels, meaning that these questionnaires measure physical function precisely regardless of how poor or good the physical function of the patient is. Completion time differed among questionnaires (p < 0.001) and was shortest for the PROMIS Physical Function and PROMIS Neuro-QoL Mobility, and longest for the TESS (Table 5). There was no difference in completion time between the PROMIS Physical Function and PROMIS Neuro-QoL Mobility questionnaires (p = 0.75), but these two questionnaires were quicker to complete than all other questionnaires (p < 0.001) (Table 5). DISCUSSION We aimed to determine if questionnaires measure the same concept of physical function in patients with bone metastases of the lower extremity and which is most useful effective, reliable, and efficient for its assessment. We found evidence that all five included questionnaires measure the same concept of physical function. The PROMIS Physical Function, LEFS, and TESS questionnaires are most reliable and have adequate coverage, but the PROMIS Physical Function is the quickest to complete, through its Computer Adaptive Testing administration. Our study has several limitations. First, the TESS items contain the option this task is not applicable for me ; we considered items missing if this option was indicated by the patient 79

82 Chapter 4 Table 4: Item completion rate, floor and ceiling effect, and score distribution per functional outcome questionnaire Questionnaire: Number of items per questionnaire Item completion rate (%) Median score (IQR) Range Possible range Floor effect n (%) Ceiling effect n (%) PROMIS Physical Function CAT* /456 (100) 36 (31-43) (0) 2 (2.0) PROMIS Neuro-QOL Mobility CAT* /491 (100) 40 (34-45) (0) 7 (7.0) Toronto Extremity Salvage Score /3000 (93.0) 75 (51-89) (0) 2 (2.0) Lower Extremity Function Scale /2000 (99.5) 51 (31-70) (0) 4 (4.0) Musculoskeletal Tumor Society Score 6 598/600 (99.7) 57 (37-70) (0) 4 (4.0) * PROMIS Physical Function v1.1 cancer item bank using Computer Adaptive Testing (CAT), PROMIS Neuro-Quality Of Life (QOL) v1.0 lower extremity (mobility) item bank using Computer Adaptive Testing (CAT), PROMIS = Patient Reported Outcomes Measurement Information System Table 5: Duration of questionnaire in seconds Multiple comparison: Median (IQR) Range p value PROMIS Physical Function CAT PROMIS Neuro- QOL Mobility CAT Toronto Extremity Salvage Score Lower Extremity Function Scale Musculo skeletal Tumor Society Score PROMIS Physical Function CAT 45 (33-62) PROMIS Neuro-QOL Mobility CAT 44 (35-63) Toronto Extremity Salvage Score 244 ( ) <0.001 < <0.001 Lower Extremity Function Scale 158 ( ) <0.001 <0.001 < Musculoskeletal Tumor Society Score 73 (55-97) <0.001 <0.001 <0.001 < CAT = Computer Adaptive Testing, QOL = Quality of life. IQR = interquartile range, PROMIS = Patient Reported Outcomes Measurement Information System 80

83 Comparison Of Questionnaires For Assessing Physical Function (7.0%) and did not include these in the overall score calculation as per the questionnaires guidelines. 6 We feel that patients were more likely to indicate this option when they were more disabled and therefore see this as an important limitation. However, we believe that this did not compromise our reliability and completion time analyses. Second, patients completed questionnaires at the outpatient clinic; we did not include patients admitted to the emergency room or ward. We might have missed a subset of more disabled patients, which could have influenced the floor effect. However, all questionnaires differentiated well among the most limited patients. Third, we did not assess test-retest reliability. We see this as a minor limitation as the concept being measured physical function might change rapidly in patients with metastatic cancer, rendering test-retest reliability analyses less valid. 23 Fourth, the survey was lengthy and questions within and between surveys seem very similar, which might have bothered patients, resulting in less representative answers. We believe this is a minor limitation as we randomized the order of surveys. Fifth, the study included a heterogeneous patient sample i.e. patients were included regardless of disease and treatment stage ; however, we see this as a strength rather than a limitation as we aimed to compare questionnaires over a wide range of functional levels. Unfortunately, this did not allow us to assess the impact of specific treatments or disease on physical function. Exploratory factor analysis can uncover the underlying concept being measured by a number of questionnaires. This statistical method describes variability among observed, correlated variables in terms of a potentially lower number of unobserved variables traits to identify relationships between measured variables. In terms of our study; are the five questionnaires comparable in what they are measuring? We demonstrated high factor loadings and substantial interquestionnaire correlations. This suggests that questionnaires measure the same concept in patients with lower extremity bone metastases despite the different conditions for which the questionnaires were originally developed. 6-8,10-13,15 Previous studies were limited comparing only two (MSTS score and TESS) of our five selected questionnaires; an interquestionnaire correlation of 0.48 was found in a sample of 57 patients with soft tissue sarcoma of the thigh; however, the MSTS score was completed by the clinician and not by the patient. 6 Tracking physical function for research studies and clinical use should be done using patient reported outcome measures, as clinicians outcome assessment tends to underestimate a patients symptoms and their impact on daily live In our study, coverage was good no floor and minimal ceiling effect for all questionnaires, except for the PROMIS Neuro-QoL Mobility demonstrating a ceiling effect of 7%. This means that the PROMIS Neuro-QoL Mobility questionnaire cannot distinguish higher functioning patients. 13 We therefore discourage the use of this questionnaire in patients with extremity bone metastatic disease. The PROMIS Physical Function, TESS, LEFS, and MSTS score differentiate well among both disabled (no floor effect) and higher functioning 4 81

84 Chapter 4 (minimal ceiling effect) patients with lower extremity bone metastases. A study by Davis et al. 29 compared the TESS and MSTS score in 97 patients with lower extremity sarcomas, and found 2% floor effect for the TESS and MSTS score (0% in our study), but a ceiling effect of 33% for the MSTS score (4% in our study). However, the MSTS score was completed by the clinician, potentially overestimating the patients function. In addition, patients were younger in comparison to our study. A study by Binkley et al. 15 demonstrated no floor or ceiling effect (4% ceiling effect in our study) of the LEFS questionnaire in a sample of patients with lower extremity conditions. The PROMIS Physical Function questionnaire had no floor and only 2% ceiling effect; this superior coverage with only 4 to 12 questions is a result of the dynamic item selection via Computer Adaptive Testing. PROMIS Physical Function questionnaires are also available as short-forms (fixed number of questions). However, these short-forms suffer from ceiling effects as demonstrated by a study in 4,840 cancer patients; 34% ceiling effect for the 4-item short-form, 25% for the 6-item short-form, 13% for the 10-item short-form, and 12% for the 16-item short form. 12 We therefore prefer to use Computer Adaptive Testing for outcome assessment. We found substantial reliability (i.e. measurement precision) of the PROMIS Physical Function, TESS, and LEFS over a large range of ability levels using Item Response Theory. Traditionally, reliability has been assessed using a single index of internal consistency: Cronbach alpha. However, the issue of a single index is that it assumes the reliability to be the same for all participants over the range of ability scores. 21 The advantage of Item Response Theory is that it demonstrates measurement precision over a range of ability scores. 21 Ability scores at the limits of a questionnaire tend to have more error than those in the middle of the ability range, which is also seen in our study. 21 This means that the precision of measurement decreases when measuring relatively disabled or higher functioning patients. In general, longer questionnaires will provide more information about a patients physical function and would therefore have less measurement error; however, the dynamic selection of items through Computed Adaptive Testing can optimize reliability while minimizing the number of items. As the MSTS score is to be completed by a clinician, it might be less well understood by patients; this could have compromised its reliability in our study. We therefore discourage the use of MSTS score to capture patient reported outcomes. The relatively high standard error of measurement of the PROMIS Neuro-QoL Mobility questionnaire in the higher functioning patients is in line with its ceiling effect: higher functioning patients are not distinguished nor measured accurately by this questionnaire. Obtaining research data in the least amount of time is key in reducing questionnaire burden and attrition of patients. We found that the PROMIS Physical Function and PROMIS Neuro-QoL Mobility are quickest to complete; 73% of the participants completed these questionnaires within a minute, compared to an average completion time of more than 4 minutes for the TESS, and almost 3 minutes for the LEFS. 82

85 Comparison Of Questionnaires For Assessing Physical Function In conclusion, this study demonstrated that the PROMIS Physical Function Cancer questionnaire is the most useful to determine physical function for patients with lower extremity bone metastases. This is due to its reliability over a wide range of ability levels, validity, brevity, and good coverage through Computer Adaptive Testing. REFERENCES 1. Bickels J, Dadia S, Lidar Z. Surgical management of metastatic bone disease. J Bone Joint Surg Am. Jun 2009; 91(6): Quinn RH, Randall RL, Benevenia J, Berven SH, Raskin KA. Contemporary management of metastatic bone disease: tips and tools of the trade for general practitioners. J Bone Joint Surg Am. Oct ; 95(20): Steensma M, Healey JH. Trends in the surgical treatment of pathologic proximal femur fractures among Musculoskeletal Tumor Society members. Clin Orthop Relat Res. Jun 2013; 471(6): Talbot M, Turcotte RE, Isler M, Normandin D, Iannuzzi D, Downer P. Function and health status in surgically treated bone metastases. Clin Orthop Relat Res. 2005; 438: Steensma M, Boland PJ, Morris CD, Athanasian E, Healey JH. Endoprosthetic treatment is more durable for pathologic proximal femur fractures. Clin Orthop Relat Res. 2012; 470: Davis AM, Wright JG, Williams JI, Bombardier C, Griffin A, Bell RS. Development of a measure of physical function for patients with bone and soft tissue sarcoma. Qual Life Res. Oct 1996; 5(5): Enneking WF, Dunham W, Gebhardt MC, Malawar M, Pritchard DJ. A system for the functional evaluation of reconstructive procedures after surgical treatment of tumors of the musculoskeletal system. Clin Orthop Relat Res. Jan 1993(286): Rebolledo DC, Vissoci JR, Pietrobon R, de Camargo OP, Baptista AM. Validation of the Brazilian version of the musculoskeletal tumor society rating scale for lower extremity bone sarcoma. Clin Orthop Relat Res. Dec 2013; 471(12): Cheng EY. Prospective quality of life research in bony metastatic disease. Clin Orthop Relat Res. Oct 2003(415 Suppl): S Garcia SF, Cella D, Clauser SB, et al. Standardizing patient-reported outcomes assessment in cancer clinical trials: a patient-reported outcomes measurement information system initiative. J Clin Oncol. Nov ; 25(32): Rose M, Bjorner JB, Becker J, Fries JF, Ware JE. Evaluation of a preliminary physical function item bank supported the expected advantages of the Patient-Reported Outcomes Measurement Information System (PROMIS). J Clin Epidemiol. Jan 2008; 61(1): Jensen RE, Potosky AL, Reeve BB, et al. Validation of the PROMIS physical function measures in a diverse US population-based cohort of cancer patients. Qual Life Res. May Gershon RC, Lai JS, Bode R, et al. Neuro-QOL: quality of life item banks for adults with neurological disorders: item development and calibrations based upon clinical and general population testing. Qual Life Res. Apr 2012; 21(3): Clayer M, Davis A. Can the Toronto Extremity Salvage Score produce reliable results when used online? Clin Orthop Relat Res. Jun 2011; 469(6):

86 Chapter Binkley JM, Stratford PW, Lott SA, Riddle DL. The Lower Extremity Functional Scale (LEFS): scale development, measurement properties, and clinical application. North American Orthopaedic Rehabilitation Research Network. Phys Ther. Apr 1999; 79(4): Revicki DA, Chen WH, Harnam N, et al. Development and psychometric analysis of the PRO- MIS pain behavior item bank. Pain. Nov 2009; 146(1-2): Amtmann D, Cook KF, Jensen MP, et al. Development of a PROMIS item bank to measure pain interference. Pain. Jul 2010; 150(1): Dancey C, Reidy J. Statistics without Maths for Psychology: using SPSS for Windows: London: Prentice Hall; Tyser AR, Beckmann J, Franklin JD, et al. Evaluation of the PROMIS physical function computer adaptive test in the upper extremity. J Hand Surg Am. Oct 2014; 39(10): e Hung M, Nickisch F, Beals TC, Greene T, Clegg DO, Saltzman CL. New paradigm for patientreported outcomes assessment in foot & ankle research: computerized adaptive testing. Foot Ankle Int. Aug 2012; 33(8): Frank B. Baker. The Basics of Item Response Theory: ERIC Clearinghouse on Assessment and Evaluation; Tighe J, McManus IC, Dewhurst NG, Chis L, Mucklow J. The standard error of measurement is a more appropriate measure of quality for postgraduate medical assessments than is reliability: an analysis of MRCP(UK) examinations. BMC Med Educ. 2010; 10: Davidshofer KR, Murphy CO. Psychological testing: principles and applications. 6 ed: Upper Saddle River, N.J.: Pearson/Prentice Hall; Nelson EC, Eftimovska E, Lind C, Hager A, Wasson JH, Lindblad S. Patient reported outcome measures in practice. BMJ. 2015; 350: g Nelson E, Conger B, Douglass R, et al. Functional health status levels of primary care patients. JAMA. Jun ; 249(24): Basch E. The missing voice of patients in drug-safety reporting. N Engl J Med. Mar ; 362(10): Basch E, Jia X, Heller G, et al. Adverse symptom event reporting by patients vs clinicians: relationships with clinical outcomes. J Natl Cancer Inst. Dec ; 101(23): Pakhomov SV, Jacobsen SJ, Chute CG, Roger VL. Agreement between patient-reported symptoms and their documentation in the medical record. Am J Manag Care. Aug 2008; 14(8): Davis AM, Bell RS, Badley EM, Yoshida K, Williams JI. Evaluating functional outcome in patients with lower extremity sarcoma. Clin Orthop Relat Res. Jan 1999(358):

87 CHAPTER 5 Predicting Pathological Fracture In Femoral Metastases Using A Clinical CT Scan Based Algorithm: A Case-Control Study 5 S.J. Janssen N.R. Paulino Pereira T.A. Meijs M.A. Bredella M.L. Ferrone C.N. van Dijk J.A.M. Bramer S.A. Lozano-Calderon J.H. Schwab Journal of Orthopaedic Science 2018 Mar; 23(2): Poster at: Musculoskeletal Tumor Society Annual Meeting 2015, Orlando, Florida, USA.

88 Chapter 5 ABSTRACT Objectives To assess if there was a difference in attenuation measurements (in Hounsfield units HU) and geometric distribution of HU between femora with metastatic lesions that fracture, and metastatic lesions that did not fracture nor underwent prophylactic fixation. Design Retrospective case-control imaging study. Setting Two tertiary care referral centers for orthopaedic oncology. Participants Nine patients who had a CT scan of a bone metastatic lesion in the femur between June 1999 and July 2013 and developed a pathological fracture within the subsequent year were matched to 27 controls who had femoral metastasis that did not fracture and did not undergo prophylactic fixation. Outcome Measures All femora were delineated in axial CT slices using a region of interest (ROI) tool; the HU within these ROIs were used to calculate: (1) the cumulative HU of the affected over the nonaffected side per slice and presented as a percentage, and (2) the cumulative HU accounting for geometric distribution (polar moment of HU). We repeated the analyses including cortical bone only (HU of 600 and above). Results CT-based calculations did not differ between patients with a lesion that fractured and those that did not fracture nor underwent prophylactic fixation when analyzing all tissue. However, when including cortical bone only, the pathological fracture group had a lower cumulative HU value compared to the no fracture and no fixation group for the weakest cross-sectional CT image (pathological fracture group, mean: 71, SD: 23 and no fracture and no prophylactic fixation group, mean: 85, SD: 18, p = 0.042) and the complete lesion analysis (pathological fracture group, mean: 78, SD: 21 and no fracture and no prophylactic fixation group, mean: 92, SD: 15, p = 0.032). Conclusions The demonstrated CT-based algorithms can be useful for predicting pathological fractures in metastatic lesions. 86

89 Predicting Pathological Fractures in Femoral Metastasis INTRODUCTION Osseous metastatic disease can weaken the bone which may lead to a pathological fracture. Historically, the fracture risk is estimated based on radiographic measurements (e.g. defect size > 30mm, more than 50% circumferential cortical destruction, lesion type) and symptoms (e.g. pain), or a combination (e.g. Mirels classification); however, the specificity of these measures in predicting fracture occurrence is low. 1-3 For example, the specificity of the Mirels classification (using a Mirels score of 9 and above) is 33%; suggesting that two-thirds of patients will undergo unnecessary prophylactic surgical stabilization using this classification. 1,3 New techniques such as finite element analysis and CT-based rigidity analysis have been developed to better predict fracture risk. 4 Finite element analysis is a technique that uses mathematical models to simulate mechanical behavior of bone under loading. 4-6 Snyder et al. 7 developed CT-based rigidity analysis using quantitative CT scans (QCT) including a hydroxyapatite bone phantom to calculate bone mineral density (BMD) from attenuation coefficients (Hounsfield units - HU), using empirically derived formulas to approximate the rigidity of the affected bone Both methods are superior to radiographic measurements in predicting the development of fractures. 4 However, these methods are labor intensive or require inclusion of a calibration phantom during the CT scan and subsequent conversion of HU into bone mineral density to approximate bone rigidity and compare the affected to the contralateral unaffected side. Several studies suggest a positive correlation between HU with BMD and the mechanical strength of bone. 11,12 Based on this knowledge, and the methods described by Snyder et al. 7, we developed an algorithm that is based on the HU in clinically obtained CT scans without inclusion of a calibration phantom. Such an algorithm would be easier to apply and more widely available. Our study aim was to describe our technique and test if this algorithm could be useful for predicting the development of a pathological fracture in patients with metastatic femoral lesions. Specifically, we assessed whether there was a difference in cumulative HU and polar moment of HU (i.e. accounting for geometric distribution of bone) between femora with metastatic lesions that fracture, and metastatic lesions that did not fracture nor underwent prophylactic fixation. In addition, we compared these measures between femora with metastatic lesions that fracture and metastatic lesions of femora that underwent prophylactic fixation to better understand the indication for prophylactic fixation in these patients. We also compared these measures between metastatic lesions that fracture and femora with no metastases to better understand possible variation in HU measurements among healthy individuals. 5 87

90 Chapter 5 METHODS Study Design Our institutional review board approved this retrospective case-control imaging study. We identified 9 patients who had a CT scan of a bone metastatic lesion in the femur at one of two tertiary care referral centers for orthopaedic oncology between June 1999 and July 2013 and developed a pathological fracture (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] codes: or ) within the subsequent year. Presence of a pathological fracture was confirmed based on clinical symptoms, imaging, and intraoperatively. We included patients with lytic lesions from bone metastasis, myeloma, and lymphoma. Exclusion criteria were: (1) sclerotic bone metastatic lesions (predominantly prostate cancer), (2) no lesion-free CT slices above and below the lesion, (3) no inclusion of the contralateral femur on the CT scan, and (4) abnormalities of the contralateral bone hindering comparison (e.g. bilateral lesions, implants). These 9 cases who developed a fracture within one year after their CT were matched on a 1:3:3:3 basis to controls who had: (control group 1) femoral metastasis that did not fracture and did not undergo prophylactic fixation and had at least the same followup duration as the matched case, (control group 2) femoral metastasis that underwent prophylactic fixation, and (control group 3) no femoral metastasis (Figure 1A). Our primary outcome was the comparison of patients who developed a fracture through a metastatic lesion (cases) to patients with femoral metastasis that did not fracture and did not undergo prophylactic fixation (control group 1). We compared the patients who developed a fracture through a metastatic lesion (cases) to patients who underwent prophylactic fixation (control group 2) and patients with no femoral metastasis (control group 3) to understand indications for prophylactic fixation and how HU measurements varies within healthy individuals. Controls were matched for sex, age (range: 10 years), and location of the lesion. No formal guidelines for prophylactic fixation exist at our clinic; In general, patients underwent prophylactic fixation of their femur when the surgeon deemed the lesion at considerable risk of fracture based on the amount of bone destruction and/or pain on load bearing in patients who were expected to live longer than 30 days. Outcome Measures CT scans were obtained for clinical purposes using different scanners (24 different scanning stations from 3 manufacturers). We retrieved the Digital Imaging and Communications in Medicine files of all CT scans through the Picture Archiving Communications System database. All CT scans were imported into OsiriX medical image viewer application (Free 32-bit open-source version 6.5.2, Bernex, Switzerland). The built-in region of interest (ROI) tool was used to delineate the femur per axial CT slice that includes the lesion (Figure 1B, red line) with a 2 cm lesion free margin proximally and distally. Subsequently, the contra- 88

91 Predicting Pathological Fractures in Femoral Metastasis 5 Figure 1A: 65 year old female with renal cell carcinoma and multiple osseous metastases. The anteroposterior radiograph demonstrates a subtrochanteric lytic lesion of the right femur. Figure 1B: Axial CT scan of the right femur including the metastatic lesion (left image) and the unaffected left femur (right image) were delineated using a region of interest (ROI) tool (red line). The pixels including their Hounsfield units within these ROIs were retained. Subsequently, square ROIs (green line) were drawn including the delineated affected and unaffected femora; all pixels within these square ROIs were exported per axial slice. lateral side including the same region was delineated (Figure 1B, red line). We corrected for unequal level between the left and right femur using the femoral head or condyles as reference; However, we were only able to correct for this within the limits of the slice thickness (range: 2 to 7.5 millimeters) and were only able to correct for height and not for rotation or angulation. Square ROIs were then drawn including the delineated affected and 89

92 Chapter 5 unaffected femoral region per axial slice (Figure 1B, green line). The HU of each pixel within the two delineated femoral ROIs were exported per slice and side into.xml (Extensible Markup Language) files. We only included pixels with positive ( 0 [= water]) Hounsfield units (i.e. excluding fat). Analysis was performed using Stata 14.0 (StataCorp LP, College Station, TX, USA) and data were transformed to allow comparison of HU and their geometric distribution between both femora per axial slice (Figure 2). We calculated a central axis of both femora per slice based on the pixels HU and their geometric distribution (Figure 2). 7,9,10 Subsequently, we calculated the following values: (1) The cumulative HU of the affected over the nonaffected side per slice and presented this as a percentage, and (2) The polar moment of HU; which is the cumulative HU accounted for geometric distribution by taking the sum of the product of the HU relative to its central axis of the affected over the nonaffected side per slice and again presented as a percentage (Figure 3). Based on these values we generated the response variables: (1) the weakest cross sectional slice based on the cumulative HU; (2) the weakest cross sectional slice based on the polar moment of HU; (3) the average cumulative HU of the complete lesion (i.e. including all axial slices demonstrating the lesion); and (4) the average polar moment of HU of the complete lesion (i.e. including all axial slices demonstrating the lesion). We repeated these four analyses including cortical bone only; all pixels with HU of 600 and above were retained, while those below this Figure 2: The delineated femora presented in Figure 1B are demonstrated using heatmaps with the Hounsfield units (HU) set to a color scale placed in the middle. The central axis per side per axial slice was calculated: x = n i = 1 xi HUi, y = n n i = 1 yi HUi. The percentage cumulative HU of the affected (right) i = 1 HUi n i = 1 HUi over the unaffected (left) side per slice: Percentage = n i = 1 HUi right * 100. The polar moment of HU (percentage cumulative HU accounted for their geometric distribution) of the affected over the unaffected n i = 1 HUi left side per slice: Percentage = n i = 1 HUi right ((xi x )2 + (y i ȳ) 2 ) n i = 1 HUi left ((xi x )2 + (y i ȳ) 2 ) *

93 Predicting Pathological Fractures in Femoral Metastasis 5 Figure 3: The percentage cumulative HU of the affected over the unaffected side (blue line) and accounted for its geometric distribution (red line) per slice. The weakest cross-section is slice 7 according to the four calculations; this axial slice is demonstrated in Figure 1B and Figure 2. threshold trabecular bone and tumor tissue were excluded. 13,14 We accounted for slice thickness in the complete lesion calculations; the median slice thickness was 5 millimeter and ranged from 2 to 7.5 millimeters. Settings of CT scans differed: the median tube voltage was 120kV (range: 100 to 140kV), the median exposure time was 500msec (range: 500 to 1401msec), and the median tube current was 229mA (range: 20 to 648mA). We also assessed conventional risk factors (i.e. risk factors based on clinical or radiological findings) that are currently most commonly used to assess fracture risk: the Mirels classification, 1,3,15 the degree of circumferential cortical destruction (greater or less than 50%), the length of cortical destruction, 16,20,21 and presence of pain on load bearing. 1,21 We used axial, coronal, and sagittal CT slices as radiographs were not always available. The Mirels classification is based on 4 aspects of the bone lesion (site of the lesion, pain, lesion type, and lesion size). 1,3,15 The score varies from 4 to 12, 4 being at low risk of fracture and 12 being at high risk of fracture. According to Mirels, a score of 9 and above is most accurate for determining an impending fracture. 1 Description of pain level in the medical record was used to assess if the pain was mild (including no pain, mild pain, or a pain score below 4 [numeric pain rating scale: 0-10]), moderate (including moderate pain or a pain score between 4 and 8), or functional (including severe pain, significant 91

94 Chapter 5 pain, a pain score of 8 and higher, or any pain that increased with loading) at time of the CT scan. Fifty percent circumferential destruction was estimated based on the axial CT The cortical length of the lesion was defined as the longest destruction of the cortical bone on the coronal or sagittal CT images in millimeters. 16,20,21 Presence of functional pain was defined in accordance with the Mirels classification. 1,21 The following explanatory variables at time of the CT scan were derived from medical records: age, sex, lesion location, primary tumor, Eastern Cooperative Oncology Group (ECOG) performance status (describing a patients level of functioning in terms of fully active (0) to completely disabled (4)), and previous radiation therapy. Tumor type, ECOG performance status, radiation therapy, functional pain, the Mirels classification, circumferential destruction, and cortical length of the lesion were only reported for the patients with osseous metastasis as these did not apply for the controls with no metastatic lesions. Statistical Analysis Categorical variables were described as frequencies with percentages. Continuous variables were described as mean with standard deviation (SD) as histograms suggested no substantial skew or outliers. We compared the 9 cases that develop a pathological fracture with each matched control group no fracture and no fixation group, prophylactic fixation group, and no lesion group in baseline using the Fisher exact test for categorical variables, and the Student T-test test for continuous variables. We used mixed-effects linear regression with random effects for the case-control matched groups and fixed effects for the group assignment to assess if there was a difference in the continuous response variables between the cases and the controls. We used mixed-effects logistic regression with random effects for the case-control matched groups and fixed effects for the group assignment to assess if there was a difference in the dichotomous response variables between the cases and the controls. Area under the curve (AUC) from receiver operator characteristic (ROC) curve analysis was used to compare performance of the CT-based calculations and conventional risk factors (Appendix 1). In addition, ROC analysis was used to determine an optimal cutoff value of the CT-based calculations for predicting development of a pathological fracture (Appendix 2). The AUC analysis was only performed for those CT-based calculations that differed between the patients with a metastatic lesion that fractured and those that did not fracture nor underwent prophylactic fixation and conventional risk factors. All statistical analyses were performed using Stata 14.0 (StataCorp LP, College Station, TX, USA). A two-tailed p value less than 0.05 was considered statistically significant. ECOG performance status was missing for 4 patients; these were not included in the baseline analyses. 92

95 Predicting Pathological Fractures in Femoral Metastasis RESULTS Baseline Characteristics The overall mean age was 63 years (SD 12), and 44% (40/90) of the patients were male. Lung and Breast carcinoma were the most common primary tumors and accounted for Table 1: Baseline characteristics Pathological fracture group (n = 9) No fracture and no fixation group (n = 27) Prophylactic fixation group (n = 27) No lesion group (n = 27) Mean (SD) Mean (SD) p value* Mean (SD) p value* Mean (SD) p value* Age in years 65 (±13) 65 (±12) (±12) (±11) Sex n (%) n (%) n (%) n (%) Men 4 (44) 12 (44) 12 (44) 12 (44) Women 5 (56) 15 (56) 15 (56) 15 (56) Lesion location Trochanteric area 4 (44) 12 (44) 12 (44) 12 (44) Subtrochanteric area 2 (22) 6 (22) (22) (22) Femoral shaft 3 (33) 9 (33) 9 (33) 9 (33) Primary tumor type Breast 0 (0) 6 (22) 7 (26) Lung 4 (44) 6 (22) 8 (30) Melanoma 2 (22) 5 (19) 1 (3.7) Renal cell carcinoma 0 (0) 2 (7.4) 2 (7.4) Multiple Myeloma 1 (11) 0 (0) 3 (11) Other 2 (22) 8 (30) 6 (22) Eastern Cooperative Oncology Group 0 2 (22) 8 (31) 1 (4) 1 5 (56) 11 (42) 15 (63) 2 1 (11) 4 (15) (21) 3 1 (11) 3 (12) 3 (13) 4 0 (0) 0 (0) 0 (0) Previous radiation therapy Yes 1 (11) 1 (4) 4 (15) No 8 (89) 26 (96) 23 (86) N/A N/A N/A N/A N/A N/A 5 N/A = not applicable. SD = Standard Deviation. * p values compare the control groups (no fracture and no fixation, prophylactic fixation, and no lesion groups) with the pathological fracture group. bold font indicates a significant difference (two-tailed p value below 0.05). ECOG status was available for 9 patients with a pathological fracture, 26 with a lytic lesion, and 24 with an impending fracture. 93

96 Chapter 5 49% of the lesions in our cohort. The mean time between the CT scan of the bone metastasis and the pathological fracture for the 9 cases was 63 days (range: 1 to 154 days). We found no difference in baseline characteristics between the pathological fracture cases and any of the control groups (Table 1). Structural CT Analysis The CT-based calculations did not differ between the patients with a metastatic lesion that fractured (pathological fracture group) and those that did not fracture nor underwent prophylactic fixation (no fracture and no fixation group) when analyzing all pixels i.e. including trabecular bone, cortical bone, and tumor tissue (Table 2, comparing first and second column). However, when including cortical bone only (retaining pixels with 600 HU), we found that the pathological fracture group had a lower cumulative HU value as compared to the no fracture and no fixation group for both the weakest cross-sectional axial CT slice (pathological fracture group, mean: 71, SD: 23 and no fracture and no fixation group, mean: 85, SD:18, p = 0.042) and the complete lesion analysis (pathological fracture group, mean: 78, SD: 21 and no fracture and no fixation group, mean: 92, SD: 15, p = 0.032). In addition, we found no difference between the metastatic lesion group that fractured (pathological fracture group) and those that underwent prophylactic fixation (prophylactic fixation group) for any of the CT-based calculations (Table 2, comparing first and third column). However, as expected we did find a difference for all CT-based calculations when comparing the pathological fracture group to the healthy controls with no metastatic femoral lesions (Table 2, comparing first and last column). Standard deviations of the structural CT scan analyses in the pathological fracture, no fracture and no fixation, and prophylactic fixation groups were high indicating substantial variation in the amount of bone destruction between patients (Table 2, first three columns). The variation was relatively low in the healthy controls without metastatic lesions (standard deviations were less than 9 for the weakest cross sectional slice) indicating that measurements are reasonably reliable as one would expect only limited variation between healthy femora within an individual (Table 2, last column). Conventional Risk Factors There was no difference in length of cortical destruction (p = 0.086), 50% circumferential destruction (p = 0.086), and rate of functional pain (p = 0.236) when comparing the patients who develop a pathological fracture and those that do not (Table 2, comparing first and second column). We did find a higher proportion of patients with a Mirels score of 9 or above in the pathological fracture group as compared to the no fracture and no fixation group (p = 0.046). In addition, there was no difference in the conventional risk factors (length of cortical destruction, 50% circumferential destruction, Mirels score, and functional pain) between 94

97 Predicting Pathological Fractures in Femoral Metastasis Table 2: Structural analysis of CT scans and conventional risk factors Structural analysis of CT scans, all pixels Pathological fracture group (n = 9) Percentage (affected over nonaffected side): No fracture and no fixation group (n = 27) Prophylactic fixation group (n = 27) No lesion group (n = 27) Weakest cross-sectional slice: Mean (SD) Mean (SD) p value* Mean (SD) p value* Mean (SD) p value* Cumulative Hounsfield units 85 (±14) 90 (±14) (±15) (±5.3) <0.001 Polar moment of Hounsfield units 85 (±12) 90 (±13) (±13) (±7.3) <0.001 Complete lesion: Cumulative Hounsfield units 91 (±13) 95 (±12) (±15) (±6.2) Polar moment of Hounsfield units 91 (±11) 96 (±13) (±11) (±8.2) Structural analysis of CT scans, cortical bone only Percentage (affected over nonaffected side): Weakest cross-sectional slice: Mean (SD) Mean (SD) p value* Mean (SD) p value* Mean (SD) p value* Cumulative Hounsfield units 71 (±23) 85 (±18) (±21) (±8.1) <0.001 Polar moment of Hounsfield units 73 (±25) 86 (±17) (±20) (±8.9) <0.001 Complete lesion: Cumulative Hounsfield units 78 (±21) 92 (±15) (±21) (±11) <0.001 Polar moment of Hounsfield units 81 (±21) 94 (±17) (±17) (±10) <0.001 Conventional risk factors: Mean (SD) Mean (SD) p value* Mean (SD) p value* Mean (SD) p value* Cortical length destruction (in mm) More than 50% circumferential destruction 35 (±20) 25 (±18) (±20) N/A N/A n (%) n (%) n (%) n (%) 4 (44) 5 (19) (37) N/A N/A Mirels score of 9 and above 7 (78) 10 (37) (96) N/A N/A Functional pain 3 (33) 4 (15) (67) N/A N/A N/A = not applicable. SD = Standard Deviation. CT = Computed Tomography * p values compare the control groups (no fracture and no fixation, prophylactic fixation, and no lesion groups) with the pathological fracture group. bold font indicates a significant difference (two-tailed p value below 0.05). Only including pixels with Hounsfield units 600 The mean Mirels score was 9.3 (±1.2) in the pathological fracture group, 8.5 (±1.1) in the no fracture and no fixation group, and 10 (±1.0) in the prophylactic fixation group. 95

98 Chapter 5 the pathological fracture group and the prophylactic fixation group (Table 2, comparing first and third column). Performance Of Predictors The AUC for differentiating metastatic lesions that fractured (pathological fracture group) from those that did not fracture nor underwent prophylactic fixation (no fracture and no fixation group) was 0.69 (95% CI: ) for the weakest cross-sectional axial CT slice analysis (cortical bone only) and 0.69 (95% CI: ) for the complete lesion analysis (cortical bone only), and 0.72 (95% CI: ) for the Mirels score; however, the performance (AUC) of the CT based calculations and conventional risk factors did not differ as demonstrated by substantial overlap of 95% confidence intervals (Table 3). When looking for an optimal cut-off (Appendix 2), we found that a threshold of 87% for the weakest cross-sectional CT slice translates into a sensitivity of 89% and a specificity of 48%; meaning that when applying this threshold to our cohort we would have missed one pathological fracture (1 out of 9), but overtreated fourteen metastatic lesions (14 out of 27). A threshold of 90% for the complete lesion analysis translates into a sensitivity of 89% (i.e. missed 1 out of 9) and a specificity of 52% (i.e. overtreated 13 out of 27). The predefined threshold of 9 for the Mirels score translates into a sensitivity of 78% (i.e. missed 2 out of 9) and a specificity of 63% (i.e. overtreated 10 out of 27). Table 3: Area under the curve for the structural analysis of CT scans and conventional risk factors Structural analysis of CT scans, cortical bone only Area under the curve (95% CI) Weakest cross-sectional slice, Cumulative Hounsfield units (in %) 0.69 ( ) Complete lesion, Cumulative Hounsfield units (in %) 0.69 ( ) Conventional risk factors: Cortical length destruction (in mm) 0.68 ( ) More than 50% circumferential destruction 0.63 ( ) Mirels score (continuous) 0.72 ( ) Functional pain 0.59 ( ) 95% CI = 95% Confidence interval Only including pixels with Hounsfield units 600 p value DISCUSSION Radiographic measures and clinical symptoms lack accuracy for predicting development of pathological fractures. 1-3 New techniques, such as finite element analysis and QCT scan based rigidity analysis have been developed to better predict fracture risk, but are labor intensive or require inclusion of a calibration phantom during the CT scan and subsequent 96

99 Predicting Pathological Fractures in Femoral Metastasis conversion of HU into bone mineral density to approximate bone rigidity. We developed and tested an algorithm that is based on the HU using clinically obtained CT scans without inclusion of a phantom. We found that our CT-based calculations did not differ between metastatic lesions that fractured and lesions that did not fracture (nor underwent prophylactic fixation) when analyzing all tissue; however, we did find a difference between these groups when including cortical bone only. Our study has limitations. First, the nine patients in our cohort that developed a pathological fracture were only a small subset of patients that might have developed a fracture if they would not have undergone prophylactic fixation; many patients undergo prophylactic fixation if a fracture seems to be impending. This selection bias is an important limitation of our study and we therefore explored the differences between the pathological fracture group and a matched group that underwent prophylactic fixation because of anticipated pathological fracture. These analyses demonstrate that baseline characteristics, our CTbased structural analysis, and the conventional risk factors did not differ between the pathological fracture and prophylactic fixation group; except for functional pain, which was more common (although not significant with current numbers) in the group that underwent prophylactic fixation (67%) as compared to the pathological fracture group (33%) (p = 0.090). This suggests that the patients who underwent prophylactic fixation are quite comparable to the pathological fracture group. Hence, in retrospect and based on the findings of our current study, prophylactic fixation seems to be justified based on the high risk of developing a pathological fracture in this group. Second, we did not perform an ante-hoc sample size calculation as we aimed to include as many cases as possible and match cases to controls on a 1:3 basis to maximize statistical power. Third, we were not able to account for other factors that might have contributed to the fracture risk, such as: a patient s level of activity, the rate of tumor growth, and the overall health status of the patient. We see this as an important limitation, as these factors might play a role in the development of a pathological fracture or in the decision to prophylactically fix an impending fracture. However, baseline characteristics did not differ between groups. Future studies might look at metabolic activity as a surrogate for tumor activity/growth measured by FDG-PET scans and its influence on fracture risk. Fourth, different CT scanners with different settings were used. We were not able to stratify or account for this in our analyses; however, the impact of these differences on our outcomes are (partially) offset by using the contralateral healthy femur as a control. Fifth, our study is uncontrolled due to its retrospective nature; we believe that a more controlled environment with standardized CT scan protocols will reduce measurement error and improve the accuracy of our algorithm. Sixth, our current methodology is limited to unilateral lesions with no implants or contralateral bone abnormalities hindering comparison. We are working on a technique that aims to estimate native bone geometry and density at the location of a metastatic lesion in terms of Hounsfield units based on other relatively unaffected bone regions in the 5 97

100 Chapter 5 same femur (above and below the lesion); this could overcome the need for a contralateral unaffected femur. In addition, we did not include sclerotic lesions. However, lytic lesions are at higher risk of pathological fracture and are more common than sclerotic lesions (sclerotic lesions are predominantly caused by prostate cancer). 16,22,23 These limitations did not compromise our current analyses, but restrict the generalizability of our current algorithms similar to the QCT based algorithms. 7 Seventh, clinical information, such as pain, 16 was not included in our algorithm as it was not possible to establish its added value in a small patient sample. Our algorithms are based on those developed by Snyder et al. 7 Their ex vivo experiments demonstrated that QCT scans can be used to predict failure load of bone with simulated lesions. 10,24 Hounsfield units in axial slices of CT scans are converted into equivalent bone densities using a phantom with hydroxyapatite cores included in the CT scan. The modulus of elasticity and shear modulus for each pixel in the axial slices was derived from these apparent bone densities using previously empirically derived formulas. Structural bone rigidity was estimated based on the cross-sectional geometric distribution of the pixels and their calculated material modulus. The weakest cross-sectional structural rigidity demonstrated a strong correlation with the mechanically tested load bearing capacity of the bone. Snyder et al. 7 subsequently applied this technique in vivo in children with benign lesions of the appendicular skeleton and compared it to radiographic measurements. The structural rigidity was expressed as a ratio of the affected bone relative to the intact bone. They found that the QCT derived parameters differed between the fracture group and the no fracture group, while the radiographic measurements did not. The combination of the structural rigidity measurements on CT scans was 97% accurate for predicting pathological fractures. A subsequent prospective study by the same group included 41 children with a benign bone lesion and used the structural rigidity of the affected relative to the unaffected bone to assess fracture risk. 9 They confirmed that the QCT rigidity analysis is superior in terms of specificity (97%) as compared to radiographic measurements (12%) for predicting the development of a pathological fracture. 9 A large multi-institutional prospective study enrolled 124 patients with 149 appendicular skeleton metastasis and assessed if the CT-based structural rigidity analysis influenced decision making of physicians and the accuracy as compared to the Mirels classification. 8 The technique was 100% sensitive and 90% specific in predicting 7 pathological fractures out of 65 metastatic lesions that had nonoperative management, while the Mirels classification was only 71% sensitive and 50% specific. CT-based structural rigidity analysis therefore seems to be a promising technique. We used comparable algorithms, but HU as unit of measurement instead of bone density. Several previous papers describe the positive correlation of HU with bone mineral density and we therefore felt that this was a reasonable surrogate. 11,12 We demonstrated 98

101 Predicting Pathological Fractures in Femoral Metastasis a difference in CT-based calculations between patients with metastatic lesions that developed a pathological fracture and those that did not when focusing on cortical bone only. However, this difference did not exist when including all pixels trabecular bone, cortical bone, and tumor tissue. We feel that cortical bone contributes more to the load bearing capacity than trabecular bone and this might explain why inclusion of trabecular bone and tumor tissue in the CT-based calculations compromises the analysis. 25 We demonstrated that a threshold of 87% for the weakest cross sectional slice of the affected femur over the unaffected femur (i.e. a 13% decrease in cumulative HU of the affected bone compared to the unaffected bone), including cortical bone only, results in a 89% sensitivity and 48% specificity. The performance of the CT-based structural rigidity analysis by Snyder s group demonstrates higher accuracy with a sensitivity of 100% and specificity of 90%. 7,26,27 The relatively poor performance of our method is a direct result of overlap in cumulative Hounsfield units between lesions that fractured and those that did not. Performance of our method could be improved by standardizing CT scan protocols including: CT scan settings, symmetric orientation of the patient (to optimize comparability of the femora), and using thin CT slices (to correct for inadvertent unequal level of the femora). Our analysis of patients with no femoral metastases demonstrates some residual variation which might be reduced by standardizing CT scans. In addition, a prospective study design that includes a more homogenous patient sample followed over time might prove our method to be more accurate than presented in the current study. It would be interesting to compare our method to the CT-based structural rigidity analysis by Snyder s group by including a hydroxyapatite phantom to establish the added value of using bone mineral density compared to using Hounsfield units. Although the accuracy (i.e. AUC) of our CT-based algorithm was comparable to the Mirels classification, the most important measure of performance in this case sensitivity was better for the CT-based algorithm. In addition, the Mirels classification is subject to interobserver variation, while the CT-based method is not. In conclusion, the presented clinical CT-based algorithms could be useful for predicting the development of pathological fractures in patients with metastatic femoral lesions. However, our results should be interpreted with care due to the biases inherent to the retrospective study design. We feel that the described technique merits additional investigation in a prospective cohort of patients. 5 REFERENCES 1. Mirels H. Metastatic disease in long bones. A proposed scoring system for diagnosing impending pathologic fractures. Clin Orthop Relat Res. Dec 1989(249): van der Linden YM, Kroon HM, Dijkstra SP, et al. Simple radiographic parameter predicts fracturing in metastatic femoral bone lesions: results from a randomised trial. Radiother Oncol. Oct 2003; 69(1):

102 Chapter 5 3. Damron TA, Morgan H, Prakash D, Grant W, Aronowitz J, Heiner J. Critical evaluation of Mirels rating system for impending pathologic fractures. Clin Orthop Relat Res. Oct 2003(415 Suppl): S Anez-Bustillos L, Derikx LC, Verdonschot N, et al. Finite element analysis and CT-based structural rigidity analysis to assess failure load in bones with simulated lytic defects. Bone. Jan 2014; 58: Spruijt S, van der Linden JC, Dijkstra PD, et al. Prediction of torsional failure in 22 cadaver femora with and without simulated subtrochanteric metastatic defects: a CT scan-based finite element analysis. Acta Orthop. Jun 2006; 77(3): Derikx LC, van Aken JB, Janssen D, et al. The assessment of the risk of fracture in femora with metastatic lesions: comparing case-specific finite element analyses with predictions by clinical experts. J Bone Joint Surg Br. Aug 2012; 94(8): Snyder BD, Hauser-Kara DA, Hipp JA, Zurakowski D, Hecht AC, Gebhardt MC. Predicting fracture through benign skeletal lesions with quantitative computed tomography. J Bone Joint Surg Am. Jan 2006; 88(1): Nazarian A, Entezari V, Zurakowski D, et al. Treatment Planning and Fracture Prediction in Patients with Skeletal Metastasis with CT-Based Rigidity Analysis. Clin Cancer Res. Feb ; 21(11): Leong NL, Anderson ME, Gebhardt MC, Snyder BD. Computed tomography-based structural analysis for predicting fracture risk in children with benign skeletal neoplasms: comparison of specificity with that of plain radiographs. J Bone Joint Surg Am. Aug ; 92(9): Hong J, Cabe GD, Tedrow JR, Hipp JA, Snyder BD. Failure of trabecular bone with simulated lytic defects can be predicted non-invasively by structural analysis. J Orthop Res. May 2004; 22(3): Pickhardt PJ, Pooler BD, Lauder T, del Rio AM, Bruce RJ, Binkley N. Opportunistic screening for osteoporosis using abdominal computed tomography scans obtained for other indications. Ann Intern Med. Apr ; 158(8): Schreiber JJ, Anderson PA, Rosas HG, Buchholz AL, Au AG. Hounsfield units for assessing bone mineral density and strength: a tool for osteoporosis management. J Bone Joint Surg Am. Jun ; 93(11): Lim Fat D, Kennedy J, Galvin R, O Brien F, Mc Grath F, Mullett H. The Hounsfield value for cortical bone geometry in the proximal humerus--an in vitro study. Skeletal Radiol. May 2012; 41(5): Aamodt A, Kvistad KA, Andersen E, et al. Determination of Hounsfield value for CT-based design of custom femoral stems. J Bone Joint Surg Br. Jan 1999; 81(1): El-Husseiny M, Coleman N. Inter- and intra-observer variation in classification systems for impending fractures of bone metastases. Skeletal Radiol. Feb 2010; 39(2): Van der Linden YM, Dijkstra PD, Kroon HM, et al. Comparative analysis of risk factors for pathological fracture with femoral metastases. J Bone Joint Surg Br. May 2004; 86(4): Hipp JA, Springfield DS, Hayes WC. Predicting pathologic fracture risk in the management of metastatic bone defects. Clin Orthop Relat Res. Mar 1995(312): Fidler M. Prophylactic internal fixation of secondary neoplastic deposits in long bones. Br Med J. Feb ; 1(5849): Fidler M. Incidence of fracture through metastases in long bones. Acta Orthop Scand. Dec 1981; 52(6):

103 Predicting Pathological Fractures in Femoral Metastasis 20. Menck H, Schulze S, Larsen E. Metastasis size in pathologic femoral fractures. Acta Orthop Scand. Apr 1988; 59(2): Beals RK, Lawton GD, Snell WE. Prophylactic internal fixation of the femur in metastatic breast cancer. Cancer. Nov 1971; 28(5): Rose PS, Laufer I, Boland PJ, et al. Risk of fracture after single fraction image-guided intensitymodulated radiation therapy to spinal metastases. J Clin Oncol. Oct ; 27(30): Randall RL. Metastatic Bone Disease: An Integrated Approach to Patient Care. New York: Springer-Verlag New York; Whealan KM, Kwak SD, Tedrow JR, Inoue K, Snyder BD. Noninvasive imaging predicts failure load of the spine with simulated osteolytic defects. J Bone Joint Surg Am. Sep 2000; 82(9): Holzer G, von Skrbensky G, Holzer LA, Pichl W. Hip fractures and the contribution of cortical versus trabecular bone to femoral neck strength. J Bone Miner Res. Mar 2009; 24(3): Snyder BD, Cordio MA, Nazarian A, et al. Noninvasive Prediction of Fracture Risk in Patients with Metastatic Cancer to the Spine. Clin Cancer Res. Dec ; 15(24): Nazarian A, Entezari V, Villa-Camacho JC, et al. Does CT-based Rigidity Analysis Influence Clinical Decision-making in Simulations of Metastatic Bone Disease? Clin Orthop Relat Res. May ; 474(3):

104 Chapter 5 Appendix 1A: Receiver operator characteristic curve of a threshold cumulative Hounsfield units (cortical bone only) of the weakest cross-sectional CT slice for estimating risk of pathological fracture. The area under the curve is 0.69 (95% confidence interval: ). Appendix 1B: Receiver operator characteristic curve of a threshold cumulative Hounsfield units (cortical bone only) of the complete lesion for estimating risk of pathological fracture. The area under the curve is 0.69 (95% confidence interval: ). 102

105 Predicting Pathological Fractures in Femoral Metastasis 5 Appendix 1C: Receiver operator characteristic curve of a threshold cortical length destruction for estimating risk of pathological fracture. The area under the curve is 0.68 (95% confidence interval: ). Appendix 1D: Receiver operator characteristic curve for 50% circumferential destruction in a crosssectional CT slice for estimating risk of pathological fracture. The area under the curve is 0.63 (95% confidence interval: ). 103

106 Chapter 5 Appendix 1E: Receiver operator characteristic curve of a threshold Mirels score (continuous) for estimating risk of pathological fracture. The area under the curve is 0.72 (95% confidence interval: ). Appendix 1F: Receiver operator characteristic curve for functional pain for estimating risk of pathological fracture. The area under the curve is 0.59 (95% confidence interval: ). 104

107 Predicting Pathological Fractures in Femoral Metastasis Appendix 2: Receiver operator characteristic (ROC) curve analysis for establishing a cut-off in CTbased calculations* Threshold Weakest cross-sectional slice: Structural CT analysis, cortical bone only: Complete lesion: Sensitivity (%) Specificity (%) Sensitivity (%) Specificity (%) < < < < < < < < < < < < * The ROC curve analysis was only performed for those CT based calculations that differed between the patients with a metastatic lesion that fractured (pathological fracture group) and those that did not fracture nor underwent prophylactic fixation (no fracture and no fixation group) (Table 2)

108

109 PART II Metastatic Humeral Lesions

110

111 CHAPTER 6 Outcome Of Operative Treatment Of Metastatic Fractures Of The Humerus: A Systematic Review Of 23 Clinical Studies S.J. Janssen T. Teunis F.J. Hornicek J.A.M. Bramer J.H. Schwab 6 International Orthopaedics 2015 Apr;39(4):735-46

112 Chapter 6 ABSTRACT Objectives To assess: (1) functional outcome, (2) local complication rate, and (3) systemic complication rate after surgery for humeral metastases. These outcomes were narratively reported for intramedullary nailing, open reduction and internal fixation with plate and screws, endoprosthetic reconstruction, and diaphysis prosthesis. Design Systematic review. Data Sources On September 5 th, 2013, we searched the Pubmed, Embase and Cochrane database for studies (published after 1980) using the keywords: pathologic and fracture and surgery, including synonyms, in title and abstract, without any limits, yielding 6,276 publications. Eligibility Criteria For Selecting Studies Studies reporting on functional outcome or complications after surgery for impending or actual pathological humerus fractures caused by metastatic disease. Exclusion criteria were: case-reports, studies with less than 10 patients within a treatment arm, reviews, letters to the editor, meeting abstracts, technique papers, revision procedures, and indiscernible treatment arms for humeral metastasis. Results Average Musculoskeletal Tumor Society score ranged from 64 to 79 (3 studies, 100 patients) after intramedullary nailing, was 90 (1 study, 24 patients) after plate-screw fixation, and 73 (1 study, 30 patients) after endoprosthetic reconstruction. Reoperation rate varied from 0 to 10% after intramedullary nailing (overall 4.4%), 5 to 14% after plate-screw fixation (overall 9.3%), 14 to 16% after diaphysis prosthesis (overall 14.6%), and 0 to 6% after endoprosthetic reconstruction (overall 2.5%). Systemic complication rate varied between 0 and 26% after intramedullary nailing (overall 2.2%), between 0 and 6% after plate-screw fixation (overall 4.8%), was 0% after endoprosthetic reconstruction, and varied between 0 and 16% after diaphysis prosthesis (overall 9.7%). Conclusions Reported complication rates help surgeons inform their patients and could aid in surgical decision making. Functional outcome, pain, and quality of life were poorly reported. Patient reported outcomes are therefore an important direction for future research. 110

113 Outcome After Operative Treatment Of Metastatic Humeral Fractures INTRODUCTION After the spine and femur, the humerus is the third most commonly affected bone by metastatic cancer. 1,2 As metastases weaken the bone, pathological fractures can occur, resulting in pain and disability. In the final stages of life, quick recovery with preservation of function is most important and can be facilitated by surgery. 1,3 In the palliative setting, the decision to undergo surgery and the surgical strategy is made by patients, their families and their doctor together. This decision depends on many factors, including: estimated survival, tumor type, extent of visceral and osseous metastasis, and the expected complications, functional outcome, reduction in pain, and quality of life. 4,5 As far as we know, no systematic review exists and small numbers of patients limit most studies focusing on impending and pathological humerus fractures resulting from bone metastases. We aimed to provide an overview of the functional outcome and local and systemic complication rate after (1) intramedullary nailing, (2) open reduction and internal fixation with plate and screws, (3) endoprosthetic reconstruction and (4) diaphysis prosthesis, to aid patients and their doctors in their decision to undergo surgery. Secondarily, we assessed pain reduction and quality of life. 6 METHODS Article Selection We report our results according to the PRISMA Statement for reporting systematic reviews. 6 Our protocol was registered on PROSPERO prior to study selection (registration number: 2013:CRD ). We searched the Pubmed, Embase, and the Cochrane database on September 5 th, 2013 using the keywords: ( pathologic* OR impending ) AND ( fracture* ) AND ( surgery OR surgeries OR operation OR operations OR operativ* OR surgical* OR intramedull* OR fixation* OR resection* OR osteosynth* OR endoprosth* OR prosth* OR arthroplas* ) in title and abstract, without limits, yielding 6,276 publications. Two reviewers (SJ, TT) independently screened the titles and abstracts and subsequently the full texts using predefined eligibility criteria. The bibliographies of included studies were checked for additional publications. Discordant judgments were resolved by consensus discussion. We included studies reporting on functional outcome, complications, pain reduction, or quality of life after surgery for humeral metastasis (including lymphoma and myeloma). We excluded meeting abstracts, reviews, indiscernible cohorts (e.g. studies mixing primary bone tumors, revision procedures or multiple treatment modalities), studies with less than ten patients within a treatment arm, revision procedures, and studies published 111

114 Chapter 6 before In case of overlapping cohorts (eight studies), 7-14 we included the largest cohort per treatment arm. 7,8,11,13 Two reviewers (S.J., T.T.) independently appraised the quality of studies using predetermined criteria and extracted data using standardized sheets. Authors were contacted if studies published after the year 2000 insufficiently reported outcomes. Of the eleven authors; 1,5,13,15-22 five responded and three provided additional data on complications 5,16 and number of patients with functional outcome 17. Outcome Measures Four studies reported on functional outcome by standardized measure, all used the Musculoskeletal Tumor Society score (MSTS). 17,20,23,24 The MSTS score ranges from 0 to 100 with a higher score indicating better function. 25 Functional outcome scores from studies utilizing a non-standardized score were not included (12 studies). 5,7,8,13,16,18,26-31 None of the studies reported preoperative function. We evaluated the following complications: reoperation, deep infection, transient or permanent nerve palsy, failure or loss of fixation, peri-implant fracture, tumor progression, systemic complication, and intraoperative mortality. Only one study reported pain reduction assessed by a standardized outcome score. 32 Fourteen studies used a non-standardized or non-numerical measure, these scores were not included. 5,7,8,11,13,16-18,23,26-31 No study reported on quality of life. Statistical Analysis Because of heterogeneity of the studies, no meta-analysis was performed and we narratively report results. Cumulative numbers of complications are mentioned per treatment and presented as percentage. Some studies do not report on every complication, cases in these studies were not included in the calculation. RESULTS Study Characteristics Twenty-three studies were included (Figure 1); 6 studies were level III, 17 were level IV evidence (Figure 2, Appendix 1). Sixty-one percent adequately reported eligibility criteria and methods of patient selection, leaving the remainder subject to selection bias. Only 13% state which complications are to be reported, leading to possible outcome bias in the other studies. Loss to followup was high or unclear in 52% of the studies, resulting in attrition bias (Figure 2, Appendix 1). 112

115 Outcome After Operative Treatment Of Metastatic Humeral Fractures 6 Figure 1: Flowchart demonstrating the literature search and the selection of publications including eligibility criteria. Patient Characteristics We included 23 studies reporting on 29 treatment arms totaling 909 fractures: 414 pathological fractures, 56 impending fractures, and in 439 cases the type of fracture was unclear. Average age varied from 58 to 78 years and the percentage men ranged between 29 and 79% (Table 1). The most common tumor types were: breast (30%), myeloma (15%), lung (15%) and kidney (13%) (Figure 3). Average survival ranged from 4 to 23 months (Table 1). The definition of an impending fracture was often unclear. Intramedullary Nailing Nineteen studies reported on intramedullary nailing in 596 fractures. 7,8,11,13,16,17,23,24,26-29,31-37 There were 265 pathological and 18 impending fractures. In 313 cases the type of fracture 113

116 Chapter 6 Level of Evidence Disclosure Patient selection Baseline reporting Outcome reporting No Attrition bias III IV 0% 20% 40% 60% 80% 100% Figure 2: Overall quality of the included studies. Properly Reported 2% 1% 10% 2% 7% 13% 3% 15% 30% 15% Breast Myeloma Lung Kidney Prostate Thyroid Lymfoma Colon/Rectum Other Unknown Figure 3: Distribution of primary tumors among the included studies. Tumors of the Breast, Myeloma, Prostate and Kidney were the most common. Appendix 2 demonstrates the distribution of tumors per study. was not specified. The use of cement varied substantially from 0 to 100%. Five hundred sixty-five (95%) fractures treated with intramedullary nails were located in the diaphysis (Appendix 3). Three studies reported on functional outcome measured by the MSTS score. 17,23,24 Mean MSTS scores reported were 64% in 20 patients after 2 months; 24 69% in 25 patients after 6 months; 23 and 79% in 55 patients after 8 months (Table 2). 17 One study reported on pain reduction assessed using the Visual Analog Scale (VAS) for pain. The mean score improved from 90 preoperative (range 80 to 100) to 15 postoperative (range 0 to 40) in 22 patients. 32 Overall, 4.4% (26/585) of the intramedullary nails required reoperation, ranging from 0 to 10% among studies (Table 3). Overall, deep infection rate was 0.68% (4/585), nerve palsy rate 1.4% (8/585), failure of fixation rate 6.0% (36/596), and peri-implant fracture 114

117 Outcome After Operative Treatment Of Metastatic Humeral Fractures rate 1.0% (5/585). Eleven studies reported on 11 cases of local tumor progression in 382 cases (2.9%); originating from lung (4 cases), kidney (3 cases), myeloma, prostate, head and neck, and in one case the tumor type was not mentioned (Appendix 4). Systemic complication rate varied between 0 and 26% (overall 2.2%, [9/406]). Table 1: Study and patient characteristics of the included studies on humeral metastasis Author, year Study design Surgical treatment modalities Number of patients Number of Mean age in Male pathological or impending fractures years (range) (%) Minimum followup (months) Alvi et al., 2013 Retrospective IMN Wedin et al., 2012 Retrospective IMN, PSF, EPR (29-87)* 86 (40) -- 9 Laitinen et al., 2011 Retrospective IMN (38-95) 18 (45) Weiss et al., 2011 Retrospective PSF (39-87) 43 (68) 1 -- Siegel et al., 2010 Retrospective PSF Piccioli et al., 2010 Retrospective IMN, EPR Pretell et al., 2010 Retrospective IMN (59) 3 23 Ofluogle et al., 2009 Sarahrudi et al., IMN (43-81)* 19 (79) 6 -- Retrospective IMN, PSF (37-89)* 15 (37) 2 4 Atesok et al., 2007 Retrospective IMN (40-86) 6 (29) 2 -- Hunt et al., 2006 Retrospective IMN (38-82)* 7 (50) 6 -- Bauze et al., 2003 Retrospective IMN (39-84) 14 (45) 6 9 Piatek et al., 2003 Prospective IMN Franck et al., IMN (45-86)* Kumta et al., IMN (39-88) 8 (42) Gebhart et al., 2001 Schurman et al., 2000 Retrospective IMN, EPR (41-83) 23 (41) -- 9 Retrospective DP (39-85) Flinkkila et al., 1998 Retrospective IMN (34-80) 4 (25) 6 5 Muller-Farber et al., 1997 Retrospective DP (50) 4 15 Dijkstra et al., 1996 Retrospective IMN, PSF (43-89) 11 (30) 6 5 Redmond et al., 1996 Retrospective IMN (54) 7 4 Ingman et al., 1994 Retrospective IMN (60-90)* 6 (40) 3 3 Lancaster et al., 1988 Retrospective IMN (34-84) 17 (33) 1 -- Mean survival (months) 6 IMN = intramedullary nailing; PSF = plate-screw open reduction and internal fixation; EPR = endoprosthetic reconstruction; DP = diaphysis prosthesis. -- = Not reported, * = Median, = Per number of fractures 115

118 Chapter 6 Open Reduction And Internal Fixation With Plate And Screws Five studies reported on plate-screw fixation in 150 fractures. 5,8,13,20,37 There were 86 pathological and 19 impending fractures, in 45 cases the type of fracture was not specified. Cement was used in the majority of cases, between 86 and 100%. Most cases (93%, 140 of 150 cases) treated with plate-screw fixation were located in the diaphysis (Appendix 3). One study reported a median MSTS score of 90 in 24 patients after 6 months and 90 in 12 patients after 12 months (Table 2). 20 Overall, 9.3% (14/150) of the plate-screw fixations required reoperation; ranging from 5 to 14% among studies (Table 3). Overall, deep infection rate was 0.79% (1/126), nerve palsy rate 5.6% (7/126), failure of fixation rate 3.3% (5/150), and peri-implant fracture rate 2.4% (3/126). Local tumor progression rate varied between 0 and 13% (overall 6.0% [9/150]); and originated from: kidney (four cases), myeloma, lymphoma, colon, unknown and in one case the tumor type was not mentioned (Appendix 4). Systemic complication rate varied between 0 and 6% (overall 4.8% [5/105]). Table 2: Functional outcome per surgical treatment modality for humeral metastasis Author, year Surgical treatment modality Followup period Patients Mean MSTSscore (%) Laitinen et al., 2011 IMN 6 months Piccioli et al., 2010 IMN 8 months Ofluogle et al., 2009 IMN 2 months Siegel et al., 2010 PSF 6 months 24 90* PSF 12 months 12 90* Piccioli et al., 2010 EPR 8 months IMN = intramedullary nailing; PSF = plate-screw open reduction and internal fixation; EPR = endoprosthetic reconstruction; MSTS = Musculoskeletal Tumor Society Score. * = Median Endoprosthetic Reconstruction Three studies reporting on endoprosthetic reconstruction in 81 proximal humerus fractures. 7,17,37 Type of fracture was not specified. Several techniques for reconstruction were used: total joint arthroplasty, hemi-arthroplasty and modular prosthesis (Appendix 3). One study reported a median MSTS score of 73 in 30 patients treated with a modular prosthesis after eight months (Table 2). 17 Overall, 2.5% (2/81) of the reconstructions required reoperation, ranging from 0 to 6% (Table 3). Overall, deep infection rate was 1.2% (1/81), nerve palsy rate 1.2% (1/81), failure of fixation rate 2.5% (2/81), and peri-implant fracture rate 1.2% (1/81). Three cases (4.6% [3/65]) with local tumor progression were reported: originating from breast (2 cases) and kidney. No case of systemic complication was described. 116

119 Outcome After Operative Treatment Of Metastatic Humeral Fractures Diaphysis Prosthesis Two studies reporting on diaphysis prosthesis in 82 fractures. 18,30 There were 63 pathological fractures and 19 impending fractures. Overall, 15% (12/82) of the diaphysis prostheses required reoperation, ranging from 14 to 16% (Table 3). Deep infection rate was 2%, nerve palsy occurred in 4% of the cases, overall failure of fixation rate was 11% (9/82), peri-implant fracture 4.8% (4/82) (Table 3). One case with local tumor progression was reported, origin of the tumor was not mentioned. Systemic complication rate ranged from 0 to 16% (overall 9.7%, [8/82]). DISCUSSION Pathological fractures of the humerus resulting from bone metastases lead to impaired function and pain, surgery is often indicated. 3,5,37 The reported complication rates help surgeons inform their patients and might aid surgical decision making. We encountered a paucity of reported functional outcome, pain, and quality of life data. Patient reported outcomes are therefore an important direction for future research. There were several limitations. First, many included studies are subject to bias. This could skew our results and overestimate the benefits of treatment. Second, functional outcome was only reported by the widely used MSTS score. The MSTS score has not been vetted for floor and ceiling effects nor has it been validated. Furthermore it is commonly administered by the treating clinician rather than directly reported by the patient. 38 Third, due to lack of uniformity in reporting of treatment characteristics we were unable to account for all confounders (e.g. adjuvant treatment). Fourth, we included unpublished data that were not validated by peer review. 5,16,17 Intramedullary nailing results in reasonable function (MSTS score 64% to 79%) and significantly reduces pain (90 preoperative to 15 postoperative). However, functional outcome is based on 100 patients with mixed followup (2-8 months), and pain reduction was measured in only 22 patients. Plate-screw fixation resulted in a MSTS score of 90 in a total of 24 patients after 6 months, and remained stable (MSTS score 90) in the 12 surviving patients. Intramedullary nailing, plate-screw fixation, and diaphysis prosthesis were most commonly used for fractures of the diaphysis. Reoperation rate was lowest after intramedullary nailing (4.4%), followed by plate-screw fixation (9.3%), and highest after diaphysis prostheses (14.6%). The systemic complication rate was also highest for patients treated with diaphysis prostheses (9.7%), followed by those treated with plate-screw fixation (4.8%), and intramedullary nailing (2.2%). However, variation in reoperation and complication rate might be explained by other factors, such as cancer status, survival, and adjuvant treatment

120 Chapter 6 Table 3: Complication and reoperation rate per included surgical treatment modality for humeral metastasis Author, year Surgical treatment modality Total Number fractures Number of pathological fractures (%) Number of impending fractures (%) Reoperation (%) Deep infection (%) Transient or permanent nerve palsy (%) Failure of fixation/loss of fixation (%) Peri-implant fracture (%) Local tumor progression (%) Systemic complications (%) Alvi et al., 2013 IMN (7) 0 Wedin et al., 2012 IMN (6) 3 (2) 3 (2) 2 (1) 1 (1) 0 -- Laitinen et al., 2011 IMN (100) 0 2 (5) 1 (3) 0 2 (5) Piccioli et al., 2010 IMN (2) 6 (11) Pretell et al., 2010 IMN (100) (26) Ofluogle et al., 2009 IMN (100) 0 2 (8) (25) 0 2 (8) 0 Sarahrudi et al., 2009 IMN (100) 0 1 (5) (11) 0 1 (5) 0 Atesok et al., 2007 IMN (83) 4 (17) 2 (8) (4) 0 1 (4) 0 Hunt et al., 2006 IMN (100) (18) Bauze et al., 2003 IMN (81) 6 (19) 2 (6) 0 1 (3) 6 (19) 3 (10) 3 (10) -- Piatek et al., 2003 IMN (10) 0 1 (10) 1 (10) 0 1 (10) 0 Franck et al., 2002 IMN (100) Kumta et al., 2002 IMN (100) (5) 0 Gebhart et al., 2001 IMN (3) (3) Flinkkila et al., 1998 IMN (100) (6) Dijkstra et al., 1996 IMN (65) 7 (5) 1 (6) (17) 1 (6) 0 1 (6) Redmond et al., 1996 IMN (94) 1 (6) 1 (6) (6) 0 Ingman et al., 1994 IMN (100) (7) Lancaster et al., 1988 IMN (9) 0 2 (4) 4 (9) 1 (2) -- 0 Wedin et al., 2012 PSF (14) (10) 2 (10) 0 -- Weiss et al., 2011 PSF (76) 15 (24) 7 (11) 1 (2) 2 (3) 1 (2) 0 5 (8) 4 (6) Siegel et al., 2010 PSF (8) (13)

121 Outcome After Operative Treatment Of Metastatic Humeral Fractures Table 3: Complication and reoperation rate per included surgical treatment modality for humeral metastasis (continued) Systemic complications (%) Local tumor progression (%) Peri-implant fracture (%) Failure of fixation/loss of fixation (%) Transient or permanent nerve palsy (%) Deep infection (%) Reoperation (%) Number of impending fractures (%) Number of pathological fractures (%) Total Number fractures Author, year Surgical treatment modality Sarahrudi et al., 2009 PSF (100) 0 1 (5) 0 4 (18) 1 (5) 1 (5) 0 0 Dijkstra et al., 1996 PSF (80) 4 (20) 1 (5) 0 1 (5) 1 (5) 0 1 (5) 1 (5) Wedin et al., 2012 EPR (6) 1 (3) 0 2 (6) 1 (3) 0 -- Piccioli et al., 2010 EPR (3) (10) 0 Gebhart et al., 2001 EPR Schurman et al., 2000 DP (80) 10 (20) 7 (14) 1 (2) 2 (4) 6 (12) 1 (2) 1 (2) 8 (16) Muller-Farber et al., 1997 DP (72) 9 (28) 5 (16) (9) 3 (9) -- 0 IMN = intramedullary nailing; PSF = plate-screw open reduction and internal fixation; EPR = endoprosthetic reconstruction. -- = Not reported 6 119

122 Chapter 6 Interestingly, although fat embolism is reported as severe complications in treatment of femoral metastasis, no cases have been reported in one of the included studies. Kidney tumors, being the fourth most common tumor (13%), were the most common tumor resulting in local tumor progression (30%, 8 of 27 cases). Renal cell bone metastasis are minimally radiosensitive and can be very aggressive; this probably explains the high recurrence rate. Future studies should focus on function, pain, and quality of life using validated patient reported outcome measures. 4 Validated questionnaires to assess function are the QuickDASH (Disabilities of Arm, Shoulder, and Hand), 42,43 the PROMIS upper extremity questionnaire (Patient-Reported Outcomes Measurement Information System), and the Toronto Extremity Salvage Score. 44,45 The latter one is the only questionnaire tested in patients with upper extremity tumors. Measuring function pre- and postoperatively would allow assessment of the influence of treatment on function. Complication reporting should be standardized to allow for comparison across studies. Several guidelines, such as the Clavien-Dindo classification, are developed to standardize reporting. 46,47 Commonly used instruments to measure pain are the visual analog scale (marking the pain level on a scale from 0 to 100) and numeric rating scale (scoring pain on an ordinal scale from 0 to 10). Both are easy to apply and provide valid outcome measures. 48 Quality of life can be measured using the EuroQol 5 dimensions (EQ-5D) questionnaire, a 5-item questionnaire validated and translated in many languages to assess quality of life. 49 Finally, providing criteria for patient selection, describing baseline characteristics and adjuvant treatment, and specifying the duration of followup and survival will further enhance quality of reporting. Again, selection bias might impact the outcomes in each of these studies and our discussion here is meant to be a starting point for further study rather than a definitive treatise comparing these methods. Conscious of these limitations, it can be concluded that several treatment options are commonly employed, depending on tumor location. Complications rates vary from study to study but intraoperative death and fat-embolism were not reported. The knowledge on complication rates helps surgeons inform their patients and guides the decision for care. Future studies should focus on functional outcome, pain reduction, and quality of life and compare treatment modalities in prospective studies. Reporting of complications should be standardized and patient selection criteria specified to allow for future comparison of treatments across studies. REFERENCES 1. Ratasvuori M, Wedin R, Keller J, et al. Insight opinion to surgically treated metastatic bone disease: Scandinavian Sarcoma Group Skeletal Metastasis Registry report of 1195 operated skeletal metastasis. Surg Oncol. 2013; 22:

123 Outcome After Operative Treatment Of Metastatic Humeral Fractures 2. Toma CD, Dominkus M, Nedelcu T, et al. Metastatic bone disease: a 36-year single centre trend-analysis of patients admitted to a tertiary orthopaedic surgical department. J Surg Oncol. Oct ; 96(5): Bickels J, Dadia S, Lidar Z. Surgical management of metastatic bone disease. J Bone Joint Surg Am. Jun 2009; 91(6): Steensma M, Healey JH. Trends in the surgical treatment of pathologic proximal femur fractures among Musculoskeletal Tumor Society members. Clin Orthop Relat Res. Jun 2013; 471(6): Weiss KR, Bhumbra R, Biau DJ, et al. Fixation of pathological humeral fractures by the cemented plate technique. J Bone Joint Surg Br. 2011; 93: Liberati A, Altman DG, Tetzlaff J, et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration. PLoS Med. Jul ; 6(7): e Gebhart M, Dequanter D, Vandeweyer E. Metastatic involvement of the humerus: a retrospective study of 51 cases. Acta Orthop Belg. 2001; 67: Dijkstra S, Stapert J, Boxma H, Wiggers T. Treatment of pathological fractures of the humeral shaft due to bone metastases: a comparison of intramedullary locking nail and plate osteosynthesis with adjunctive bone cement. Eur J Surg Oncol. 1996; 22: Dijkstra S, Wiggers T, van Geel BN, Boxma H, Dijstra S. Impending and actual pathological fractures in patients with bone metastases of the long bones. A retrospective study of 233 surgically treated fractures. Eur J Surg. 1994; 160: Flinkkila T, Hyvonen P, Lakovaara M, Linden T, Ristiniemi J, Hamalainen M. Intramedullary nailing of humeral shaft fractures. A retrospective study of 126 cases. Acta Orthop Scand. 1999; 70: Flinkkila T, Hyvonen P, Leppilahti J, Hamalainen M. Pathological fractures of the humeral shaft. Ann Chir Gynaecol. 1998; 87: Sarahrudi K, Hora K, Heinz T, Millington S, Vecsei V. Treatment results of pathological fractures of the long bones: a retrospective analysis of 88 patients. Int Orthop. 2006; 30: Sarahrudi K, Wolf H, Funovics P, Pajenda G, Hausmann JT, Vecsei V. Surgical treatment of pathological fractures of the shaft of the humerus. J Trauma. 2009; 66: Vandeweyer E, Gebhart M. Treatment of humeral pathological fractures by internal fixation and methylmetacrylate injection. Eur J Surg Oncol. 1997; 23: Bickels J, Kollender Y, Wittig JC, Meller I, Malawer MM. Function after resection of humeral metastases: analysis of 59 consecutive patients. Clin Orthop Relat Res. 2005: Kumta SM, Quintos AD, Griffith JF, Chow LTC, Wong KC. Closed retrograde nailing of pathological humeral fractures. Int Orthop. 2002; 26: Piccioli A, Maccauro G, Rossi B, Scaramuzzo L, Frenos F, Capanna R. Surgical treatment of pathologic fractures of humerus. Injury. 2010; 41: Schurmann M, Gradl G, Andress HJ, Kauschke T, Hertlein H, Lob G. Metastatic lesions of the humerus treated with the isoelastic diaphysis prosthesis. Clin Orthop Relat Res. 2000: Scotti C, Camnasio F, Peretti GM, Fontana F, Fraschini G. Modular prostheses in the treatment of proximal humerus metastases: review of 40 cases. J Orthop Traumatol. 2008; 9: Siegel HJ, Lopez-Ben R, Mann JP, Ponce BA. Pathological fractures of the proximal humerus treated with a proximal humeral locking plate and bone cement. J Bone Joint Surg Br. 2010; 92:

124 Chapter Talbot M, Turcotte RE, Isler M, Normandin D, Iannuzzi D, Downer P. Function and health status in surgically treated bone metastases. Clin Orthop Relat Res. 2005; 438: Thai DM, Kitagawa Y, Choong PF. Outcome of surgical management of bony metastases to the humerus and shoulder girdle: a retrospective analysis of 93 patients. Int Semin Surg Oncol. 2006; 3: Laitinen M, Nieminen J, Pakarinen TK. Treatment of pathological humerus shaft fractures with intramedullary nails with or without cement fixation. Arch Orthop Trauma Surg. 2011; 131: Ofluoglu O, Erol B, Ozgen Z, Yildiz M. Minimally invasive treatment of pathological fractures of the humeral shaft. Int Orthop. 2009; 33: Enneking WF, Dunham W, Gebhardt MC, Malawar M, Pritchard DJ. A system for the functional evaluation of reconstructive procedures after surgical treatment of tumors of the musculoskeletal system. Clin Orthop Relat Res. Jan 1993(286): Atesok K, Liebergall M, Sucher E, Temper M, Mosheiff R, Peyser A. Treatment of pathological humeral shaft fractures with unreamed humeral nail. Ann Surg Oncol. 2007; 14: Franck WM, Olivieri M, Jannasch O, Hennig FF. An expandable nailing system for the management of pathological humerus fractures. Arch Orthop Trauma Surg. 2002; 122: Ingman AM, Waters DA. Locked intramedullary nailing of humeral shaft fractures. Implant design, surgical technique, and clinical results. J Bone Joint Surg Br. 1994; 76: Lancaster JM, Koman LA, Gristina AG, et al. Pathologic fractures of the humerus. South Med J. 1988; 81: Muller-Farber J, Muller KH. The treatment of metastatic humeral lesions with the diaphyseal prosthesis. Aktuelle Traumatol. 1997; 27: Redmond BJ, Biermann JS, Blasier RB. Interlocking intramedullary nailing of pathological fractures of the shaft of the humerus. J Bone Joint Surg Am. 1996; 78: Pretell J, Rodriguez J, Blanco D, Zafra A, Resines C. Treatment of pathological humeral shaft fractures with intramedullary nailing. A retrospective study. Int Orthop. 2010; 34: Alvi HM, Damron TA. Prophylactic stabilization for bone metastases, myeloma, or lymphoma: Do we need to protect the entire bone? Tumor. Clin Orthop Relat Res. 2013; 471: Bauze AJ, Clayer MT. Treatment of pathological fractures of the humerus with a locked intramedullary nail. J Orthop Surg (Hong Kong). 2003; 11: Hunt KJ, Gollogly S, Randall RL. Surgical fixation of pathologic fractures: an evaluation of evolving treatment methods. Bull Hosp Jt Dis. 2006; 63: Piatek S, Westphal T, Bischoff J, Schubert S, Holmenschlager F, Winckler S. [Intramedullary stabilisation of metastatic fractures of long bones]. Zentralbl Chir. 2003; 128: Wedin R, Hansen BH, Laitinen M, et al. Complications and survival after surgical treatment of 214 metastatic lesions of the humerus. J Shoulder Elbow Surg. 2012; 21: Cheng EY. Prospective quality of life research in bony metastatic disease. Clin Orthop Relat Res. Oct 2003(415 Suppl): S Gibbons CER, Pope SJ, Murphy JP, Hall AJ. Femoral metastatic fractures treated with intramedullary nailing. Int Orthop. 2000; 24: Kerr PS, Jackson M, Atkins RM. Cardiac arrest during intramedullary nailing for femoral metastases. J Bone Joint Surg Br. Nov 1993; 75(6): Peter RE, Schopfer A, Le Coultre B, Hoffmeyer P. Fat embolism and death during prophylactic osteosynthesis of a metastatic femur using an unreamed femoral nail. J Orthop Trauma. Apr 1997; 11(3):

125 Outcome After Operative Treatment Of Metastatic Humeral Fractures 42. Beaton DE, Wright JG, Katz JN. Development of the QuickDASH: comparison of three itemreduction approaches. J Bone Joint Surg Am. May 2005; 87(5): Kennedy CA, Beaton DE, Smith P, et al. Measurement properties of the QuickDASH (Disabilities of the Arm, Shoulder and Hand) outcome measure and cross-cultural adaptations of the QuickDASH: a systematic review. Qual Life Res. Mar Davis AM, Wright JG, Williams JI, Bombardier C, Griffin A, Bell RS. Development of a measure of physical function for patients with bone and soft tissue sarcoma. Qual Life Res. Oct 1996; 5(5): Clayer M, Davis A. Can the Toronto Extremity Salvage Score produce reliable results when used online? Clin Orthop Relat Res. Jun 2011; 469(6): Slankamenac K, Graf R, Barkun J, Puhan MA, Clavien PA. The comprehensive complication index: a novel continuous scale to measure surgical morbidity. Ann Surg. Jul 2013; 258(1): Clavien PA, Barkun J, de Oliveira ML, et al. The Clavien-Dindo classification of surgical complications: five-year experience. Ann Surg. Aug 2009; 250(2): Hawker GA, Mian S, Kendzerska T, French M. Measures of adult pain: Visual Analog Scale for Pain (VAS Pain), Numeric Rating Scale for Pain (NRS Pain), McGill Pain Questionnaire (MPQ), Short-Form McGill Pain Questionnaire (SF-MPQ), Chronic Pain Grade Scale (CPGS), Short Form-36 Bodily Pain Scale (SF-36 BPS), and Measure of Intermittent and Constant Osteoarthritis Pain (ICOAP). Arthritis Care Res (Hoboken). Nov 2011; 63 Suppl 11: S Group. E. EuroQol--a new facility for the measurement of health-related quality of life. Health Policy. Dec 1990; 16(3):

126 Chapter 6 Appendix 1: Critical appraisal Author, year Level of evidence Disclosure Patient selection Baseline reporting Outcome reporting No attrition bias Alvi et al., Wedin et al., Laitinen et al., Weiss et al., Siegel et al., Piccioli et al., Pretell et al., Ofluogle et al., Sarahrudi et al., Atesok et al., Hunt et al., Bauze et al., Piatek et al., Franck et al., Kumta et al., Gebhart et al., Schurman et al., Flinkkila et al., Muller-Farber et al., Dijkstra et al., Redmond et al., Ingman et al., Lancaster et al., Level of evidence Disclosure + Reported - Not reported Patient selection + Eligibility criteria, sources and methods of selection of patients described - Potential selection bias/not described Baseline reporting + Detailed baseline characteristics per surgical treatment modality (age, gender, anatomical site and primary tumor type). - Mixed or unspecified baseline characteristics Outcome reporting + Clear definition of which complications are to be reported - Complications to be reported not specified No attrition bias (Completeness of outcome data) + <20% lost to follow up, and appropriate patient selection - >20% Patients were lost to followup, withdrawn or missing for outcome assessment or not mentioned 124

127 Outcome After Operative Treatment Of Metastatic Humeral Fractures Appendix 2: Distribution of primary tumors per study Unknown (%) Other (%) Colon/ Rectum (%) Lymphoma (%) Thyroid (%) Prostate (%) Kidney (%) Lung (%) Myeloma (%) Breast (%) Number of pathological or impending fractures Number of patients Author, year Wedin et al., (31) 24 (12) 26 (13) 30 (14) 21 (10) 0 (0) 0 (0) 0 (0) 43 (21) 0 (0) Laitinen et al., (43) 4 (10) 4 (10) 4 (10) 7 (18) 0 (0) 2 (5) 0 (0) 1 (3) 1 (3) Weiss et al., (3) 14 (22) 13 (21) 13 (21) 1 (2) 4 (6) 1 (2) 1 (2) 6 (10) 8 (13) Siegel et al., (21) 4 (17) 7 (29) 4 (17) 0 (0) 2 (8) 2 (8) 0 (0) 0 (0) 0 (0) Piccioli et al., (26) 8 (9) 17 (19) 13 (14) 5 (5) 7 (8) 5 (5) 1 (1) 8 (9) 3 (3) Pretell et al., (9) 9 (39) 0 (0) 1 (4) 2 (9) 0 (0) 0 (0) 3 (13) 6 (26) 0 (0) Ofluogle et al., (22) 2 (9) 12 (52) 3 (13) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 1 (4) Sarahrudi et al., (44) 3 (8) 4 (10) 4 (10) 4 (10) 0 (0) 0 (0) 0 (0) 7 (18) 0 (0) Atesok et al., (52) 2 (10) 0 (0) 2 (10) 1 (5) 1 (5) 1 (5) 2 (10) 0 (0) 1 (5) Hunt et al., (29) 3 (21) 0 (0) 5 (36) 0 (0) 0 (0) 0 (0) 1 (7) 1 (7) 0 (0) Bauze et al., (26) 9 (29) 1 (3) 4 (13) 0 (0) 0 (0) 2 (6) 1 (3) 6 (19) 0 (0) Franck et al., (23) 10 (45) 7 (32) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) Kumta et al., (26) 0 (0) 5 (26) 2 (11) 2 (11) 0 (0) 0 (0) 3 (16) 0 (0) 2 (11) Gebhart et al., (45) 10 (18) 6 (11) 5 (9) 0 (0) 0 (0) 0 (0) 0 (0) 10(18) 0 (0) Schurman et al., (36) 14 (28) 5 (10) 6 (12) 2 (4) 2 (4) 0 (0) 0 (0) 3 (6) 0 (0) Flinkkila et al., (31) 2 (13) 2 (13) 2 (13) 0 (0) 1 (6) 0 (0) 0 (0) 2 (13) 2 (13) Muller-Farber et al., (34) 1 (3) 5 (16) 5 (16) 4 (13) 0 (0) 1 (3) 1 (3) 2 (6) 2 (6) Dijkstra et al., (50) 2 (5) 3 (8) 7 (18) 1 (3) 0 (0) 1 (3) 0 (0) 4 (11) 1 (3) Redmond et al., (15) 4 (31) 1 (8) 1 (8) 2 (15) 0 (0) 1 (8) 0 (0) 1 (8) 1 (8) Ingman et al., (27) 3 (20) 2 (13) 0 (0) 3 (20) 0 (0) 1 (7) 0 (0) 0 (0) 2 (13) Lancaster et al., (38) 17 (33) 4 (8) 3 (6) 1 (2) 1 (2) 2 (4) 2 (4) 1 (2) 1 (2) Alvi et al & Piatek et al did not report primary tumor distribution of operated humeral metastasis cases. -- = not reported, = per fracture 6 125

128 Chapter 6 Appendix 3: Surgical treatment characteristics Author, year Surgical treatment modality Total (Impending) Pathological Fractures Cement Antegrade nailing Proximal Humerus Diaphysis Distal Humerus Type of implants Alvi et al., 2013 IMN Locked intramedullairy nail Wedin et al., 2012 IMN (30) Locked intramedullairy nail Laitinen et al., 2011 IMN (53) 40 (100) Synthes UHN Piccioli et al., 2010 IMN (84) 57 (100) Synthes UHN Pretell et al., 2010 IMN 23 0 (0) 23 (100) Acumed Polarus nail, Synthes UHN, Smith & Nephew TriGen nail Ofluogle et al., 2009 IMN (100) Acumed Polarus nail, Synthes PHN, Hipokrat C-75 nail Sarahrudi et al., 2009 IMN 19 0 (0) Synthes UHN, Howmedica Seidel nail, Sanatmetal AR-nail Atesok et al., 2007 IMN 24 5 (21) 24 (100) Synthes UHN Hunt et al., 2006 IMN 11 6 (55) Locked intramedullairy nail Bauze et al., 2003 IMN 31 4 (13) 24 (77) Austofix nail Piatek et al., 2003 IMN (80) Russell-Taylor, bundle nail Franck et al., 2002 IMN 23 0 (0) 15 (65) Fixion expandable nail Kumta et al., 2002 IMN (100) 0 (0) Ender rods Gebhart et al., 2001 IMN (100) Hackethal nail, Rush rods, Howmedica Seidel nail Flinkkila et al., 1998 IMN (100) Howmedica Seidel nail, Kuntscher nail, St-Pro nail Dijkstra et al., 1996 IMN 18 0 (0) 11 (61) Locked intramedullairy nail Redmond et al., 1996 IMN 16 0 (0) 16 (100) Biomet Uniflex nail Ingman et al., 1994 IMN 15 0 (0) 3 (20) Locked intramedullairy nail Lancaster et al., 1988 IMN (67) Kuntscher nail, Rush rods, Ender rods Wedin et al., 2012 PSF (86) NA Plate-screw fixation Weiss et al., 2011 PSF (100) NA Plate-screw fixation Siegel et al., 2010 PSF (100) NA Synthes proximal humeral locking plate 126

129 Outcome After Operative Treatment Of Metastatic Humeral Fractures Appendix 3: Surgical treatment characteristics (continued) Type of implants Diaphysis Distal Humerus Proximal Humerus Cement Antegrade nailing Total (Impending) Pathological Fractures Surgical treatment modality Author, year Sarahrudi et al., 2009 PSF (100) NA Synthes DCP, Synthes LCP, Synthes Philos, Y-plate Synthes Dijkstra et al., 1996 PSF (100) NA Plate-screw fixation Wedin et al., 2012 EPR NA Hemi-arthroplasty, total joint arthroplasty, modular tumor prosthesis Piccioli et al., 2010 EPR NA Modular tumor prosthesis Gebhart et al., 2001 EPR NA Proximal endoprosthesis Schurman et al., 2000 DP NA Diaphyse prothesis Muller-Farber et al., 1997 DP (100) NA Diaphyse prothesis IMN = intramedullary nailing; PSF = plate-screw open reduction and internal fixation; EPR = endoprosthetic reconstruction; DP = diaphysis prosthesis; NA = not applicable, UHN = unreamed humeral nail, PHN = proximal humeral nail; DCP = Dynamic Compression Plate; LCP = Locking Compression Plate. -- = not reported 6 127

130 Chapter 6 Appendix 4: Local tumor progression primary tumor types Author, year Surgical treatment modality Number of impending or pathological fractures Local tumor progression (%) Origin of primary tumor Alvi et al., 2013 IMN 14 1 (7) Head and Neck tumor Wedin et al., 2012 IMN Piccioli et al., 2010 IMN Ofluogle et al., 2009 IMN 24 2 (8) Kidney, Lung Sarahrudi et al., 2009 IMN 19 1 (5) -- Atesok et al., 2007 IMN 24 1 (4) Kidney Bauze et al., 2003 IMN 31 3 (10) Kidney, Myeloma, Prostate Piatek et al., 2003 IMN 10 1 (10) Lung Kumta et al., 2002 IMN 21 1 (5) Lung Dijkstra et al., 1996 IMN Redmond et al., 1996 IMN 16 1 (6) Lung Wedin et al., 2012 PSF Weiss et al., 2011 PSF 63 5 (8) 3 x Kidney, Colon, Unknown Siegel et al., 2010 PSF 24 3 (13) Kidney, Myeloma, Lymphoma Sarahrudi et al., 2009 PSF Dijkstra et al., 1996 PSF 20 1 (5) -- Wedin et al., 2012 EPR Piccioli et al., 2010 EPR 30 3 (10) Kidney, 2 x Breast Schurman et al., 2000 DP 50 1 (2) -- IMN = intramedullary nailing; PSF = plate-screw open reduction and internal fixation; EPR = endoprosthetic reconstruction; DP = diaphysis prosthesis. -- = not reported 128

131 CHAPTER 7 Complications After Surgery For Metastatic Humeral Lesions S.J. Janssen M. van Dijke S.A. Lozano-Calderon J.E. Ready K.A. Raskin M.L. Ferrone F.J. Hornicek J.H. Schwab 7 Journal of Shoulder and Elbow Surgery 2016 Feb;25(2): Presented at: American Academy of Orthopaedic Surgeons Annual Meeting 2015, Las Vegas, Nevada, USA.

132 Chapter 7 ABSTRACT Objectives To establish surgical outcome after treatment of humeral metastases. Design Retrospective cohort study. Setting Two tertiary care referral centers for orthopaedic oncology. Participants 295 consecutive patients with humeral metastases who underwent surgery between 1998 and Interventions Proximal lesions were treated by intramedullary nailing (43%, n = 57), endoprosthetic reconstruction (34%, n = 46), open reduction and internal fixation using plate and screws (22%, n = 30), and a combination (n = 1). Diaphyseal lesions were treated by intramedullary nailing (69%, n = 91), and open reduction and internal fixation using plate and screws (30%, n = 39), and a combination (n = 1). Distal lesions were treated by open reduction and internal fixation using plate and screws (97%, n = 29) and intramedullary nailing (3.3%, n = 1). Outcome Measures Primary outcome measures were reoperations and 30-day systemic complications. Secondary outcome measures were total estimated blood loss, anesthesia time, duration of hospital admission, and 30-day survival. Results We found 25 (8.5%) reoperations and 17 (5.8%) patients had 18 systemic complications; pneumonia (3.7%, n = 11), pulmonary embolism (1.3%, n = 4), sepsis (0.68%, n = 2), and fat embolism (0.34%, n = 1). No factors were independently associated with reoperation. Logistic regression analysis demonstrated that favorable cancer status (i.e. a higher modified Bauer score: OR 0.48, 95% CI: , p = 0.005) was independently associated with a decreased systemic complication rate. Conclusion Poor cancer status is an independent predictor of postoperative systemic complications. This could help inform the patient and anticipate postoperative problems. 130

133 Complications After Surgery For Metastatic Humeral Lesions INTRODUCTION Indications for surgery of a metastasis, myeloma, or lymphoma lesion of the humerus vary from a completed pathological fracture, through a lesion at risk for fracture (i.e. impending fracture), to a solitary lesion. The decision for surgery is not always clear and many factors are considered including expected survival, systemic load, anatomical location, tumor type, size of the lesion, fracture risk, and expected outcome. 1-4 Several criteria are proposed to assess the fracture risk, of which the Mirels classification is most commonly used. 5 Metastasectomy is occasionally warranted in patients with a solitary metastasis as some studies suggest that this improves survival in patients with renal cell carcinoma; 6,7 however, this finding is contradicted by others. 8,9 Surgical treatment for metastatic humerus lesions is often palliative and aims to stabilize the bone for the remaining lifetime of the patient to preserve quality of life while minimizing the risk of complications. 2,4,10 Most previous studies are relatively small and focus on a single surgical technique. 2,11,12 This study aims to assess the outcome reoperation and 30-day systemic complication rate of surgery in a larger cohort of patients with metastatic humerus lesions treated in various ways. Our null hypothesis was that there are no factors associated with reoperation and complications among patients surgically treated for metastatic humerus lesions. Additionally, we assessed differences in estimated blood loss, anaesthesia time, duration of hospital admission, and 30-day survival among surgical strategies. Knowledge of complication rate and its predictors can help inform the patient and aid surgical decision making. 7 METHODS Study Design We assessed reoperations and 30-day systemic complications, and risk factors for these outcomes, in patients who underwent surgery for metastatic humerus lesions. Medical record data was obtained of patients who had an International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code for a metastatic fracture (code: ) or a Current Procedural Terminology (CPT) code (code: and 24498) for prophylactic fixation of the humerus between 1998 and 2013 at two tertiary care referral centers for orthopaedic oncology. We reviewed all medical records of identified patients to assess eligibility. A final consecutive series of 295 patients with a metastatic humerus lesion impending or pathological fractures was included. We included only the first surgery per patient if patients had bilateral lesions (12 patients) so as to not violate the assumption of independence. 13 Inclusion criteria were patients older than 18 years who underwent intramedullary nailing, open reduction and internal fixation with plate and 131

134 Chapter 7 screws, endoprosthetic reconstruction, or a combination. No patients underwent distal or intercalary endoprosthetic reconstruction. We included patients regardless of followup duration as we considered all reoperations, short-term and long-term, to be relevant. We included metastases from solid tumors, multiple myeloma, and lymphoma. We excluded patients who underwent only revision surgery at our institutions or who underwent fixation with non-interlocking nails (4 patients were treated with Rush rods). Surgical Strategies The surgeon, together with the patient, decided for surgery and selected the surgical strategy on the basis of life expectancy, systemic load, tumor type, location, and size of the lesion. Orthopaedic oncologists performed 286 of the 295 (97%) surgeries, the remainder were performed by trauma surgeons. There were 237 (80%) pathological fractures and 58 (20%) impending fractures. Proximal lesions without significant involvement of the humeral head were treated by intramedullary nailing (43%, 57 of 134 cases; 45 pathological and 12 impending fractures), plate-screw fixation (22%, 30 of 134 cases; 23 pathological and 7 impending fractures), or a combination of these techniques (0.75%, 1 of 134 cases; 1 pathological fracture). Cement was used in 7 of 57 intramedullary nailing cases (12%): in 1 cases after curettage of the tumor and in 6 cases to create a strong construct for the proximal interlocking screws in the humeral head. Cement was used in 25 of 30 plate-screw fixations (83%), an osteoarticular allograft was used in combination with plate-screw fixation after proximal humerus resection in one patient with renal cell carcinoma (3.3%, 1 of 30 cases). Endoprosthetic reconstruction was used for 46 (34%, 46 of 134 cases; 43 pathological and 3 impending fractures) proximal humeral lesions after resection of the humeral head (17 humeral head hemiarthroplasties) or proximal humerus (29 modular tumor prostheses). Five of the 17 (29%) humeral head replacements were combined with a proximal intercalary allograft (3 renal cell, 2 breast carcinomas). Cement was used in 43 of 46 (93%) endoprosthetic reconstruction cases. Diaphyseal lesions were treated by non-cemented intramedullary nailing (69%, 91 of 131 cases; 70 pathological and 21 impending fractures), plate-screw fixation (30%, 39 of 131; cases 28 pathological and 11 impending fractures), or a combination of these techniques (0.76%, 1 of 131 cases; 1 impending fracture). Cement was used in 19 of 39 (49%) plate-screw fixation cases. Six patients (2 renal cell, 1 breast, 1 lung carcinoma, and 2 multiple myeloma) underwent a segmental resection of the humeral shaft followed by intercalary allograft placement and plate screw fixation (15%, 6 of 39 cases). Distal lesions were mainly treated by plate-screw fixation (97%, 29 of 30 cases; 26 pathological and 3 impending fractures); cement was used in 24 cases (83%). The remaining case (3.3%; 1 pathological fracture) was treated with an non-cemented intramedullary nail. 132

135 Complications After Surgery For Metastatic Humeral Lesions The type of operation, as outlined above, varied on the basis of the location of the lesion (p < 0.001, by Fisher exact test) and the type of fracture (p = 0.037, by Fisher exact test). All intramedullary nails were interlocking and inserted in an antegrade fashion. Postoperative care and rehabilitation varied among patients depending on their health status. Outcome Measures Our primary outcome was reoperation in one of the two hospitals. Only the first reoperation was taken into account. Two research fellows (S.J., M.v.D.) independently screened all reports of subsequent procedures to identify reoperations. The secondary outcome was non pre-existing systemic complications pneumonia, pulmonary embolism, fat/cement embolism, myocardial infarction, and sepsis within 30 days after index surgery. These systemic complications were identified through disease specific ICD-9-CM codes (Appendix 1). Medical records of patients with one of these codes were reviewed independently by two research fellows (S.J., M.v.D.) to assess if the complication fulfilled the predefined criteria: (1) pneumonia was defined as symptoms clinically consistent with pneumonia and a positive sputum culture or empirical start of antibiotics; (2) the diagnosis pulmonary, fat or cement embolism was based on computed tomography or ventilation/perfusion scan; (3) the diagnosis myocardial infarction was based on electrocardiography or echocardiography; and (4) sepsis was defined as Systemic Inflammatory Response Syndrome requiring intensive care unit admission with a positive culture. Any discordant judgments were resolved by consensus discussion. We obtained data on the following variables through chart review: age, body mass index, sex, comorbidities, cancer status, tumor type, systemic load, type of fracture (pathological or impending), location of the lesion, preoperative local radiation and systemic therapy, postoperative local radiation and systemic therapy, type of operation, allograft use, reaming, cementation, concomitant procedures (e.g. femoral fixation during same surgery), white blood cell count (in K/uL), duration of hospital admission, anesthesia time, and estimated blood loss. We scored bone metastatic lesions using Mirels classification to demonstrate the fracture risk of the impending fractures. The total score, based on 4 aspects of the bone lesion (site of the lesion [all upper extremity], pain, lesion type, and lesion size), varies from 4 to 12; 4 being at low risk of fracture, and 12 being at high risk of fracture. 5 According to Mirels, a score of 8 and above is suggestive of an impending fracture. 5,14,15 We reviewed radiographs to establish the lesion type (blastic/mixed/lytic) and size (<1/3, 1/3-2/3, >2/3). 5,14,15 Description of pain level in the medical record was used to assess if the pain was mild (including no pain, mild pain, or a pain score below 4 [numeric pain rating scale: 0-10]), moderate (including moderate pain or a pain score between 4 and 8), or functional (including severe pain, significant pain, a pain score of 8 and higher, or any pain that increased with loading). 5,14,

136 Chapter 7 The modified Charlson Comorbidity Index was used to assess comorbidity status and determined through an algorithm based on ICD-9-CM codes given prior to surgery (Appendix 2). The Modified Charlson Comorbidity Index is based on 12 comorbidities and the score ranges from 0 to 24 with a higher score indicating more severe comorbidity status We used the modified Bauer score as a surrogate marker for cancer status. 1,19 The modified Bauer score is most commonly used to estimate survival and is a summary score of four prognostic factors: (1) no visceral metastases, (2) no lung cancer, (3) breast, renal cell, multiple myeloma, or lymphoma as primary tumor type, and (4) a solitary skeletal metastasis. 1,19 The score ranges from 0 to 4 with a lower score indicating worse prognosis. 1,19 A patient with breast cancer with a single bone metastatic lesion and no visceral metastases is assigned a score of 4 (indicating relatively good prognosis). We categorized tumor type into thyroid, renal cell, and other; as the former two are relative radioresistant tumors and have a relatively good prognosis and are therefore potentially more likely to recur leading to failure of the fixation mandating reoperation. 4,20 Followup The median followup was 4 months (interquartile range [IQR] 1 to 14 months, range 0 months to 10 years). One hundred and sixty-four (56%) patients were alive at followup after 3 months, 79 (27%) were deceased and 52 (18%) were lost to followup. After 6 months; 130 (44%) patients were still in followup, 116 (39%) deceased, and 49 (17%) were lost to followup. One year after surgery, 85 (29%) were still in followup, 150 (51%) patients were deceased, and 60 (20%) were lost to followup. Statistical Analysis Variables were presented with frequencies and percentages for categorical variables and as mean with standard deviation (SD) for continuous variables. Cases were not included in the respective analyses that had missing values for one of these variables: body mass index (18%, 54 of 295 cases), estimated blood loss (11%, 31 of 295 cases), white blood cell count (9.5%, 28 of 295 cases), duration of hospital admission (0.68%, 2 of 295 cases), and anesthesia time (26%, 77 of 295 cases). The association between the response variable reoperation and explanatory variables was assessed using univariate Cox regression analysis. Unadjusted hazard ratios (HR) with 95% confidence intervals (CI) indicating the relative likelihood of reoperation in one group as compared to another group are presented to quantify the association of explanatory variables with reoperation. Multivariable Cox regression analysis including all variables marginally associated with reoperation (i.e. p < 0.10) in bivariate analysis was used to assess if variables were independently associated with reoperation. Cox regression analysis accounts for variable lengths of followup and survival. We used Kaplan-Meier curves to 134

137 Complications After Surgery For Metastatic Humeral Lesions plot probability of survival and reoperation in our cohort. Testing the Schoenfeld residuals indicated no violation of the proportional hazard assumption. The association between the response variable 30-day systemic complication and explanatory variables was demonstrated using unadjusted odds ratios (OR) with 95% confidence intervals derived from logistic regression analysis. Multivariable logistic regression analysis including all variables with a p value below 0.10 in bivariate analysis was used to assess if variables were independently associated with systemic complications. Postoperative adjuvant treatment was not included in these analyses as it is often given after the 30-day period. Variation in blood loss, anesthesia time, and duration of hospital admission between surgical strategies was tested using one-way analysis of variance (ANOVA). Difference in 30-day survival was tested using the Fisher exact test. All statistical analyses were performed using Stata 13.0 (StataCorp LP, College Station, TX, USA). A two-tailed p value less than 0.05 was considered statistically significant. RESULTS 7 Baseline Characteristics Among the 295 cases, 143 were men (48%) and the mean age was 63 years (Table 1). Of the 58 cases with an impending fracture; 1 (1.7%) had a Mirels score of five, 3 (5.2%) had a score of seven, 11 (19%) had a score of eight, 28 (48%) a score of nine, and 13 (22%) a score of ten. Establishing the Mirels score was not possible in two cases (3.4%, 2 of 58 cases) due to missing radiographs. Most bone metastases included lung (24%), breast (17%) and renal cell carcinoma (15%) (Table 2). Reoperations We found 25 reoperations (8.5%, 25 of 295 cases) (Table 3). Primary reasons for the 25 reoperation in our cohort were: deep infection (2.0% [6 of 295 cases]), nonunion (2.0% [6 of 295 cases]), peri-implant fracture (1.4% [4 of 295 cases]), allograft nonunion (1.0% [3 of 295 cases]), massive hematoma (0.68% [2 of 295 cases]), tumor progression (0.68% [2 of 295 cases]), compartment syndrome (0.34% [1 of 295 cases]), and early plate loosening in an intercalary allograft construct (0.34% [1 of 295 cases]). There were 5 local progressions (3 renal cell, 1 lung, and 1 colorectal metastasis) of which 3 coincided with other reasons for reoperation (Table 3). Acknowledging the variation in indications among the surgical strategies, we found that 5 of the 46 (11%) patients who underwent endoprosthetic reconstruction required reoperation; 3 cases for deep infection and 2 cases for massive hematoma. The reoperation 135

138 Chapter 7 Table 1: Demographics and baseline characteristics Mean (± Standard Deviation) Age (in years) 63 (±12) Body mass index* 28 (±6.0) Modified Charlson Comorbidity Index 6.5 (±2.1) Modified Bauer score 2.1 (±1.1) White blood cell count (in K/uL)* 9.4 (±5.0) Estimated blood loss (in ml)* 323 (±438) Duration of hospital admission (in days)* 6.4 (±5.9) Anesthesia time (in minutes)* 209 (±70) n = 295 n (%) Men 143 (48) Visceral metastases 124 (42) Multiple bone metastases 222 (75) Pathological fracture 237 (80) Anatomical location Proximal humerus 134 (45) Diaphyseal humerus 131 (44) Distal humerus 30 (10) Previous local radiotherapy 64 (22) Previous systemic therapy 177 (60) Operation Intramedullary nailing 149 (51) Plate-screw fixation 98 (33) Proximal endoprosthetic reconstruction 46 (16) Intramedullary nailing + plate screw fixation 2 (0.68) Allograft 12 (4.1) Reaming 121 (41) Cement 119 (40) Concomitant other surgery 22 (7.5) * Body mass index was available in 241 (82%) cases, white blood cell count in 267 (91%), estimated blood loss in 264 (89%), duration of hospital admission in 293 (99%), and anesthesia time in 218 (74%). 20 patients underwent concomitant fixation of a femur metastatic lesion, 1 for a tibia metastatic lesion, and 1 patient underwent concomitant fixation of a femur metastatic lesion and decompression of the L3 nerve root due to spine metastases. rate was 10% (10 of 98 cases) for patients who underwent plate-screw fixation and 6.7% (10 of 149) for those who underwent intramedullary nailing. 136

139 Complications After Surgery For Metastatic Humeral Lesions Table 2: Tumor distribution n = 295 Bone metastases: n (%) Lung 70 (24) Breast 50 (17) Renal cell 43 (15) Unknown 17 (5.8) Prostate 14 (4.7) Thyroid 10 (3.4) Esophageal 6 (2) Melanoma 6 (2) Colorectal 4 (1.4) Adenocarcinoma of unknown origin 3 (1.02) Bladder 2 (0.68) Hepatocellular 2 (0.68) Pancreatic 2 (0.68) Gastric 1 (0.34) Primary bone tumors: n (%) Myeloma 54 (18) Lymphoma 11 (3.7) 7 The proportion of patients who required reoperation increased considerably with time from 2.6% at 1 month up to 19% at two years (Figure 1). Survival is longer in patients who underwent a reoperation (median: 60 months, IQR 22 to 110 months) compared with those who did not (median: 9 months, IQR 3 to 36 months; p = 0.002, by log-rank test), indicating that longer survival is associated with a higher risk of reoperation. We found that higher white blood cell count (HR 1.11, 95% CI: 1.02 to 1.20, p = 0.010), higher body mass index (HR 1.06, 95% CI: 1.00 to 1.13, p = 0.078), and longer anesthesia time (HR 1.01, 95% CI: 1.00 to 1.01, p = 0.057) were marginally associated with a higher risk of reoperation in bivariate analyses (Appendix 3). None of the other explanatory variables (e.g. sex, age) were associated with reoperation (Appendix 3). Multivariable Cox regression analysis including white blood cell count, body mass index, and anesthesia time demonstrated that none of the factors were independently associated with the outcome reoperation: white blood cell count (p = 0.73), body mass index (p = 0.21), and anesthesia time (p = 0.35). 137

140 Chapter 7 Table 3: Reoperations Sex, age Fracture type Humerus Location Tumor type Operation Implant type Reason for reoperation Months, Reoperation Male, 65 Pathological Proximal Prostate IMN Interlocking nail Deep infection 0, Debridement Male, 68 Pathological Diaphysis Lymphoma IMN Interlocking nail Peri-implant fracture 1, Additional plate fixation with 2 plates Female, 58 Pathological Proximal Lung IMN Interlocking nail Compartment syndrome 1, Fasciotomy Female, 42 Pathological Proximal Breast IMN Interlocking nail Nonunion 9, Plate fixation and bone graft Male, 61 Pathological Diaphysis Renal cell IMN Interlocking nail Nonunion, tumor recurrence 10, Tumor resection and placement modular prosthesis Female, 69 Pathological Diaphysis Breast IMN Interlocking nail Nonunion 14, Nail replacement and bone grafting Female, 78 Pathological Proximal Breast IMN Interlocking nail Nonunion 17, Nail removal and bone grafting Female, 62 Pathological Proximal Lung IMN Interlocking nail Peri-implant fracture 20, Additional plate fixation and bone graft Female, 61 Pathological Diaphysis Lung IMN Interlocking nail Nonunion and nail loosening 34, Nail removal, plate fixation with bone graft Male, 72 Pathological Proximal Renal cell IMN Interlocking nail Tumor recurrence 58, Tumor resection and allograft with head prosthesis Male, 29 Pathological Proximal Myeloma Prox EPR Modular prosthesis Hematoma, compartment syndrome 0, Debridement Female, 52 Pathological Proximal Breast Prox EPR Humeral head prosthesis Hematoma, followed by deep infection 0, Debridement Female, 53 Pathological Proximal Lymphoma Prox EPR Humeral head prosthesis Deep infection 1, Debridement Female, 66 Pathological Proximal Breast Prox EPR Humeral head prosthesis Deep infection 17, Debridement, prosthesis removal, placement spacer Male, 72 Pathological Proximal Myeloma Prox EPR Modular prosthesis Deep infection 38, Debridement, and replacement prosthesis Female, 44 Pathological Proximal Lung PSF 1 plate Peri-implant fracture 0, Plate removal and placement of modular prosthesis 138

141 Complications After Surgery For Metastatic Humeral Lesions Table 3: Reoperations (continued) Tumor type Operation Implant type Reason for reoperation Months, Reoperation Sex, age Fracture type Humerus Location Male, 62 Pathological Proximal Breast PSF 1 plate Deep infection 1, Debridement 1, Refixation with 2 new plates distally Hardware loosening distal allograft Male, 79 Pathological Diaphysis Renal cell PSF 1 plate + intercalary allograft Female, 93 Pathological Distal Lymphoma PSF 1 plate Peri-implant fracture 1, Refixation with additional plate 3, Debridement and removal of hardware Female, 49 Impending Proximal Colorectal PSF 2 plates Deep infection, tumor recurrence Female, 77 Pathological Distal Breast PSF 1 plate Nonunion 12, Hardware removal, refixation 2 plates and bone graft Allograft nonunion 14, Humeral head prosthesis with new allograft Male, 57 Pathological Proximal Renal cell PSF 1 plate + osteoarticular allograft 17, Allograft and tumor resection, new allograft + 1 plate Allograft nonunion, tumor recurrence Female, 54 Impending Proximal Renal cell PSF 1 plate + intercalary allograft 20, Revision plate fixation with bone graft Allograft nonunion and broken plate Female, 43 Pathological Proximal Breast PSF 1 plate + intercalary allograft 23, Curettage, plate refixation and cement Male, 80 Pathological Diaphysis Lung PSF 1 plate Plate loosening, tumor recurrence Prox EPR = proximal endoprosthetic reconstruction, IMN = intramedullary nailing, PSF = plate-screw fixation 7 139

142 Chapter 7 Reoperation Probability of reoperation N at risk: Months Figure 1: Kaplan-Meier failure plot demonstrating the probability of reoperation among all patients (solid black line) with 95% Confidence interval (CI) (gray dashed line). The probability of reoperation was 2.6% at 1 month (95% CI: 1.3% to 5.4%), 4.5% at 3 months and 6 months (95% CI: 2.5 to 8.1%), 7.6% at one year (95% CI: 4.3% to 13%), and 19% at two years (95% CI: 12% to 29%). Systemic Complications Seventeen patients (5.8%) had 18 postoperative systemic complications: pneumonia (3.7% [11 of 295]), pulmonary embolism (1.3% [4 of 295]), sepsis (0.68% [2 of 295]), and fat embolism (0.34% [1 of 295]) (Appendix 4). There was no case of myocardial infarction and no patient died intraoperatively. The fat embolism occurred while reaming the humeral shaft for intramedullary nailing resulting in a transient drop in saturation and blood pressure (Appendix 4). Bivariate analysis demonstrated that poor cancer status (modified Bauer score: OR 0.47, 95% CI: 0.29 to 0.75, p = 0.001) and longer duration of hospital admission (OR 1.08, 95% CI: 1.02 to 1.14, p = 0.008) were associated with an increased systemic complication rate. The complication rate was lower among patients who underwent plate-screw fixation as compared to those who underwent intramedullary nailing (OR 0.22, 95% CI: 0.05 to 0.99, p = 0.048). None of the other included variables were associated with complications in bivariate analyses. However, when controlling for possible confounding in a multivariable logistic regression analysis we only found the modified Bauer score to be independently associated with the outcome systemic complications (OR 0.48, 95% CI: 0.29 to 0.80, p = 0.005) (plate-screw fixation as compared to intramedullary nailing [p = 0.47], hospital duration [p = 0.11], and use of cement [p = 0.26]). This means that having a relatively poor prognosis as indicated by a lower Bauer score is associated with a higher rate of systemic complications. 140

143 Complications After Surgery For Metastatic Humeral Lesions The complication rate per modified Bauer score category was 18% (5 of 28) for score zero, 8.6% (5 of 58) for score one, 5.8% (5 of 86) for score two, 2.1% (2 of 95) for score three, and 0% (0 of 28) for score four. This further explains its association with complications. Additional outcomes The mean estimated blood loss differed between treatment types and was 165mL (SD: 174) after intramedullary nailing, 472mL (SD: 530) after plate screw fixation, and 600mL (SD: 640) after endoprosthetic reconstruction (p < 0.001). The mean anesthesia time did not differ between treatments (p = 0.32) and was 201 (SD: 68) minutes after intramedullary nailing, 212 (SD: 71) minutes after plate-screw fixation, and 227 (SD: 70) minutes after endoprosthetic reconstruction. The mean duration of hospital admission did not differ between treatments (p = 0.07): 6.4 (SD: 6.5) days after intramedullary nailing, 5.6 (SD: 4.5) days after plate-screw fixation, and 8.0 (SD: 6.5) after endoprosthetic reconstruction. By the last moment of followup, 220 (75%) patients were deceased; the median survival was 11 months (IQR: 3 to 40 months) (Figure 2). The 30-day survival was 92% and did not differ between surgical strategies (p = 0.99): 92% after intramedullary nailing, 92% after plate-screw fixation, and 93% after endoprosthetic reconstruction. 7 Survival Probability N at risk: Months Figure 2: Kaplan-Meier survival plot demonstrating probability of survival (solid black line) with 95% Confidence interval (CI) (gray dashed line). Median survival is 321 days (95% CI: 239 to 423), with an interquartile range from 82 (95% CI: 62 to 97) to 1228 (95% CI: 930 to 1810) days. 141

144 Chapter 7 DISCUSSION Understanding outcomes after surgical treatment of metastatic humerus lesions could help inform the patient and aid surgical decision-making. Previous studies are limited by their relatively small sample size. We aimed to assess reoperation and systemic complication rates and assessed factors associated with these outcomes in a relatively large cohort of patients from two institutions. We found an overall reoperation rate of 8.5%, and a systemic complication rate of 5.8%. None of the included variables were found to be independently associated with the outcome reoperation. We found that a lower modified Bauer score indicating relatively poor prognosis was independently associated with postoperative systemic complications. There were several limitations. First, there were no uniform criteria for surgical treatment owing to the retrospective study design (i.e. selection bias). Decision for surgical strategy depends on many factors, and could have differed among surgeons or changed over time and might therefore have influenced the occurrence of reoperations. Second, reoperations and complications would have been missed if a patient went to a different hospital or deceased outside of the hospital because of a complication. The complication rates found in our study might therefore be an underestimation of the true complication rate. Third, the relatively low number of reoperations (n = 25 and systemic complications (18 in 17 patients) limited the number of variables we could include in the multivariable analyses. Therefore it was not possible to account for all potential confounders. Furthermore, a larger sample size might have resulted in more statistical power to detect subtle but relevant risk factors (type II error). For example, bivariate analysis demonstrated a trend towards a higher reoperation rate (HR 3.15, 95% CI: 0.74 to 13.42, p = 0.12) in patients with a pathological fracture as compared to those with an impending fracture; a larger sample size might have identified a difference. Fourth, followup and survival varied substantially among patients due to the severity of the underlying disease. We accounted for this by using time-dependent Cox regression analysis for reoperations and using a relatively short timeframe (30-days) for systemic complications. The overall reoperation rate found in our study (8.5%) is comparable with reoperation rates reported in previous studies (ranging from 0 to 9%). 2 A study by Wedin et al. presented a large cohort of 214 surgically treated metastatic humerus lesions and found a reoperation rate (6.1%) after intramedullary nailing comparable to ours (6.7%). 21 However, we found a higher reoperation rate (11% vs. 6%) with endoprosthetic reconstruction and a lower rate with plate-screw fixation (10% vs. 14%). 21 Our reoperation rate (10%) after plate-screw fixation is comparable to the rate (11%) found in a large study by Weiss et al. focusing on plate-screw fixation (n = 63) for metastatic humerus disease. 12 Differences in reoperation rates might be explained by different indications for surgery and variation in technique. Consistent with previous studies we found that the proportion of patients 142

145 Complications After Surgery For Metastatic Humeral Lesions who require reoperation increased considerably with time (Figure 3). 12 The finding that reoperations occur more often in patients who live longer emphasizes the importance of estimating life expectancy in these patients. 1,3,22 In further exploring reasons for reoperation, it seems that nonunion is the main reason for late reoperation (6 months after the index surgery); about two-thirds (64%, 9 of 14 cases) of the late reoperations in our cohort were performed for nonunion, comparable to the results presented by Wedin et al. (67%, 6 of 9 late reoperations). 21 Peri-implant fractures and deep infections were more evenly distributed among early and late reoperations, while massive hematoma and compartment syndrome only occurred in the early postoperative period. 21 These findings might help the surgeon to anticipate these local complications. The association of high white blood cell count with reoperation in bivariate analysis did not hold when accounting for potential confounding in multivariable analysis. We found no independent predictors of reoperation. This could be a result of the variation in reasons for reoperation (e.g. peri-implant fracture, nonunion, deep infection), which might have other underlying causes and corresponding risk factors. The study by Wedin et al. studied predictors of reoperations and found a higher rate among distal humerus lesions and tumors originating from the prostate A B C Figure 3: An 80 year old man presented with this (Figure 3A) pathological humerus fracture after lifting an object from the floor. Work-up revealed a non-small-cell lung carcinoma. Open reduction internal fixation was done using plate-screw fixation with cement augmentation after curettage of gross tumor (Figure 3B). The patient presented 2 years later with progressive arm pain, imaging demonstrated tumor recurrence and screw loosening (Figure 3C). 143

146 Chapter 7 The systemic complication rate found in our study (5.8%) is also in line with previously published studies (0 to 26%). 2 Difference in the definition of systemic complication probably explains the substantial variation in reported rates. No intraoperative death occurred, which is in line with previous studies. 2 One patient had a transient intraoperative saturation and blood pressure drop during reaming of the humeral shaft; this was ascribed to a fat embolism. This potentially severe complication has been reported during fixation of metastatic femoral lesions; 23,24 however, it is very rare in treatment of metastatic humerus disease. 2 Relatively poor cancer status (indicated by a low modified Bauer score) was the only independent predictor of developing systemic complications. Previous studies demonstrated that cancer status as measured by the modified Bauer score was associated with survival; 1,19 however, its association with complications as far as we know has not been shown before. This means that patients with a poor cancer status not only have a decreased life-expectancy, they also have a higher risk of postoperative systemic complications. Type of fixation seemed to be associated with complication risk in bivariate analysis; however, this association was confounded by a difference in Bauer score among surgical strategies as can be gleaned from the multivariable analysis. The Bauer score could be used for preoperative risk stratification to identify patients at higher risk for postoperative systemic complications. In conclusion, we found a reoperation rate of 8.5% and a systemic complication rate of 5.8%. Poor cancer status was an independent predictor of postoperative systemic complications. This could help inform the patient and anticipate postoperative problems. REFERENCES 1. Bauer HC, Wedin R. Survival after surgery for spinal and extremity metastases. Prognostication in 241 patients. Acta Orthop Scand. Apr 1995; 66(2): Janssen SJ, Teunis T, Hornicek FJ, Bramer JA, Schwab JH. Outcome of operative treatment of metastatic fractures of the humerus: a systematic review of twenty three clinical studies. Int Orthop. Nov Nathan SS, Healey JH, Mellano D, et al. Survival in patients operated on for pathologic fracture: implications for end-of-life orthopedic care. J Clin Oncol. Sep ; 23(25): Quinn RH, Randall RL, Benevenia J, Berven SH, Raskin KA. Contemporary management of metastatic bone disease: tips and tools of the trade for general practitioners. J Bone Joint Surg Am. Oct ; 95(20): Mirels H. Metastatic disease in long bones. A proposed scoring system for diagnosing impending pathologic fractures. Clin Orthop Relat Res. Dec 1989(249): Alt AL, Boorjian SA, Lohse CM, Costello BA, Leibovich BC, Blute ML. Survival after complete surgical resection of multiple metastases from renal cell carcinoma. Cancer. Jul ; 117(13):

147 Complications After Surgery For Metastatic Humeral Lesions 7. Baloch KG, Grimer RJ, Carter SR, Tillman RM. Radical surgery for the solitary bony metastasis from renal-cell carcinoma. J Bone Joint Surg Br. Jan 2000; 82(1): Fuchs B, Trousdale RT, Rock MG. Solitary bony metastasis from renal cell carcinoma: significance of surgical treatment. Clin Orthop Relat Res. Feb 2005(431): Lin PP, Mirza AN, Lewis VO, et al. Patient survival after surgery for osseous metastases from renal cell carcinoma. J Bone Joint Surg Am. Aug 2007; 89(8): Bickels J, Dadia S, Lidar Z. Surgical management of metastatic bone disease. J Bone Joint Surg Am. Jun 2009; 91(6): Laitinen M, Nieminen J, Pakarinen TK. Treatment of pathological humerus shaft fractures with intramedullary nails with or without cement fixation. Arch Orthop Trauma Surg. 2011; 131: Weiss KR, Bhumbra R, Biau DJ, et al. Fixation of pathological humeral fractures by the cemented plate technique. J Bone Joint Surg Br. 2011; 93: Bryant D, Havey TC, Roberts R, Guyatt G. How many patients? How many limbs? Analysis of patients or limbs in the orthopaedic literature: a systematic review. J Bone Joint Surg Am. Jan 2006; 88(1): El-Husseiny M, Coleman N. Inter- and intra-observer variation in classification systems for impending fractures of bone metastases. Skeletal Radiol. Feb 2010; 39(2): Evans AR, Bottros J, Grant W, Chen BY, Damron TA. Mirels rating for humerus lesions is both reproducible and valid. Clin Orthop Relat Res. Jun 2008; 466(6): Quan H, Li B, Couris CM, et al. Updating and validating the Charlson comorbidity index and score for risk adjustment in hospital discharge abstracts using data from 6 countries. Am J Epidemiol. Mar ; 173(6): Quan H, Parsons GA, Ghali WA. Validity of information on comorbidity derived rom ICD-9- CCM administrative data. Med Care. Aug 2002; 40(8): Roche JJ, Wenn RT, Sahota O, Moran CG. Effect of comorbidities and postoperative complications on mortality after hip fracture in elderly people: prospective observational cohort study. BMJ. Dec ; 331(7529): Leithner A, Radl R, Gruber G, et al. Predictive value of seven preoperative prognostic scoring systems for spinal metastases. Eur Spine J. Nov 2008; 17(11): Les KA, Nicholas RW, Rougraff B, et al. Local progression after operative treatment of metastatic kidney cancer. Clin Orthop Relat Res. Sep 2001(390): Wedin R, Hansen BH, Laitinen M, et al. Complications and survival after surgical treatment of 214 metastatic lesions of the humerus. J Shoulder Elbow Surg. 2012; 21: Janssen SJ, van der Heijden AS, van Dijke M, et al Marshall Urist Young Investigator Award: Prognostication in Patients With Long Bone Metastases: Does a Boosting Algorithm Improve Survival Estimates? Clin Orthop Relat Res. Jul Gibbons CER, Pope SJ, Murphy JP, Hall AJ. Femoral metastatic fractures treated with intramedullary nailing. Int Orthop. 2000; 24: Kerr PS, Jackson M, Atkins RM. Cardiac arrest during intramedullary nailing for femoral metastases. J Bone Joint Surg Br. Nov 1993; 75(6):

148 Chapter 7 Appendix 1: International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) Codes Used to Flag Systemic Complications Complication Codes Flagged Pulmonary embolism , , 451.2, , 451.9, , , 453.8, 453.9, 453, 453.4, 415.1, , , 453.9, , , , , 416.2, 444.1, , , , , 444.9, , , , , 453.2, 453.3, , , , , , , 453.6, , , , , , , , , , , , , , , , , Pneumonia 481, 482.0, 482.1, 482.2, , , , , , , , , , , , , 482.9, 483.8, 485, 486, 495.7, 507 Sepsis 038.0, , , , 038.3, , , , , , , 038.8, 038.9, Myocardial infarction 427.5, , , , , , , , , , , , , , , , , , , , Appendix 2: Modified Charlson Comorbidity Index Algorithm Based on International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) Codes Comorbidity Weight* Codes AIDS/HIV Any malignancy, including leukemia and lymphoma* , , , , , , , 189.9, , , , , , , , , , , 230.8, 231.2, 231.9, , 233.0, 233.1, , , 233.4, 233.7, , 235.7, 235.8, 236.2, 236.4, 236.5, , , 237.6, , , , 239.6, 239.7, , Chronic pulmonary disease , 416.9, , , , 506.4, 508.1, Congestive heart failure , , , , , , , , , , , Dementia 2 290, 290.0, 290.3, , 294.1, , Diabetes with chronic complications , Hemiplegia or paraplegia , , Metastatic solid tumor* , , , Mild liver disease* , , , , , , 070.6, 070.9, 570, 570.1, 573.3, 573.4, 573.8, 753.9, V42.7 Moderate or severe liver disease* , Renal disease , , , , , , , , , , , 588.0, V42.0, V45.1, V56-V56.8 Rheumatologic disease , , , 714.8, 725 *The following comorbidities were mutually exclusive: mild liver disease and moderate or severe liver disease, and any malignancy and metastatic solid tumor. For example, a patient with a metastatic solid tumor received 6 points total (not 6 points for metastatic solid tumor and 2 points for any malignancy). 146

149 Complications After Surgery For Metastatic Humeral Lesions Appendix 3: Bivariate analysis reoperations and 30 day systemic complications Reoperations Unadjusted hazard ratio (95% confidence interval) p value n = day Systemic complications Unadjusted odds ratio (95% confidence interval) p value Age 0.99 ( ) ( ) 0.77 Men 0.72 ( ) ( ) 0.12 Body mass index* 1.06 ( ) ( ) 0.61 Modified Charlson Comorbidity Index 0.94 ( ) ( ) 0.94 Modified Bauer score 1.10 ( ) ( ) < White blood cell count* 1.11 ( ) ( ) 0.63 Estimated blood loss 1.00 ( ) ( ) 0.32 Duration of hospital admission 1.02 ( ) ( ) 0.01 Anesthesia time* 1.01 ( ) ( ) 0.85 Tumor type Other tumor types reference value reference value reference value reference value Thyroid 1.00 ( ) ( ) 0.99 Renal cell 1.32 ( ) ( ) 0.99 Pathological fracture 3.15 ( ) ( ) 0.30 Anatomical location Proximal humerus reference value reference value reference value reference value Diaphyseal humerus 0.58 ( ) ( ) 0.77 Distal humerus 0.88 ( ) ( ) 0.99 Preoperative local radiotherapy 1.28 ( ) ( ) 0.96 Preoperative systemic therapy 0.58 ( ) ( ) 0.90 Postoperative local radiotherapy 0.63 ( ) 0.26 Postoperative systemic therapy 0.79 ( ) 0.60 Operation Intramedullary nailing reference value reference value reference value reference value Plate-screw fixation 1.32 ( ) ( ) Proximal endoprosthetic reconstruction 1.24 ( ) ( ) 0.34 Intramedullary nailing + plate-screw fixation 1.00 ( ) ( ) 0.99 Allograft 2.37 ( ) ( ) 0.99 Reaming 1.13 ( ) ( ) 0.61 Cement 1.07 ( ) ( ) 0.06 Concomitant other surgery 1.42 ( ) ( ) bold font indicates a significant difference (two-tailed p value below 0.05). * Body mass index was available in 241 (82%) cases, white blood cell count in 267 (91%), estimated blood loss in 264 (89%), duration of hospital admission in 293 (99%), and anesthesia time in 218 (74%). 147

150 Chapter 7 Postoperative adjuvant treatment is not included in the analysis for systemic complications as it is frequently started after the 30 day period. Unadjusted hazard ratios with 95% confidence interval and p value were derived from univariate Cox regression analysis. Unadjusted odds ratios with 95% confidence interval and p value were derived from univariate logistic regression analysis. Appendix 4: Systemic complications Sex, age Fracture type Humerus Location Tumor type Operation Implant type Complication within 30 days Female, 61 Pathological Diaphysis Lung IMN Interlocking nail Pulmonary embolism Male, 62 Pathological Diaphysis Lung IMN Interlocking nail Pneumonia Male, 68 Pathological Diaphysis Lymphoma IMN Interlocking nail Pneumonia Male, 58 Pathological Proximal Myeloma IMN Interlocking nail Pneumonia Female, 65 Pathological Proximal Lung IMN Interlocking nail Pneumonia Female, 39 Impending Diaphysis Breast IMN Interlocking nail Pneumonia Male, 66 Impending Proximal Prostate IMN Interlocking nail Pneumonia Male, 86 Pathological Diaphysis Lung IMN Interlocking nail Pneumonia Female, 62 Pathological Proximal Lung IMN Interlocking nail Pneumonia Female, 69 Pathological Proximal Lymphoma IMN Interlocking nail Pneumonia Female, 71 Pathological Diaphysis Lung IMN Interlocking nail Sepsis Female, 66 Pathological Diaphysis Pancreatic IMN Interlocking nail Pulmonary embolism, sepsis Female, 61 Impending Diaphysis Lung IMN Interlocking nail Fat embolism Female, 68 Impending Proximal Lung Prox EPR Humeral head Pulmonary embolism Female, 55 Pathological Proximal Lung Prox EPR Humeral head Pulmonary embolism Female, 51 Impending Diaphysis Lung PSF 1 Plate Pneumonia Female, 44 Pathological Proximal Lung PSF 1 Plate Pneumonia n = 17 Prox EPR = proximal endoprosthetic reconstruction, IMN = intramedullary nailing, PSF = plate-screw fixation 148

151 CHAPTER 8 Management Of Metastatic Humeral Fractures: Variations According To Orthopedic Subspecialty, Tumor Characteristics S.J. Janssen J.A.M. Bramer T.G. Guitton F.J. Hornicek J.H. Schwab 8 Orthopaedics & Traumatology: Surgery & Research 2018 Feb; 104(1):59-65 Presented at: European Musculoskeletal Oncology Society Annual Meeting 2016, La Baule, France. Musculoskeletal Tumor Society Annual Meeting 2016, Detroit, Michigan, USA.

152 Chapter 8 ACKNOWLEDGMENTS: The authors thank the Skeletal Oncology Research Group* and Science Of Variation Group * members of the Skeletal Oncology Research Group: Albert J. Aboulafia; John A. Abraham; Dimosthenis Andreou; Luis A. Aponte-Tinao; Robert U. Ashford; Tessa Balach; Nicholas M. Bernthal; David J. Biau; Susan V. Bukata; George T. Calvert; Felix H. Cheung; Paul W. Clarkson; Matthew W. Colman; Judd E. Cummings; Francis Cyran; Timothy A. Damron; Will Eward; German L. Farfalli; Peter C. Ferguson; Marco Ferrone; Joerg Friesenbichler; Philipp T. Funovics; Mark C. Gebhardt; Michelle Ghert; Parker Gibbs; Francois Gouin; Robert J. Grimer; James B. Hayden; John H. Healey; Eric Henderson; Robert M. Henshaw; Marc H. Isler; Shintaro Iwata; Lee M. Jeys; Muchael J. Joyce; Johnny Keller; Piya Kiatisevi; David M. King; Nitin Kukkar; Minna K. Laitinen; Andreas Leithner; Adam S. Levin; Edward A. Levine; Dieter Lindskog; Santiago A. Lozano Calderon; Joel Mayerson; Corey O. Montgomery; Carol D. Morris; John C. Neilson; Lukas Nystrom; Vania Oliveira; Eduardo J. Ortiz-Cruz; Joshua C. Patt; Robert H. Quinn; Rajiv Rajani; Lor R. Randall; John E. Ready; Peter S. Rose; Matt Steensma; Gwen M.L. Sys; Richard M. Terek; Frank Traub; Takafumi Ueda; Rikard Wedin; Kurt R. Weiss; Jay Wunder; Daniel Wurtz; and Rosanna L. Wustrack members of the Science of Variation Group: Jeffrey Abrams; Lars E. Adolfsson; Michael H. Amini; Paul Appleton; George C. Babis; Efstathios G. Ballas; Flamarion Dos Santos Batista; Grant J. Bayne; Thierry Begue; Arne Berner; Jan Biert; Lars C. Borris; Maarten W.G.A. Bronkhorst; Richard Buckley; Tim Chesser; Joseph M. Conflitti; Brett D. Crist; Brian J. Cross; Thomas DeCoster; Derek F.P. van Deurzen; Thomas Dienstknecht; Percy V. van Eerten; Nelson Elias; Denise Eygendaal; Christos Garnavos; Peter Giannoudis; Vincenzo Giordano; Hans Goost; Roger G. Hackney; Jeremy Hall; Eric Mark Hammerberg; Ian Harris; Steven J. Hattrup; Robert Haverlag; Huub van der Heide; Svenhjalmar van Helden; Edward F. Ibrahim; Prashanth Inna; Vishwanath M. Iyer; Richard Jenkinson; Kyle Jeray; Clifford Jones; Axel Jubel; Koroush Kabir; Nikolaos K. Kanakaris; Matej Kastelec; Jack Elias Kazanjian; Cyrus Klostermann; Georges Kohut; Peter Krause; Anze Kristan; Paul Levin; Iain McGraw; Matthew Menon; Milind Merchant; Ladislav Mica; Thomas Mittlmeier; Jesus Moreta; Matt Mormino; Betsy M. Nolan; Timothy Omara; Richard S. Page; Emilija Stojkovska Pemovska; Rodrigo Pesantez; Marinis Pirpiris; Lodewijk M.S.J. Poelhekke; Michael Prayson; Michael Quell; Radzeli Mohd Ramli; Martin Richardson; Edward K. Rodriguez; Juan Miguel Rodríguez Roiz; Nigel Rossiter; Parag Sancheti; Michiel A.J. van de Sande; Wojciech Satora; Peter Schandelmaier; John A. Scolaro; Adam B. Shafritz; Raymond Malcolm Smith; Andy B. Spoor; Marc Swiontkowski; Lisa Taitsman; Theodoros Tosounidis; Michael Verhofstad; Anne J.H. Vochteloo; Richard Wallensten; Mohammad Waseem; Yoram Weil; and David Weiss 150

153 Management Of Metastatic Humeral Fractures ABSTRACT Objectives To assess: (1) if there is a difference between orthopaedic oncology surgeons and other orthopaedic surgeons in their recommendation for treatment of metastatic humeral lesions, and (2) what patient and tumor characteristics were associated with the decision for treatment. Design Cross sectional survey study. Setting Online survey sent to two international groups of orthopaedic surgeons. Participants 78 (48%) orthopaedic oncologists and 83 (52%) orthopaedic surgeons that were not regularly involved in the treatment of bone tumors. Outcome Measures Recommendation for treatment. Twenty-four case scenarios were created by combining: tumor type, life expectancy, fracture type, and anatomical location. Participants were asked for every case: what treatment would you recommend? 8 Results There was a difference between orthopaedic oncologists and other subspecialty surgeons in recommendation for specific treatments: intramedullary nailing was less often recommended by orthopaedic oncologists (53%, 95% CI: 47-59) compared to other surgeons (62%, 95% CI: 57-67) (p = 0.023); while endoprosthetic reconstruction (orthopaedic oncologists: 8.8% [95% CI: ], other surgeons: 3.6%[95% CI: ], p < 0.001) and plate-screw fixation (orthopaedic oncologists: 19%[95% CI: 14-25], other surgeons: 9.5%[95% CI: ], p = 0.003) were more often recommended by orthopaedic oncologists. There was no difference in recommendation for nonoperative management. There were differences in recommendation for specific treatments based on tumor type, life expectancy, and anatomical location, but not fracture type. Conclusions Subspecialty training and patient and tumor characteristics influence the decision for surgical management and the decision for a specific implant in metastatic humeral fractures. 151

154 Chapter 8 INTRODUCTION Treatment of metastatic humeral lesions is a challenge as indications vary, several implant options exist, different types of adjuvant treatment are available, and many patient and tumor factors need to be considered. Commonly used implants are intramedullary nails, endoprostheses, and plate and screws. 1,2 Case-series and retrospective studies support the use of each of these implants and high quality comparative studies are lacking. 1 Commonly cited patient and tumor factors that are considered in surgical decision making are: tumor type, life expectancy, location of the tumor, and fracture type. 2-5 Several orthopaedic subspecialties take care of patients with metastatic humeral lesions, including orthopaedic oncology surgeons, trauma surgeons, and general orthopaedic surgeons. As training among these subspecialties differs, their surgical approach might differ as well. We therefore investigated if orthopaedic oncology surgeons approach metastatic humeral lesions differently than surgeons who are not regularly involved in the treatment of bone tumors. It is unclear to what extend these patient, tumor, and surgeon factors influence surgical decision making. Better understanding of what treatment is recommended and the factors that have the greatest influence on recommending specific treatment for metastatic humeral lesions would help: development of guidelines, highlight areas that require further (comparative) study, create techniques to improve measurement of these criteria (e.g. techniques to estimate life expectancy or fracture risk), and educate trainees. Specifically, we assessed: (1) if there was a difference between orthopaedic oncology surgeons and other orthopaedic surgeons in their recommendation for treatment, and (2) what patient and tumor characteristics were associated with the decision for treatment. 6 METHODS Study Design Our institutional review board exempted this study from review and informed consent was not needed. We created 24 case scenarios by combining four categorical variables: tumor type (breast, renal cell, lung; 8 cases each), estimated life expectancy (<3, >3 months; 12 cases each), fracture type (displaced pathological, impending; 12 cases each), and anatomical location (proximal, diaphyseal; 12 cases each). Humeral metastases from breast, renal cell, and lung tumors were chosen because these are most common. 1 We explained for every case that the patient had activity related pain, that the tumor was widely metastatic, that the patient walked with a walker, and that he or she has not had radiotherapy. We searched our institutions humeral metastases database for the first random radiograph that matched the constructed case scenario (i.e. matched tumor type, fracture type, and anatomical location). 152

155 Management Of Metastatic Humeral Fractures We used SurveyMonkey (Palo Alto, CA, USA), a web-based assessment tool, to develop a survey including the 24 case scenarios. For all cases, we asked: what treatment would you recommend: Intramedullary nailing, Endoprosthetic reconstruction, Plate-screw fixation, or Nonoperative management? In addition, we asked for the 12 impending fracture cases: What is the fracture risk on a scale from 0 to 100%? We collected the following demographics from participants: year finished residency, sex, country, subspecialties, and the proportion of practice dedicated to oncology. This was the first study of a new collaborative in musculoskeletal oncology, named the Skeletal Oncology Research Group (SORG). The idea to develop this collaborative was based on the existing Science of Variation Group (SOVG, an international group of upper extremity and trauma surgeons). 7 The objective of this new group is to study variation in interpretation and management of musculoskeletal tumors. We invited people to this new group by ing the members of two professional organizations (the European MusculoSkeletal Oncology Society [n = 156] and the Connective Tissue Oncology Society [n = 783]) and we also reached out to our colleagues, friends, and acquaintances (n = 83); we welcome all interested physicians involved in treatment of musculoskeletal tumors who wish to join. Eighty-five people subscribed (8.3% [85/1,022]) and 71 orthopaedic oncology surgeons completed this study. We also invited all trauma surgeons and shoulder and elbow surgeons of the SOVG (n = 441) and specifically asked them to only complete the survey if they treat metastatic humeral lesions. One hundred thirty (29%) members responded: 22 indicated that they do not treat this condition, and 18 did not complete all questions, leaving 90 (20%) complete surveys. However, 7 SOVG members indicated that orthopaedic oncology was one of their subspecialties and we therefore grouped them with the orthopaedic oncologists from the SORG; of the 161 participants in total, 78 (48%) were orthopaedic oncologists (surgeons), the remaining 83 (52%) participants were not. 8 Statistical Analysis An ante-hoc power calculation determined that 122 participants (61 per group) would provide 80% statistical power (beta 0.20; alpha 0.05) to detect a difference in proportion of recommendation for a specific treatment of 20% assuming a proportion of 10% in one group, and 30% in the other. We assessed if there was a difference in baseline characteristics between groups using a Fisher exact test for categorical variables and an unpaired T-test for continuous variables. We calculated treatment scores per specific treatment per surgeon by dividing the amount of cases they recommend a specific treatment for by the number 24 (i.e. the total number of cases) and presented this as a percentage. The score ranges from 0 to 100% with a higher score indicating a higher likelihood of choosing a specific treatment. We used an unpaired T-test to compare the scores for the four specific treatments between the two 153

156 Chapter 8 study groups. We also calculated the observed proportion of agreement and did a Fleiss kappa analysis for choosing: (1) operative treatment versus nonoperative treatment, and (2) specific implants. Kappa is a quantitative measure of agreement among observers and is adjusted for the amount of agreement that can be expected to occur by chance alone. 8,9 Bootstrapping (resamples = 1,000) was used to calculate a standard error, z statistic, p value, and 95% confidence interval. 10 Subsequently, we calculated treatment scores per specific treatment per case by dividing the amount of surgeons choosing a treatment by the total number of surgeons within the study group. We used an unpaired T-test to assess the difference in treatment score based on dichotomous patient characteristics and one-way analysis of variance (ANOVA) for categorical patient characteristics. Intraclass correlation coefficient (ICC) was used to demonstrate interobserver agreement for the estimated fracture risk. The ICC was calculated though a two-way mixed-effects model with absolute agreement. Absolute agreement assesses how much each fracture risk estimate per observer differs from the other observers. As with kappa, a score of 1 reflects perfect agreement in ICC, whereas 0 reflects no agreement. Fisher s z transformation was used to calculate p values. 11 All statistical analyses were performed using Stata 14.0 (StataCorp LP, College Station, TX, USA). A two-tailed p value less than 0.05 was considered statistically significant. RESULTS Baseline Characteristics Of the 161 participants, 149 (93%) were men and the mean years in practice was 15 (Table 1). There was no difference in sex and years in practice between groups. Most participants were from North America (49%) and Europe (39%). There was a difference in location of practice between the two study groups; most participating orthopaedic oncology surgeons were from North America, while the majority of the other orthopaedic surgeons was from Europe (p = 0.001). We also found as expected differences in subspecialty training between both groups (Table 1). Among those not involved in treating bone tumors (other subspecialty orthopaedic surgeons); the majority was subspecialized in trauma surgery (88%). Recommendation For Specific Treatment Based On Subspecialty We found a difference between orthopaedic oncology surgeons and other subspecialty surgeons in recommendation for specific treatments: endoprosthetic reconstruction and plate-screw fixation were more often recommended by orthopaedic oncology surgeons, while intramedullary nailing was more often recommended by other subspecialty surgeons. 154

157 Management Of Metastatic Humeral Fractures Table 1: Baseline characteristics of participating surgeons per group Group 1: Orthopaedic oncology surgeons (n = 78) Mean (±SD) Group 2: Other subspecialty orthopaedic surgeons (n = 83) Mean (±SD) p value Years in practice 15 (10) 16 (8.9) Sex n (%) n (%) Men 70 (90) 79 (95) Women 8 (10) 4 (4.8) Location of practice North America 50 (64) 29 (35) Europe 22 (28) 40 (48) Asia 3 (3.8) 6 (7.2) Autralia 0 (0) 5 (6) South America 3 (3.8) 3 (3.6) Percentage of practice dedicated to orthopaedic oncology surgery 0-25% 8 (10) % 6 (7.7) % 15 (19) % 49 (63) - Surgical Subspecialties* Orthopaedic oncology 78 (100) 0 (0) <0.001 Trauma 8 (10) 73 (88) <0.001 Arthroplasty 19 (24) 15 (18) Shoulder and elbow 2 (2.6) 28 (34) <0.001 General Orthopaedics 5 (6.4) 14 (17) Hand and wrist 1 (1.3) 14 (17) Foot and ankle 3 (3.8) 3 (3.6) Spine 8 (10) 5 (6) Sports 1 (1.3) 10 (12) Paediatric 5 (6.4) 4 (4.8) Other 2 (2.6) 1 (1.2) bold font indicates a significant difference (two-tailed p value below 0.05). SD = standard deviation. - indicates not applicable. *Participants can have multiple subspecialties. Fifty-three percent (SD±26%, 95% CI: 47 to 59) of orthopaedic oncologists recommended intramedullary nailing compared to 62% (SD±24%, 95% CI: 57 to 67) of other subspecialty surgeons (p = 0.023); endoprosthetic reconstruction was recommended by 8.8% (SD±9.8%, 95% CI: 6.6 to 11) of orthopaedic oncologists compared to 3.6% (SD±5.7%, 95% CI: 2.3 to 4.8) of other subspecialty surgeons (p < 0.001); and plate-screw fixation 155

158 Chapter 8 was recommended by 19% (SD±24%, 95% CI: 14 to 25) of orthopaedic oncologists compared to 9.5% (SD±16%, 95% CI: 5.9 to 13) of other subspecialty surgeons (p = 0.003). There was no difference in recommendation for nonoperative management between both groups. Overall, intramedullary nailing was the most commonly recommended treatment (58%), followed by nonoperative management (22%), plate-screw fixation (14%), and endoprosthetic reconstruction (6.1%). Interobserver agreement for recommending operative versus nonoperative management was poor among both orthopaedic oncologists (kappa: 0.23, 95% CI: 0.15 to 0.30, proportion of agreement: 76%) and other subspecialty surgeons (kappa: 0.25, 95% CI: 0.18 to 0.33, proportion of agreement: 72%) and did no differ between these groups (p = 0.642). When assessing interobserver agreement for recommending specific treatments, we found poor agreement among both orthopaedic oncologists (kappa: 0.16, 95% CI: 0.11 to 0.21, proportion of agreement: 46%) and other subspecialty surgeons (kappa: 0.19, 95% CI: 0.14 to 0.24, proportion of agreement: 56%) and again no difference between groups (p = 0.369). Recommendation For Specific Treatment Based On Patient Characteristics There were differences in recommendation for specific treatments based on tumor type, life expectancy, and anatomical location, but not fracture type (Table 2). Among orthopaedic oncology surgeons, we found that plate-screw fixation was more often recommended for renal cell carcinoma (25%), as compared to breast (14%) and lung carcinoma (18%) (p = 0.038). Both endoprosthetic reconstruction (>3 months: 15%, <3 months: 2.7%, p = 0.024) and plate-screw fixation (>3 months: 24%, <3 months: 14%, p = 0.004) were more often recommended for patients who had a life expectancy over 3 months. Nonoperative management was rarely recommended in patients who were expected to live longer than 3 months (>3 months: 3.1%, <3 months: 35%, p < 0.001). Recommendation for a specific implant; intramedullary nailing (proximal: 38%, diaphyseal: 68%, p < 0.001), endoprosthetic reconstruction (proximal: 14%, diaphyseal: 3.3%, p = 0.046), and plate-screw fixation (proximal: 24%, diaphyseal: 15%, p = 0.014) also differed based on the anatomical location of the metastatic lesion (Table 2). Among other subspecialty surgeons, we found that the decision for intramedullary nailing (>3 months: 73%, <3 months: 50%, p = 0.011), plate-screw fixation (>3 months: 14%, <3 months: 5.1%, p < 0.001), and nonoperative management (>3 months: 6.5%, <3 months: 43%, p < 0.001) differed based on life expectancy. Recommendation for a specific implant; intramedullary nailing (proximal: 49%, diaphyseal: 75%, p = 0.003), endoprosthetic reconstruction (proximal: 6.5%, diaphyseal: 0.6%, p = 0.031), and platescrew fixation (proximal: 13%, diaphyseal: 5.6%, p = 0.004) also differed based on the anatomical location of the metastatic lesion (Table 2). 156

159 Management Of Metastatic Humeral Fractures Table 2: Variation in management of metastatic humerus lesions based on case characteristics Tumor type Intramedullary nail Mean (±SD) Breast carcinoma 54 (31) p value Group 1: Orthopaedic oncology surgeons Endoprosthetic reconstruction Mean (±SD) 9.0 (17) p value Plate-screw fixation Mean (±SD) 14 (7.9) p value Nonoperative management Mean (±SD) 23 (22) Renal cell carcinoma 47 (16) (15) (8.9) (15) Lung carcinoma 57 (16) 3.5 (4.8) 18 (8.1) 21 (21) Estimated life expectancy < 3 months 48 (20) 2.7 (4.8) 14 (8.1) 35 (15) > 3 months 58 (24) 15 (17) 24 (7.5) 3.1 (4.2) Fracture type Displaced Pathological fracture 53 (25) (15) (6.9) (17) Impending fracture 53 (20) 6.2 (12) 19 (11) 22 (22) p value < Anatomical location Proximal 38 (20) 14 (17) 24 (7.7) 24 (22) < Diaphysis 68 (12) 3.3 (4.5) 15 (8.7) 14 (16) Tumor type Intramedullary nail Mean (±SD) Breast carcinoma 58 (29) p value Group 2: Other subspecialty orthopaedic surgeons Endoprosthetic reconstruction Mean (±SD) 4.8 (11) p value Plate-screw fixation Mean (±SD) 9.0 (7.1) p value Nonoperative management Mean (±SD) 28 (25) Renal cell carcinoma 65 (20) (4.4) (8.5) (21) Lung carcinoma 63 (21) 2.7 (4.3) 7.7 (5.5) 27 (24) Estimated life expectancy < 3 months 50 (20) 1.0 (2.5) 5.1 (3.7) 43 (16) <0.001 > 3 months 73 (21) 6.1 (8.8) 14 (6.9) 6.5 (7.2) Fracture type Displaced Pathological fracture 64 (26) (8.8) (6.6) (22) Impending fracture 60 (21) 2.1 (4.0) 9.9 (7.7) 28 (24) Anatomical location Proximal 49 (22) 6.5 (8.8) 13 (7.6) 31 (25) Diaphysis 75 (16) 0.60 (0.96) 5.6 (3.5) 19 (19) p value < bold font indicates a significant difference (two-tailed p value below 0.05). The mean indicates the average percentage of surgeons recommending a specific treatment, SD = standard deviation. 157

160 Chapter 8 Interobserver Agreement For Assessing Fracture Risk The interobserver agreement for fracture risk estimation was poor among both orthopaedic oncologists (ICC: 0.053, 95% CI: to 0.15) and other subspecialty surgeons (ICC: 0.042, 95% CI: to 0.12) and did not differ between groups (p = 0.945). DISCUSSION Treatment of metastatic humeral lesions is a challenge and many factors are considered in surgical decision making. It is unclear to what extend tumor, patient, and surgeon characteristics influence the decision for surgical strategy. We found that orthopaedic oncology surgeons are less likely to choose intramedullary nailing, but more likely to use endoprosthetic reconstruction and plate-screw fixation as compared to colleagues that are not regularly involved in treatment of bone tumors. We also found that tumor type, life expectancy, and location of the fracture influence the decision for surgical strategy, but fracture type did not. This study has limitations. First, this study only assessed the decision for operative versus nonoperative management and specific implants, but did not assess adjuvants (e.g. radiation therapy). There is substantial variation in use of these adjuvants. 1 Difference in decision for a specific implant may stem from differences in local treatment of the metastasis (e.g. curettage or resection). Second, only 6.9% of those invited for the SORG and 20% of the members of the SOVG completed the survey. Furthermore, participants are a subgroup within the larger community of orthopaedic surgeons that treat metastatic humeral lesions. The majority of participants are Western orthopaedic surgeons and this might limit the studies generalizability. In addition, when comparing baseline characteristics we found a difference in location of practice between both study groups. This might confound the association of subspecialty with the decision for implants. However, stratifying participants from North America and Europe demonstrated the same trends in use of implants when comparing orthopaedic oncologists and other subspecialty surgeons. In addition, years in practice and sex did not differ between both groups. We therefore feel that the difference in decision for a specific implant based on subspecialty training was not confounded by years in practice, sex, or location of practice. Third, the analyses of interobserver agreement might have been subject to the so-called kappa paradox because the kappa values were considerably lower than the observed proportion of agreement, especially in the decision for operative versus nonoperative management (e.g. the kappa among orthopaedic oncologists was 0.23 [poor agreement], while the overall observed proportion of agreement was 76%). 8,9 The kappa paradox describes that if the prevalence of an outcome is low, it causes an imbalance in the marginal totals, generating a low kappa. 8 This might have been the case in calculating the kappa values in our study 158

161 Management Of Metastatic Humeral Fractures and we therefore also presented the observed proportions of agreement. We see this as a minor limitation as a comparison of kappa values or observed proportions of agreement did not result in a difference in interobserver agreement. Our finding that orthopaedic oncology surgeons recommend endoprosthetic reconstruction and plate-screw fixation more often, and less commonly recommend intramedullary nailing as compared to other subspecialty surgeons probably has its root in their training. Orthopaedic oncologists are trained in the treatment of bone and soft tissue neoplasms, developmental dysplasias, tumor-like conditions, and major skeletal defects. 12 As such, they might be more comfortable with creating a bony defect by resection of a metastatic lesion and subsequently reconstructing this defect using a prosthesis or plate-screw fixation with cement augmentation or allograft to fill the defect. In addition, they might be more aware of options for adjuvant treatment than other subspecialists and therefore base their decision for an implant on its combination with adjuvant treatment. Furthermore, orthopaedic oncologists might be more often confronted with the local complications of surgical management (such as tumor progression) and therefore be more inclined to treat metastatic lesions aggressively. It would be interesting to compare surgical and oncological outcomes between orthopaedic oncologists and orthopaedic surgeons not regularly involved in treatment of bone tumors as their surgical decision making differs. We also studied the impact of commonly cited patient and tumor factors that are considered in surgical decision making and found that estimated life expectancy is a strong 8 Figure 1A: An example of a case that demonstrated much variability in recommendation for treatment (Case 5B). Case description: A 59-year-old female has activity related pain in her arm. The patient has widely metastatic renal cell carcinoma. The patient walks with a walker, and has not had radiotherapy. The estimated life expectancy is more than 3 months. Figure 1B: Bar graphs demonstrating the treatment recommendations among orthopaedic oncology surgeons (left) and other subspecialty surgeons (right) for the case presented in Figure 1A. 159

162 Chapter 8 driver of recommending operative versus nonoperative management; operative treatment was recommended for almost all patients with more than 3 months life expectancy, while nonoperative management was recommended for over one-third of those with life expectancy less than 3 months. The strong influence of life expectancy on decision making emphasizes the need for an accurate method to estimate survival. 13 Tumor type, life expectancy, and anatomical location influences the decision for a specific implant. For example, intramedullary nailing was most commonly recommended for metastatic lesions located in the diaphysis, while endoprosthetic reconstruction and plate-screw fixation were more commonly recommended for proximal metastatic lesions. Trends in how these factors influence decision making were comparable between orthopaedic oncologists and other subspecialty surgeons. Some specific cases demonstrated more variability in recommendation for treatment (Figure 1) than others (Figure 2); participants seemed to be more likeminded when choosing treatment for diaphyseal lesions as compared to proximal lesions. Our findings can help to create guidelines by reaching consensus; this might reduce variation, improve quality of care, and decrease costs. 14,15 However, our findings do not imply that one approach is superior to another. Studies that directly compare outcomes between different treatments, preferably in a prospective experimental study design, are needed. Especially the areas that demonstrated most controversy (e.g. Figure 1) benefit from further comparative study. Figure 2A: An example of a case that demonstrated little variability in recommendation for treatment (Case 3B). Case description: A 62-year-old female has activity related pain in her arm. The patient has widely metastatic breast carcinoma. The patient walks with a walker, and has not had radiotherapy. The estimated life expectancy is more than 3 months. Figure 2B: Bar graphs demonstrating the treatment recommendations among orthopaedic oncology surgeons (left) and other subspecialty surgeons (right) for the case presented in Figure 2A. 160

163 Management Of Metastatic Humeral Fractures The interobserver agreement for assessment of fracture risk was poor. Fracture risk is an important factor in deciding whether to treat an impending fracture surgically or not. However, the accuracy of current classification systems (e.g. the Mirels classification; including the site, lesion type, size of the lesion, and pain) in predicting the risk of fracture is relatively poor A Mirels score of 9 or higher is commonly used as a cut off for recommending prophylactic fixation. New techniques, such as finite element analyses and CT-based rigidity analysis demonstrate promising results Including data from such techniques might improve fracture risk estimation. 22 In conclusion, subspecialty training and patient and tumor characteristics influence the decision for surgical management and the decision for a specific implant in metastatic humeral lesions. Our findings can be used when developing guidelines and demonstrate where additional comparative study is most needed. It could be valuable to discuss metastatic humeral lesions cases preoperatively in a team with an orthopaedic oncologist because of their expertise in reconstructing major skeletal defects and knowledge about adjuvant treatments. REFERENCES 1. Janssen SJ, Teunis T, Hornicek FJ, Bramer JA, Schwab JH. Outcome of operative treatment of metastatic fractures of the humerus: a systematic review of twenty three clinical studies. Int Orthop. Apr 2015; 39(4): Wedin R, Hansen BH, Laitinen M, et al. Complications and survival after surgical treatment of 214 metastatic lesions of the humerus. J Shoulder Elbow Surg. 2012; 21: Piccioli A, Maccauro G, Rossi B, Scaramuzzo L, Frenos F, Capanna R. Surgical treatment of pathologic fractures of humerus. Injury. 2010; 41: Bickels J, Dadia S, Lidar Z. Surgical management of metastatic bone disease. J Bone Joint Surg Am. Jun 2009; 91(6): Quinn RH, Randall RL, Benevenia J, Berven SH, Raskin KA. Contemporary management of metastatic bone disease: tips and tools of the trade for general practitioners. Instr Course Lect. 2014; 63: Steensma M, Healey JH. Trends in the surgical treatment of pathologic proximal femur fractures among Musculoskeletal Tumor Society members. Clin Orthop Relat Res. Jun 2013; 471(6): Guitton TG, Ring D. Interobserver reliability of radial head fracture classification: twodimensional compared with three-dimensional CT. J Bone Joint Surg Am. Nov ; 93(21): Feinstein AR, Cicchetti DV. High agreement but low kappa: I. The problems of two paradoxes. J Clin Epidemiol. 1990; 43(6): Cicchetti DV, Feinstein AR. High agreement but low kappa: II. Resolving the paradoxes. J Clin Epidemiol. 1990; 43(6): Lee J, Fung KP. Confidence interval of the kappa coefficient by bootstrap resampling. Psychiatry Res. Oct 1993; 49(1):

164 Chapter Bond C, Jr., Richardson K. Seeing the FisherZ-transformation. Psychometrika. 2004/06/ ; 69(2): White J, Toy P, Gibbs P, Enneking W, Scarborough M. The current practice of orthopaedic oncology in North America. Clin Orthop Relat Res. Nov 2010; 468(11): Nathan SS, Healey JH, Mellano D, et al. Survival in patients operated on for pathologic fracture: implications for end-of-life orthopedic care. J Clin Oncol. Sep ; 23(25): Every NR, Hochman J, Becker R, Kopecky S, Cannon CP. Critical pathways: a review. Committee on Acute Cardiac Care, Council on Clinical Cardiology, American Heart Association. Circulation. Feb ; 101(4): Kirkpatrick DH, Burkman RT. Does standardization of care through clinical guidelines improve outcomes and reduce medical liability? Obstet Gynecol. Nov 2010; 116(5): Mirels H. Metastatic disease in long bones. A proposed scoring system for diagnosing impending pathologic fractures. Clin Orthop Relat Res. Dec 1989(249): Damron TA, Morgan H, Prakash D, Grant W, Aronowitz J, Heiner J. Critical evaluation of Mirels rating system for impending pathologic fractures. Clin Orthop Relat Res. Oct 2003(415 Suppl): S Van der Linden YM, Dijkstra PD, Kroon HM, et al. Comparative analysis of risk factors for pathological fracture with femoral metastases. J Bone Joint Surg Br. May 2004; 86(4): Anez-Bustillos L, Derikx LC, Verdonschot N, et al. Finite element analysis and CT-based structural rigidity analysis to assess failure load in bones with simulated lytic defects. Bone. Jan 2014; 58: Derikx LC, van Aken JB, Janssen D, et al. The assessment of the risk of fracture in femora with metastatic lesions: comparing case-specific finite element analyses with predictions by clinical experts. J Bone Joint Surg Br. Aug 2012; 94(8): Snyder BD, Hauser-Kara DA, Hipp JA, Zurakowski D, Hecht AC, Gebhardt MC. Predicting fracture through benign skeletal lesions with quantitative computed tomography. J Bone Joint Surg Am. Jan 2006; 88(1): Nazarian A, Entezari V, Villa-Camacho JC, et al. Does CT-based Rigidity Analysis Influence Clinical Decision-making in Simulations of Metastatic Bone Disease? Clin Orthop Relat Res. May

165

166

167 PART III Survival

168

169 CHAPTER 9 Metastasectomy, Intralesional Resection, Or Stabilization Only In The Treatment Of Bone Metastases From Renal Cell Carcinoma D.W.G. Langerhuizen S.J. Janssen Q.M.J. van der Vliet K.A. Raskin M.L. Ferrone F.J. Hornicek J.H. Schwab S.A. Lozano-Calderon 9 Journal of Surgical Oncology 2016 Aug;114(2): Poster at: Massachusetts General Hospital Clinical Research Day 2015, Boston, Massachusetts, USA.

170 Chapter 9 ABSTRACT Objectives To assess differences in: (1) local tumor recurrence, (2) reoperation, and (3) survival between metastasectomy, intralesional curettage, and stabilization only for renal cell carcinoma metastases to the appendicular skeleton. Secondarily, we evaluated whether there is a difference in these outcomes based on tumor margin status. Design Retrospective cohort study. Setting Two tertiary care referral centers for orthopaedic oncology. Participants 183 consecutive patients with renal cell carcinoma bone metastasis in the appendicular skeleton who underwent surgery between 1990 and Interventions Metastasectomy (48%, n=88, margins: 64 negative; 20 positive; 4 unclear), intralesional curettage (30%, n=54), and stabilization only (22%, n=41). Outcome measures Local recurrence, reoperation, and survival. Results The recurrence rate differed and was highest after stabilization only (39%), followed by intralesional curettage (22%), and metastasectomy (12%) (p = 0.003). However, we found no difference in reoperation rate (p = 0.847). Survival was better in patients who underwent metastasectomy (p = 0.020). The recurrence rate was lower in patients who had a negative margin (5%) as compared to those with a positive margin (26%) (p < 0.001). However, we found no difference in reoperation rate (p = 0.97). Negative margins showed better survival (p < 0.001). Conclusions Our findings emphasize the importance of obtaining negative margins in patients with a good life expectancy, as lower recurrence rate can be attained at a comparable risk for reoperation, with a potential impact on survival. 168

171 Treatment Of Bone Metastases From Renal Cell Carcinoma INTRODUCTION Renal cell carcinoma (RCC) bone metastases are often hypervascular and relatively radioresistant. 1 Patients have a favorable prognosis as compared to patients with other primary tumors; their overall one-year survival rate ranges from 47% to 88%. 2-5 RCC bone metastases are challenging to treat given their radioresistant nature. The mainstay of treatment is surgery with indications for surgical intervention ranging from a pathological fracture, through a lesion at risk of fracture, to a solitary lesion. 6 Surgical techniques are usually grouped into metastasectomy, intralesional treatment, and stabilization only. 7,8 Factors considered in the decision to operate and the selection of surgical strategy include: location and size of the lesion, oncologic burden (i.e. visceral and other bone metastases), presence of a fracture, and life expectancy. 9,10 It is unclear how oncological outcome survival and tumor recurrence are affected by these different surgical strategies. Several studies suggest improved survival after metastasectomy as compared to patients who underwent intralesional treatment; however, these findings are not confirmed by others. 4,7,8,11 Tumor recurrence and subsequent reoperations are less commonly described and compared between surgical strategies. 8,12 We aim to compare different surgical techniques for renal cell carcinoma bone metastases. Specifically, we assessed differences in: (1) local tumor recurrence, (2) reoperation, and (3) survival between metastasectomy, intralesional curettage, and stabilization only for renal cell carcinoma metastasis to the appendicular skeleton. Secondarily, we evaluated whether there is a difference in these outcomes based on tumor margin status. 9 METHODS Study Design Our institutional review board approved this retrospective study. We used our orthopaedic oncology database including data on all patients with musculoskeletal tumors from two tertiary care referral centers for orthopaedic oncology and queried pathology and operative reports using a combination of the words renal cell carcinoma and bone and metastases, including synonyms and misspellings between January 1990 and January We identified 864 patients. We reviewed the medical records of all identified patients and included patients 18 years of age and older who underwent surgical treatment for a renal cell metastasis in the appendicular skeleton. Exclusion criteria were revision surgery and biopsies. We only included the first surgery per patient so as to not violate the assumption of independence. 13 All remaining 183 eligible patients were included. We included all patients, regardless of followup duration, and accounted for this in our statistical analyses. 169

172 Chapter 9 Surgical Strategies The surgeon decided together with the patient on the need for surgery and the surgical strategy. Eighty patients (44%) were treated in hospital 1 and 103 (56%) in hospital 2. One hundred seventy-four (95%) of the 183 surgeries were performed by 10 orthopaedic oncologists; the remainder by trauma surgeons. We divided the surgical approach into: metastasectomy, intralesional curettage, and stabilization only. In general, a higher proportion of patients with a solitary metastasis or less disseminated disease resulting in a relatively good prognosis and low metastatic load underwent metastasectomy. These procedures were not limited by the location of the tumor; enbloc resections were performed in the extremities, the shoulder girdle, and the pelvis. An endoprosthetic reconstruction was almost always performed after resection; however, in some pelvic and scapular lesions, reconstruction was not deemed necessary (Appendix 1). Postoperative radiation therapy was not routinely performed after metastasectomy. Stabilization by intramedullary nailing or open reduction and internal fixation using plate and screws with or without intralesional curettage and cement packing was more commonly done in patients with a worse prognosis and higher oncologic burden. Patients receiving intralesional curettage with subsequent fixation with plates and screws or intramedullary nail devices, and patients whom received surgical stabilization only, were more often treated with postoperative radiation. There was heterogeneity in postoperative care between patients due to differences in the degree of illness. Outcome Measures Our outcome measures were (1) local tumor recurrence/progression, (2) reoperation, and (3) survival. Local recurrence was defined as identification of metastasis in radiographs or CT scan after resection with negative margins or resection with positive margins (intralesional curettage and enbloc resections with positive margins). 14 Tumor progression, grouped with local tumor recurrence, was defined as identification of metastasis increasing in size in radiographs or CT scan in patients receiving surgical stabilization only. 14 A patient with no signs of recurrence or progression at the last imaging was considered censored, meaning that the outcome did not occur by the last moment of followup. Reoperations were identified by two research fellows (D.L., Q.v.d.V.) who independently checked all reports of surgeries performed after the index procedure. A patient who did not undergo a reoperation before the last moment of clinical followup was considered censored. Survival was defined as the period between the date of surgery and date of death. We used the Social Security Death Index to determine the date of death; March 18 th 2015 was the final moment of followup to establish the date of death. 15 We only considered the first notice of local recurrence and reoperation in our time-to-event analyses. Fourteen (8%) patients had no followup imaging and were not included in the analysis of local recurrence, 1 (0.5%) 170

173 Treatment Of Bone Metastases From Renal Cell Carcinoma patient had no clinical followup and was not included in the reoperation analysis, and all patients had complete followup for date of death. We obtained data on the following explanatory variables through chart review of electronic medical records by two research fellows (D.L., Q.v.d.V.): age, sex, body mass index, smoking status, preoperative hemoglobin level (in g/dl), preoperative white blood cell count (in K/uL), preoperative albumin level (in g/dl), preoperative calcium level (in mg/ dl), histological subtype, time diagnosis primary tumor until surgery, time first notice of a metastasis until surgery, presence of a solitary metastasis, presence of visceral metastases, presence of multiple bone metastases, Eastern Cooperative Oncology Group (ECOG) performance status, modified Charlson Comorbidity Index, additional comorbidities, previous systemic therapy, previous conventional radiation, postoperative conventional radiation within two months, anesthesia type, preoperative embolization, anatomic location of surgery, presence of a pathological fracture, surgical strategy, implant type, cement use during surgery, margin status, estimated blood loss, duration of surgery, length of hospital stay, reason for reoperation, type of reoperation, and last date of followup. The surgical technique, anatomic location of surgery, and the implant type were derived from operative reports (Appendix 1). The surgical strategy to remove the tumor was divided into: metastasectomy, intralesional curettage, and stabilization only. Metastasectomy was defined as an intended enbloc resection with or without reconstruction. 6 Intralesional curettage with cement packing was often combined with intramedullary nailing or open reduction and internal fixation with plate and screws. Stabilization only was almost always done using an intramedullary nail. Pathology reports were reviewed to assess the margin status, which was considered to be positive in case of any residual tumor after enbloc resection. Margins were also considered positive in all intralesional curettage and stabilization only patients. 4 We also reviewed the pathology reports to assess the histological subtype of the tumor lesion. Tumor margin was not reported for four (2%) patients, we did not include these in our analysis. The Eastern Cooperative Oncology Group (ECOG) performance status was retrieved from chart review. This instrument describes a patient level of function ranging from fully active (0) to completely disabled (4). We dichotomized this into two groups: and 3-4. An algorithm based on the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes was used to calculate the Modified Charlson Comorbidity Index score at the time of the index surgery (Appendix 2). 16 The index ranges from 0 to 24 with a higher score indicating more severe comorbidity status. Additional comorbidity was defined as a patient who had any of the comorbidities in addition to the metastatic tumor. We compared outcomes between patients who underwent metastasectomy, intralesional curettage, and stabilization only. We also compared outcomes between patients with a negative (enbloc resection with negative margin) and positive margin (intralesional curettage, stabilization only, and enbloc resection with positive margins)

174 Chapter 9 We did four additional subanalyses: assessing survival based on resection type and margin status in patients with a solitary metastases only (24% [44/183]), and survival based on resection type and margin status in patients with multiple visceral or bone metastases (76% [139/183]). Followup The median followup was 10 months (interquartile range: 2 to 26 months, range 0 months to 17 years). One hundred fourteen (62%) patients were alive at clinical followup after 6 months, 62 (34%) were deceased and 7 (4%) were lost to followup. After 12 months: 84 (46%) patients were still in clinical followup, 87 (48%) were deceased, and 12 (7%) were lost to followup. Two years after surgery, 49 (27%) were still in clinical followup, 112 (61%) were deceased, and 22 (12%) were lost to followup. Five years after surgery, 10 (5%) were still in clinical followup, 137 (75%) were deceased, and 36 (20%) were lost to followup. Statistical Analysis We presented continuous variables as a mean with standard deviation (SD) and categorical variables as frequencies with percentages. Baseline characteristics were compared in bivariate analysis to assess differences between the group of patients who underwent metastasectomy, intralesional curettage, and stabilization only. We used a Fisher exact test for categorical variables and a one-way analysis of variance (ANOVA) for continuous variables. We used log-rank analysis to compare local tumor recurrence, reoperation, and survival between the three groups. We used Kaplan-Meier plots to demonstrate the outcomes over time and calculate one-year and five-year probabilities with 95% confidence intervals (95% CI). Visual inspection of log-log plots suggest no violation of the proportional hazards assumption. 17 All statistical analyses were performed using Stata 13.0 (StataCorp LP, College Station, TX, USA). A two-tailed p value less than 0.05 was considered statistically significant. There were missing values for some variables (Table 1); these were not included in the respective analyses. RESULTS Baseline Characteristics One-hundred and twenty-seven patients were men (69%); the overall mean age was 65 years (SD 10) (Table 1). Eighty patients (44%) had a pathological fracture and 103 patients (56%) had an impending fracture. Eighty-eight (48%) patients underwent a metastasectomy, 54 (30%) underwent intralesional curettage, and 41 (22%) had stabilization only. Of the 172

175 Treatment Of Bone Metastases From Renal Cell Carcinoma 88 patients who underwent metastasectomy; 64 (73%) had negative margins, 20 (23%) had positive margins, and margin status was unclear in four. The majority had a clear cell histological subtype (86% [140/163]), and the histological subtype was unknown in 20 (11%) patients. The mean estimated blood loss available in 163 patients was 772 ml (SD: 1199 ml), the mean length of hospital stay available in 177 patients was 7.3 days (SD: 6.0 days). When comparing baseline characteristics between groups we found that (Table 2): visceral metastases were more common in the intralesional curettage group (70%) and stabilization only group (71%) as compared to the metastasectomy group (42%) (p = 0.001); multiple bone metastases were most common in the stabilization only group (73%), and less common in the intralesional curettage (48%) and metastasectomy group (47%) (p = 0.012). This results in a higher proportion of patients with solitary metastasis (34%) in the metastasectomy group as compared to the other groups (p = 0.001). Postoperative radiation within two months (p < 0.001) was more commonly done after stabilization only and intralesional curettage. Cement was predominantly used in the intralesional curettage and metastasectomy groups (p < 0.001). The type of reconstruction also varied between the groups; endoprosthetic reconstruction was the most common after metastasectomy, open reduction internal fixation after intralesional curettage, and intramedullary nailing after stabilization only (p < 0.001). Table 1: Baseline characteristics Patient characteristics Mean (±Standard deviation) Age (in years) 65 ± 10 Body mass index* (in kg/m²) 29 ± 7.1 Modified Charlson Comorbidity Index 6.7 ±1.5 Time since primary tumor diagnosis (in days)* 1306 ± 1641 Time since diagnosis of metastasis (in days)* 463 ± 889 Hemoglobin (in g/dl)* 11 ± 1.9 White blood cell count (in K/uL)* 9.9 ± 4.7 Albumin (in g/dl)* 3.8 ± 0.6 Calcium (in mg/dl)* 8.9 ± 0.8 n (%) Men 127 (69) Smoking* Never smoked 45 (30) Quit > 1 year ago 82 (55) Current smoker 23 (15) Additional comorbidity 59 (33) n =

176 Chapter 9 Table 1: Baseline characteristics (continued) Eastern Cooperative Oncology Group Performance status* Mean (±Standard deviation) (92) (8) Tumor characteristics n (%) Pathological fracture 80 (44) Solitary metastasis 44 (24) Visceral metastases None 77 (42) Lung/liver 82 (45) Brain 6 (3) Lung/liver and brain 18 (10) Multiple bone metastases 97 (53) Histology* Clear cell 140 (86) Non clear cell* 23 (14) Treatment characteristics n (%) Operation Endoprosthetic reconstruction 70 (38) Intramedullary nailing 52 (28) Open reduction internal fixation 39 (21) Resection/curretage only 22 (12) Type of resection Metastasectomy (En bloc resection) 88 (48) Intralesional curettage 54 (30) Stabilization only 41 (22) Cement use 101 (55) Negative margins 65 (36) Previous local radiotherapy 50 (27) Postoperative radiotherapy within two months 49 (27) Previous systemic therapy 56 (31) Preoperative embolization 87 (48) Body mass index was available in 111 (61%) cases, modified Charlson Comorbidity Index in 181 (99%), time since primary tumor diagnosis in 179 (98%), time since diagnosis of metastasis in 177 (97%), hemoglobin in 178 (97%), white blood cell count in 179 (98%), albumin in 100 (55%), calcium in 167 (91%), smoking in 150 (81%), Eastern Cooperative Oncology Group Performance Status in 107 (58), histology in 163 (89) (Non clear cell histology: Sacromatoid (7), clear cell + sarcomatoid (5) clear cell + granular (2), poorly differentiated (2), chromophobe (1), adenocarcinoma (1), clear cell + papillary + oxyphilic (1), clear cell + papillary + rhabdoid (1), clear cell + sarcomatoid + papillary (1), granular (1), sarcomatoid + unclassified (1) 174

177 Treatment Of Bone Metastases From Renal Cell Carcinoma Table 2: Bivariate analyses Patient characteristics Metastasectomy (n = 88) Mean (±Standard deviation) Intralesional curettage (n = 54) Mean (±Standard deviation) Stabilization only (n = 41) Mean (±Standard deviation) n = 183 p value Age (in years) 63 ±10 66 ±11 67 ± Body mass index (in kg/m 2 )* 29 ± ± ± Modified Charlson Comorbidity Index* 6.6 ± ± ± Time since primary tumor diagnosis (in days)* 1479 ± ± ± Time since diagnosis of metastasis (in days)* 432 ± ± ± Hemoglobin (in g/dl)* 11 ± ± ± White blood cell count (in K/uL)* 10 ± ±5.6 9 ± Albumin (in g/dl)* 3.9 ± ± ± Calcium (in mg/dl)* 8.9 ± ± ± n (%) n (%) n (%) Men 62 (70) 39 (72) 26 (63) Smoking* Never smoked 24 (36) 12 (27) 9 (23) Quit > 1 year ago 36 (54) 20 (45) 26 (67) Current smoker 7 (10) 12 (27) 4 (10) Additional comorbidity 27 (31) 15 (29) 17 (41) Eastern Cooperative Oncology Group Performance status* (93) 29 (85) 29 (97) (7) 5 (14) 1 (3) Tumor characteristics n (%) n (%) n (%) Pathological fracture 40 (45) 24 (44) 16 (39) Solitary metastasis 30 (34) 12 (22) 2 (5) Visceral metastases 37 (42) 38 (70) 29 (71) Multiple bone metastases 41 (47) 26 (48) 30 (73) Histology* Clear cell 65 (83) 41 (87) 34 (89) Non clear cell 13 (17) 6 (13) 4 (11) Treatment characteristics n (%) Operation Endoprosthetic reconstruction 61 (69) 9 (17) 0 (0) Intramedullary nailing 0 13 (24) 39 (95) Open reduction and internal fixation 13 (15) 24 (44) 2 (5) Resection/curettage 14 (16) 8 (15) 0 (0) <

178 Chapter 9 Table 2: Bivariate analyses (continued) Metastasectomy (n = 88) Intralesional curettage (n = 54) Stabilization only (n = 41) p value Cement use 57 (65) 43 (80) 1 (2) <0.001 Previous local radiotherapy 29 (33) 13 (24) 8 (20) Postoperative radiotherapy within two months 7 (8) 18 (33) 24 (59) <0.001 Previous systemic therapy 25 (28) 18 (33) 13 (32) Preoperative embolization 49 (56) 22 (41) 16 (39) Body mass index was available in 111 (61%) cases, modified Charlson Comorbidity Index in 181 (99%), time since primary tumor diagnosis in 179 (98%), time since diagnosis of metastasis in 177 (97%), hemoglobin in 178 (97%), white blood cell count in 179 (98%), albumin in 100 (55%), calcium in 167 (91%), smoking in 150 (82%), Eastern Cooperative Oncology Group Performance Status in 107 (58), histology in 163 (89) (Non clear cell histology: sarcomatoid (7), clear cell + sarcomatoid (5) clear cell + granular (2), chromophobe (1), adenocarcinoma (1), clear cell + papillary + oxyphilic (1), clear cell + papillary + rhabdoid (1), clear cell + sarcomatoid + papillary (1), granular (1), poorly differentiated (1), sarcomatoid + unclassified (1). Bold indicates significant difference (two-tailed p value below 0.05). Local Tumor Recurrence, Reoperation, And Survival Between Treatment Groups The local recurrence rate varied between groups and was highest after stabilization only (39%, [15/38]), followed by intralesional curettage (22%, [11/50]), and metastasectomy (12%, [10/81]) (p = 0.003) (Figure 1). The oneyear local recurrence probability was 0.13 (95% CI: 0.06 to 0.26) after metastasectomy, 0.32 (95% CI: 0.17 to 0.55) after intralesional curettage, and 0.20 (95% CI: 0.09 to 0.42) after stabilization only. The five-year local recurrence probability was 0.26 (95% CI: 0.14 to 0.44) after metastasectomy, 0.77 (95% CI: 0.40 to 0.98) after intralesional curettage, and 0.85 (95% CI: 0.63 to 0.97) after stabilization only. However, we found no difference, with the numbers evaluated, in overall reoperation rate between the metastasectomy (19%, [17/88]), intralesional curettage (19%, [10/53]), and surgical stabilization groups (17%, [7/41]) (p = 0.847) (Figure 2). The one-year reoperation probability was 0.21 (95% CI: 0.12 to 0.34) after metastasectomy, 0.16 (95% CI: 0.06 to 0.32) after intralesional curettage, and 0.07 (95% CI: 0.02 to 0.27) after stabilization only. The five-year reoperation probability was 0.32 (95% CI: 0.21 to 0.48) after metastasectomy, 0.63 (95% CI: 0.41 to 0.93) after intralesional curettage, and 0.47 (95% CI: 0.26 to 0.87) after stabilization only. The overall most common reasons for reoperation were: recurrence (11%, [20/182]) and deep infection (4.9%, [9/182]). After metastasectomy 6.8% (6/88) had a reoperation for a deep infection and 6.8% (6/88) for recurrence, while the rates were 15% (8/53) for recurrence and 5.7% (3/53) for deep infection after intralesional curettage, and 15% (6/41) for recurrence in the stabilization only group without any deep infection described (Table 3). 176

179 Treatment Of Bone Metastases From Renal Cell Carcinoma Figure 1: Local recurrence probability in the metastasectomy group (dashed dot line), intralesional curettage group (solid line), and stabilization only group (dashed line) are shown. 9 Figure 2: Reoperation probability in the metastasectomy group (dashed dot line), intralesional curettage group (solid line), and stabilization only group (dashed line) are shown. Survival was better in patients who underwent metastasectomy as compared with those who underwent intralesional curettage or stabilization only (p = 0.020) (Figure 3). However, when we focused on patients with solitary bone metastases (24%, [44/183]) we found no difference, with the numbers evaluated, in survival between treatment groups (p = 0.997). Focusing on patients with multiple metastases (76%, [139/183]) we also found no difference in survival comparing the different surgical techniques (p = 0.099). 177

180 Chapter 9 Table 3: Reason for reoperation and surgical strategy Sex, Age Path. fract. Histology Anatomic location Type of surgery F, 71 No CC Acetabulum Intralesional curettage M, 70 No CC F, 46 No CC Tibia Ulna/ metacarpal M, 72 No CC Pelvis M, 62 No CC Femur M, 67 Yes NCC Humerus M, 69 No CC Femur M, 72 Yes CC Humerus M, 49 No CC Humerus M, 72 Yes CC Tibia Intralesional curettage Intralesional curettage Intralesional curettage Intralesional curettage Intralesional curettage Intralesional curettage Intralesional curettage Intralesional curettage Intralesional curettage Implant ORIF ORIF Curettage only Curettage only ORIF ORIF ORIF IMN Curettage only IMN M, 63 No CC Femur Stabilization only IMN Reason for reoperation Deep infection Deep infection Pain and recurrence Recurrence Recurrence Recurrence Recurrence Recurrence Recurrence Recurrence & deep infection Peri-implant fracture M, 61 Yes CC Humerus Stabilization only IMN Recurrence Months, reoperation 0, ID 1, ID 3, Tumor resection, MTP - knee + cement 13, Resection soft tissue recurrence 16, Revision curettage, new plate-screw fixation + cement 19, Tumor resection, new plate-screw fixation + cement 52, Revision curettage, new plate-screw fixation + cement 58, Tumor resection, nail removal, hemia long-stem (shoulder) 6, Tumor resection + osteoarticular allograft and plate 11, ID, tumor resection, AB impregnated spacer 1, Refixation nail 10, Tumor resection, nail removal, MTP shoulder M, 55 No CC Femur Stabilization only IMN Recurrence 18, Tumor resection, THA M, 65 Yes CC Humerus Stabilization only IMN Recurrence M, 48 Yes CC Humerus Stabilization only IMN Recurrence M, 63 Yes CC Femur Stabilization only IMN Recurrence M, 67 No CC Femur Stabilization only IMN Recurrence M, 44 No CC Femur Metastasectomy ORIF M, 57 Yes CC Humerus Metastasectomy ORIF Allograft fracture Allograft nonunion 20, Tumor resection, nail removal, intercalary allograft + new nail 24, Tumor resection, nail removal, intercalary prosthesis + cement 25, Tumor resection, nail removal, MTP - with cup (hip) 29, Tumor resection, total femur replacement (hinge knee + bipolar hip) 6, Revision ORIF 14, Removal allograft, new allograft with hemia longstem (shoulder) 178

181 Treatment Of Bone Metastases From Renal Cell Carcinoma Table 3: Reason for reoperation and surgical strategy (continued) Sex, Age Path. fract. Histology Anatomic location Type of surgery Implant M, 76 Yes CC Femur Metastasectomy EPR M, 70 Yes CC Femur Metastasectomy EPR F, 67 No CC Pelvis Metastasectomy ORIF M, 63 Yes CC Humerus Metastasectomy EPR F, 86 No CC Pelvis Metastasectomy Resection only F, 54 No CC Tibia Metastasectomy EPR Reason for reoperation Deep infection Deep infection Deep infection Deep infection Deep infection Deep infection Months, reoperation 1, ID 1, ID 1, ID 2, ID + head exchange F, 55 No CC Pelvis Metastasectomy EPR Dislocation 1, Open reduction M, 57 No CC Femur Metastasectomy EPR M, 79 Yes CC Humerus Metastasectomy ORIF M, 63 No CC Scapula Metastasectomy Resection only Loosening prosthesis Loss of distal fixation Recurrence 4, ID 7, ID 3, Revision femoral component 0, Revision plate fixation 10, Resection soft tissue recurrence M, 71 Yes NCC Tibia Metastasectomy ORIF Recurrence 14, Knee disarticulation F, 53 No CC Pelvis Metastasectomy Resection only Recurrence F, 54 Yes CC Humerus Metastasectomy ORIF Recurrence 14, Tumor resection hemipelvis 17, Removal allograft, tumor resection, new allograft + orif M, 55 Yes CC Humerus Metastasectomy EPR Recurrence 9, Above elbow amputation F, 50 No NCC Femur/ acetabulum Metastasectomy EPR Recurrence & dislocation 10, Open reduction 9 IMN = intramedullary nailing, EPR = endoprosthetic replacement, ORIF = open reduction and internal fixation, MTP = modular total prosthesis, ID = Incision & Debridement. Path. Fract. = Pathological fracture, CC = Clear cell, NCC = Non clear cell Local Tumor Recurrence, Reoperation, And Survival Based On Margin Status The local recurrence rate was lower in patients who had a negative margin as compared to those with a positive margin (p < 0.001) (Figure 4). The one-year recurrence probability was 0.03 (95% CI: 0.00 to 0.21) when a patient had negative margins and 0.25 (95% CI: 0.16 to 0.39) when a patient had positive margins. The five-year recurrence probability was 0.11 (95% CI: 0.04 to 0.31) when a patient had negative margins and 0.87 (95% CI: 0.60 to 0.99) when a patient had positive margins. However, we found no difference in overall reoperation rate between patients who had a negative margin as compared to those with a positive margin (p = 0.975) (Figure 5). 179

182 Chapter 9 Figure 3: Survival probability in the metastasectomy group (dashed dot line), intralesional curettage group (solid line), and stabilization only group (dashed line) are shown. Figure 4: Local recurrence probability in the negative margins group (solid line) and positive margins group (dashed line) are shown. The one-year reoperation probability was 0.19 (95% CI: 0.11 to 0.33) when a patient had negative margins and 0.11 (95% CI: 0.05 to 0.21) when a patient had positive margins. The five-year reoperation probability was 0.23 (95% CI: 0.15 to 0.44) when a patient had negative margins and 0.65 (95% CI: 0.36 to 0.92) when a patient had positive margins. We found a better survival among patients who had negative margins (p < 0.001) (Figure 6). However, when we focused on patients with solitary bone metastases (24%, [44/183]) 180

183 Treatment Of Bone Metastases From Renal Cell Carcinoma Figure 5: Reoperation probability in the negative margins group (solid line) and positive margins group (dashed line) are shown. 9 Figure 6: Survival probability in the negative margins group (solid line) and positive margins group (dashed line) are shown. we found no difference, with the numbers evaluated, in survival between patients with negative margins and those with positive margins (p = 0.35). We did find a difference in survival based on margin status (p = 0.015) in patients with multiple metastases (76%, [139/183]); patients with negative margins had better survival than patients with positive margins. 181

184 Chapter 9 DISCUSSION The influence of surgical intervention on survival in patients with bone metastatic disease from renal cell carcinoma is previously described but studies are contradictory. 4,7,8,11 Conversely, the difference in local recurrence and reoperation rates after different surgical modalities are less commonly studied. 8,12 We therefore aimed to assess difference in these outcomes between patients who underwent metastasectomy, intralesional curettage, and surgical stabilization only for renal cell carcinoma bone metastases in a relatively large cohort of patients. We found a difference in local recurrence rate based on surgical strategy; recurrence rate was highest after stabilization only, followed by intralesional curettage, and metastasectomy. However, we found no difference in overall reoperation rates between strategies. Overall survival was better in patients who underwent metastasectomy and those who had negative surgical margins; however, this difference did not hold when focusing on patients with solitary bone metastases. Albeit, our sample size of patients with solitary metastasis might have been too small to detect a difference in this subanalysis. A study by Evenski et al. 12 showed higher local recurrence rate after intralesional procedures (27%) as compared to wide resections (5%) in 86 patients with a renal cell metastasis of the appendicular skeleton. Another study by Les et al. 8 demonstrated a reoperation rate for tumor progression of 41% after internal fixation (with or without curettage and cement augmentation) as compared to 2.7% after marginal or wide resection (with or without reconstruction) in 78 patients with osseous renal cell metastasis. These previously reported differences are comparable to our findings; we found a 39% local recurrence rate in the stabilization only group, 22% in the intralesional curettage group, and 12% in the metastasectomy group. Residual tumor after stabilization only and intralesional curettage most likely explains the higher local recurrence (or progression) rate as compared to enbloc resection. This is supported by our comparison of recurrence rate based on margin status: the recurrence rate was 5% (3/64) in patients with negative margins, while it was 26% (30/115) in those with a positive margin. Four recurrences (20%, [4/20]) occurred in the enbloc metastasectomy group with positive margins. Focusing on reoperations, we found no difference with the numbers evaluated in reoperation rates between the surgical strategies. The previous studies by Evenski et al. 12 and Les et al. 8 reported on reoperation rates for local recurrence, but did not report other reasons for reoperation. Our study demonstrated that reoperations in the metastasectomy group were mainly performed for deep infection and tumor recurrence, while recurrence was the main reason for reoperation in the intralesional curettage and stabilization only groups. Deep infections seem to occur early (Table 3), while local recurrence tends to occur late. This emphasizes the importance of incorporating life expectancy in surgical decision making. Metastasectomy with negative margins in particular reduces the chance of local recurrence and subsequent reoperation, but seems to come at the cost of a slightly higher 182

185 Treatment Of Bone Metastases From Renal Cell Carcinoma early deep infection rate. In addition, the surgeon needs to consider the potential morbidity (e.g. other complications, blood loss, hospital stay, rehabilitation, disability) caused by large resections that aim to achieve negative margins and subsequently require complex reconstructions (e.g. pelvic resection followed by saddle prosthesis). Our overall one-year survival probability of 62% is comparable with previous studies. 2,4 In line with previous investigations, we found better survival in patients who underwent metastasectomy as compared to those who underwent intralesional curettage or stabilization only. 8,18,19 However, this analysis is confounded by the difference in metastatic load between groups as demonstrated by our subanalysis in patients with solitary metastasis. To avoid this confounding effect, previous studies focused on survival in patients with solitary metastases only. Ratasvuori et al. 20 found a median survival of 50 months (95% CI: 16 to 84 months) in 21 patients who underwent enbloc resection as compared to a median survival of 15 months (95% CI: 8.8 to 21 months) in 34 patients who underwent intralesional surgery for solitary renal cell metastasis; however, confidence intervals overlapped indicating no statistical significance. Lin et al. 4 found no significant difference (p = 0.52) between a median survival of 45 months in 33 patients who underwent enbloc resection and a median survival of 22 months following curettage in 15 patients for solitary renal cell metastasis. Fuchs et al. 7 also found no difference in survival when comparing 13 patients with metastasectomy and 20 patients with intralesional curettage for solitary renal cell metastases. In line with these trends, we found a median survival of 43 months in the 30 patients who underwent metastasectomy, a median survival of 20 months in 12 patients who underwent intralesional curettage, and a median survival of 27 months in 2 patients who underwent stabilization only; however, this difference was also not significant. All studies suggest a trend towards better survival in the metastasectomy group as compared to intralesional treatment, but sample sizes are relatively small compromising the power of the statistical analyses. Based on our findings, we feel that patients with a relatively good prognosis life expectancy of more than one year would benefit from enbloc resection with negative margins to improve local tumor control in the long term. In addition, metastasectomy aiming to achieve negative margins should be discussed with patients who have solitary or oligometastatic disease and have a reasonable prognosis as it might improve survival. Several models to estimate life expectancy in patients with bone metastatic disease have been developed; estimates from these models can be incorporated in surgical decision making However, a prognostication model that is specific for patients with bone metastases for renal cell carcinoma might be more accurate and useful. A future study that develops and externally validates such as model might as well incorporate a treatment algorithm based on life expectancy estimates and metastatic load. Our study has limitations. First, the retrospective methodology jeopardizes the quality of collected data. We might have missed patients because of misspellings or simply missing 9 183

186 Chapter 9 information; however, we used common synonyms and misspellings and queried both pathology and operation reports to minimize this. In addition, we used the social security death index to establish date of death and inaccuracies in this database exist. However, the database is most accurate for older and oncology patients. 15 We see these as minor limitations as most likely, these occurred irrespective of surgical strategy and outcome or margin status and therefore, did not substantially influence our results. The retrospective design caused variation in followup. We tried to account for this using time-dependent analysis; however we were not able to track if patients went to a different institution for a local recurrence or reoperation. We see this as a minor limitation as we believe that this occurred irrespective of the surgical strategy and therefore did not compromise our comparison. Furthermore, our two-year loss to clinical followup rate was relatively low at 12%. In addition, there is heterogeneity as surgeries were performed by ten different orthopaedic oncologists over a 25-year period. Indications for the surgical techniques differed, could have varied between surgeons, and preferences might have changed over time. We see this as an important limitation of our study and therefore, we explored differences in baseline characteristics between the surgical strategies and ran subanalyses on survival to understand how this selection bias might have affected our outcome measures. Nevertheless, medical treatment has changed over time and the impact of medical treatment in survival cannot be taken into consideration in this study due to its retrospective nature. Furthermore, we feel that the effect of postoperative radiation had a limited or rather minimal effect on tumor recurrence and reoperation analysis. Postoperative radiation was more commonly given after intralesional curettage (33%), and stabilization only (59%), as compared to metastasectomy (8%). Nevertheless, recurrence rates remain higher in the first two groups. Lastly, our sample size could have been insufficient. We believe that the observed better survival after metastasectomy when compared to intralesional curettage and stabilization only might be confounded by the lower metastatic load (i.e. higher proportion of patients with solitary metastases) in the metastasectomy group. We therefore deployed subanalyses in patients with solitary metastasis, in which no difference in survival was demonstrated when comparing surgical strategies. However, post-hoc power analysis demonstrated that this subanalysis is underpowered due to its relatively small sample size (n = 44); the achieved power for this analysis was only and we would have needed 45,997 patients to find a difference assuming a similar effect. A post-hoc power analysis for survival based on margin status in patients with solitary metastasis demonstrated that the achieved power was 0.21 and we would have needed 249 patients to demonstrate better survival among patients who had negative margins compared to positive margins. In the same line, the number of patients who underwent a reoperation (19%) was relatively small and could have resulted in insufficient power to detect differences among surgical strategies. 184

187 Treatment Of Bone Metastases From Renal Cell Carcinoma In conclusion, the risk of developing a local recurrence is higher among patients who undergo intralesional curettage or stabilization only as compared to patients who undergo metastasectomy. Enbloc resection with negative margins results in the lowest risk of local recurrence. The overall risk of reoperation does not differ between surgical strategies nor does it differ based on margin status; However, tumor recurrence is the main reason for reoperation after intralesional curettage and stabilization only, while early deep infections requiring reoperation occur more often after metastasectomy. Overall survival was better after metastasectomy and resection with negative margins, but these findings were confounded by the difference in metastatic load. A larger sample size, perhaps by combining the data from multiple institutions, is needed to determine the benefit of metastasectomy in patients with solitary or oligometastatic bone disease. Our findings emphasize the importance of obtaining negative margins in patients with a relatively good life expectancy (i.e. in patients with a low oncologic burden), as a lower recurrence rate can be attained at a comparable risk for reoperation, with a potential impact on survival. REFERENCES 1. Motzer RJ, Bacik J, Murphy BA, Russo P, Mazumdar M. Interferon-alfa as a comparative treatment for clinical trials of new therapies against advanced renal cell carcinoma. J Clin Oncol. Jan ; 20(1): Baloch KG, Grimer RJ, Carter SR, Tillman RM. Radical surgery for the solitary bony metastasis from renal-cell carcinoma. J Bone Joint Surg Br. Jan 2000; 82(1): Flanigan RC, Campbell SC, Clark JI, Picken MM. Metastatic renal cell carcinoma. Curr Treat Options Oncol. Oct 2003; 4(5): Lin PP, Mirza AN, Lewis VO, et al. Patient survival after surgery for osseous metastases from renal cell carcinoma. J Bone Joint Surg Am. Aug 2007; 89(8): Onufrey V, Mohiuddin M. Radiation therapy in the treatment of metastatic renal cell carcinoma. Int J Radiat Oncol Biol Phys. Nov 1985; 11(11): Kollender Y, Bickels J, Price WM, et al. Metastatic renal cell carcinoma of bone: indications and technique of surgical intervention. J Urol. Nov 2000; 164(5): Fuchs B, Trousdale RT, Rock MG. Solitary bony metastasis from renal cell carcinoma: significance of surgical treatment. Clin Orthop Relat Res. Feb 2005(431): Les KA, Nicholas RW, Rougraff B, et al. Local progression after operative treatment of metastatic kidney cancer. Clin Orthop Relat Res. Sep 2001(390): Alvi HM, Damron TA. Prophylactic stabilization for bone metastases, myeloma, or lymphoma: Do we need to protect the entire bone? Tumor. Clin Orthop Relat Res. 2013; 471: Talbot M, Turcotte RE, Isler M, Normandin D, Iannuzzi D, Downer P. Function and health status in surgically treated bone metastases. Clin Orthop Relat Res. Sep 2005; 438: Hoshi M, Takada J, Ieguchi M, Takahashi S, Nakamura H. Prognostic factors for patients with solitary bone metastasis. Int J Clin Oncol. Feb 2013; 18(1):

188 Chapter Evenski A, Ramasunder S, Fox W, Mounasamy V, Temple HT. Treatment and survival of osseous renal cell carcinoma metastases. J Surg Oncol. Dec 2012; 106(7): Bryant D, Havey TC, Roberts R, Guyatt G. How many patients? How many limbs? Analysis of patients or limbs in the orthopaedic literature: a systematic review. J Bone Joint Surg Am. Jan 2006; 88(1): Harada H, Katagiri H, Kamata M, et al. Radiological response and clinical outcome in patients with femoral bone metastases after radiotherapy. J Radiat Res. 2010; 51(2): Huntington JT, Butterfield M, Fisher J, Torrent D, Bloomston M. The Social Security Death Index (SSDI) most accurately reflects true survival for older oncology patients. Am J Cancer Res. 2013; 3(5): Janssen SJ, Braun Y, Wood KB, Cha TD, Schwab JH. Allogeneic blood transfusions and postoperative infections after lumbar spine surgery. Spine J. May ; 15(5): Allison PD. Survival Analysis Using SAS: A Practical Guide. 2 ed Fottner A, Szalantzy M, Wirthmann L, et al. Bone metastases from renal cell carcinoma: patient survival after surgical treatment. BMC Musculoskelet Disord. 2010; 11: Jung ST, Ghert MA, Harrelson JM, Scully SP. Treatment of osseous metastases in patients with renal cell carcinoma. Clin Orthop Relat Res. Apr 2003(409): Ratasvuori M, Wedin R, Hansen BH, et al. Prognostic role of en-bloc resection and late onset of bone metastasis in patients with bone-seeking carcinomas of the kidney, breast, lung, and prostate: SSG study on 672 operated skeletal metastases. J Surg Oncol. Sep 2014; 110(4): Sorensen MS, Gerds TA, Hindso K, Petersen MM. Prediction of survival after surgery due to skeletal metastases in the extremities. Bone Joint J. Feb 2016; 98-B(2): Janssen SJ, van der Heijden AS, van Dijke M, et al Marshall Urist Young Investigator Award: Prognostication in Patients With Long Bone Metastases: Does a Boosting Algorithm Improve Survival Estimates? Clin Orthop Relat Res. Oct 2015; 473(10): Nathan SS, Healey JH, Mellano D, et al. Survival in patients operated on for pathologic fracture: implications for end-of-life orthopedic care. J Clin Oncol. Sep ; 23(25):

189 Treatment Of Bone Metastases From Renal Cell Carcinoma Appendix 1: Implant type per anatomic area n = 183 Anatomic location Operation type Implant type n (%) Femur (n = 75) IMN (n = 32) Intramedullary nailing 32 (43) EPR (n = 31) MTP Bipolar hemiarthroplasty 18 (24) MTP - Knee (Hinge) 6 (8) MTP With Cup 4 (5) Intercalary prosthesis 1 (1) Link Saddle Prosthesis 1 (1) Total Hip Arthroplasty 1 (1) ORIF (n = 10) Plate screw fixation 10 (13) Resection (n = 2) Girdlestone 1 (1) Curettage only 1 (1) Humerus (n = 52) IMN (n = 16) Intramedullary nailing 16 (31) EPR (n = 19) MTP - Shoulder hemiarthroplasty 12 (23) Hemiarthroplasty long-stem (head) 5 (10) Hemiarthroplasty (head) 1 (2) Total elbow 1 (2) ORIF (n = 16) Plate screw fixation 16 (31) Resection (n = 1) Curettage only 1 (2) Acetabulum (n = 11) EPR (n = 9) Total hip arthroplasty 6 (55) MTP With Cup 2 (18) Link Saddle Prosthesis 1 (9) ORIF (n = 2) Plate screw fixation 2 (18) Pelvis (n = 11) EPR (n = 1) Link Saddle Prosthesis 1 (9) ORIF (n = 3) Plate screw fixation 3 (27) Resection (n = 7) Resection only 7 (64) Tibia (n = 8) IMN (n = 3) Intramedullary nailing 3 (38) EPR (n = 2) MTP - Knee (Hinge) 2 (25) ORIF (n = 2) Plate screw fixation 2 (25) Resection (n = 1) Curettage only 1 (13) Femur and Acetabulum (n = 8) EPR (n =8) MTP - Hemiarthroplasty 5 (63) Total Hip Arthroplasty 2 (25) MTP With Cup 1 (13) Scapula (n = 5) Resection (n = 5) Resection only 5 (100) Ulna (n = 3) IMN (n =1) Intramedullary nailing 1 (33) ORIF (n = 2) Plate screw fixation 2 (67) Clavicle (n = 3) Resection (n = 3) Resection only 3 (100) Calcaneus (n = 3) ORIF (n = 1) Plate screw fixation 1 (33) Resection (n = 2) Curettage only 2 (67) 9 187

190 Chapter 9 Appendix 1: Implant type per anatomic area (continued) Anatomic location Operation type Implant type n (%) Radius (n = 1) ORIF (n = 1) Plate screw fixation 1 (100) Metatarsal (n = 1) ORIF (n = 1) Plate screw fixation 1 (100) Metacarpal (n = 1) Resection (n = 1) Curettage only 1 (100) Ulna and Metacarpal (n = 1) ORIF (n = 1) Plate screw fixation, both 1 (100) n = 183 IMN = intramedullary nailing, EPR = endoprosthetic reconstruction, ORIF = open reduction and internal fixation, MTP = modular tumor prosthesis Appendix 2: Modified Charlson Comorbidity Index Algorithm Based on International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) Codes Comorbidity Weight* Codes AIDS/HIV Any malignancy, including leukemia and lymphoma* , , , , , , , 189.9, , , , , , , , , , , 230.8, 231.2, 231.9, , 233.0, 233.1, , , 233.4, 233.7, , 235.7, 235.8, 236.2, 236.4, 236.5, , , 237.6, , , , 239.6, 239.7, , Chronic pulmonary disease , 416.9, , , , 506.4, 508.1, Congestive heart failure , , , , , , , , , , , Dementia 2 290, 290.0, 290.3, , 294.1, , Diabetes with chronic complications , Hemiplegia or paraplegia , , Metastatic solid tumor* , , , Mild liver disease* , , , , , , 070.6, 070.9, 570, 570.1, 573.3, 573.4, 573.8, 753.9, V42.7 Moderate or severe liver disease* , Renal disease , , , , , , , , , , , 588.0, V42.0, V45.1, V56-V56.8 Rheumatologic disease , , , 714.8, 725 *The following comorbidities were mutually exclusive: mild liver disease and moderate or severe liver disease, and any malignancy and metastatic solid tumor. For example, a patient with a metastatic solid tumor received 6 points total (not 6 points for metastatic solid tumor and 2 points for any malignancy). 188

191 CHAPTER 10 Prognostication In Patients With Long Bone Metastases: Does A Boosting Algorithm Improve Survival Estimates? S.J. Janssen A.S. van der Heijden M. van Dijke J.E. Ready K.A. Raskin M.L. Ferrone F.J. Hornicek J.H. Schwab 10 Clinical Orthopaedics and Related Research 2015 Oct;473(10): Award: Marshall Urist Young Investigator Award 2015 Presented at: Association of Bone and Joint Surgeons Annual Meeting 2015, Eugene, Oregon, USA. Musculoskeletal Tumor Society Annual Meeting 2015, Orlando, Florida, USA. Poster at: Massachusetts General Hospital Clinical Research Day 2014, Boston, Massachusetts, USA.

192 Chapter 10 ABSTRACT Objectives To assess factors independently associated with decreased survival in patients with a fracture through a long bone metastasis and (2) compare the accuracy of a classic scoring system, nomogram, and boosting algorithms in predicting 30-, 90-, and 365-day survival. Design Retrospective cohort study. Setting Two tertiary care referral centers for orthopaedic oncology. Participants 927 consecutive patients who underwent surgery for a pathological or impending fracture through a long bone metastasis between 1999 and Outcome Measures Overall survival. Results The following factors were associated with a decreased likelihood of survival after surgical treatment of a long bone metastasis, after controlling for relevant confounding variables: older age (hazard ratio [HR], 1.0; 95% CI, ; p < 0.001), additional comorbidity (HR, 1.2; 95% CI, ; p = 0.034), BMI less than 18.5 kg/m² (HR, 2.0; 95% CI, ; p = 0.011), tumor type with poor prognosis (HR, 1.8; 95% CI, ; p < 0.001), multiple bone metastases (HR, 1.3; 95% CI, ; p = 0.008), visceral metastases (HR, 1.6; 95% CI, ; p < 0.001), and lower hemoglobin level (HR, 0.91; 95% CI, ; p < 0.001). The survival estimates by the nomogram were moderately accurate for predicting 30-day (area under the curve [AUC], 0.72), 90-day (AUC, 0.75), and 365-day (AUC, 0.73) survival and remained stable after correcting for optimism through fivefold cross validation. Boosting algorithms were better predictors of survival on the training datasets, but decreased to a performance level comparable to the nomogram when applied on testing datasets for 30-day (AUC, 0.69), 90-day (AUC, 0.75), and 365-day (AUC, 0.72) survival prediction. Performance of the classic scoring system was lowest for all prediction periods. Conclusions Comorbidity status and BMI are newly identified factors associated with decreased survival and should be taken into account when estimating survival. Performance of the 190

193 Prognostication In Patients With Long Bone Metastases boosting algorithms and nomogram were comparable on the testing datasets. However, the nomogram is easier to apply and therefore more useful to aid surgical decision making in clinical practice. INTRODUCTION Estimated survival is an important factor in the decision to operate and surgical strategy in patients with bone metastasis. 1,2 Physicians often estimate survival based on their clinical assessments and previously described risk factors. Several tools in the form of scoring systems have been developed to assist clinicians with their estimation. 1-5 However, these tools lack accuracy and identification of additional and more-specific risk factors might improve survival estimation. 2,6 Historically, tools like scoring systems that are used to provide survival probability are based on a summary score of weighted clinical or laboratory factors. 1,2,5 The Bauer score is such a classic scoring system commonly used for estimation of survival in patients with bone metastases. 1 It is a summary score of five prognostic factors: (1) no pathological fracture, (2) no visceral or brain metastases, (3) a solitary bone metastasis, (4) no lung cancer, and (5) multiple myeloma, lymphoma, breast or kidney carcinoma. Fulfilling four to five criteria corresponded to a 1-year survival probability of 0.5, two to three criteria to a 1-year survival probability of 0.25, and all patients who fulfilled none or only one criterion were deceased within 6 months after surgery. 1 Another frequently used tool to estimate survival in patients with cancer is the nomogram, which is a simple figure that generates an individualized numerical probability of survival based on a patient s unique set of characteristics; a number of points is assigned to each prognostic factor, which can be read from the nomogram and the sum of these points corresponds to a survival probability. 7-9 The nomogram can be seen as an extension of the classic scoring system. Advances in computer science has led to the development of more sophisticated boosting (machine learning) algorithms. 10,11 Machine learning is a method of automatically developing and constantly adjusting computer algorithms to recognize patterns in data and improve predictions (e.g. filtering spam ). 11 Boosted regression emerged from this field and is a method that iteratively applies classifiers (variables) in a sequential way each step building on the previous step aiming to fit the residuals and subsequently combines them to obtain predictions. 10,11 This can improve the accuracy of predicting an outcome based on weak learners (i.e. classifiers that are only slightly better than random guessing). A boosting algorithm provides outcome probabilities based on every possible combination of variables. 10 We aimed to assess factors associated with survival in patients with long bone fracture through a bone metastasis. Based on those factors, we created a classic scoring algorithm,

194 Chapter 10 a nomogram, and boosting algorithms to estimate survival. Specifically, we sought to (1) assess factors independently associated with decreased survival in patients with long bone metastasis, and (2) compare the accuracy of a classic scoring system, nomogram, and boosting algorithms in predicting 30-, 90-, and 365-day survival. METHODS Study Design This retrospective study was approved by our institutional review board and a waiver of informed consent was granted. To identify patients with a long bone fracture through a bone metastasis, we retrieved all medical record data of patients who had an International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code for a long bone fracture through a bone metastasis or a Current Procedural Terminology (CPT) code for prophylactic fixation of a long bone fracture at two tertiary care referral centers for orthopaedic oncology (Appendix 1). After operative report and medical record screening, we included all 927 patients older than 18 years who had surgery for a pathological or impending long bone fracture between January 1999 and December We included only the first surgery per patient if patients underwent multiple operations on different occasions so as to not violate the statistical assumption of independence. 12 We defined long bones as the femur, humerus, tibia, fibula, radius, and ulna; multiple long bones operated on during the same procedure were categorized separately. Metastatic disease included, in addition to metastases from solid organs, multiple myeloma and lymphoma. We included patients regardless of followup duration. Exclusion criteria were (1) revision procedures, (2) metastatic involvement of the acetabulum or pelvis requiring reconstruction, and (3) surgical treatments other than endoprosthetic reconstruction, intramedullary nailing, open reduction and internal fixation with plate and screws or a dynamic hip screw. The decision to operate and the selection of surgical strategy were made by the surgeon together with the patient. These decisions were based on factors including type of fracture, primary tumor type, extent of the metastatic lesion, level of disability and pain, and the estimated survival. Outcome Measures Our primary outcome was survival, defined as the time from surgical treatment until death resulting from any cause. Date of death was extracted from the medical records and Social Security Death Index database (database of death records created from the US Social Security Administration), last updated February 24, This date also provided the moment of final followup for all patients alive in our study. Median followup was 8 192

195 Prognostication In Patients With Long Bone Metastases 1.00 Probability of Survival Survival in Months Number of Patients at Risk Figure 1: The Kaplan-Meier plot shows the probability of survival (solid line) with 95% CI (dashed line). The median survival is 268 days (95% CI, ), with an interquartile range from 84 (95% CI, 72-97) to 1089 days (95% CI, ). months (interquartile range, 3 to 25 months). All patients who were alive at 30 days were still in followup (n = 853), 673 of 678 patients (99%) who were alive at 3 months were still in followup, and 369 of 412 (90%) patients who were alive at 1 year were still in followup (Figure 1). We selected the following explanatory variables based on the existing studies or theoretical association with survival: age, sex, BMI, comorbidity status, primary tumor type, type of fracture, anatomic location of fracture, time from diagnosis of the primary tumor until surgical treatment, other bone metastases, other previous pathological fractures, visceral metastases, previous systemic therapy, previous local radiation therapy of the affected long bone, and preoperative hemoglobin level, platelet level, white blood cell count, creatinine, and serum calcium level. 2,5,14 We categorized BMI into: less than 18.5 kg/m² (underweight), between 18.5 and 30 kg/ m² (normal weight), and 30 kg/m² or greater (obese) based on previously published cutoff points as we expect a nonlinear association of BMI with survival. 15,16 We used the modified Charlson Comorbidity Index to indicate comorbidity status. 17,18 This index provides a score ranging from 0 to 24 with a higher score representing more severe comorbidity status based on 12 weighted comorbidities (congestive heart failure, dementia, chronic pulmonary disease, rheumatologic disease, mild liver disease, diabetes

196 Chapter 10 with chronic complications, hemiplegia or paraplegia, renal disease, any malignancy, moderate or severe liver disease, metastatic solid tumor, and HIV/AIDS). We determined the modified Charlson Comorbidity Index through a previously described algorithm based on ICD-9-CM codes given before the day of surgery (Appendix 2) The modified Charlson Comorbidity Index was dichotomized into any additional comorbidity (additional to the malignancy and metastatic disease) or none. Based on a study by Katagiri et al. 5, we dichotomized primary tumor types into tumors with a relatively good prognosis (breast, kidney, prostate, thyroid, myeloma, and lymphoma) and tumors with a poor prognosis (lung and all other tumor types). Fracture type was classified as pathological or impending. The latter was defined as bone with no visible fracture line, loss of height, rotation, or angulation, but the degree of destruction did mandate, in the surgeon s opinion, surgical treatment. Previous pathological fractures or prophylactically treated impending fractures were categorized into none, previous long bone fracture, and previous spine fracture (with or without previous long bone fracture). We extracted the presence of bone metastases from bone scan, CT, and other imaging reports. Bone metastases were categorized into single bone metastasis, multiple bone metastases without spinal involvement, and multiple bone metastases with spinal involvement. The presence of visceral metastases was derived from CT and positron emission tomography scan reports. We regarded lung, liver, and brain metastases as visceral metastases and grouped lung and/or liver metastases together; brain metastases (with or without lung/liver metastases) were categorized separately. We used laboratory values measured within 7 days before surgical treatment. Statistical Analysis Variables are presented with frequencies and percentages for categorical variables and as mean with SD for continuous variables. In bivariate analyses, the association between the response variable survival and the explanatory variables was assessed using Cox regression analysis (Appendix 3). The proportional hazards assumption was tested using Schoenfeld residuals and verified by assessing if Kaplan-Meier survival curves crossed. Our exploratory analysis identified the following variables: age (p < 0.001), BMI less than 18.5 kg/m² (p < 0.001), additional comorbidity (p < 0.001), multiple long bones surgically treated during same procedure (p = 0.077), poor prognosis tumor type (lung and all other tumor types) (p < 0.001), multiple bone metastases without spinal involvement (p = 0.096) and with spinal involvement (p = 0.014), lung and/or liver metastasis (p < 0.001) and brain metastasis (p < 0.001), previous systemic therapy (p = 0.057), hemoglobin level (p = 0.001), and platelet level (p = 0.004), which then were incorporated in our multivariable model (Appendix 3). We recategorized bone metastases as single and multiple metastases as exploratory analysis showed no difference in hazard ratios (HR) between patients with 194

197 Prognostication In Patients With Long Bone Metastases multiple bone metastases without spinal involvement (HR, 1.2; 95% CI, ) and those with spinal involvement (HR, 1.3; 95% CI, ; p = 0.67). Visceral metastases also was recategorized as none and any visceral metastases as there was no difference in hazard ratios between lung/liver metastases (HR, 1.9; 95% CI, ) and brain metastases (HR, 1.8; 95% CI, ; p = 0.82) (Appendix 3). We entered these explanatory variables with a p value less than 0.10 on bivariate analysis in a backward stepwise multivariable Cox regression analysis to assess the independent association with survival. 22,23 We did not test interactions of variables. Hazard ratios and beta regression coefficients are presented to quantify the association of explanatory variables with survival. The HR indicates the relative likelihood of death in one group compared with another group. Hazard ratios are adjusted for all explanatory variables included in the multivariable Cox regression analysis. We assume missing values BMI (21%, 197 of 927 patients) and hemoglobin (6%, 59 of 927 patients) to be at random and used multiple imputation to replace missing values 40 times based on the remaining explanatory variables. 24 All statistical analyses were performed using Stata 13.0 (StataCorp LP, College Station, TX, USA). A two-tailed p value less than 0.05 was considered statistically significant. Development Of A Classic Scoring System, Nomogram, And Boosting Algorithms We developed a classic scoring system to estimate survival probability by assigning a weighted score to every factor independently associated with survival by rounding its HR to the nearest integer. 1,5,14 To allow for scoring of continuous variables, we dichotomized age (65 years and older) and hemoglobin level (10 g/dl and less) and rounded the HRs of the mean difference between the dichotomized groups to the nearest integer (mean difference in age, 20 years [HR, 1.34] and mean difference in hemoglobin level 2.6 g/ dl [HR, 1.27]). 2 The total score of the classic scoring system ranges from 0 to 10 (Table 1). We categorized scores as: good prognosis, (0 2 points), intermediate prognosis (3 5 points), and poor prognosis (6 10 points) based on the survival probability curves (Figure 2). Survival probability was demonstrated per prognostic group for each prediction period (30, 90, and 365 days; Table 2). 1,5 We developed a nomogram by ranking the effect estimates (β regression coefficients) of all factors independently associated with survival to a scale ranging from 0 to 100 points. 8,9 The predicted probability of 30-, 90-, and, 365-day survival were calculated for each patient using the multivariable Cox regression model underlying the nomogram. 8,25 Boosting algorithms to predict 30-, 90-, and 365-day survival were developed using the multiple additive regression trees gradient boosting technique implemented in Stata ,11 Boosting is a machine learning technique that produces a prediction algorithm based on additive decision trees to classify outcome (30-, 90-, and 365-day survival) in a

198 Chapter 10 Table 1: Classic scoring system Variable Points Age 65 years or older 1 Additional comorbidity 1 Body mass index below 18.5 kg/m² 2 Tumor type other than: breast, kidney, prostate, thyroid, myeloma, and lymphoma 2 Multiple Bone metastases 1 Visceral metastases * 2 Hemoglobin level 10 or below (g/dl) 1 * Visceral metastases include lung, liver, and/or brain metastases. Table 2: Classic scoring system: survival probability per prognostic group and prediction period (n = 927) * Prognostic groups Good prognosis (score 0-2) Probability of survival (95% confidence interval) 30 days 0.98 ( ) 90 days 0.93 ( ) 365 days 0.66 ( ) Intermediate prognosis (score 3-5) 30 days 0.92 ( ) 90 days 0.73 ( ) 365 days 0.39 ( ) Poor prognosis (score 6-10) 30 days 0.84 ( ) 90 days 0.49 ( ) 365 days 0.17 ( ) * Missing values for body mass index and hemoglobin level were imputed using multiple imputation. stepwise fashion. 10,11 We included the same set of factors independently associated with survival for development of the boosting algorithms. Each boosting algorithm allowed for two-way interactions. The algorithm provides an estimated survival probability for every possible combination (permutation) of the included variables for each prediction period, thereby taking into account the interaction of the included variables. 10 All three prediction models were compared using fivefold cross validation on the 40 multiple imputed datasets, meaning that the models were created five times on randomly selected training subsets (80%) of the data and tested on the remaining 20%. 26,27 The average performance (the ability of a model to separate patients with different outcomes) was calculated over the five training and testing repetitions per multiple imputed dataset for all three models and prediction periods and subsequently pooled. Performance was 196

199 Prognostication In Patients With Long Bone Metastases Figure 2: The Kaplan-Meier plot shows the probability of survival per prognostic group of the classic scoring algorithm: good prognosis (0 2; dashed line), intermediate prognosis (3 5; dotted line), and poor prognosis (6 10; solid line). assessed using receiver operating characteristic (ROC) curves ROC curves are made by plotting the rate of false positives (1 specificity) on the x-axis and the rate of true positives (sensitivity) on the y-axis for all threshold values. The area under the ROC curve (AUC) represents its discriminatory power; an AUC of 1.0 indicates perfect discrimination (100% sensitivity and 100% specificity), whereas an AUC of 0.50 represents no discriminatory power (a coin toss). The final classic scoring system, nomogram, and boosting algorithms were developed on the 40 multiple imputed datasets and results were pooled. 26,27 10 RESULTS Baseline Characteristics Among the 927 patients, 401 (43%) were men, and the mean age of the patients was 62 years (Table 3). There were 515 (56%) pathological fractures and 412 (44%) impending fractures. The femur (70%; 646 of 927 patients) and humerus (23%; 210 of 927 patients) were most commonly affected. Most tumors originated from the breast (23%; 216 of 927 patients), lung (23%; 215 of 927 patients), myeloma (16%; 148 of 927 patients), kidney (9%; 87 of 927 patients), and prostate (5%; 48 of 927 patients) (Table 4). Median overall survival was 9 months (Figure 1). Ninety-two percent of the patients survived 30 days 197

200 Chapter 10 Table 3: Baseline characteristics Demographics Mean (± SD) Age (years) 62 (± 13) Body mass index (kg/m²) * 27 (± 5.8) Modified Charlson Comorbidity Index 6.7 (± 2.0) n (%) Men 401 (43) Pathological fracture 515 (56) Anatomical location Femur 646 (70) Humerus 210 (23) Tibia 31 (3.3) Radius 3 (0.32) Ulna 2 (0.22) Multiple locations 35 (3.8) Oncological status n (%) Bone metastases Single bone metastasis 202 (22) Multiple bone metastases without spinal involvement 185 (20) Multiple bone metastases with spinal involvement 540 (58) Previous pathological fractures None 662 (71) Previous long bone (impending) pathological fracture 79 (8.5) Previous spine (impending) pathological fracture 186 (20) Visceral metastases None 504 (54) Lung and/or liver metastases 273 (29) Brain metastases (with or without lung/liver metastases) 150 (16) Previous systemic therapy 577 (62) Previous local radiotherapy of the affected long -bone 170 (18) Laboratory values Mean (± SD) Hemoglobin (g/dl) 11 (± 1.6) Platelets (1000/mm³) 254 (± 116) White blood cell count (1000/mm³) 9.7 (± 4.9) Creatinine (mg/dl) 0.94 (± 0.81) Calcium (mg/dl) 8.8 (± 0.93) n = 927 * Body mass index was available in 730 cases; Multiple metastatic fracture locations undergoing fixation during the same procedure were: bilateral femur (12 patients), femur and humerus (16 patients), bilateral humerus (2), tibia and femur (1), tibia and humerus (1), fibula and tibia (1), radius and humerus (1), ulna and radius (1); with or without previous long bone pathological fracture; hemoglobin level was available in 868 cases, platelet level in 866 cases, white blood cell count in 867 cases, creatinine in 812 cases, and calcium in 654 cases. 198

201 Prognostication In Patients With Long Bone Metastases Table 4: Origin of primary tumors Tumor distribution n (%) Breast 216 (23) Lung 215 (23) Myeloma 148 (16) Kidney 87 (9) Prostate 48 (5) Lymphoma 43 (5) Melanoma 25 (3) Esophagus 18 (2) Colorectal 16 (2) Thyroid 15 (2) Hepatocellular 12 (1) Bladder 10 (1) Other * 35 (4) Unknown 39 (4) * Neuroendocrine (n = 6), salivary gland (n = 5), nasopharyngeal squamous cell carcinoma (n = 5), pancreas (n = 4), ovaries (n = 4), endometrium (n = 3), skin squamous cell carcinoma (n = 3), stomach (n = 2), vulva (n = 2), and mesothelioma (n = 1). (853 of 927 patients), 73% (676 of 922 patients) survived 90 days, and 42% (368 of 884 patients) survived 365 days. The median time from diagnosis of the primary tumor until surgical treatment of the fracture was 21 months. Five-hundred sixty (60%) patients underwent intramedullary nailing, 209 (23%) had endoprosthetic reconstruction, 140 (15%) had open reduction and internal fixation with plate and screws or a dynamic hip screw in 18 (2%). Explanatory Variables Associated With Survival The following factors were associated with a decreased likelihood of survival after surgical treatment of a long bone metastasis, after controlling for relevant confounding variables: older age (HR, 1.0; 95% CI, ; p < 0.001), additional comorbidity (HR, 1.2; 95% CI, ; p = 0.034), BMI less than 18.5 kg/m² (HR, 2.0; 95% CI, ; p = 0.011), tumor type with poor prognosis (HR, 1.8; 95% CI, ; p < 0.001), multiple bone metastases (HR, 1.3; 95% CI, ; p = 0.008), visceral metastases (HR, 1.6; 95% CI, ; p < 0.001), and lower hemoglobin level (HR, 0.91; 95% CI, ; p < 0.001) (Table 5)

202 Chapter 10 Table 5: Hazard ratios for survival from stepwise backward multivariable Cox regression analysis Explanatory variables β regression coefficient Standard error Age (in years) Additional comorbidity Body mass index Below 18.5 kg/m² * Hazard ratio (95% confidence interval) (1.008, 1.021) (1.013, 1.372) (1.175, 3.480) p value < Between 18.5 and 30 kg/m² * Reference Reference Reference Reference Above 30 kg/m² * Tumor type other than: breast, kidney, prostate, thyroid, myeloma, and lymphoma Multiple Bone metastases Visceral metastases Hemoglobin level (g/dl) * (0.549, 1.288) (1.574, 2.150) (1.069, 1.555) (1.412, 1.926) (0.867, 0.956) < < < bold font indicates a significant difference (two-tailed p value below 0.05); * body mass index was available in 730 cases; and hemoglobin level in 868 cases; missing values were imputed using multiple imputation. Variables were selected using stepwise backward multivariable Cox regression analysis retaining variables with a p value below Table 6: Area under the curve (AUC) for the classic scoring algorithm and boosting algorithm from receiver operating characteristic analysis Prediction period Training subsets Classic Scoring System, AUC (95% confidence interval) Nomogram, AUC (95% confidence interval) Boosting Algorithm, AUC (95% confidence interval) p value 30 days 0.66 ( ) 0.72 ( ) 0.83 ( ) < days 0.70 ( ) 0.76 ( ) 0.81 ( ) < days 0.68 ( ) 0.73 ( ) 0.78 ( ) < Testing subsets 30 days 0.67 ( ) 0.72 ( ) 0.69 ( ) days 0.70 ( ) 0.75 ( ) 0.75 ( ) days 0.68 ( ) 0.73 ( ) 0.72 ( ) bold font indicates a significant difference (two-tailed p value below 0.05) 200

203 Prognostication In Patients With Long Bone Metastases Comparing Performance Of The Classic Scoring, Nomogram, And Boosting Algorithms The survival estimates by the nomogram were moderately accurate for predicting 30-day (AUC, 0.72), 90-day (AUC, 0.75), and 365-day (AUC, 0.73) survival and remained stable after correcting for optimism through fivefold cross validation (Table 6). Boosting algorithms were better predictors of survival at all prediction periods on the training datasets; however, after applying these to the testing datasets we found that accuracy of the boosting algorithms decreased substantially for the 30-day (AUC, 0.83 to 0.69), 90-day (AUC, 0.81 to 0.75), and 365-day (AUC, 0.78 to 0.72) prediction periods resulting in a performance comparable to the that of the nomogram (Table 6). Performance of the classic scoring system was lowest for all prediction periods. DISCUSSION Expected survival is an important factor in the decision to operate and in the selection of a surgical strategy for patients with long bone metastasis and a fracture. 1,2 Previous studies developed scoring algorithms to estimate survival, 1-5,14 but survival estimates remain imprecise. 2 In an attempt to improve survival estimation, we assessed which clinical factors and laboratory values were independent predictors of survival. Additionally, we developed a classic scoring system, nomogram, and boosting algorithms to estimate 30-, 90-, and 365-day survival and compared the accuracy of these methods. We found that older age, additional comorbidity, BMI less than 18.5 kg/m², primary tumor type with poor prognosis, multiple bone metastases, visceral metastases, and lower hemoglobin level, were independently associated with decreased likelihood of survival. The survival estimates by the boosting algorithm were most accurate on the training datasets, but comparable to those derived from the nomogram when applied to the testing datasets for 30-, 90-, and 365-day survival. We emphasize the use of the nomogram (Figure 3) for estimating survival as it is simpler to use in clinical practice. This study has some limitations. First, there were no uniform criteria for surgical treatment because the study was retrospective. This might have resulted in selection bias and potentially influenced accuracy of the prediction models; for example, patients with a very poor overall health status might not have been considered for surgical treatment. Although this might limit the usefulness of the algorithms in these patients, we believe that this did not compromise the comparison of performance of the algorithms in our study. Second, we used diagnostic and billing codes to identify potentially eligible patients. We might have missed patients using this methodology; however, we expect this number to be low and therefore not influence our conclusions. Third, although we internally validated the algorithms through fivefold cross validation, predictive performance can worsen substantially

204 Chapter 10 Age (in years) Hemoglobin (in g/dl) Other Bone Metastases Visceral/Brain Metastases Primary Tumor Type Other Comorbidity Body Mass Index Category 18 No No 1 No Yes Yes 14 Yes Points Total Points day Survival Probability day Survival Probability day Survival Probability Figure 3: The nomogram for prediction of 30-, 90-, and 365-day survival is shown. Locate the patients age on the age axis and draw a straight line to the points axis. Repeat this process for all variables and sum the points obtained for each predictor. Locate the total points on the total points axis and draw a line straight down to find the 30-, 90-, and 365-day survival probabilities. BMI categories are: (1) less than 18.5 kg/m², (2) between 18.5 and 30 kg/m², and (3) 30 kg/m² or greater (obese). The primary tumor Group 1 includes breast, kidney, prostate, thyroid, myeloma, and lymphoma; Group 2 includes lung and all other primary tumor types. This nomogram is not applicable to a patient who otherwise is not a candidate for surgical treatment of a long bone metastasis. The outcome is a point estimate and the nomogram does not include the uncertainty of the estimate. on external validation. External validation should be performed before widespread use of a prediction algorithm. 31,32 Fourth, we did not assess how discriminant the predicted probabilities by the different models were. We see this as a minor limitation and emphasized performance of the models as better performance (higher AUC) implies less uncertainty more precision of the points estimate of the predicted probability. Fifth, because the study was retrospective, we could not include performance status of the patient. Including this might have improved the predictive accuracy of our algorithms as previous studies showed a strong association with survival. 2,5 Sixth, we decided to select variables for inclusion in the boosting algorithms based on theory. Including all explanatory variables using a kitchen-sink approach (having the algorithm select variables) could improve its accuracy. However, this might have resulted in a large number of factors to consider when estimating survival, making it less useful in clinical practice, and potentially worsening its external validity. We aimed to compare the performance of prediction models based on the 202

205 Prognostication In Patients With Long Bone Metastases same set of variables supported by theory. Seventh, we did not define minimum followup. We see this as a minor limitation as we used Cox regression analysis to account for right censoring (loss to followup) and followup was 90% for patients alive at 1 year. Previous studies in patients with fractures through bone metastasis found that visceral metastases, primary tumor type, number of bone metastases, time from diagnosis of primary tumor to surgery, fracture type, performance status, previous chemotherapy, and hemoglobin level were independently associated with survival. 1,2,5,14,33 We identified additional factors associated with survival in patients with long bone metastasis, namely, comorbidity status and BMI. The association of comorbidity status with overall and cancer-specific survival has been shown in patients with primary malignancies but not in patients with fracture through a long bone metastasis. BMI can be considered a surrogate marker of cancer severity because low body weight often is associated with more advanced cancer. Previous studies showed an association of BMI with survival in patients with cancer. 37,38 Furthermore, we divided multiple bone metastases into those with and those without spinal involvement and visceral metastases into lung and/or liver metastases and brain metastases to explore differences in their association with survival. However, we found no differences between these groups in terms of survival nor did we find an association of previous pathological fractures with survival. The difference in survival between impending and pathological fractures, found by Bauer and Wedin, 1 was not found in our study. Based on our findings, future studies should explore how specific comorbidities influence survival in patients with fractures and if optimizing perioperative nutritional status improves survival in these patients. Although the boosting algorithm was most accurate in estimating survival on the training samples, its performance decreased when applied to the testing subsets of the data. This might be a result of overfitting of the boosting algorithm on the training data. Performance of the boosting algorithm was comparable to that of the nomogram when applied to the testing subsets. We therefore prefer using the nomogram in estimating survival as it is simpler to use in daily practice. However, the nomogram does not make treatment recommendations, it simply provides estimated survival probabilities and can enable a more informed decision-making process. Thirty-, 90-, and 365-day survival probabilities are based on the sum score of the points assigned to the prognostic factors of an individual patient (Figure 3). For example, a 77-year old patient with breast carcinoma, normal BMI, multiple bone metastases, but no visceral metastases, diabetes with chronic complications (additional comorbidity), and a preoperative hemoglobin of 9.4 g/dl gets assigned 157 points which corresponds to a 30-day survival probability of 0.93, 90-day survival probability of 0.74, and 365-day survival probability of The mean total points in our cohort was 156 points (SD, 39). Forsberg et al. 3,4 described the development and testing of machine learning algorithms in determining survival of patients with long bone metastasis. The algorithms they developed, based on a prospective cohort of 189 patients, had an

206 Chapter 10 AUC of 0.85 for 3-month survival and 0.83 for 1-year survival, indicating a higher accuracy compared with our algorithms. 3 The prospective collection of their data might have been more accurate and less prone to bias, resulting in higher accuracy. Future studies should externally validate survival prognostication models and assess which algorithm is most accurate in predicting survival for patients with extremity metastases. Comorbidity status and BMI are two additional factors associated with survival and should be taken into account when estimating survival. These factors should be incorporated in survival prediction models. The nomogram remained most accurate in predicting survival after correcting for optimism and could be used on paper; however, its estimates will be more precise when implemented in an application. The nomogram could be made available on interfaces convenient in clinical practice (such as smartphone applications) to aid surgical decision making. We are working on external validation of the developed algorithms and aim to develop an online tool to estimate survival for use in clinical practice. REFERENCES 1. Bauer HC, Wedin R. Survival after surgery for spinal and extremity metastases. Prognostication in 241 patients. Acta Orthop Scand. Apr 1995; 66(2): Nathan SS, Healey JH, Mellano D, et al. Survival in patients operated on for pathologic fracture: implications for end-of-life orthopedic care. J Clin Oncol. Sep ; 23(25): Forsberg JA, Eberhardt J, Boland PJ, Wedin R, Healey JH. Estimating survival in patients with operable skeletal metastases: an application of a bayesian belief network. PLoS One. 2011; 6(5): e Forsberg JA, Wedin R, Bauer HC, et al. External validation of the Bayesian Estimated Tools for Survival (BETS) models in patients with surgically treated skeletal metastases. BMC Cancer. 2012; 12: Katagiri H, Takahashi M, Wakai K, Sugiura H, Kataoka T, Nakanishi K. Prognostic factors and a scoring system for patients with skeletal metastasis. J Bone Joint Surg Br. May 2005; 87(5): Henderson R, Keiding N. Individual survival time prediction using statistical models. J Med Ethics. Dec 2005; 31(12): Oberije C, De Ruysscher D, Houben R, et al. A Validated Prediction Model for Overall Survival From Stage III Non-Small Cell Lung Cancer: Toward Survival Prediction for Individual Patients. Int J Radiat Oncol Biol Phys. Apr Iasonos A, Schrag D, Raj GV, Panageas KS. How to build and interpret a nomogram for cancer prognosis. J Clin Oncol. Mar ; 26(8): Kattan MW, Eastham JA, Stapleton AM, Wheeler TM, Scardino PT. A preoperative nomogram for disease recurrence following radical prostatectomy for prostate cancer. J Natl Cancer Inst. May ; 90(10): Schonlau M. Boosted regression (boosting): An introductory tutorial and a Stata plugin. Stata Journal. 2005; 5(3):

207 Prognostication In Patients With Long Bone Metastases 11. Hastie T, Tibshirani R, Friedman J. The Elements of Statistical Learning. 2 ed. New York: Springer-Verlag New York; Bryant D, Havey TC, Roberts R, Guyatt G. How many patients? How many limbs? Analysis of patients or limbs in the orthopaedic literature: a systematic review. J Bone Joint Surg Am. Jan 2006; 88(1): Huntington JT, Butterfield M, Fisher J, Torrent D, Bloomston M. The Social Security Death Index (SSDI) most accurately reflects true survival for older oncology patients. Am J Cancer Res. 2013; 3(5): Ratasvuori M, Wedin R, Keller J, et al. Insight opinion to surgically treated metastatic bone disease: Scandinavian Sarcoma Group Skeletal Metastasis Registry report of 1195 operated skeletal metastasis. Surg Oncol. Jun 2013; 22(2): US National Heart Lung and Blood Institute. The Practical Guide: Identification, Evaluation, and Treatment of Overweight and Obesity in Adults. Vol No : National Institutes of Health; Goodwin PJ, Ennis M, Pritchard KI, Koo J, Trudeau ME, Hood N. Diet and breast cancer: evidence that extremes in diet are associated with poor survival. J Clin Oncol. Jul ; 21(13): Quan H, Li B, Couris CM, et al. Updating and validating the Charlson comorbidity index and score for risk adjustment in hospital discharge abstracts using data from 6 countries. Am J Epidemiol. Mar ; 173(6): Poeran J, Rasul R, Suzuki S, et al. Tranexamic acid use and postoperative outcomes in patients undergoing total hip or knee arthroplasty in the United States: retrospective analysis of effectiveness and safety. BMJ. 2014; 349: g Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol. Jun 1992; 45(6): Quan H, Parsons GA, Ghali WA. Validity of information on comorbidity derived rom ICD-9- CCM administrative data. Med Care. Aug 2002; 40(8): Voskuijl T, Hageman M, Ring D. Higher Charlson Comorbidity Index Scores are associated with readmission after orthopaedic surgery. Clin Orthop Relat Res. May 2014; 472(5): Hosmer D, Lemeshow S. Applied Survival Analysis: John Wiley & Sons, INC.; Allison PD. Survival Analysis Using SAS: A Practical Guide. 2 ed Mackinnon A. The use and reporting of multiple imputation in medical research - a review. J Intern Med. Dec 2010; 268(6): Kim MS, Lee SY, Lee TR, et al. Prognostic nomogram for predicting the 5-year probability of developing metastasis after neo-adjuvant chemotherapy and definitive surgery for AJCC stage II extremity osteosarcoma. Ann Oncol. May 2009; 20(5): Harrell FE, Jr., Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med. Feb ; 15(4): Steyerberg EW, Harrell FE, Jr., Borsboom GJ, Eijkemans MJ, Vergouwe Y, Habbema JD. Internal validation of predictive models: efficiency of some procedures for logistic regression analysis. J Clin Epidemiol. Aug 2001; 54(8): Cleves MA. From the help desk: Comparing areas under receiver operating characteristic curves from two or more probit or logit models. Stata Journal. 2002; 2(3):

208 Chapter Gupta S, Tran T, Luo W, et al. Machine-learning prediction of cancer survival: a retrospective study using electronic administrative records and a cancer registry. BMJ Open. 2014; 4(3): e Pepe M, Longton G Fau - Janes H, Janes H. Estimation and Comparison of Receiver Operating Characteristic Curves. Stata Journal ( X (Print)). 31. Siontis GC, Tzoulaki I, Castaldi PJ, Ioannidis JP. External validation of new risk prediction models is infrequent and reveals worse prognostic discrimination. J Clin Epidemiol. Jan 2015; 68(1): Collins GS, Reitsma JB, Altman DG, Moons KG. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement. BMJ. 2015; 350: g Schneiderbauer MM, von Knoch M, Schleck CD, Harmsen WS, Sim FH, Scully SP. Patient survival after hip arthroplasty for metastatic disease of the hip. J Bone Joint Surg Am. Aug 2004; 86-A(8): Edwards BK, Noone AM, Mariotto AB, et al. Annual Report to the Nation on the status of cancer, , featuring prevalence of comorbidity and impact on survival among persons with lung, colorectal, breast, or prostate cancer. Cancer. May ; 120(9): Sogaard M, Thomsen RW, Bossen KS, Sorensen HT, Norgaard M. The impact of comorbidity on cancer survival: a review. Clin Epidemiol. 2013; 5(Suppl 1): Patnaik JL, Byers T, Diguiseppi C, Denberg TD, Dabelea D. The influence of comorbidities on overall survival among older women diagnosed with breast cancer. J Natl Cancer Inst. Jul ; 103(14): McTiernan A, Irwin M, Vongruenigen V. Weight, physical activity, diet, and prognosis in breast and gynecologic cancers. J Clin Oncol. Sep ; 28(26): Jung SH, Yang DH, Ahn JS, et al. Decreased body mass index is associated with poor prognosis in patients with multiple myeloma. Ann Hematol. May 2014; 93(5):

209 Prognostication In Patients With Long Bone Metastases Appendix 1: International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes and Current Procedural Terminology (CPT) codes for identification of patients with pathologic or impending fractures Category Code Code description ICD-9-CM codes Pathologic fracture, unspecified site Pathologic fracture of humerus Pathologic fracture of radius and ulna Pathologic fracture of neck of femur Pathologic fracture of other specified part of femur Pathologic fracture of tibia or fibula Pathologic fracture of other specified site CPT codes Prophylactic treatment (nailing, pinning, plating or wiring) with or without methylmethacrylate; proximal humerus Prophylactic treatment (nailing, pinning, plating or wiring), with or without methylmethacrylate, humeral shaft Prophylactic treatment (nailing, pinning, plating or wiring) with or without methylmethacrylate; radius Prophylactic treatment (nailing, pinning, plating or wiring) with or without methylmethacrylate; ulna Prophylactic treatment (nailing, pinning, plating or wiring) with or without methylmethacrylate; radius and ulna Prophylactic treatment (nailing, pinning, plating or wiring) with or without methylmethacrylate, femoral neck and proximal femur Prophylactic treatment (nailing, pinning, plating or wiring) with or without methylmethacrylate, femur Prophylactic treatment (nailing, pinning, plating or wiring) with or without methylmethacrylate, tibia

210 Chapter 10 Appendix 2: Modified Charlson Comorbidity Index Algorithm Based on International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) Codes Comorbidity Weight* Codes AIDS/HIV Any malignancy, including leukemia and lymphoma* , , , , , , , 189.9, , , , , , , , , , , 230.8, 231.2, 231.9, , 233.0, 233.1, , , 233.4, 233.7, , 235.7, 235.8, 236.2, 236.4, 236.5, , , 237.6, , , , 239.6, 239.7, , Chronic pulmonary disease , 416.9, , , , 506.4, 508.1, Congestive heart failure , , , , , , , , , , , Dementia 2 290, 290.0, 290.3, , 294.1, , Diabetes with chronic complications , Hemiplegia or paraplegia , , Metastatic solid tumor* , , , Mild liver disease* , , , , , , 070.6, 070.9, 570, 570.1, 573.3, 573.4, 573.8, 753.9, V42.7 Moderate or severe liver disease* , Renal disease , , , , , , , , , , , 588.0, V42.0, V45.1, V56-V56.8 Rheumatologic disease , , , 714.8, 725 *The following comorbidities were mutually exclusive: mild liver disease and moderate or severe liver disease, and any malignancy and metastatic solid tumor. For example, a patient with a metastatic solid tumor received 6 points total (not 6 points for metastatic solid tumor and 2 points for any malignancy). 208

211 Prognostication In Patients With Long Bone Metastases Appendix 3: Hazard ratios for survival from bivariate Cox regression analysis* Explanatory variables β regression coefficient Standard error Hazard ratio (95% CI) p value Age (years) ( ) < BMI Less than 18.5 kg/m² ( ) < Between 18.5 and 30 kg/m² Reference Reference Reference Greater than 30 kg/m² ( ) Additional comorbidity ( ) < Men ( ) Pathological fracture ( ) Multiple locations operated during same procedure ( ) Race White Reference Reference Reference Other ( ) Tumor type other than breast, kidney, prostate, thyroid, myeloma, and lymphoma Time between primary tumor diagnosis and surgery for metastatic lesion Bone metastases ( ) < ( ) Single bone metastasis Reference Reference Reference Multiple bone metastases without spinal involvement ( ) Multiple bone metastases with spinal involvement ( ) Previous pathological fractures None Reference Reference Reference Previous long bone (impending) pathological fracture ( ) Previous spine (impending) pathological fracture ( ) Visceral metastases None Reference Reference Reference Lung and/or liver metastases ( ) < Brain metastases (with or without lung/liver metastases) ( ) < Previous local radiotherapy of the affected long bone ( ) Previous systemic therapy ( ) Hemoglobin (g/dl) ( ) Platelets (1000/mm³) ( ) White blood cell count (1000/mm³) ( ) Creatinine (mg/dl) ( ) Calcium (mg/dl) ( ) Total number of patients = 927; *patients with missing values were excluded per bivariate analysis: BMI was available in 730 patients; hemoglobin level in 868, platelet level in 866, white blood cell count in 867, creatinine in 812, and calcium in 654; with or without previous long bone pathological fracture; variables with a p value less than 0.10 in bivariate analysis were included in multivariable Cox regression analysis

212

213 CHAPTER 11 Are Allogeneic Blood Transfusions Associated With Decreased Survival After Surgery For Long Bone Metastatic Fractures? S.J. Janssen Y. Braun J.E. Ready K.A. Raskin M.L. Ferrone F.J. Hornicek J.H. Schwab Clinical Orthopaedics and Related Research 2015 Jul;473(7): Award: Poster of Distinction Massachusetts General Hospital Clinical Research Day 2014 Presented at: Association of Bone and Joint Surgeons Annual Meeting 2015, Eugene, Oregon, USA. Poster at: Massachusetts General Hospital Clinical Research Day 2014, Boston, Massachusetts, USA. Scientific Advisory Committee conference 2015, Boston, Massachusetts, USA. Commentary by: L.R. Randall in Clinical Orthopaedics and Related Research 2015 Jul;473(7):

214 Chapter 11 ABSTRACT Objectives To assess whether perioperative allogeneic blood transfusion is associated with decreased survival after surgical treatment of long bone metastatic fractures after accounting for clinical, laboratory, and treatment factors. Design Retrospective cohort study. Setting Two tertiary care referral centers for orthopaedic oncology. Participants 789 patients who underwent surgery at two institutions for a pathological or impending fracture through a long bone metastasis between 1999 and Interventions Allogeneic blood transfusion within 7 days before until 7 days after surgery. Considering transfusion as an exposure, and assessing a dose-response relationship per unit of blood transfused. Outcome Measures Overall survival. Results Comparing patients who received transfusions (exposure) with those who did not, we found that blood transfusion was not associated with decreased survival after accounting for all explanatory variables (hazard ratio [HR] 1.06; 95% CI, ; p = 0.57). Evaluating transfusion in terms of dose-response, we found that patients who received more transfusions had lower survival compared with those who had fewer transfusions after accounting for all explanatory variables (HR per unit of blood transfused, 1.07; 95% CI, ; p = 0.005). Conclusions We found that exposure to perioperative allogeneic blood transfusion does not decrease survival, with the numbers available. However, our sample size might have been insufficient to reveal a small but potentially relevant effect. Our results do suggest a dose-response relationship; patients who received more transfusions had lower survival compared with those with fewer transfusions. Risk of death increased by 7% per unit of blood transfused. 212

215 Blood Transfusions In Patients With Long Bone Metastases INTRODUCTION Perioperative allogeneic blood transfusion has been associated with tumor recurrence and decreased survival in patients with numerous primary malignancies, including colon, esophageal, hepatic, bladder, and endometrial cancers. 1-4 This detrimental effect of blood transfusion is explained by the immunomodulating effects of allogeneic blood transfusions. 1-4 The immunomodulating effect was first hypothesized in 1973 in a study showing improved graft survival in patients who received allogeneic blood transfusion prior to kidney transplantation. 5 This led to the idea by Gantt 6 in 1981 that in patients with malignant tumors, blood transfusions would give the tumor a better chance to survive. 2,6 Although the exact mechanism is still unclear, 7 several subsequent clinical and laboratory studies confirmed the immunosuppressive effect of blood transfusions in patients with cancer. 2,7,8 Approximately ⅔ of patients with metastatic breast or prostate cancer, and ⅓ with metastatic lung or gastrointestinal cancer have bone metastases develop Bone metastatic disease adversely affects quality of life and survival, 9,12 and many patients with bone metastases undergo surgery to treat or prevent a pathological fracture. 9,12,13 Blood loss during surgical treatment of bone metastatic lesions can be substantial and blood transfusions frequently are administered. 14 Although studies suggest that transfusions increase the risk of metastasis and death, 1-4 we do not know whether allogeneic blood transfusions influence survival in patients with already disseminated cancer. We therefore sought to determine whether perioperative allogeneic blood transfusion is associated with decreased patient survival after surgical treatment of long bone metastatic lesions, accounting for clinical, laboratory, and treatment factors. Secondarily, we aimed to identify potential factors associated with decreased survival. METHODS Study Design Our retrospective study was approved by our institutional review board and a waiver of informed consent was obtained. We included all 789 eligible patients with an impending or pathological fracture through a metastatic long bone lesion undergoing surgery between 1999 and 2013 at two tertiary care referral centers for orthopaedic oncology. Metastatic disease included, in addition to metastases from solid organs, multiple myeloma and lymphoma; we regarded the femur, humerus, tibia, fibula, radius, and ulna as long bones. Medical records of patients with a billing or diagnostic code for a pathological long bone fracture or prophylactic treatment of an impending fracture were reviewed to assess eligibility (Appendix 1). Exclusion criteria were: (1) revision procedures; (2) metastatic involvement of the acetabulum or pelvis requiring additional reconstruction; (3) metastatic fractures

216 Chapter 11 in multiple bones requiring surgery; and (4) surgical treatment other than endoprosthetic reconstruction, intramedullary nailing, open reduction and internal fixation with plate and screws or a dynamic hip screw. The decision to operate and the surgical strategy were based on estimated survival, location and size of the metastatic lesion, primary tumor type, level of disability, and pain. Postoperative care and rehabilitation varied among patients owing to disease severity. Outcome Measures Survival was our primary outcome measure. We used the Social Security Death Index (a database of death records created from the United States Social Security Administration) throughout February 24, 2014 to establish date of death. 15 February 24, 2014 was considered the final moment of followup for assessment of the outcome. Six hundred thirty-seven patients (81%) were deceased by the final moment of followup. We defined perioperative allogeneic blood transfusion as transfusion of packed red blood cells within 7 days before until 7 days after surgery. Transfusion volume was expressed as the number of units transfused; one unit of blood contains approximately 330 ml, but can vary from 300 ml to 360 ml (the hematocrit per unit of blood can vary from 55% to 58%). Patients in the no transfusion group either had no perioperative transfusion or only autologous blood transfusion. We did not account for transfusion of other blood products, including fresh frozen plasma or platelets. Seventy percent (914 of 1298) of the packed red blood cells transfused in this study were leukoreduced (leucocytes removed from packed red blood cells). We found no difference in survival between patients who received only leukoreduced blood units (n = 288) and patients who received at least one nonleukoreduced blood unit (n = 155) (p = 0.68, by log-rank analysis). Transfusion guidelines in both hospitals did not change during our study period. The blood transfusion threshold guidelines at one hospital (Hospital 1) during the 15-year study period were: hematocrit less than 24% for patients younger than 40 years; less than 27% for patients between 40 and 60 years old; and less than 30% for patients 60 years or older. At the other hospital (Hospital 2), the hematocrit thresholds used were: 26% for patients with cancer, who were pregnant, or who had preoperative anemia; 30% for patients with acute coronary syndrome or major thoracic surgery; and 21% for patients who were normovolemic and non-bleeding for which the previously mentioned thresholds did not apply. Adherence to these transfusion guidelines was not documented; however, we assessed the trigger for transfusion with time and graphed pretransfusion hematocrit levels (Figure 1). Linear regression analysis showed no change in patient hematocrit level resulting in transfusion with time (p = 0.12). We measured the following explanatory patient variables: age at time of surgery, sex, comorbidity status, BMI in Kg/m², primary tumor type, fracture type, fracture location, presence of other bone metastatic lesions, presence of visceral metastases, previous 214

217 Blood Transfusions In Patients With Long Bone Metastases Hematocrit (%) Year Figure 1: The change in hematocrit level triggering perioperative allogeneic blood transfusion in our cohort with time is shown. The decrease was not significant as assessed using multivariate linear regression analysis (p = 0.12). The 95% CIs are indicated in gray. The hematocrit to hemoglobin ratio was approximately 3:1. local radiation therapy, previous systemic therapy, preoperative embolization of the tumor, preoperative hemoglobin level, surgical treatment type, anesthesia time in minutes, total estimated blood loss during surgery in ml, duration of hospital admission in days, year of surgery, and hospital (1 or 2). Patient comorbidity status was assessed using the modified Charlson Comorbidity Index, 16,17 which provides a score ranging from 0 to 24, with a higher score representing more severe comorbidity status, based on 12 weighted comorbidities. 16 We determined the modified Charlson Comorbidity Index through a previously described algorithm based on International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9- CM) codes given to the patient before the day of surgery (Appendix 2) Based on the existing literature, we categorized primary tumor type into relatively good prognosis (breast, kidney, prostate, thyroid, myeloma, and lymphoma) and poor prognosis (all other tumor types) Fracture type was impending or pathological. Impending fractures were defined as a destructive bone lesion with no visible fracture line, loss of height, rotation, or angulation. The degree of destruction and intractable pain did mandate, in the surgeon s opinion, surgical treatment to preclude a pathological fracture. We extracted the presence of other bone metastatic lesions from bone scans, CT scans, and other imaging reports, and categorized combined results into either multiple- or singlebone metastatic lesion(s). The presence of visceral metastases was derived from CT and positron emission tomography reports. We regarded lung, liver, and brain metastases as visceral metastases

218 Chapter 11 We defined previous local radiation therapy as radiation to the area of surgery, and previous systemic therapy as any type of hormonal therapy, immunotherapy, or chemotherapy for the primary tumor. Preoperative embolization was performed in 21 cases, of which most were renal cell carcinomas (15 of 20 [75%]), to reduce tumor vascularity and decrease intraoperative blood loss. Preoperative hemoglobin was extracted when measured within 7 days before surgery; we used the last available hemoglobin measurement before surgery. Types of surgical treatment were endoprosthetic reconstruction, intramedullary nailing, plate-screw fixation, or dynamic hip screw. We considered anesthesia time as a surrogate marker for surgical treatment time and measured this time in minutes from the moment the patient entered the operating room until the patient left the operating room. Statistical Analysis Variables are presented as frequencies with percentages for categorical variables and median with interquartile range (IQR) for continuous variables. Bivariate analysis was used to compare explanatory variables between the no transfusion group (no exposure) and the perioperative allogeneic blood transfusion group (exposure) by Fisher exact test for categorical variables and Mann-Whitney U test (also known as the rank-sum test) for continuous variables. We used nonparametric analysis for continuous variables as visual inspection of histograms suggested non-normal distributions. Multivariate Cox regression analysis was used to assess the relationship of perioperative allogeneic blood transfusion with survival, adjusted for all explanatory variables included in the study. We compared the exposure with the no exposure group, and assessed a dose-response relationship. Hazard ratios (HRs) the relative likelihood of death in the exposure group versus the no exposure group and per unit of blood transfused with 95% confidence intervals were provided to quantify the association with survival. To preclude overfitting of the model, we reduced the number of variables by including year of surgery as a continuous variable and categorized location of the fracture into upper or lower extremity. We used multiple imputation for missing values (number of imputations was set to 40; i.e. the missing values were 40 times replaced by values imputed based on all other explanatory variables): hemoglobin level was missing in 49 of 789 (6.2%) cases, total estimated blood loss was missing in 117 of 789 (15%) cases, anesthesia time was missing in 85 of 789 (11%) cases, and BMI was missing in 179 of 789 (23%) cases. All statistical analyses were performed using Stata 13.0 (StataCorp LP, College Station, TX, USA). A two-tailed p value less than 0.05 was considered statistically significant. 216

219 Blood Transfusions In Patients With Long Bone Metastases RESULTS Baseline Characteristics Study patients included 351 (44%) men and 438 (56%) women, with a median age of 64 years (IQR, 54 to 72; Table 1). Median preoperative hemoglobin was 11 g/dl (available for 740 patients; IQR, 10 to 12) and median perioperative estimated blood loss was 200 ml (available for 672 patients; IQR, 100 to 400). There were 341 (43%) impending and 448 (57%) pathological fractures. Most fractures were in the femur (574 of 789; 73%) and Table 1: Patient- and treatment characteristics for no transfusion group and allogeneic transfusion group Variable No transfusion (no exposure) (n = 346) Median (interquartile range) Perioperative allogeneic blood transfusion* (exposure) (n = 443) Median (interquartile range) (n = 789) p value Age (years) 62 (53-70) 64 (55-73) Modified Charlson Comorbidity Index 6 (6-7) 6 (6-8) < Preoperative hemoglobin level (g/dl ) 12 (11-13) 11 (10-12) < Total estimated blood loss during surgery (ml ) 200 ( ) 300 ( ) < Anesthesia time (minutes ) 185 ( ) 198 ( ) < Duration of hospital admission (days) 5 (3-7) 7 (5-11) < BMI (kg/m²) 27 (24-31) 26 (22-29) < Sex n (%) n (%) Men 180 (51) 171 (49) Women 166 (38) 272 (62) Primary tumor type Good prognosis 186 (40) 281 (60) Poor prognosis 160 (50) 162 (50) Fracture type Impending fracture 178 (52) 163 (48) Pathological fracture 168 (37.5) 280 (62.5) Fracture location Femur 215 (37) 359 (63) Humerus 113 (62) 69 (38) Tibia 17 (61) 11 (39) Radius 0 (0) 3 (100) Ulna 1 (50) 1 (50) < < <

220 Chapter 11 Table 1: Patient- and treatment characteristics for no transfusion group and allogeneic transfusion group (continued) Variable Number of bone metastatic lesions No transfusion (no exposure) (n = 346) Median (interquartile range) Single 103 (55) 84 (45) Multiple 243 (40) 359 (60) Visceral metastases No 195 (47) 224 (53) Yes (lung, liver, or brain) 151 (41) 219 (59) Previous local radiotherapy No 282 (44) 359 (56) Yes 64 (43) 84 (57) Previous systemic therapy No 148 (49) 152 (51) Yes 198 (40) 291 (60) Preoperative embolization of tumor No 338 (44) 431 (56) Yes 8 (40) 12 (60) Surgical treatment type Year Intramedullary nailing 221 (47) 246 (53) Endoprosthetic reconstruction 55 (30) 129 (70) Plate-screw fixation 61 (49) 63 (51) Dynamic hip screw 9 (64) 5 (36) (39) 138 (61) (43) 137 (57) (48) 168 (52) Hospital # Hospital (45) 188 (55) Hospital (43) 255 (57) Perioperative allogeneic blood transfusion* (exposure) (n = 443) Median (interquartile range) p value < < bold font indicates a significant difference (two-tailed p value below 0.05); *allogeneic blood transfusion within 7 days before until 7 days after surgery; preoperative hemoglobin level was available for 740 patients (309 in no transfusion group, 431 in transfusion group), total estimated blood loss in 672 patients (293 in no transfusion group, 379 in transfusion group), anesthesia time for 704 patients (315 in no transfusion group, 389 in transfusion group), and BMI for 610 patients (284 in no transfusion group, 326 in transfusion group); primary tumor type with good prognosis includes breast, kidney, prostate, thyroid, myeloma and lymphoma and with poor prognosis includes all other tumor types; # see text for transfusion guidelines per hospital

221 Blood Transfusions In Patients With Long Bone Metastases Table 2: Origin of primary tumors Tumor distribution n (%) Lung 191(24) Breast 181 (23) Myeloma 116 (15) Kidney 74 (9.4) Prostate 45 (5.7) Lymphoma 36 (4.6) Melanoma 20 (2.5) Colorectal 16 (2.0) Esophagus 15 (1.9) Thyroid 15 (1.9) Hepatocellular 11 (1.4) n = 789 Other * 37 (4.7) Unknown 39 (4.1) * Bladder (n = 8), neuroendocrine (n = 6), salivary gland (n = 5), adenocarcinoma of unknown origin (n = 5), pancreas (n = 4), ovaries (n = 4), stomach (n = 1), vulva (n = 2), and endometrium (n=2). humerus (182 of 789; 23%). The five most common primary tumor types were lung (191 of 789; 24%), breast (181 of 789; 23%), myeloma (116 of 789; 15%), kidney (74 of 789; 9.4%), and prostate (45 of 789; 5.7%; Table 2). The median number of allogeneic blood units transfused among patients (443 of 789 patients; 56%) who received a transfusion was two (IQR, 2 to 4; range, 1 to 15). Three hundred forty-six (44%) patients did not have a perioperative transfusion. Median overall survival was 254 days (95% CI, 221 to 288) and 42% of the patients survived one year (331 of 789). Factors associated with perioperative allogeneic blood transfusion in bivariate analyses included older age (p = 0.025), more severe comorbidity status (p < 0.001), lower preoperative hemoglobin level (p < 0.001), more blood loss during surgery (p < 0.001), longer anesthesia time (p < 0.001), longer hospital stay (p < 0.001), lower BMI (p < 0.001), female sex (p < 0.001), primary tumor type with relatively good prognosis (p = 0.007), pathological fracture (p < 0.001), multiple metastatic bone lesions (p < 0.001), and previous systemic therapy (p = 0.018; Table 1). Furthermore, there was a difference between the no transfusion and the transfusion groups among location of the fracture (p < 0.001) and type of surgical treatment (p < 0.001; Table 1)

222 Chapter 11 Association Of Transfusion With Survival Considering transfusion as an exposure and comparing patients who received transfusions with those who did not, we found that blood transfusion was not associated with decreased survival after surgery of a long bone metastasis. The hazard ratio after accounting for all explanatory variables in multivariable Cox regression analysis was 1.06 with a 95% CI of 0.87 to 1.30 (p = 0.57), indicating no difference in survival between patients exposed to blood transfusions and those who were not exposed (Table 3). Evaluating transfusion in terms of dose-response, we found that patients who received more transfusions during their skeletal stabilization procedures had poorer survival compared with those who had fewer transfusions. Multivariable Cox regression analysis showed an hazard ratio of 1.07 (95% CI, 1.02 to 1.12; p = ) per unit of blood transfused after accounting for all explanatory variables, meaning a 7% higher risk of death per unit of blood transfused (Table 3). Risk Factors For Decreased Survival We found that older age, more severe comorbidity status, longer duration of hospital stay, poor prognosis tumor type, and visceral metastases were associated with decreased survival after accounting for likely confounding variables. Variables independently associated with decreased survival from the multivariable Cox regression analysis including blood transfusion units as a continuous variable were: age (HR, 1.02; 95% CI, 1.01 to 1.02; p < 0.001), modified Charlson Comorbidity Index (HR, 1.06; 95% CI, 1.01 to 1.10; p = 0.014), duration of hospital stay (HR, 1.02; 95% CI, 1.00 to 1.03; p = 0.021), tumor type (HR, 1.71; 95% CI, 1.44 to 2.03; p < 0.001), and visceral metastases (HR, 1.59; 95% CI, 1.34 to 1.88; p < 0.001; Table 3). DISCUSSION Perioperative blood transfusion increases cancer recurrence and decreases patient survival after resection of primary malignancies. 1-4 It is unclear if the same is true for patients with cancer already disseminated to bone. After controlling for likely confounding variables, we found that exposure to perioperative allogeneic blood transfusion was not associated with decreased survival in patients undergoing surgery for metastatic bone lesions. However, when evaluating a dose-response relationship, we found that patients who received more transfusions had lower survival compared with those with fewer transfusions. Risk of death increased by 7% per unit of blood transfused. Our study has several limitations. First, with our sample size we found no effect of exposure to blood transfusion on survival; however, a larger sample size might have resulted in a significant difference in survival between patients exposed to blood transfusions and those 220

223 Blood Transfusions In Patients With Long Bone Metastases Table 3: Multivariate Cox regression analysis assessing the influence of perioperative allogeneic blood transfusion and the number of units transfused on survival. Blood transfusion (exposure/no exposure)* Blood transfusion (per unit transfused)* p value Hazard ratio (95%CI) Standard error p value Hazard ratio (95% CI) Standard error Age (years) ( ) < ( ) < Modified Charlson Comorbidity Index ( ) ( ) Preoperative hemoglobin level (g/dl ) ( ) ( ) Total estimated blood loss during surgery (ml ) < ( ) < ( ) Anesthesia time (minutes ) < ( ) < ( ) Duration of hospital admission (days) ( ) ( ) BMI (kg/m² ) ( ) ( ) Male sex ( ) ( ) Poor prognosis tumor type ( ) < ( ) < Pathological fracture ( ) ( ) Lower extremity ( ) ( ) Other bone metastatic lesions ( ) ( ) Visceral metastases ( ) < ( ) < Previous local radiotherapy ( ) ( ) Previous systemic therapy ( ) ( ) Preoperative embolization of tumor ( ) ( ) Surgical treatment type Reference value Reference value Reference value Reference value Reference value Reference value Intramedullary nailing Endoprosthetic reconstruction ( ) ( )

224 Chapter 11 Table 3: Multivariate Cox regression analysis assessing the influence of perioperative allogeneic blood transfusion and the number of units transfused on survival. (continued) Blood transfusion (exposure/no exposure)* Blood transfusion (per unit transfused)* Standard error Hazard ratio (95% CI) p value Standard error Hazard ratio (95%CI) p value Plate-screw fixation ( ) ( ) Dynamic hip screw ( ) ( ) Year of surgery ( ) ( ) Hospital 1 # ( ) ( ) Blood transfusion* (exposure versus no exposure) ( ) Blood transfusion* (per unit transfused) ( ) bold font indicates a significant difference (two-tailed p value below 0.05); *allogeneic blood transfusion within 7 days before until 7 days after surgery; preoperative hemoglobin level was available in 740 patients, total estimated blood loss in 672 patients, anesthesia time in 704 patients, and BMI in 610 patients. Missing values were imputed using multiple linear imputation including all variables (the number of imputations was set to 40); primary tumor type with good prognosis includes: breast, kidney, prostate, thyroid, myeloma and lymphoma; # see text for transfusion guidelines per hospital (Hospital 2 is the reference group). 222

225 Blood Transfusions In Patients With Long Bone Metastases not exposed. To achieve a power of 0.80 and assuming a similar hazard ratio, variability, and covariate correlation, we would have needed 17,058 patients to show a difference in survival between the patients exposed to blood transfusion and those not exposed (with alpha of 0.05). With the current sample size, variability, and covariate correlation, we would have found a difference in survival between the exposure group and non-exposure group when the hazard ratio exceeded 1.31 (alpha of 0.05 and power of 0.80). The relatively small sample size is a limitation as even a small effect would be clinically relevant; a large sample size might have revealed such a small effect. Second, there were no uniform criteria for surgical treatment owing to the retrospective design of the study; however, we see this as a minor limitation as most surgeons used similar approaches during the study period when deciding whether to stabilize a pathological fracture or intervene in the face of an impending fracture. In general, a pathological fracture was stabilized when the patient was expected to live longer than 30 days. An impending fracture was treated surgically when there was pain on load-bearing or in case of substantial bone destruction in a load-bearing bone. Third, although transfusion guidelines did not change during our study period, the exact indication in every patient was not documented; however, we believe this is a minor limitation as we found no change in the hematocrit level during time that was associated with transfusion. Fourth, we assessed overall and not cancer-specific survival. We could not determine how many of the deceased had died from their malignancy as opposed to from other causes; however, as the majority of patients who have skeletal metastases eventually die from their malignancies, we do not believe this is a severe limitation. We found that allogeneic blood transfusion did not affect survival of patients undergoing surgery for metastatic long bone fractures with the numbers available. Clausen et al. 24 showed that perioperative allogeneic blood transfusion in patients with spine metastases was not associated with 3-month survival; they found an increased 12-month survival after 1 to 2 units of allogeneic blood transfusion compared with no transfusion (odds ratio, 2.6; 95% CI, 1.0 to 6.8; p = 0.049). Increased survival might be explained by a difference in indication for blood transfusion or anatomic site studied, or because of the few confounding factors taken into account (preoperative hemoglobin, age, sex, Tokuhashi score [survival prognostication score], and number of instrumented levels). 24 Previous studies in visceral metastatic disease have shown a negative effect of perioperative blood transfusion on survival: 25,26 Katoh et al. 25 found that perioperative transfusion was independently associated with worse survival after resection of stage IV (disseminated disease) colorectal cancer (n = 162) after accounting for clinical and treatment parameters; Margulis et al. 26 demonstrated that survival of patients undergoing cytoreductive nephrectomy for metastatic renal cell carcinoma was also negatively influenced by intraoperative blood transfusion. The decrease in observed survival after perioperative blood transfusions in primary malignancies was not apparent in our study on metastatic long bone lesions and might be explained by the already widely disseminated disease and poor overall survival, potentially attenuating the effect of blood transfusions

226 Chapter 11 We found that older age, more severe comorbidity status, longer duration of hospital stay, poor prognosis tumor type, and visceral metastases were associated with decreased patient survival. Results from previous studies support the association of tumor type and visceral metastases with survival Bauer et al. 23 found a survival rate of 0.30 at 1 year in a retrospective cohort of 241 patients who had surgery for bone metastatic disease; they found decreased survival among patients with a pathological fracture, visceral metastases, multiple bone metastases, and lung cancer. Katagiri et al. 22 reported a survival rate of 0.48 at 1 year in a retrospective cohort of 350 patients with bone metastatic disease. Poor performance status, specific tumor types (hepatocellulair, gastric, and lung carcinoma), visceral metastases, previous chemotherapy, and multiple skeletal metastases were risk factors for decreased patient survival. A prospective cohort study by Nathan et al. 21 showed a median survival of 8 months for 191 patients who had surgery for bone metastases. Tumor type, performance status, number of bone metastases, visceral metastases, hemoglobin count, and the surgeon s estimate of survival were predictors of patient survival. Our results showed an association of age and comorbidity status with survival. These prognostic factors, in addition to those already known, should be considered when estimating life expectancy. Estimated survival is an important factor in the decision to operate and the selection of surgical strategy in patients with metastatic bone disease Future studies should incorporate these factors in prognostication models to improve prediction accuracy. We found that exposure to perioperative allogeneic blood transfusion does not decrease survival with the numbers available. However, our sample size might have been insufficient to reveal a small but potentially relevant effect. Our results do suggest a dose-response relationship; patients who received more transfusions had lower survival compared with those with fewer transfusions. Risk of death increased by 7% per unit of blood transfused. REFERENCES 1. Linder BJ, Frank I, Cheville JC, et al. The impact of perioperative blood transfusion on cancer recurrence and survival following radical cystectomy. Eur Urol. May 2013; 63(5): Wu HS, Little AG. Perioperative blood transfusions and cancer recurrence. J Clin Oncol. Aug 1988; 6(8): Uccella S, Ghezzi F, Cromi A, et al. Perioperative allogenic blood transfusions and the risk of endometrial cancer recurrence. Arch Gynecol Obstet. May 2013; 287(5): Amato A, Pescatori M. Perioperative blood transfusions for the recurrence of colorectal cancer. Cochrane Database Syst Rev. 2006(1): CD Opelz G, Sengar DP, Mickey MR, Terasaki PI. Effect of blood transfusions on subsequent kidney transplants. Transplant Proc. Mar 1973; 5(1): Gantt CL. Red blood cells for cancer patients. Lancet. Aug ; 2(8242): Buddeberg F, Schimmer BB, Spahn DR. Transfusion-transmissible infections and transfusionrelated immunomodulation. Best Pract Res Clin Anaesthesiol. Sep 2008; 22(3):

227 Blood Transfusions In Patients With Long Bone Metastases 8. Heiss MM, Mempel W, Jauch KW, et al. Beneficial effect of autologous blood transfusion on infectious complications after colorectal cancer surgery. Lancet. Nov ; 342(8883): Quinn RH, Randall RL, Benevenia J, Berven SH, Raskin KA. Contemporary management of metastatic bone disease: tips and tools of the trade for general practitioners. Instr Course Lect. 2014; 63: Mundy GR. Metastasis to bone: causes, consequences and therapeutic opportunities. Nat Rev Cancer. Aug 2002; 2(8): Coleman RE. Clinical features of metastatic bone disease and risk of skeletal morbidity. Clin Cancer Res. Oct ; 12(20 Pt 2): 6243s-6249s. 12. Bickels J, Dadia S, Lidar Z. Surgical management of metastatic bone disease. J Bone Joint Surg Am. Jun 2009; 91(6): Capanna R, Campanacci DA. The treatment of metastases in the appendicular skeleton. J Bone Joint Surg Br. May 2001; 83(4): Ward WG, Holsenbeck S, Dorey FJ, Spang J, Howe D. Metastatic disease of the femur: surgical treatment. Clin Orthop Relat Res. 2003: S Huntington JT, Butterfield M, Fisher J, Torrent D, Bloomston M. The Social Security Death Index (SSDI) most accurately reflects true survival for older oncology patients. Am J Cancer Res. 2013; 3(5): Quan H, Li B, Couris CM, et al. Updating and validating the Charlson comorbidity index and score for risk adjustment in hospital discharge abstracts using data from 6 countries. Am J Epidemiol. Mar ; 173(6): Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987; 40(5): Voskuijl T, Hageman M, Ring D. Higher Charlson Comorbidity Index Scores are associated with readmission after orthopaedic surgery. Clin Orthop Relat Res. May 2014; 472(5): Quan H, Parsons GA, Ghali WA. Validity of information on comorbidity derived rom ICD-9- CCM administrative data. Med Care. Aug 2002; 40(8): Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol. Jun 1992; 45(6): Nathan SS, Healey JH, Mellano D, et al. Survival in patients operated on for pathologic fracture: implications for end-of-life orthopedic care. J Clin Oncol. Sep ; 23(25): Katagiri H, Takahashi M, Wakai K, Sugiura H, Kataoka T, Nakanishi K. Prognostic factors and a scoring system for patients with skeletal metastasis. J Bone Joint Surg Br. May 2005; 87(5): Bauer HC, Wedin R. Survival after surgery for spinal and extremity metastases. Prognostication in 241 patients. Acta Orthop Scand. Apr 1995; 66(2): Clausen C, Lonn L, Morgen SS, et al. Perioperative blood transfusion does not decrease survival after surgical treatment of spinal metastases. Eur Spine J. May Katoh H, Yamashita K, Kokuba Y, et al. Surgical resection of stage IV colorectal cancer and prognosis. World J Surg. Jun 2008; 32(6): Margulis V, Shariat SF, Rapoport Y, et al. Development of accurate models for individualized prediction of survival after cytoreductive nephrectomy for metastatic renal cell carcinoma. Eur Urol. May 2013; 63(5):

228 Chapter 11 Appendix 1: International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes and Current Procedural Terminology (CPT) codes for identification of patients with pathologic or impending fractures Category Code Code description ICD-9-CM codes Pathologic fracture, unspecified site Pathologic fracture of humerus Pathologic fracture of radius and ulna Pathologic fracture of neck of femur Pathologic fracture of other specified part of femur Pathologic fracture of tibia or fibula Pathologic fracture of other specified site CPT codes Prophylactic treatment (nailing, pinning, plating or wiring) with or without methylmethacrylate; proximal humerus Prophylactic treatment (nailing, pinning, plating or wiring), with or without methylmethacrylate, humeral shaft Prophylactic treatment (nailing, pinning, plating or wiring) with or without methylmethacrylate; radius Prophylactic treatment (nailing, pinning, plating or wiring) with or without methylmethacrylate; ulna Prophylactic treatment (nailing, pinning, plating or wiring) with or without methylmethacrylate; radius and ulna Prophylactic treatment (nailing, pinning, plating or wiring) with or without methylmethacrylate, femoral neck and proximal femur Prophylactic treatment (nailing, pinning, plating or wiring) with or without methylmethacrylate, femur Prophylactic treatment (nailing, pinning, plating or wiring) with or without methylmethacrylate, tibia 226

229 Blood Transfusions In Patients With Long Bone Metastases Appendix 2: Modified Charlson Comorbidity Index Algorithm Based on International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) Codes Comorbidity Weight* Codes AIDS/HIV Any malignancy, including leukemia and lymphoma* , , , , , , , 189.9, , , , , , , , , , , 230.8, 231.2, 231.9, , 233.0, 233.1, , , 233.4, 233.7, , 235.7, 235.8, 236.2, 236.4, 236.5, , , 237.6, , , , 239.6, 239.7, , Chronic pulmonary disease , 416.9, , , , 506.4, 508.1, Congestive heart failure , , , , , , , , , , , Dementia 2 290, 290.0, 290.3, , 294.1, , Diabetes with chronic complications , Hemiplegia or paraplegia , , Metastatic solid tumor* , , , Mild liver disease* , , , , , , 070.6, 070.9, 570, 570.1, 573.3, 573.4, 573.8, 753.9, V42.7 Moderate or severe liver disease* , Renal disease , , , , , , , , , , , 588.0, V42.0, V45.1, V56-V56.8 Rheumatologic disease , , , 714.8, 725 *The following comorbidities were mutually exclusive: mild liver disease and moderate or severe liver disease, and any malignancy and metastatic solid tumor. For example, a patient with a metastatic solid tumor received 6 points total (not 6 points for metastatic solid tumor and 2 points for any malignancy).*the following comorbidities were mutually exclusive: mild liver disease and moderate or severe liver disease, and any malignancy and metastatic solid tumor. For example, a patient with a metastatic solid tumor received 6 points total (not 6 points for metastatic solid tumor and 2 points for any malignancy)

230

231 CHAPTER 12 Summary 12

232 Chapter 12 Appendicular bone metastastases are most frequent in the femur and the humerus. These lesions can weaken the bone and often lead to pathological fractures. An increase in the number of patients living with bone metastatic disease will most likely result in a larger number of patients developing pathological fractures. Pathological fractures negatively impact quality of life, physical function, and survival. Surgical intervention is often indicated to treat a pathological fracture in a long bone and aims to optimize quality of life by providing a stable construct that outlives the patient. The purpose of this PhD thesis is to: develop tools for better patient selection for surgery, improve implant selection based on patient- and tumor characteristics, identify risk factors for adverse outcomes, and evaluate outcome after treatment for patients with long bone metastases. PART I: METASTATIC FEMORAL LESIONS This part starts with a retrospective study (Chapter 2) comparing outcomes after surgical management of pathological and impending fractures of the proximal femur using data from two affiliated orthopaedic oncology centers. This study included 417 consecutive patients with proximal femoral metastases that were treated by intramedullary nailing, endoprosthetic reconstruction, and open reduction internal fixation. Overall reoperation rate did not differ significantly between these strategies and were 5.3% after intramedullary nailing, 11% after endoprosthetic reconstruction, and 13% after open reduction and internal fixation. However, we did find differences in reoperation rates when focusing on specific reasons for reoperation: fixation failure and deep infection. The fixation failure rate was highest after open reduction internal fixation (13%), followed by intramedullary nailing (3%), and none after endoprosthetic reconstruction. The deep infection rate also differed and was highest after endoprosthetic reconstruction (8.6%), followed by intramedullary nailing (2%), and none after open reduction internal fixation. Deep infections occur in the first months after surgery, while fixation failure accumulates over time. Overall, 30-day systemic complication rates (9.6%) and 30-day survival (93%) did not differ between surgical strategies. Intraoperative blood loss was almost twice as high and anesthesia time was about 40 minutes longer for endoprosthetic reconstruction when compared with intramedullary nailing and open reduction internal fixation. The subsequent systematic review (Chapter 3) puts these findings into perspective and included 40 studies describing endoprosthetic reconstruction (n=1,461), intramedullary nailing (n=1,054), and open reduction internal fixation (n=233) for proximal femoral metastases. Breast (35%), lung (15%), prostate (10%), and renal cell (8.2%) are the most common primary cancers. Oneyear survival ranged from 0 to 62%; variation in baseline (e.g. primary tumor type) explains the substantial variation in oneyear survival. The pooled overall reoperation rate differed significantly between surgical strategies and was 5.2% after endoprostheses, 4.2% after intramedullary nailing, and 14% after open reduction 230

233 Summary internal fixation. Again, when focusing on reasons for reoperation, we noticed a difference. The pooled fixation failure rate was 0.4% after endoprostheses, 2.8% after intramedullary nailing, and 10% after open reduction internal fixation. The pooled deep infection rate was 0.68% for endoprostheses, 0.04% for intramedullary nailing, and none for open reduction internal fixation. We were not able to pool and compare functional outcome and systemic complications between surgical strategies due to varying definitions and heterogeneity in reporting; functional outcome has been scarcely reported. Subsequently, we established which patient reported outcome questionnaire was the most useful in terms of effectiveness, reliability, and efficiency when measuring physical function in patients with lower extremity bone metastases (Chapter 4). One hundred of 115 invited patients with lower extremity metastases participated in this prospective study and completed five questionnaires in random order: PROMIS-CAT Physical Function Cancer, PROMIS-CAT Neuro-QoL Mobility, Toronto Extremity Salvage Score (TESS), Lower Extremity Function Score (LEFS), and Musculoskeletal Tumor Society score (MSTS). All questionnaires measured the same concept; this was demonstrated by substantial correlation of every questionnaire with the underlying trait (factor scores >0.7) and by high interquestionnaire correlations (>0.7). Floor effect (i.e. score at lowest limit of questionnaire) was absent, while ceiling effect (i.e. score at upper limit of questionnaire) was present in all, but highest for the PROMIS Neuro-QoL Mobility (7%) indicating poor coverage. The standard error of measurement (i.e. a measure of precision of a questionnaire) was below the threshold indicating reliability over a wide range of ability levels for the PROMIS Physical Function Cancer, TESS, and LEFS. Completion time differed between questionnaires, and was shortest for the two PROMIS questionnaires. The PROMIS Physical Function Cancer questionnaire is the most useful questionnaire to measure physical function in patients with lower extremity bone metastases due to its reliability over a wide range of ability levels, validity, brevity, and good coverage. Finally, we introduced and tested an algorithm to establish fracture risk in lytic femoral metastases using CT scans (Chapter 5). The algorithm uses attenuation measurements of metastases (measured in Hounsfield units) in CT scans compared to the unaffected contralateral side. We tested this method by comparing the cumulative Hounsfield units and the geometric distribution of the Hounsfield units between femora with metastatic lesions that fractured and femora with metastatic lesions that did not fracture nor underwent prophylactic fixation. We found no difference between these two groups when analyzing all tissue trabecular and cortical bone, and tumor tissue. However, when including cortical bone only, we found that the femora that fractured had lower cumulative Hounsfield units as compared to those that did not fracture. The diagnostic accuracy of this method was expressed by an area under the curve (AUC) of 0.69; a threshold of <87% resulted in a sensitivity of 89% and a specificity of 48%. Our findings can be useful for predicting the development of a pathological fracture

234 Chapter 12 PART II: METASTATIC HUMERAL LESIONS This part starts with an overview of the current literature (systematic review) on outcomes after surgical treatment of metastatic humeral fractures (Chapter 6). Breast (30%), multiple myeloma (15%), lung (15%), and renal cell (13%) are the most common metastatic humeral lesions. The average postoperative survival of the patients was poor and varied from 4 to 23 months; variation in baseline (e.g. primary tumor type) explains the substantial variation in average survival. We found that among 23 studies including 909 patients overall reoperation rates were 4.4% after intramedullary nailing (range: 0 to 10%), 9.3% after plate-screw fixation (range: 5 to 14%), 2.5% after endoprosthetic reconstruction (range: 0 to 6%), and 15% after diaphysis prosthesis (range: 14 to 16%). Although indications for these surgical strategies vary and preclude comparison, the presented findings can be used to inform patients and their caregivers. In addition, we found that the quality of reporting was poor, definitions of systemic complications varied or were not described, and patient reported outcome data were scarce. Subsequently, we reviewed reoperations and systemic complications after surgical treatment of 295 metastatic humeral lesions from two large affiliated orthopaedic oncology centers (Chapter 7). The median postoperative survival was 11 months. We found that the reoperation rates were 6.7% after intramedullary nailing, 10% after plate-screw fixation, and 11% after endoprosthetic reconstruction. Deep infection, nonunion, peri-implant fracture, and tumor progression were the most common reasons for reoperation. No patient or tumor characteristics were associated with the risk of undergoing reoperation; however, we did find that the proportion of patients who underwent reoperation increased considerably with longer survival from 2.6% at 1 month up to 19% at 2 years. In addition, we found a postoperative systemic complication rate of 5.8%, and demonstrated that a worse modified Bauer score (an indicator of survival prognosis that combines cancer type and degree of dissemination) was associated with a higher risk of developing systemic complications. Finally, we surveyed 161 orthopaedic surgeons and assessed how they would treat metastatic humeral lesions in a cross-sectional survey study (Chapter 8). We created 24 fictional case scenarios by combining: tumor type, life expectancy, fracture type, and anatomical location of the lesion. For every case, all 161 participants (78 orthopaedic oncologists, 83 trauma / general orthopaedic surgeons) answered the question What treatment would you recommend?. Orthopaedic oncologists were more likely to recommend endoprosthetic reconstruction and plate-screw fixation and less likely to recommend intramedullary nailing compared to other subspecialty surgeons. Recommendation for nonoperative management did not differ. In addition, we found that recommendation for specific treatments differed based on tumor type, life expectancy, and anatomical location, but not fracture type. In some cases, surgeons agreed about the treatment strategy, while in other cases, there was substantial variability. 232

235 Summary PART III: SURVIVAL This part starts with a comparison of resection strategies for bone metastases from renal cell carcinoma (Chapter 9). It remains unclear how different types of tumor resection affect oncological outcome survival and tumor recurrence. We assessed differences in local tumor recurrence/progression, reoperation, and survival between three different surgical strategies: metastasectomy, intralesional curettage, and stabilization only in this retrospective study. Secondarily, we assessed whether the margin status affected outcomes. We included 183 patients from two affiliated orthopaedic oncology centers: 48% underwent metastasectomy (margins: 72% negative; 23% positive; 5% unclear), 30% intralesional curettage (margins: 100% positive), and 22% stabilization only (margins: 100% positive). The local recurrence/progression rate differed between surgical strategies and was highest after stabilization only (39%), followed by intralesional curettage (22%), and metastasectomy (12%). However, reoperation rates were comparable between surgical strategies. Survival was best in patients who underwent metastasectomy; however, this difference did not hold when stratifying patients based on metastatic load. Local recurrence/progression rate was lower in patients who had a negative margin (5%) compared to those who had a positive margin (26%). However, we found no difference in overall reoperation rate based on margin status. Survival was best in patients who had negative margins. The next chapter (Chapter 10) aims to develop an algorithm to estimate survival in patients with long bone metastases. We assessed what factors were independently associated with decreased survival, and used these risk factors to build three algorithms: a classic scoring system, a nomogram, and a boosting algorithm. After construction of each algorithm, we compared their accuracies at predicting 30-, 90-, and 365-day survival. This retrospective study included 927 patients who underwent surgery for a metastatic long bone lesion. We found that the following factors are independently associated with decreased survival: older age (hazard ratio [HR]: 1.02), additional comorbidity (HR: 1.2), body mass index less than 18.5 kg/m 2 (HR: 2.0), tumor type with poor prognosis (HR: 1.8), presence of multiple bone metastases (HR: 1.3), presence of visceral metastases (HR: 1.6), and low hemoglobin level (HR: 0.91). Survival estimates by the nomogram were moderately accurate for predicting 30-day (area under the curve [AUC], 0.72), 90-day (AUC, 0.75), and 365-day (AUC, 0.73) survival, and remained stable after correcting for optimism through fivefold cross validation. Boosting algorithms were better predictors of survival on the training datasets, but decreased to a performance level comparable to that of the nomogram when applied on testing datasets for all three timepoints. Accuracy of the classic scoring system was lowest for all prediction periods. Comorbidity status and body mass index are newly identified risk factors associated with survival, and should be accounted for when estimating survival. We recommend use of the nomogram due to its simplicity

236 Chapter 12 Finally, we studied the impact of perioperative blood transfusion on survival in patients who undergo surgery for long bone metastases (Chapter 11). Several studies demonstrate that perioperative allogeneic blood transfusions increase the risk of tumor recurrence and decrease survival after surgery for primary cancers. It is unclear if this association exists in patients who undergo surgery for long bone metastases. We included 789 patients who underwent surgery for a long bone metastatic lesion in this retrospective study. We assessed exposure versus no exposure as well as a dose-response relationship per unit of blood transfused and its association to survival while controlling for possible confounding variables. Exposure to blood transfusion (HR: 1.06) was not associated with worse survival; however, we did find a dose-response relationship with an increased risk of death per unit of blood transfused (HR: 1.07). Our sample size might have been insufficient to demonstrate an effect of exposure versus no exposure. Perioperative blood transfusion in patients who undergo surgery for long bone metastases might negatively influence survival, although the effect is small. 234

237 Summary SUMMARIZED ANSWERS TO PRIMARY STUDY QUESTIONS Part I: Metastatic Femoral Lesions Is there a difference in outcome physical function, reoperations, and complications between endoprosthetic reconstruction, intramedullary nailing, and open reduction internal fixation for proximal femoral metastasis? Overall reoperation rates are comparable for endoprosthetic reconstruction and intramedullary nailing, but is higher after open reduction internal fixation (Chapter 3). Fixation failure leading to reoperation is more common after open reduction internal fixation and intramedullary nailing, while deep infection leading to reoperation occurs more often after endoprosthetic reconstruction (Chapter 2 & 3). Systemic complication rates do not differ between surgical strategies (Chapter 2). All three surgical strategies result in reasonable function on average. However, wide ranges indicate that both poor and good functional levels are obtained. In addition, functional outcome was only scarcely reported and difference in their use preclude pooling of results and comparison between surgical strategies (Chapter 3). What questionnaire is most useful for measurement of physical function in patients with lower extremity bone metastasis? The PROMIS Physical Function Cancer questionnaire is the most useful questionnaire due to its reliability over a wide range of ability levels, validity, brevity, and good coverage (Chapter 4). Can a CT-scan based algorithm predict occurrence of a pathological fracture through a metastatic femoral lesion? An attenuation coefficients (Hounsfield units) based algorithm can predict occurrence of a pathological fracture when focusing on cortical bone. However, the accuracy of this method is moderate and comparable to radiographic measurements and symptoms (Chapter 5)

238 Chapter 12 Part II: Metastatic Humeral Lesions What outcome physical function, reoperations, and complications can be expected after surgical treatment of humeral metastasis? In our systematic review, reoperation rates were 0 to 10% after intramedullary nailing, 5 to 14% after plate-screw fixation, 0 to 6% after endoprosthetic reconstruction, and 14 to 16% after diaphysis prosthesis (Chapter 6). In our cohort study, we found reoperation rates of 6.7% after intramedullary nailing, 10% after plate-screw fixation, and 11% after endoprosthetic reconstruction. Deep infection, nonunion, peri-implant fracture, and tumor progression are the most common reasons for reoperation (Chapter 7). In our systematic review, definitions of systemic complications varied substantially among studies or were poorly reported (Chapter 6). In our cohort study, we found a systemic complication rate of 5.8% (Chapter 7). In our systematic review, we found that patient reported outcome measures such as physical function have not been reported (Chapter 6). What factors are associated with reoperations and systemic complications after surgical treatment of humeral metastasis? No factors are independently associated with reoperation. However, the probability of undergoing reoperation increases considerably with longer survival (Chapter 7). A prognostic survival score (the modified Bauer score, including tumor type and degree of tumor dissemination) is an independent predictor of developing postoperative systemic complications (Chapter 7). Is there a difference in surgical decision making for humeral metastasis based on physician, patient, or tumor characteristics? Orthopaedic oncologists are more likely to recommend endoprosthetic reconstruction and plate-screw fixation, and less likely to recommend intramedullary nailing compared to other subspecialty surgeons (Chapter 8). Recommendation for specific implants differ based on tumor type, life expectancy, and anatomical location, but does not differ based on fracture type (Chapter 8). 236

239 Summary Part III: Survival Is there a difference in local tumor recurrence, reoperation, and survival between metastasectomy, intralesional resection, and stabilization only for renal cell metastasis? The local recurrence rate differs between types of treatment and is highest after stabilization only, followed by intralesional curettage, and lowest after metastasectomy (Chapter 9). Reoperation rates do not differ between stabilization only, intralesional curettage, and metastasectomy (Chapter 9). Survival is best in patients who underwent metastasectomy, when compared to stabilization only and intralesional curettage; however, this difference does not hold when stratifying patients based on metastatic load (Chapter 9). What factors are associated with worse survival among patients who underwent surgery for long bone metastases? Older age, comorbidity, low body mass index, tumor type with poor prognosis, presence of multiple bone metastases, presence of visceral metastases, and low hemoglobin level are independently associated with worse survival (Chapter 10). What type of algorithm is most accurate for predicting survival probability after surgery for long bone metastases? The nomogram is moderately accurate for predicting 30-day, 90-day, and 365-day survival (Chapter 10). Are allogeneic blood transfusions associated with worse survival after surgery for long bone metastases? Exposure to blood transfusion is not associated with worse survival; however, we did find a dose-response relationship with an increased risk of death per unit of blood transfused (Chapter 11)

240

241 CHAPTER 13 General Discussion 13

242 Chapter 13 The introduction of this PhD thesis describes the unique features of managing bone lesions secondary to metastasis. Numerous aspects that need to be considered when deciding upon a surgical strategy have been addressed in the previous chapters. Specific study aims concerning surgical management of bone metastasis were divided into three main parts; metastatic femoral lesions, metastatic humeral lesions, and survival. PART I: METASTATIC FEMORAL LESIONS The proximal femur is the most commonly affected long bone by metastasis. Breast, lung, prostate, and renal cell cancer are the most common bone metastases in the femur, and together account for 68% of all femoral metastases. 1 General consensus is that surgical treatment is indicated in case of a pathological fracture or when a fracture is impending and the estimated life expectancy exceeds the recovery time. Several surgical options to treat metastatic lesions of the proximal femur exist and are often categorized into endoprosthetic reconstruction, intramedullary nailing, and open reduction internal fixation with plate and screws. There is substantial controversy among orthopaedic oncologists about when which surgical strategy is recommended. 2 This is attributed to the low level of evidence with inherent biases and potential for confounding, small sample sizes lacking statistical power, and lack of studies measuring patient reported outcomes. We found in our relatively large (n = 417) retrospective cohort of patients that underwent surgery for proximal femoral metastasis (excluding the femoral head) no difference in overall reoperation rate between endoprosthetic reconstruction, intramedullary nailing, and open reduction internal fixation (Chapter 2). However, when focusing on specific reasons for reoperation, we found that fixation failure most commonly occurred after open reduction internal fixation (13% versus 3.0% after intramedullary nailing versus 0% after endoprosthetic reconstruction), whereas deep infection was most common after endoprosthetic reconstruction (8.6% versus 2.0% after intramedullary nailing versus 0% after open reduction and internal fixation). Systemic complication rate and 30-day mortality did not differ between surgical strategies. Intraoperative blood loss was almost twice as high and anesthesia time was about 40 minutes longer for endoprosthetic reconstruction when compared with intramedullary nailing and open reduction internal fixation. Our additional study finding that a worse prognostic survival score (i.e. a lower modified Bauer score [a score that combines cancer type and degree of dissemination]) is associated with a higher risk of systemic complications could be used to anticipate postoperative problems, and identify patients that might benefit from preoperative optimization. The overall reoperation rate in our study (7.2%) is consistent with reoperation rates reported in the two largest previous studies that directly compared all three surgical strategies by Steensma et al. (6.4%) 3 and Wedin et al. (10%) 4. Our results are also consistent with these two 240

243 Discussion previous studies in terms of durability of endoprosthetic reconstructions (i.e. low fixation failure rates): Steensma et al. 3 found that 0.5% of endoprosthetic reconstructions required implant exchange after fixation failure compared to 6.1% after intramedullary nailing and 42% after open reduction internal fixation. Wedin et al. 4 demonstrated an 8.3% fixation failure rate requiring reoperation after endoprosthetic reconstruction compared with 13% after intramedullary nailing and 25% after the use of a dynamic hip screw. This cohort study is followed by our systematic review of the literature (Chapter 3) in which we demonstrated that the overall reoperation rate was highest after open reduction internal fixation (14%), and comparable for intramedullary nailing (4.2%) and endoprosthetic reconstruction (5.2%). Separating reasons for reoperation, we again identified that the fixation failure rate was highest after open reduction internal fixation (10%), followed by intramedullary nailing (2.8%), and endoprosthetic reconstruction (0.4%); whereas, deep infection rate was highest after endoprosthetic reconstruction (0.68%) as compared to intramedullary nailing (0.04%) and open reduction internal fixation (0.00%). Systemic complications were inconsistently reported among studies and therefore not comparable. Patient reported outcomes such as function and pain were scarcely reported according to our systematic review and we therefore could not draw any conclusions on these outcomes. Findings from our retrospective cohort study and systematic review suggest that extramedullary fixation in the form of open reduction internal fixation with plate and/or screws or a dynamic hip screw is an inferior option for treatment of metastatic proximal femoral lesions unless the medullary canal is not accessible (e.g. due to already present implants or osteoblastic intramedullary tumor depositions). Endoprosthetic reconstruction and intramedullary nailing seem to be comparable in terms of overall reoperation rate; however endoprosthetic reconstructions more commonly develop deep infections (often an early complication), whereas intramedullary nail fixation seems to fail more often over time (often a late complication). These implant-specific reasons for reoperation as well as their timing should be considered when deciding between endoprosthetic reconstruction and intramedullary nailing. We therefore argue that life expectancy is an important factor for establishing which surgical strategy is most optimal for proximal femoral metastases. Future study on the impact of these local complications (deep infection and fixation failure) on recovery and functional outcome would be valuable. 5 The difference in surgical time and blood loss, and no difference in systemic complications and 30-day mortality does, in general, not influence our recommendation for a surgical strategy. Our findings can be used to inform future patients and caregivers. The most important limitations of the systematic literature review are that all studies were retrospective cohort studies and indications for surgical strategies as well as definition of outcomes to be reported varied among studies or were not clearly described which could have lead to bias. Future studies need to: use standardized methods of reporting by adhering to guidelines to improve study quality (e.g. STROBE guidelines) 6, report definitions of complications including timing and treatment consequences to allow comparison

244 Chapter 13 across studies 7, and incorporate patient reported outcome measures to better establish the impact of treatment modalities on patients function and quality of life. 8,9 Furthermore, multi-institutional collaboration by sharing databases are needed to obtain large patient samples as adverse outcomes such as reoperations are relatively uncommon and analyses often lack sufficient power to detect differences. A multi-center prospective randomized controlled trial comparing long-stem cemented hemiarthroplasty with intramedullary nailing for proximal femoral metastases is ongoing and aims to further determine which treatment is optimal in terms of quality of life and physical function. 10 In addition, it would be valuable to understand the rapidity of postoperative recovery in terms of improvement of quality of life and physical function over time in order to determine the minimal estimated life expectancy that justifies surgical intervention. Based on our retrospective cohort study and systematic review, a treatment scheme to guide surgical decision making for proximal femoral metastasis is proposed (Figure 1). Many studies underline the importance of incorporating functional outcome measures; however, only few attempted to do so as can be concluded from our systematic reviews. 1,11 Some studies use the clinician reported Musculoskeletal Tumor Society (MSTS) score six items evaluating function through the eyes of the clinician ; however, doctors consistently overestimate a patients function and mental health. This discrepancy can lead to: less satisfaction among patients as their expectations are not met or their recovery is slower than expected, patients might feel misunderstood or unheard by their physician which could in turn lead to less compliance with treatment recommendations, distorted study results when physicians evaluate the outcome of certain treatments towards they are biased, an attitude of complacency and inertia among clinicians that precludes further improvement, and finally third-party payers that may use reported (overestimated) outcomes to dissuade further innovation and research Use of clinician reported outcomes should therefore be discouraged and obtaining patient reported outcome measures is important We assessed which patient reported outcome measure is most useful efficient, reliable, and covers the full spectrum of functional levels to measure physical function among patients with lower extremity metastases (Chapter 4). We approached 115 patients with lower extremity metastases from two orthopaedic oncology clinics to participate in this study. One hundred patients completed five questionnaires and we found that the PROMIS Physical Function, the Toronto Extremity Salvage Score, and the Lower Extremity Function Score had good coverage and high reliability over a wide range of functioning; however, the PROMIS Physical Function questionnaire is quickest to complete and therefore deemed the most useful. The high rate of participation (100 out of 115 invited patients) indicates that patients with bone metastases are open for participation in patient reported outcome studies despite their relatively poor prognosis. PROMIS questionnaires (Patient-Reported Outcomes Measurement Information System) are developed standardized item banks to measure physical, mental, and social health and is funded by the National Institute of 242

245 Discussion Health. PROMIS questionnaires allow for Computer Adaptive Testing (CAT) a dynamic selection of items wherein the response to each item guides the system s choice of the next item resulting in a tailored series of questionnaires reducing questionnaire burden, while maintaining precision and limiting floor and ceiling effect. Most PROMIS questionnaires are currently only available in English language, but efforts to translate and standardize item banks for use in other countries are undertaken. The accuracy of establishing if a fracture is impending i.e. predicting occurrence of a pathological fracture using historical radiographic measures (e.g. >50% circumferential cortical destruction, defect size >30mm), symptoms (e.g. pain on weight bearing), or a combination thereof (e.g. Mirels 19 score) are not sufficiently accurate New techniques such as finite element analysis and CT-based rigidity analysis have been developed to better predict fracture occurrence; however, these methods are complex and relatively labor intensive We found that our newly developed algorithm using attenuation coefficients (Hounsfield units) in clinically obtained CT-scans; a simplification of the CT-based rigidity analysis as described by Snyder et al can predict occurrence of a pathological fracture when focusing on cortical bone only (Chapter 5). Our CT-based calculations did not differ when including all tissue (i.e. trabecular bone, cortical bone, and tumor tissue). This could be explained by the fact that cortical bone contributes more to the load bearing capacity than trabecular bone and tumor tissue. 29 However, the accuracy of our method is only moderate and comparable to radiographic measurements and symptoms. A more controlled study design (prospective followup of patients) with standardized CT-scan protocols might improve the accuracy of our method and merits further study. On the other hand, accurately predicting pathological fracture risk based on a CT-scan only might simply be impossible due to the multitude of other factors that might contribute to pathological fracture occurrence, such as: tumor growth/activity, response to systemic therapy or radiation therapy, activity of the patient, risk of falling, etc. Future studies might test the influence of such factors and study if a combination of lesion characteristics (based on CT imaging) and clinical factors might improve prediction accuracy. More precise determination of fracture risk would reduce the potential for over- and undertreatment of non-fractured metastatic lesions. PART II: METASTATIC HUMERAL LESIONS 13 After the axial skeleton and proximal femur, the humerus is the third most commonly affected bone by metastatic cancer. Breast, myeloma, lung, and renal cell cancer are the most common bone metastases in the humerus, and together account for 73% of all humeral metastases. 11 As with femoral metastases, surgical fixation is indicated in case of a pathological fracture or when a fracture seems to be impending; However, the threshold 243

246 Chapter 13 Head/Neck EPR* (Hemi-A/MTP) Proximal Femoral Metastasis Life expectancy <3 months IMN Impending pathological fracture IMN Inter-/Subtrochanteric Life expectancy 3-12 months Adequate bone stock IMN Pathological fracture Inadequate bone stock EPR (MTP/Hemi-A) Life expectancy >12 months EPR (MTP/Hemi-A) Figure 1: Surgical treatment algorithm for metastatic lesions of the proximal femur. EPR = Endoprosthetic reconstruction, IMN = Intramedullary nailing, MTP = Modular Tumor Prosthesis, Hemi-A = Hemi-arthroplasty. *For head/neck lesions and lesions that minimally affect the trochanteric region, a cemented hemi-arthroplasty is the first choice; however, in case of tumor involvement of the trochanteric region, a modular tumor prosthesis is most likely needed to reconstruct the defect. Bone stock: adequate bone stock means a small solitary lesion in otherwise healthy bone, inadequate bone stock means multiple lesions and/or diffusely weakened bone. Extramedullary fixation in the form of open reduction internal fixation with plate and/or screws or a dynamic hip screw is generally not recommended for metastatic proximal femoral lesions unless the medullary canal is not accessible (e.g. due to already present implants or osteoblastic intramedullary tumor depositions). The entire bone needs to be imaged at least in two directions preoperatively to evaluate for additional metastatic lesions. The implant needs to span at least all metastatic lesions. This algorithm does not pertain to concomitant metastatic involvement of the acetabular region, it also does not include nonoperative management of for example metastatic lesions of the proximal femoral region that are not at risk of fracture or patients with a life expectancy considered to poor to undergo surgery. for surgery of a non-fractured metastatic lesion in a non-load bearing bone (such as the humerus) is higher as compared to load bearing bones (e.g. the proximal femur). Surgical options for treating metastasis to the humerus can be divided into: intramedullary nailing, open reduction internal fixation with plate and screws, and endoprosthetic reconstruction of the proximal, shaft (diaphyseal/intercalary), distal (total elbow), or complete humerus. Our systematic review narratively reports outcomes after these treatment modalities without performing a meta-analysis as indications for surgical strategies vary widely, definition of outcomes vary substantially, and studies are heterogeneous (Chapter 6). None of the included studies described patient reported outcomes, only five studies demonstrated functional outcome as determined by the clinician. We found that intra- 244

247 Discussion medullary nailing, plate-screw fixation, and proximal endoprosthetic reconstruction seem to result in reasonable to good function on average. Overall reoperation rates were 4.4% after intramedullary nailing (range: 0 to 10%), 9.3% after plate-screw fixation (range: 5 to 14%), 2.5% after proximal endoprosthetic reconstruction (range: 0 to 6%), and 15% after diaphysis prosthesis (range: 14 to 16%). Although indications for these treatment modalities vary, intramedullary nailing, plate-screw fixation, and diaphyseal prosthesis were all predominantly used for diaphyseal pathological fractures; the diaphysis prosthesis seems to have the highest risk of reoperation due to failure of fixation and peri-implant fracture. This finding might be affected by confounding (e.g. diaphysis prosthesis might be used in larger defects, and/or in patients with better prognosis). 30,31 Intercalary diaphyseal prosthesis should probably be reserved for substantial diaphyseal segmental bone destruction that cannot be easily spanned with a plate-screw construct or intramedullary nail combined with cement augmentation or structural bone grafting in a patient with an otherwise relatively good life expectancy (e.g. in case of a solitary metastasis). 32,33 Methodological quality of future studies should be improved by using standardized methods of reporting (e.g. STROBE guidelines) 6 and by reporting definitions of complications including timing and treatment consequences. 7 In addition, it would be valuable to study patient reported outcome measures to better establish the impact of surgical treatment on patients function and quality of life. Furthermore, multi-institutional collaborations by sharing databases are needed to obtain sufficiently large samples to compare adverse outcomes such as reoperation in relatively homogenous populations (i.e. per anatomical area). To better understand why implants for metastatic humeral lesions undergo reoperation and why patients develop postoperative complications, we studied a relatively large (n = 295) retrospective cohort of patients that underwent intramedullary nailing, open reduction internal fixation with plate and screws, and proximal endoprosthetic reconstruction of a metastatic humeral lesion (Chapter 7). We found that 8.5% underwent reoperation, predominantly for deep infection (2.0%), nonunion (2.0%), tumor recurrence (1.6%), and peri-implant fracture (1.4%). None of the clinical, tumor, or treatment characteristics were associated with risk of reoperation, except for duration of survival: the risk of reoperation increases from 2.6% at one month up to 19% at two years. This emphasizes the importance of incorporating life expectancy in surgical decision making and underlines the need for durable constructs in patients who are expected to live longer than one year. Based on our systematic review and this retrospective cohort study we cannot determine which implant is most durable; However, complete resection of tumor and reconstruction, or stable fixation combined with radiation of a radiosensitive tumor is probably the most durable in terms of fixation failure. Yet, the evidence supporting radiation therapy to improve local tumor control after surgical fixation for long bone metastasis is limited. 34 Our finding that none of the other factors were associated with reoperation might be explained by the variation in reasons for reoperation (e.g. infection, peri-implant fracture, nonunion

248 Chapter 13 etc.) which might have other underlying causes and corresponding risk factors. Our cohort was not large enough to assess risk factors for specific local complications leading to reoperation. Systemic complications occurred in 5.8% of the patients and the modified Bauer score, as with our proximal femoral metastasis study, is the only factor associated with developing postoperative complications. This score facilitates risk stratification and might help identify patients at unacceptable high risk for postoperative complications and patients that might benefit from preoperative optimization. Evidence supports the use of intramedullary nailing, open reduction internal fixation, and endoprosthetic reconstructions for metastatic humeral lesions. 11 We therefore explored when surgeons (n = 161) used which implant in a cross sectional survey study to improve our understanding of surgical decision making and explore areas of controversy (Chapter 8). Orthopaedic oncologists are more likely to choose plate-screw fixation or endoprosthetic reconstruction, and less likely to choose intramedullary nailing when compared to orthopaedic surgeons not daily involved in skeletal oncological surgery (predominantly trauma surgeons). Recommendation for nonoperative management does not differ between groups. This difference in preference for implants probably has it roots in the difference in training as orthopaedic oncologists are trained in treatment of bone and soft tissue neoplasms, developmental dysplasias, tumor-like conditions, and major skeletal defects. 35 As such, they might be more comfortable with creating bony defects by resection and subsequent reconstruction with a prosthesis or plate-screw fixation with cement augmentation and therefore base their decision for an implant on its combination with adjuvant treatment. It is therefore valuable to discuss metastatic humeral fracture cases preoperatively in a team with an orthopaedic oncologist, medical oncologist, and radiation oncologist. In addition to subspecialty training, we found that: tumor type, estimated life expectancy and anatomical location influenced the recommendation for treatment, while fracture type did not. Nonoperative management was recommended for slightly over one-third of patients with an estimated life expectancy less than 3 months, while surgical management was recommended for almost all patients with a life expectancy over 3 months. Accurate prognostication is important and requires further study as life expectancy is as our findings demonstrate an important factor in deciding whether to operate or not. The recommendation for a specific implant was strongly influenced by anatomical location of the lesion: intramedullary nailing (68%) was predominantly recommended for diaphyseal lesions, followed by plate-screw fixation (15%); while all three techniques were recommended for proximal humeral lesions (intramedullary nailing: 38%, endoprosthetic reconstruction: 14%, and plate-screw fixation: 24%). Future studies that directly compare these techniques preferably in an experimental study design per anatomical area (Diaphyseal lesions: intramedullary nail versus plate-screw construct. Proximal lesions: endoprosthesis versus plate-screw construct versus intramedullary nail) are needed to elucidate which implant is most optimal in which situation. 246

249 Discussion Based on these studies, a treatment scheme to guide surgical decision making for humeral metastasis is proposed; however, scientific equipoise results in a large variety of valid options (Figure 2). Life expectancy <3 months IMN EPR* PSF Proximal Life expectancy 3-12 months Adequate bone stock IMN PSF EPR* Inadequate bone stock EPR PSF Humeral Metastasis Life expectancy >12 months EPR Life expectancy <3 months IMN Shaft Life expectancy 3-12 months Adequate bone stock IMN PSF Life expectancy >12 months Inadequate bone stock DP PSF IMN Distal PSF EPR* Figure 2: Surgical treatment algorithm for metastatic lesions of the humerus. EPR = Endoprosthetic reconstruction (proximal or distal), IMN = Intramedullary nailing, PSF = Plate-screw fixation, DP = diaphysis prosthesis. *Endoprosthetic reconstruction is recommended in case of articular involvement. Bone stock: adequate bone stock means a small solitary lesion in otherwise healthy bone, inadequate bone stock means multiple lesions and/or diffusely weakened bone. The entire bone needs to be imaged at least in two directions preoperatively to evaluate for additional metastatic lesions. The implant needs to span at least all metastatic lesions. This algorithm does not include nonoperative management of for example metastatic lesions that are not at risk of fracture or patients with a life expectancy considered to poor to undergo surgery

250 Chapter 13 PART III: SURVIVAL Survival of patients with bone metastases that require surgical treatment is generally poor as more than half of the patients decease within one year. 1,11 Life expectancy is considered a key parameter in deciding upon treatment as surgeons need to balance life expectancy against rehabilitation time, longevity of different implants, and risk of tumor recurrence. Treatment of bone metastases is often considered palliative, aiming to providing pain relief and optimize quality of life. However, some previous studies suggest that surgical management of bone metastases could improve survival in specific populations Patients with bone metastasis from renal cell carcinoma have a relatively favorable prognosis and some studies suggest improved survival after metastasectomy (i.e. en bloc resection) versus intralesional treatment in patients with a solitary metastasis We therefore assessed difference in local tumor recurrence/progression, reoperation, and survival based on type of resection and margin status in patients (n = 183) undergoing surgery for bone metastasis from renal cell carcinoma (Chapter 9). Our retrospective study demonstrates that the risk of developing local recurrence is higher among patients who undergo intralesional curettage and stabilization only as compared to patients who undergo metastasectomy; and that negative margins results in the lowest risk of local recurrence. Two previous studies support this finding of higher local recurrence rate after intralesional procedures compared with wide resection. 39,40 Residual tumor after intralesional treatment most likely explains this higher risk of local recurrence as compared to metastasectomy (with negative margins). The overall risk of reoperation does not differ between surgical strategies nor does it differ based on margin status. This is explained by the finding that about half of the reoperations in the metastasectomy group were for deep infection, and the other half were for tumor recurrence; while the majority of reoperations in the intralesional treatment group and the stabilization only group were for tumor recurrence. Survival in our study was better after metastasectomy especially when negative margins were obtained, but these findings were confounded by the difference in metastatic load and our sample size was to small to demonstrate a difference in survival among patients with solitary bone metastasis only. Three previous studies addressed differences in survival based on resection type among patients with solitary renal cell bone metastasis and all demonstrated in line with our study a trend towards better survival in patients who undergo en bloc resection versus intralesional treatment, but small sample sizes fail to achieve sufficient power to demonstrate significance in all three A larger sample size, perhaps by combining the data from multiple institutions, is needed to determine the benefit in terms of survival for metastasectomy in patients with solitary or oligometastatic bone disease. In conclusion, based on our findings, patients with a relatively good life expectancy more than one year should undergo metastasectomy with negative margins to improve local tumor control in the long term at a non-significant additional risk for reoperation, and with a potential impact on survival. 248

251 Discussion Life expectancy for patients with bone metastasis is often estimated using prognostication algorithms. Numerous algorithms exist and can be categorized into: simple classic scoring systems, nomograms, and computer or application-based machine learning algorithms However, accuracy of currently existing algorithms in predicting survival is moderate and any improvement in performance would be clinically relevant. We assessed what factors were independently associated with survival among patients (n = 927) with long bone metastases undergoing surgery (Chapter 10). We identified two new factors associated with survival: comorbidity status and body mass index. These newly identified prognostic factors should be tested in future studies and incorporated in prognostication algorithms. Many other risk factors have been identified previously and most are confirmed in multiple subsequent studies, including: primary tumor type, performance status, presence of visceral and/or brain metastases, number of metastases, previous systemic therapy, age, hemoglobin level, sex, presence of a pathological fracture, the surgeons estimate of survival, presence of pain More recently, emphasis has been put on further tumor characterization (e.g. specific gene mutations such as epidermal growth factor receptor for lung cancer) and biochemical variables (e.g. neutrophil count, alkaline phosphatase, C- reactive protein, lactate dehydrogenase, albumin, calcium, platelet level, bilirubin) in order to identify more prognostic factors Incorporating a multitude of prognostic factors in a survival algorithm increases model performance (i.e. more information leads to more precise and more accurate life expectancy estimates); however, this comes at the cost of increased model complexity and more burden to the person completing the algorithm due to the number of variables that need to be known and entered. Balancing model performance and simplicity is key for clinical use. Web-based applications such as pathfx. org 51 and optimal 52 help estimate life expectancy, but still need to be completed manually. Future lies in electronic medical records perhaps connected on national or international level that automatically provide survival estimates based on diagnostic and billing coding, tumor characteristics, biochemical variables, and natural language processing of physician notes. Ideally, these algorithms would continuously update through machine learning. 53 The latter is not only important for identifying subtle prognostic factors, but also to stay upto-date with regards to the rapidly developing changes in systemic oncological therapy. Until then, our developed nomogram and web-based applications balance simplicity and stable model performance. Comparison and adjustment of prognostication tools would ideally be done in a new patient population to assess external validity and to compare accuracy of these algorithms. Patients with bone metastasis are often chronically anemic. Blood loss during surgery for bone metastasis can worsen this anemia and often leads to blood transfusion. As the relationship between perioperative blood transfusions and adverse oncological outcomes (survival and tumor recurrence) has been demonstrated in resection of primary malignancies, the question arises how to manage perioperative anemia in patients undergoing surgery for

252 Chapter 13 bone metastasis. We therefore studied the impact of perioperative blood transfusions on survival among patients (n = 789) who undergo surgery for long bone metastases (Chapter 11). We found that perioperative blood transfusion in patients who undergo surgery for long bone metastases might negatively influence survival, although the effect is small. Other studies on surgery for spine metastasis demonstrated contradicting results; one found increased 12-months survival among patients who received 1 to 2 units of blood, while another study demonstrated no effect of exposure to blood transfusion nor a dose-response relationship. 54,55 We subsequently studied the effect of perioperative blood transfusion on infection within 90 days after surgical treatment for bone metastases. 54 We found no effect of exposure to blood transfusion, nor did we find a dose-response relationship (per unit of blood transfused) on the risk of developing a postoperative infection. 54 The negative impact of perioperative blood transfusion on postoperative complications, overall survival, and tumor recurrence has been demonstrated clearly in patients who undergo curative resection for colorectal cancer, but also in patients who undergo resection of colorectal liver metastases Given these findings, one needs to balance the risk of acute anemia versus the potential immunomodulation and its possible consequences imposed by allogeneic blood transfusions. Recommendations regarding cut-off points for transfusion cannot be given based on these studies. Future study could consider applying patient reported outcome measures such as quality of life to establish the impact of blood transfusions. CONCLUSIONS For patients with proximal femoral metastasis, intramedullary nailing and endoprosthetic reconstruction are preferred over open reduction internal fixation (Chapter 2 & 3). Endoprosthetic reconstruction and intramedullary nailing for proximal femoral metastasis seem to be comparable in terms of overall risk for reoperation (Chapter 2 & 3). Implant-specific reasons for reoperation and their timing should be considered when deciding between endoprosthetic reconstruction and intramedullary nailing for proximal femoral metastasis (Chapter 2 & 3). The PROMIS Physical Function Cancer questionnaire is recommended for assessment of functional outcome in patients with lower extremity bone metastasis (Chapter 4). A clinical CT scan based algorithm can be useful for predicting occurrence of a pathological fracture through a femoral metastasis (Chapter 5). Endoprosthetic reconstruction, intramedullary nailing, and open reduction internal fixation seem to be reasonable options for treatment of humeral metastasis. Variation in indications and reporting of outcomes, and small and heterogeneous patient samples preclude direct comparison of surgical strategies (Chapter 6). 250

253 Discussion The probability of undergoing reoperation increases considerably with longer survival in patients with humeral metastasis (Chapter 7). A prognostic survival score (modified Bauer score) could be useful for risk stratification of postoperative systemic complications in patients with femoral and humeral metastasis (Chapter 2 & 7). Subspecialty training of the surgeon, as well as tumor type, life expectancy, and anatomical location of a metastatic humeral lesions are important drivers of treatment variation (Chapter 8). Metastasectomy with negative margins for renal cell bone metastasis improves local tumor control at a comparable risk for reoperation, and with a potential impact on survival (Chapter 9). Comorbidity status and body mass index are two newly identified prognostic factors for survival in patients with long bone metastasis (Chapter 10). A nomogram is the preferred method for predicting survival probabilities in patients with long bone metastasis (Chapter 10). Perioperative allogeneic blood transfusions in patients with long bone metastasis are a potential modifiable risk factor as it might negatively affect survival; although the effect is small (Chapter 11). REFERENCES 1. Janssen SJ, Teunis T, Hornicek FJ, van Dijk CN, Bramer JA, Schwab JH. Outcome after fixation of metastatic proximal femoral fractures: A systematic review of 40 studies. J Surg Oncol. Sep 2016; 114(4): Steensma M, Healey JH. Trends in the surgical treatment of pathologic proximal femur fractures among Musculoskeletal Tumor Society members. Clin Orthop Relat Res. Jun 2013; 471(6): Steensma M, Boland PJ, Morris CD, Athanasian E, Healey JH. Endoprosthetic treatment is more durable for pathologic proximal femur fractures. Clin Orthop Relat Res. Mar 2012; 470(3): Wedin R, Bauer HC. Surgical treatment of skeletal metastatic lesions of the proximal femur: endoprosthesis or reconstruction nail? J Bone Joint Surg Br. Dec 2005; 87(12): Forsberg JA, Wedin R, Bauer H. Which implant is best after failed treatment for pathologic femur fractures? Clin Orthop Relat Res. Mar 2013; 471(3): von Elm E, Altman DG, Egger M, Pocock SJ, Gotzsche PC, Vandenbroucke JP. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. BMJ. Oct ; 335(7624): Clavien PA, Barkun J, de Oliveira ML, et al. The Clavien-Dindo classification of surgical complications: five-year experience. Ann Surg. Aug 2009; 250(2): van der Vliet QM, Paulino Pereira NR, Janssen SJ, et al. What Factors are Associated With Quality Of Life, Pain Interference, Anxiety, and Depression in Patients With Metastatic Bone Disease? Clin Orthop Relat Res. Feb 2017; 475(2):

254 Chapter Piccioli A. CORR Insights(R): What Factors are Associated With Quality Of Life, Pain Interference, Anxiety, and Depression in Patients With Metastatic Bone Disease? Clin Orthop Relat Res. Feb 2017; 475(2): Healey JH, Boland PJ. A Randomized Trial to Assess Patient Quality of Life and Function After Alternative Surgeries for Pathologic Fractures of the Femur. 2016; https: //clinicaltrials.gov/ct2/ show/nct Accessed 2/17/17, Janssen SJ, Teunis T, Hornicek FJ, Bramer JA, Schwab JH. Outcome of operative treatment of metastatic fractures of the humerus: a systematic review of twenty three clinical studies. Int Orthop. Apr 2015; 39(4): Rosenberger PH, Jokl P, Cameron A, Ickovics JR. Shared decision making, preoperative expectations, and postoperative reality: differences in physician and patient predictions and ratings of knee surgery outcomes. Arthroscopy. May 2005; 21(5): Francis V, Korsch BM, Morris MJ. Gaps in doctor-patient communication. Patients response to medical advice. N Engl J Med. Mar ; 280(10): Coran JJ, Koropeckyj-Cox T, Arnold CL. Are physicians and patients in agreement? Exploring dyadic concordance. Health Educ Behav. Oct 2013; 40(5): Janssen SJ, van Rein EA, Paulino Pereira NR, et al. The Discrepancy between Patient and Clinician Reported Function in Extremity Bone Metastases. Sarcoma. 2016; 2016: Cheng EY. Prospective quality of life research in bony metastatic disease. Clin Orthop Relat Res. Oct 2003(415 Suppl): S Nelson EC, Eftimovska E, Lind C, Hager A, Wasson JH, Lindblad S. Patient reported outcome measures in practice. BMJ. Feb ; 350: g Nelson E, Conger B, Douglass R, et al. Functional health status levels of primary care patients. JAMA. Jun ; 249(24): Mirels H. Metastatic disease in long bones. A proposed scoring system for diagnosing impending pathologic fractures. Clin Orthop Relat Res. Dec 1989(249): Dijkstra PD. Pathological fractures of long bones due to bone metastases: Orthopaedic Surgery, Erasmus Universiteit Rotterdam; Van der Linden YM, Dijkstra PD, Kroon HM, et al. Comparative analysis of risk factors for pathological fracture with femoral metastases. J Bone Joint Surg Br. May 2004; 86(4): Anez-Bustillos L, Derikx LC, Verdonschot N, et al. Finite element analysis and CT-based structural rigidity analysis to assess failure load in bones with simulated lytic defects. Bone. Jan 2014; 58: Derikx LC, van Aken JB, Janssen D, et al. The assessment of the risk of fracture in femora with metastatic lesions: comparing case-specific finite element analyses with predictions by clinical experts. J Bone Joint Surg Br. Aug 2012; 94(8): Leong NL, Anderson ME, Gebhardt MC, Snyder BD. Computed tomography-based structural analysis for predicting fracture risk in children with benign skeletal neoplasms: comparison of specificity with that of plain radiographs. J Bone Joint Surg Am. Aug ; 92(9): Nazarian A, Entezari V, Villa-Camacho JC, et al. Does CT-based Rigidity Analysis Influence Clinical Decision-making in Simulations of Metastatic Bone Disease? Clin Orthop Relat Res. May ; 474(3): Nazarian A, Entezari V, Zurakowski D, et al. Treatment Planning and Fracture Prediction in Patients with Skeletal Metastasis with CT-Based Rigidity Analysis. Clin Cancer Res. Feb ; 21(11):

255 Discussion 27. Snyder BD, Hauser-Kara DA, Hipp JA, Zurakowski D, Hecht AC, Gebhardt MC. Predicting fracture through benign skeletal lesions with quantitative computed tomography. J Bone Joint Surg Am. Jan 2006; 88(1): Spruijt S, van der Linden JC, Dijkstra PD, et al. Prediction of torsional failure in 22 cadaver femora with and without simulated subtrochanteric metastatic defects: a CT scan-based finite element analysis. Acta Orthop. Jun 2006; 77(3): Holzer G, von Skrbensky G, Holzer LA, Pichl W. Hip fractures and the contribution of cortical versus trabecular bone to femoral neck strength. J Bone Miner Res. Mar 2009; 24(3): Muller-Farber J, Muller KH. The treatment of metastatic humeral lesions with the diaphyseal prosthesis. Aktuelle Traumatol. 1997; 27: Schurmann M, Gradl G, Andress HJ, Kauschke T, Hertlein H, Lob G. Metastatic lesions of the humerus treated with the isoelastic diaphysis prosthesis. Clin Orthop Relat Res. 2000: Damron TA, Leerapun T, Hugate RR, Shives TC, Sim FH. Does the second-generation intercalary humeral spacer improve on the first? Clin Orthop Relat Res. Jun 2008; 466(6): Damron TA, Sim FH, Shives TC, An KN, Rock MG, Pritchard DJ. Intercalary spacers in the treatment of segmentally destructive diaphyseal humeral lesions in disseminated malignancies. Clin Orthop Relat Res. Mar 1996(324): Willeumier JJ, van der Linden YM, Dijkstra PD. Lack of clinical evidence for postoperative radiotherapy after surgical fixation of impending or actual pathologic fractures in the long bones in patients with cancer; a systematic review. Radiother Oncol. Oct 2016; 121(1): White J, Toy P, Gibbs P, Enneking W, Scarborough M. The current practice of orthopaedic oncology in North America. Clin Orthop Relat Res. Nov 2010; 468(11): Fuchs B, Trousdale RT, Rock MG. Solitary bony metastasis from renal cell carcinoma: significance of surgical treatment. Clin Orthop Relat Res. Feb 2005(431): Lin PP, Mirza AN, Lewis VO, et al. Patient survival after surgery for osseous metastases from renal cell carcinoma. J Bone Joint Surg Am. Aug 2007; 89(8): Ratasvuori M, Wedin R, Hansen BH, et al. Prognostic role of en-bloc resection and late onset of bone metastasis in patients with bone-seeking carcinomas of the kidney, breast, lung, and prostate: SSG study on 672 operated skeletal metastases. J Surg Oncol. Sep 2014; 110(4): Les KA, Nicholas RW, Rougraff B, et al. Local progression after operative treatment of metastatic kidney cancer. Clin Orthop Relat Res. Sep 2001(390): Evenski A, Ramasunder S, Fox W, Mounasamy V, Temple HT. Treatment and survival of osseous renal cell carcinoma metastases. J Surg Oncol. Dec 2012; 106(7): Willeumier JJ, van der Linden YM, van de Sande MA, Dijkstra PS. Treatment of pathological fractures of the long bones. EFORT Open Reviews. 2016; 1(5): Bauer HC, Wedin R. Survival after surgery for spinal and extremity metastases. Prognostication in 241 patients. Acta Orthop Scand. Apr 1995; 66(2): Forsberg JA, Eberhardt J, Boland PJ, Wedin R, Healey JH. Estimating survival in patients with operable skeletal metastases: an application of a bayesian belief network. PLoS One. 2011; 6(5): e Forsberg JA, Wedin R, Bauer HC, et al. External validation of the Bayesian Estimated Tools for Survival (BETS) models in patients with surgically treated skeletal metastases. BMC Cancer. Oct ; 12:

256 Chapter Nathan SS, Healey JH, Mellano D, et al. Survival in patients operated on for pathologic fracture: implications for end-of-life orthopedic care. J Clin Oncol. Sep ; 23(25): Katagiri H, Takahashi M, Wakai K, Sugiura H, Kataoka T, Nakanishi K. Prognostic factors and a scoring system for patients with skeletal metastasis. J Bone Joint Surg Br. May 2005; 87(5): Ratasvuori M, Wedin R, Keller J, et al. Insight opinion to surgically treated metastatic bone disease: Scandinavian Sarcoma Group Skeletal Metastasis Registry report of 1195 operated skeletal metastasis. Surg Oncol. Jun 2013; 22(2): Willeumier JJ, van der Hoeven NM, Bollen L, et al. Epidermal growth factor receptor mutations should be considered as a prognostic factor for survival of patients with pathological fractures or painful bone metastases from non-small cell lung cancer. Bone Joint J. Apr 2017; 99-B(4): Sorensen MS, Hovgaard TB, Hindso K, Petersen MM. Prognostic value of biochemical variables for survival after surgery for metastatic bone disease of the extremities. J Surg Oncol. Mar 2017; 115(4): Katagiri H, Okada R, Takagi T, et al. New prognostic factors and scoring system for patients with skeletal metastasis. Cancer Med. Oct 2014; 3(5): PATHFx.org. 2017; Accessed 04/30/2017, optimal-study. 2017; Accessed 4/30/2017, Gupta S, Tran T, Luo W, et al. Machine-learning prediction of cancer survival: a retrospective study using electronic administrative records and a cancer registry. BMJ Open. Mar ; 4(3): e Paulino Pereira NR, Langerhuizen DW, Janssen SJ, et al. Are perioperative allogeneic blood transfusions associated with 90-days infection after operative treatment for bone metastases? J Surg Oncol. Dec 2016; 114(8): Clausen C, Lonn L, Morgen SS, et al. Perioperative blood transfusion does not decrease survival after surgical treatment of spinal metastases. Eur Spine J. Aug 2014; 23(8): Amato A, Pescatori M. Perioperative blood transfusions for the recurrence of colorectal cancer. Cochrane Database Syst Rev. Jan (1): CD Bennett S, Baker LK, Martel G, et al. The impact of perioperative red blood cell transfusions in patients undergoing liver resection: a systematic review. HPB (Oxford). Apr 2017; 19(4): Rosen CB, Nagorney DM, Taswell HF, et al. Perioperative blood transfusion and determinants of survival after liver resection for metastatic colorectal carcinoma. Ann Surg. Oct 1992; 216(4): ; discussion Kooby DA, Stockman J, Ben-Porat L, et al. Influence of transfusions on perioperative and long-term outcome in patients following hepatic resection for colorectal metastases. Ann Surg. Jun 2003; 237(6): ; discussion Hallet J, Tsang M, Cheng ES, et al. The Impact of Perioperative Red Blood Cell Transfusions on Long-Term Outcomes after Hepatectomy for Colorectal Liver Metastases. Ann Surg Oncol. Nov 2015; 22(12):

257 CHAPTER 14 Summary In Dutch (Samenvatting In Het Nederlands) 14

258 Chapter 14 Botmetastasen (uitzaaiingen van kanker in bot) komen na het axiale skelet (o.a. wervelkolom) het meest frequent voor in het femur (dijbeen) en de humerus (opperarmbeen). Deze laesies verzwakken het bot en kunnen leiden tot een pathologische fractuur (botbreuk). De toename in het aantal patiënten met botmetastasen zal waarschijnlijk leiden tot een groter aantal patiënten dat een pathologische fractuur ontwikkelt. De overleving van patiënten met botmetastasen en een pathologische fractuur is slecht: meer dan de helft overlijdt binnen een jaar. Behandeling is met name gericht op palliatie (verzachten van het lijden), en zelden op curatie (genezing). Een pathologische fractuur vermindert de kwaliteit van leven en iemands zelfstandigheid. Chirurgische fixatie beoogt de pathologische fractuur of een dreigende pathologische fractuur te stabiliseren om kwaliteit van leven en zelfstandigheid te behouden. Echter, een (dreigende) pathologische fractuur wordt niet altijd veroorzaakt door een botmetastase en zorgvuldige diagnostiek (ter uitsluiting van een primaire bottumor) dient vooraf te gaan aan chirurgische fixatie. Veel voorkomende methodes van chirurgische behandeling voor de lange pijpbeenderen zijn: intramedullaire (merg) penfixatie, gedeeltelijke botresectie gevolgd door reconstructie met endoprothese, en open reductie en interne fixatie middels plaat- en schroefosteosynthese. Meerdere factoren spelen een rol bij de keuze voor een behandelstrategie bij een patiënt met botmetastasen; onder andere de primaire tumor, de levensverwachting, de locatie van de botmetastase, de aanwezigheid van een pathologische fractuur, en de aanwezigheid van andere bot- danwel viscerale (orgaan) metastasen. Het onderzoek in dit proefschrift heeft als doel: patiënt-selectie voor chirurgische behandeling te verbeteren, nauwkeurigere selectie van chirurgische implantaten gebaseerd op patiënt- en tumorkarakteristieken, het identificeren van risicofactoren voor complicaties, en het evalueren van uitkomsten na chirurgische behandeling. Dit proefschrift bestaat uit drie delen: (1) botmetastasen van het femur, (2) botmetastasen van de humerus, en (3) overleving. DEEL 1: BOTMETASTASEN VAN HET FEMUR Dit deel begint met een retrospectieve studie (Chapter 2) die de uitkomst na verschillende chirurgische behandelmethodes voor (dreigende) pathologische fracturen van het proximale femur vergelijkt. Er zijn 417 opeenvolgende patiënten geïncludeerd met botmetastasen van het proximale femur welke behandeld zijn middels: intramedullaire penfixatie, endoprothese, en plaatfixatie. De primaire uitkomstmaat, re-operatie, verschilde niet significant tussen de drie behandelmethodes. Echter, subanalyses naar type complicatie leidend tot re-operatie (diepe infectie en falen van fixatie) laten wel een verschil zien. Falen van fixatie resulterend in re-operatie komt het meest frequent voor na plaatosteosynthese (13%), gevolgd door intramedullaire penfixatie (3%), en komt niet 256

259 Samenvatting voor na endoprothese. Diepe infectie resulterend in re-operatie komt het meest frequent voor na endoprothese (8.6%), gevolgd door intramedullaire penfixatie (2%), en komt niet voor na plaatosteosynthese. Diepe infecties komen het meest frequent voor in de eerste maanden na operatieve behandeling, terwijl het risico op falen van fixatie toeneemt op de lange termijn. Systemische complicaties en 30-dagen mortaliteit verschilden niet tussen de verschillende behandelmethodes. Peroperatief bloedverlies was tweemaal zo hoog en de operatieduur 40 minuten langer voor endoprothese in vergelijking met intramedullaire penfixatie en plaatosteosynthese. De hieropvolgende systematische review (Chapter 3) plaatst onze bevindingen in een breder perspectief. Veertig studies (met in totaal 2,748 patiënten) die operatieve behandeling van proximale femur metastasen beschrijven zijn geïncludeerd: endoprothese (1,461 patiënten), intramedullaire penfixatie (1,054 patiënten), en plaatosteosynthese (233 patiënten). Borst (35%), long (15%), prostaat (10%), en niercel (8.2%) zijn de meest voorkomende primaire tumoren. De éénjaarsoverleving varieerde van 0 tot 62%. De gepoolde kans op re-operatie verschilde significant tussen behandelmethodes en was 5.2% na endoprothese, 4.2% na intramedullaire penfixatie, en 14% na plaatosteosynthese. Opnieuw werd een duidelijk verschil gezien in type complicatie leidend tot re-operatie. De gepoolde kans op re-operatie door falen van fixatie was 0.4% na endoprothese, 2.8% na intramedullaire penfixatie, en 10% na plaatosteosynthese. De gepoolde kans op re-operatie na diepe infectie was 0.68% na endoprothese, 0.04% na intramedullaire penfixatie, en 0% na plaatosteosynthese. Heterogeniteit, variatie in definitie, en slechts sporadische rapportage van functionele resultaten en systemische complicaties maakten het niet mogelijk deze uitkomstmaten te vergelijken. Bevindingen uit chapter 2 en 3 suggereren dat plaatosteosynthese geen geschikte optie is voor de behandeling van (dreigende) pathologische fracturen van het proximale femur ten gevolge van botmetastasen. Endoprothese en intramedullaire penfixatie zijn met het oog op algemeen re-operatie risico vergelijkbaar maar kennen andere onderliggende oorzaken van re-operatie (diepe infectie versus falen van fixatie). Deze complicaties en diens timing maken dat levensverwachting een belangrijke rol speelt in de besluitvorming voor één van deze behandelmethodes. Een belangrijk aandachtspunt voor toekomstige studies is het meten van patiënt gerapporteerde uitkomsten zoals kwaliteit van leven en functionele resultaten. Dit gaat ons helpen om chirurgische technieken verder te vergelijken en daarnaast helpt het ons begrijpen hoe snel een patiënt herstelt en wat de impact is van verschillende complicaties op het postoperatief herstel. Vervolgens hebben wij prospectief onderzocht welke vragenlijst het meest bruikbaar is in termen van: effectiviteit, betrouwbaarheid, en efficiëntie voor het meten van functionele resultaten bij patiënten met botmetastasen van de onderste extremiteit (Chapter 4). Honderd van de 115 uitgenodigde patiënten met botmetastasen in de onderste extremiteit hebben vijf veelgebruikte vragenlijsten ingevuld in random volgorde: PROMIS-CAT Physical Function Cancer, PROMIS-CAT Neuro-QoL Mobility, Toronto Extremity Salvage Score

260 Chapter 14 (TESS), Lower Extremity Function Score (LEFS), en de Musculoskeletal Tumor Society score (MSTS). Wij vonden dat deze vijf vragenlijsten hetzelfde concept (in dit geval: functie) meten; dit werd duidelijk door de hoge correlatie (factor scores >0.7) van alle vragenlijsten met het onderliggende wiskundig afgeleide concept, en de hoge inter-vragenlijst correlaties (>0.7). Vloer effect (d.w.z. laagst mogelijke score op vragenlijst) was afwezig voor alle vijf vragenlijsten, terwijl alle vijf vragenlijsten in enige mate een plafond effect (d.w.z. hoogst mogelijke score op vragenlijst) lieten zien. Het plafond effect was het grootst voor de PROMIS Neuro-QoL mobility (7%) vragenlijst. De standaardmeetfout (d.w.z. de mate van precisie van een vragenlijst) was beneden het afkappunt en derhalve betrouwbaar over een grote functionele range voor de PROMIS Physical Function Cancer, TESS, en de LEFS. Voltooiingstijd verschilde tussen de vijf vragenlijsten en was het kortst voor de twee PROMIS vragenlijsten. Wij maken hieruit op dat de PROMIS Physical Function Cancer vragenlijst het meest bruikbaar is voor het meten van functionele resultaten bij patiënten met botmetastasen van de onderste extremiteit vanwege de hoge betrouwbaarheid, validiteit, en beknoptheid. Tot slot hebben wij een nieuwe methode ontwikkeld en vervolgens getest om het risico op het ontwikkelen van een pathologische fractuur door een botmetastase in het femur in te schatten (Chapter 5). Voorheen werd gebruik gemaakt van metingen op röntgenfoto s, symptomen, en combinaties hiervan. Deze historische methodes zijn echter niet voldoende accuraat. Voor ons nieuw ontwikkelde algoritme hebben wij gebruik gemaakt van de mate van röntgenstraling-verzwakking (uitgedrukt in Hounsfield units) en de geometrische verdeling van deze waardes op CT scans van femora met botmetastasen die een pathologische fractuur ontwikkelden en femora met botmetastasen die geen pathologische fractuur ontwikkelden. Wij vonden geen verschil tussen deze twee groepen wanneer wij al het weefsel (corticaal bot, beenmerg, en tumor weefsel) gebruikten in de berekeningen. Echter, wanneer wij alleen corticaal bot analyseerden vonden wij dat de femora die een pathologische fractuur ontwikkelde meer aangetast waren door de botmetastase dan de femora die geen pathologische fractuur ontwikkelden. Vergeleken met de historische methodes bleek deze methode net zo accuraat te zijn. Desondanks denken wij dat het waardevol is om onze methode in een meer gecontroleerde omgeving te testen omdat accuratere inschatting van fractuurrisico de potentiële over- en onderbehandeling van dreigende pathologische fracturen sterk kan terugdringen. DEEL 2: BOTMETASTASEN VAN DE HUMERUS Dit deel begint met een overzicht van de huidige literatuur (systematische review) betreffende resultaten na operatieve behandeling van botmetastasen van de humerus (Chapter 6). Borst (30%), multipel myeloom (15%), long (15%), en niercel (13%) zijn de meest 258

261 Samenvatting voorkomende primaire tumoren. De gemiddelde postoperatieve overleving was slecht en varieerde tussen studies van 4 tot 23 maanden. Wij vonden, in 23 studies met in totaal 909 patiënten, een gemiddelde kans op re-operatie van 4.4% na intramedullaire penfixatie (range: 0 tot 10%), 9.3% na plaatosteosynthese (range: 5 tot 14%), 2.5% na proximale humerus reconstructie met endoprothese (range: 0 tot 6%), en 15% na schacht reconstructie met een diafyse (schacht) endoprothese (range: 14 tot 16%). Echter, indicaties voor deze technieken variëren en derhalve zijn technieken niet direct vergelijkbaar. Daarnaast vonden wij dat kwaliteit van geïncludeerde studies beperkt was, de definitie van systemische complicaties varieerde, en er waren geen studies die patiënt gerapporteerde uitkomsten beschreven. Samenwerking tussen meerdere orthopedisch oncologische centra is nodig om lokale en systemische complicaties in voldoende grote en relatief homogene groepen te vergelijken. Vervolgens hebben wij een retrospectieve studie uitgevoerd waarbij wij naar alle reoperaties en systemische complicaties hebben gekeken na chirurgische behandeling van botmetastasen in de humerus (Chapter 7). De mediane overleving was 11 maanden. Wij vonden een kans op re-operatie van 6.7% na intramedullaire penfixatie, 10% na plaatosteosynthese, en 11% na proximale humerus reconstructie met endoprothese. Diepe infectie (2.0%), non-union (2.0%), tumor progressie (1.6%), en fractuur rond het implantaat (1.4%) zijn de meest voorkomende oorzaken van re-operatie. Geen patiënt- of tumorkarakteristiek was geassocieerd met een hoger risico op re-operatie. Echter, wij vonden dat de proportie patiënten die een re-operatie onderging aanzienlijke toenam bij langere overleving van de patiënt; 2.6% onderging re-operatie één maand postoperatief, terwijl 19% re-operatie onderging 2 jaar postoperatief. Dit benadrukt dat levensverwachting een belangrijke factor is in chirurgische besluitvorming. Daarnaast geeft het aan dat het belangrijk is om duurzame reconstructies na te streven in patiënten die naar verwachting langer dan één jaar zullen leven. Tevens vonden wij dat het postoperatieve risico op systemische complicaties (gemiddeld: 5.8%) was geassocieerd met een prognostische score voor overleving (de modified Bauer score). Dit laatste kan gebruikt worden voor betere preoperatieve patiëntselectie. Tot slot hebben wij 161 orthopedisch chirurgen, waarvan 78 orthopedisch oncologisch chirurgen, en 83 traumachirurgen, benaderd om een vragenlijst over behandeling van botmetastasen in de humerus te completeren (Chapter 8). Wij hebben 24 fictieve casus gemaakt door verschillende karakteristieken te combineren: type primaire tumor, levensverwachting, type fractuur, anatomische locatie van de laesie in de humerus. Voor elke casus stelden wij de vraag: welke behandeling adviseer je? Orthopedisch oncologisch chirurgen waren vaker geneigd tot reconstructie middels endoprothese en plaatosteosynthese en minder vaak geneigd tot intramedullaire penfixatie in vergelijking met traumachirurgen. Daarnaast vonden wij dat keuze voor behandeling sterk afhankelijk was van het type primaire tumor, levensverwachting, en anatomische locatie. Keuze van behandeling

262 Chapter 14 was niet afhankelijk van aanwezigheid van een pathologische fractuur. Dit bevestigt de aanzienlijke controverse onder orthopedisch chirurgen ten aanzien van de meest optimale operatieve strategie. DEEL 3: OVERLEVING Dit deel begint met een vergelijking van type chirurgische resecties van botmetastasen uitgaande van de niercel (Chapter 9). Het is onduidelijk of verschillende type chirurgische tumor resectie de oncologische uitkomst overleving en tumor progressie beïnvloedt. Wij hebben gekeken naar lokale tumor progressie, re-operatie, en overleving en vergeleken drie chirurgische technieken: metastasectomie, intralesionale excisie, en chirurgische fixatie zonder resectie. Wij hebben 183 patiënten vanuit twee geaffilieerde orthopedisch oncologische centra in deze retrospectieve studie geïncludeerd: 48% onderging metastasectomie (marges: 72% negatief, 23% positief, 5% onzeker), 30% intralesionale excisie (marges: 100% positief), en 22% chirurgische fixatie zonder resectie (marges: 100% positief). De kans op lokale tumor progressie verschilde en was het hoogst na chirurgische fixatie zonder resectie (39%), gevolgd door intralesionale excisie (22%), en metastasectomie (12%). Echter, de kans op re-operatie was vergelijkbaar tussen de drie chirurgische behandelingen. Overleving was het beste na metastasectomie; echter, dit verschil verdween bij stratificeren op basis van uitgebreidheid van metastasering. Lokale tumor progressie was lager in patiënten met negatieve tumormarges (5% versus 26%). Echter, wij vonden geen verschil in re-operatie op basis van tumormarges. Overleving was beter in patiënten met negatieve tumormarges. Op basis van deze bevindingen concluderen wij dat patiënten met een niercel botmetastase en een redelijke levensverwachting (meer dan één jaar) metastasectomie dienen te ondergaan waarbij gestreefd wordt naar negatieve tumormarges om het risico op lokale tumorprogressie te reduceren, zonder additioneel risico op re-operatie, en met een potentieel effect op overleving. Vervolgens hebben wij algoritmes ontwikkeld om levensverwachting in te schatten (Chapter 10). Eerst hebben wij gekeken welke factoren onafhankelijk geassocieerd waren met kortere overleving. Deze factoren hebben wij gebruikt om een klassiek algoritme, nomogram, en een boosting algoritme te ontwikkelen. Hierna hebben wij de accuraatheid van deze drie algoritmes vergeleken voor het voorspellen van 30-, 90-, en 365-dagen overleving. Deze retrospectieve studie omvatte 927 patiënten die een operatie voor een botmetastase in een lang pijpbeen hebben ondergaan. Wij identificeerde de volgende risicofactoren voor slechtere overleving: hogere leeftijd, comorbiditeit, body mass index < 18.5 kg/m 2, primaire tumor met slechte prognose, aanwezigheid van meerdere botmetastasen, aanwezigheid van viscerale metastasen, en een laag hemoglobine. Het nomogram was redelijk accuraat voor het voorspellen van 30-, 90-, en 365-dagen overleving. Het 260

263 Samenvatting boosting algoritme was accurater bij ontwikkeling, maar de accuraatheid was vergelijkbaar met het nomogram op een test dataset. Het klassieke algoritme was het minst accuraat voor alle drie de tijdpunten. Comorbiditeit en body mass index zijn nieuw geïdentificeerde risicofactoren voor overleving. Wij adviseren het gebruik van het nomogram voor het inschatten van overleving. Tot slot hebben wij gekeken naar de invloed van peri-operatieve bloedtransfusies op overleving in patiënten die operaties voor botmetastasen in de lange pijpbeenderen ondergaan (Chapter 11). Verschillende studies hebben laten zien dat peri-operatieve bloedtransfusies het risico op tumorprogressie en overleving verslechtert bij operatie voor primaire tumoren. Het is onbekend of deze associatie ook bestaat voor patiënten die operaties voor botmetastasen ondergaan. We hebben 789 patiënten geïncludeerd in deze retrospectieve studie die een operatie ondergingen voor een botmetastase in een lang pijpbeen. Wij hebben gekeken naar een verschil in het krijgen van een bloedtransfusie versus het niet krijgen van een bloedtransfusie; ook hebben wij gekeken naar een dosisresponsrelatie per eenheid getransfundeerd bloed. Het krijgen van een bloedtransfusie was niet geassocieerd met slechtere overleving; echter, er was wel sprake van een dosisresponsrelatie met een verhoogd risico op overlijden bij een toename in aantal eenheden getransfundeerd bloed. Concluderend is er mogelijk sprake van een associatie tussen bloedtransfusie en slechtere overleving, echter dit effect is klein. CONCLUSIES Intramedullaire penfixatie en endoprothese hebben de voorkeur boven open reductie en interne fixatie voor proximale femur metastasen (Chapter 2 & 3). Intramedullaire penfixatie en endoprothese hebben een vergelijkbaar risico op reoperatie voor proximale femur metastasen (Chapter 2 & 3). Implantaat-specifieke oorzaken voor re-operatie en diens timing moeten in acht worden genomen bij de keuze tussen intramedullaire penfixatie en endoprothese voor proximale femur metastasen (Chapter 2 & 3). De PROMIS Physical Function Cancer questionnaire is de aanbevolen vragenlijst voor het bepalen van functionele uitkomst in patiënten met botmetastasen van de onderste extremiteit (Chapter 4). CT scan gebaseerde algoritmes kunnen bruikbaar zijn voor het inschatten van het risico op het ontwikkelen van een pathologische fractuur door een botmetastase in het femur (Chapter 5). Endoprothese, intramedullaire penfixatie, en open reductie en interne fixatie zijn acceptabele opties voor de behandeling van botmetastasen in de humerus. Variatie in

264 Chapter 14 indicatie en rapportage van uitkomsten, en kleine en heterogene studiegroepen maken het echter niet mogelijk deze technieken te vergelijken (Chapter 6). De kans op re-operatie neemt aanzienlijk toe bij langere overleving van patiënten met humerus metastasen (Chapter 7). Een prognostische score voor overleving (modified Bauer score) is bruikbaar voor risico stratificatie van postoperatieve systemische complicaties in patiënten met femur of humerus metastasen (Chapter 2 & 7). Specialisatie van de orthopedisch chirurg, evenals type tumor, levensverwachting, en anatomische locatie van humerus metastasen beïnvloeden variatie in behandeling (Chapter 8). Metastasectomie met negatieve tumormarges voor niercel botmetastasen vermindert het risico op lokale tumorprogressie met een vergelijkbaar risico op re-operatie, en een potentiële impact op overleving (Chapter 9). Comorbiditeit en body mass index zijn twee nieuwe prognostische factoren voor overleving in patiënten met botmetastasen (Chapter 10). Een nomogram is de methode van voorkeur voor het voorspellen van overleving in patiënten met botmetastasen (Chapter 10). Peri-operatieve bloedtransfusie in patiënten met botmetastasen is een mogelijke risico factor voor slechtere overleving, echter het effect is klein (Chapter 11). 262

265 Abbreviations 95%CI = 95% confidence interval AMC = Academisch Medisch Centrum (Amsterdam) AUC = area under curve BC = before Christ BI = Beth Israel hospital BMD = bone mineral density BMI = body mass index BWH = Brigham and Women s Hospital (Boston) CAT = computerized adaptive testing CCI = Charlson comorbidity index CPT = current procedural terminology code CT = computed tomography scan DHS = dynamic hip screw DP = diaphysis prosthesis ECOG = Eastern Cooperative Oncology Group score EPR = endoprosthetic reconstruction EQ-5D = EuroQol 5 Dimensions FDG-PET = fluorodeoxyglucose positron emission tomography scan HR = hazard ratio HU = Hounsfield Units Hemi-A = hemiarthroplasty HMS = Harvard Medical School (Boston) I&D = incision & debridement ICC = intraclass correlation coefficient ICD-9-CM = international classification of diseases, ninth revision, clinical modification IMN = intramedullary nail IQR = interquartile range IRT = item response theory LEFS = lower extremity functional scale MGH = Massachusetts General Hospital (Boston) MRI = magnetic resonance imaging scan MSTS = musculoskeletal tumor society MTP = modular tumor prosthesis OR = odds ratio 263

266 Abbreviations ORIF = open reduction internal fixation PMMA = polymethyl methacrylate bone cement PROMIS = patient-reported outcomes measurement information system PSF = plate-screw fixation QOL = quality of life QuickDASH = disabilities of the arm, shoulder and hand questionnaire RCC = renal cell carcinoma ROC = receiver operating characteristic ROI = region of interest SD = standard deviation SEM = standard error of the mean SF-36 = short form health survey SORG = skeletal oncology research group SOVG = science of variation group SRE = skeletal related event TESS = Toronto extremity salvage score THA = total hip arthroplasty WBC = white blood cell count 264

267 Acknowledgements Prof. van Dijk, beste Niek, dank voor alle adviezen, steun, en vertrouwen. Uw toewijding en inventiviteit zijn inspirerend. Het was mooi om samen door Boston te lopen, en te fietsen via de Mont Ventoux naar Barcelona. Dr. Bramer, beste Jos, dank voor je ideeën, meedenken, en het ondersteunen van de brug naar Boston. De onderzoeks-meetings in het AMC, afgewisseld met kliniek, congressen en een biertje waren bijzonder leerzaam en gezellig. Ik kijk uit naar verdere samenwerking op het gebied van de orthopedische oncologie. Drs. Schwab, dear Joe, thank you for your endless support and confidence. The diversity of both your clinical and scientific work is truly unbelievable. The meetings we had were inspirational in many ways. Besides work; drinks at Harvard Gardens, traveling around the US for conferences, and of course the weekends in New Hampshire have been at least as inspiring and fun. One of your commonly used quotes: If I have seen further it is by standing on the shoulders of giants (Newton 1675). To me, you are one of those. You are a great person, and I am looking forward to further collaboration. Prof. Ring, dear David, your scientific mindset is extraordinary. You are an incredible mentor, thank you for your continuing guidance. Leescommissie, dank voor jullie interesse in het lezen van dit proefschrift. Teun, paranimf, je visie en tomeloze discipline zijn bewonderenswaardig. Tegelijkertijd met jou op uitzending naar Boston was een prachtige ervaring. Peter-Paul, paranimf, je bevlogenheid en een vlijmscherp denkvermogen zelfs na meer dan een paar biertjes zijn indrukwekkend. Mooi om te zien dat onze interesses in wetenschap en gezondheidszorg gelijk opgaan, ik kijk uit naar het volgende boek. Nuno Rui Paulino Pereira, Peña etc., gaaf dat je bent aangesloten in de onderzoekslijn naar botmetastasen en deze succesvol hebt voortgezet, helemaal prachtig dat we hier deze dag samen staan. Pils. 265

268 Acknowledgements Olivier, Ollie, etc., wat hebben we onwijs veel gelachen, en mooie dingen geklust in New Hampshire en op de Harvard Street. Je interesse in techniek vertaalt zich in krachtige biomechanische studies. Nuno mag ons dankbaar zijn dat wij hem nooit stoorden tijdens werk. Pils. Bart, Lubberts, je vermogen om te werken en sporten gaan hand in hand en zijn ongeëvenaard, het was prachtig om samen op de Harvard Street te wonen. Dirk, nooit is een idee te gek in jouw ogen, en vaak weet je deze ideeën ook nog te realiseren. Fisher-Price speelgoed en een banaan in de CT scan, een weekendje Amherst, etc. Sjoerd, Nota, levensgenieter, een gesprek of discussie met jou verveelt nooit. Bedankt dat je mij hebt betrokken bij Schwab s lab. Michiel, dank voor je enthousiasme en steun bij het opstarten van de eerste onderzoeksprojecten in Boston. David, dank voor je bevlogen en persisterende inzet in diverse studies. Nieuwe ideeën krijgen vorm op de racefiets, ook bij wat tegenwind. Wouter, mooi om samen in Boston en MGH te hebben gewoond. Wandelend van of naar het ziekenhuis, pils bij de Whiskey s, film bij AMC Loews, gezellig en meestal zelfs dan productief. Bastiaan, prachtig om nog een paar maanden samen in Boston te hebben gezeten. Jouw enthousiasme, maar juist ook je kritische houding naar je eigen werk heeft tot succesvol onderzoek geleid. Prof. Hornicek, Drs. Raskin, Dr. Lozano-Calderon, and Drs. Ferrone, it has been a pleasure and tremendous honor to work with you. Thank you for facilitating and encouraging ongoing research at MGH and BWH. Colleagues of the MGH Science Factory: Stephanie Becker, Reinier Beks, Marijn van Berckel, Johann Blauth, Niels Bosma, Yvonne Braun, Pim van Dijk, Eva van Dijk, Maarten van Dijke, Tessa Drijkoningen, Thierry Guitton, Stefan Hartveldt, Andrea van der Heijden, Hugo Hermanussen, Prakash Jayakumar, Amir Kachooei, Joost Kortlever, Michael Kuntz, Amanda Lans, Sjoerd Meijer, Timion Meijs, Jos Mellema, Mariano Menendez, Amin Mohamadi, Jeroen Molleman, Ali Moradi, Kamil Oflazoglu, Paul Ogink, Eva van Rein, Nicky Stoop, Matthew Tarabochia, Nina Theyskens, Bianca Verbeek, and Dafang Zhang: thank you all for the fruitful collaboration in many projects! 266

269 Acknowledgements Pap en Mam, oneindig veel dank voor jullie onvoorwaardelijke liefde en steun. Fijn dat we elkaar ondanks de afstand toch met enige regelmaat hebben opgezocht, inclusief twee mooie reizen door New England en de Southeast. Ruby, Rubs, al eenendertig jaar mijn grootste steun en toeverlaat. Ik bewonder je zorgzaamheid en geduld met de drie kleintjes. Loes, lieve Loes, een half jaar Boston werden er twee, en dat terwijl we elkaar maar net hadden ontmoet. Vijfduizend kilometer water heeft ons gelukkig niet kunnen scheiden. Je bent de allerliefste en ik kijk uit naar ons volgende avontuur. 267

270

271 Portfolio PhD TRAINING Research Courses Year Workload Introduction meeting graduate school and APROVE - AMC Practical Biostatistics AMC Basic Biostatistics for Clinical Research MGH/HMS Problem-Based Biostatistics for the Clinical Investigator MGH/HMS Introduction to Survey Design MGH/HMS Applied Biostatistics for Clinical Trials MGH/HMS Certificate in Applied Biostatistics HMS Seminars, workshops and master classes Year Workload Journal clubs Orthopaedic Hand Service MGH/BWH/BI Journal clubs Orthopaedic Spine Service MGH/BWH Journal clubs Orthopaedic Hand Service MGH/BWH/BI Grand Rounds Orthopaedics MGH Workshop and cadaver lab: Management Spine Tumors Depuy, Raynham Brian Silber Spine Oncology Symposium MGH Fractures of the distal radius webinar OTA Tibial plateau fracture webinar OTA Workshop and cadaver lab: elbow arthroscopy Smith & Nephew, Boston Journal clubs Orthopaedic Spine Service MGH/BWH Journal clubs Orthopaedic Hand Service MGH/BWH/BI Grand Rounds Orthopaedics MGH Workshop and cadaver lab: Thoracolumbar spine trauma Depuy, Raynham Workshop and cadaver lab: Spinal deformity Depuy, Raynham Workshop and cadaver lab: Cervical spine trauma Depuy, Raynham AO Trauma Basic Principles Dallas Medical Business Masterclass Amsterdam Workshop: Trauma resident days Amersfoort Workshop and cadaver lab: Basics spine surgery Maastricht Presentations (see report of scholarship for list of presentations) Year Workload 13 Podium presentations (co-author on another 19 podium presentations) Poster presentations (co-author on another 16 poster presentations) (Inter)national conferences Year Workload Smith hand surgery day Boston

272 Portfolio Jesse Jupiter Hand Course Boston Harvard Orthopaedic Trauma day Boston th Annual International Spine symposium Boston Musculoskeletal Tumor Society Annual Meeting Houston American Academy of Orthopaedic Surgeons Annual Meeting Las Vegas Smith hand surgery day Boston Harvard Orthopaedic Trauma day Boston Association of Bone and Joint Surgeons Annual Meeting Oregon Musculoskeletal Tumor Society Annual Meeting Orlando European MusculoSkeletal Oncology Society Annual Meeting La Baule Edinburgh Instructional Trauma Course Edinburgh Dutch Trauma Conference Amsterdam Surgical resident trauma conference - Soestduinen European MusculoSkeletal Oncology Society Annual Meeting Budapest Dutch Trauma Conference Amsterdam European MusculoSkeletal Oncology Society Annual Meeting Amsterdam TEACHING Lecturing Year Workload Basic statistics and STATA 13 courses Stein Janssen & Teun Teunis Basic statistics and STATA 13 courses Stein Janssen Supervising, tutoring, and mentoring Year Workload Jeroen Molleman Orthopaedic Hand Service MGH Niels Bosma Orthopaedic Hand Service MGH Bart Lubberts Orthopaedic Hand Service MGH Marijn van Berckel Orthopaedic Hand Service MGH Joost Kortlever Orthopaedic Oncology/Spine/Hand Service MGH Andrea van der Heijden Orthopaedic Oncology/Spine Service MGH Maarten van Dijke Orthopaedic Oncology/Spine Service MGH Eva van Dijk Orthopaedic Oncology/Spine Service MGH Timion Meijs Orthopaedic Oncology/Spine Service MGH Nicky Stoop Orthopaedic Hand Service MGH Amanda Lans Orthopaedic Oncology/Spine Service MGH Matthew Tarabochia Orthopaedic Hand Service MGH Olivier van Wulfften Palthe Orthopaedic Oncology/Spine Service MGH Nuno Rui Paulino Pereira Orthopaedic Oncology/Spine Service MGH Wouter van Leeuwen Orthopaedic Hand Service MGH Sjoerd Meijer Orthopaedic Hand Service MGH Hugo Hermanussen Orthopaedic Hand Service MGH

273 Portfolio Reinier Beks Orthopaedic Hand Service MGH Quirine van der Vliet Orthopaedic Oncology/Spine Service MGH Stefan Hartveldt Orthopaedic Oncology/Spine Service MGH David Langerhuizen Orthopaedic Oncology/Spine Service MGH Eva van Rein Orthopaedic Oncology/Spine Service MGH Nina Theyskens Orthopaedic Oncology/Spine Service MGH Paul Ogink Orthopaedic Oncology/Spine Service MGH Bianca Verbeek Orthopaedic Oncology/Spine Service MGH PARAMETERS OF ESTEEM Grants The Prof. Michaël-van Vloten Fund Research fellowship Orthopaedic Oncology. MGH/HMS, Boston, USA Fund from the Anna Fonds foundation Research fellowship Orthopaedic Oncology. MGH/HMS, Boston, USA The De Drie Lichten fund Research fellowship Orthopaedic Oncology. MGH/HMS, Boston, USA 2013 KWF Kankerbestrijding Research fellowship Orthopaedic Oncology. MGH/HMS, Boston, USA 2014 Awards and Prizes Orthopaedics Departmental Award MGH Clinical Research Day Sacral Insufficiency Fractures after High-Dose Radiation with and without Partial Sacrectomy: A Twenty Year Retrospective. Poster of Distinction MGH Clinical Research Day Association of Perioperative Allogeneic Blood Transfusion with Survival in Patients with Metastatic Long Bone Fractures. Marshall Urist Young Investigator Award (Association of Bone and Joint Surgeons) Prognostication in Patients With Long Bone Metastases: Does a Boosting Algorithm Improve Survival Estimates? Year Year PUBLICATIONS Peer reviewed (see report of scholarship for list of publications) 64 peer reviewed publications Other Reviewer: The Bone & Joint Journal Reviewer: Clinical Orthopaedics and Related Research (Top 50 reviewer 2016) Reviewer: The Journal of Bone & Joint Surgery (Elite reviewer 2017) AMC = Academisch Medisch Centrum (Amsterdam) MGH = Massachusetts General Hospital (Boston) HMS = Harvard Medical School (Boston) BWH = Brigham and Women s Hospital (Boston) BI = Beth Israel hospital *Workload in ECTS 271

274

275 Report Of Scholarship PUBLICATIONS, PEER-REVIEWED S.J. Janssen, D.P. ter Meulen, M.G.J.S. Hageman, B.E. Earp, D. Ring. Quantitative 3-dimensional CT analyses of fractures of the middle phalanx base. HAND 2014 S.J. Janssen, D.P. ter Meulen, S.P.F.T. Nota, M.G.J.S. Hageman, D. Ring. Does Verbal and Nonverbal Communication of Pain Correlate With Disability? Psychosomatics 2014 S.J. Janssen, T. Teunis, D.P. ter Meulen, M.G.J.S. Hageman, D. Ring. Estimation of Base of Middle Phalanx Size Using Anatomical Landmarks. J Hand Surg Am 2014 S.J. Janssen, T. Teunis, F.J. Hornicek, J.A.M. Bramer, J.H. Schwab. Outcome of operative treatment of metastatic fractures of the humerus: a systematic review of twenty three clinical studies. Int Orthop 2014 J.J. Mellema, S.J. Janssen, T.G. Guitton, D. Ring. Quantitative 3-Dimensional Computed Tomography Measurements of Coronoid Fractures. J Hand Surg Am 2014 J. Molleman, S.J. Janssen, C.L. Overbeek, D. Ring. A threshold disability score corresponds with an estimated diagnosis of clinical depression in patients with upper extremity disease. HAND 2014 S.J. Janssen, Y. Braun, J.E. Ready, K.A. Raskin, M.L. Ferrone, F.J. Hornicek, J.H. Schwab. Are Allogeneic Blood Transfusions Associated With Decreased Survival After Surgery for Long-bone Metastatic Fractures? Clin Orthop Relat Res 2015 A.R. Kachooei, A. Moradi, S.J. Janssen, D. Ring. The influence of dominant limb involvement on DASH and QuickDASH. HAND 2015 S.J. Janssen, Y. Braun, K.B. Wood, T.D. Cha, J.H. Schwab, Allogeneic blood transfusions and postoperative infections after lumbar spine surgery. The Spine Journal 2015 S.J. Janssen, A.L. Bredenoord, W. Dhert, M. de Kleuver, F. Cumhur Oner, J.J. Verlaan, Potential Conflicts of Interest of Editorial Board Members from Five Leading Spine Journals. Plos One 2015 T. Teunis, S.J. Janssen, T.G. Guitton, A.M. Vranceanu, B. Goos, D. Ring. Surgeon personality is associated with recommendation for operative treatment. HAND 2015 M.T. Kuntz, S.J. Janssen, D. Ring. Incidental signal changes in the extensor carpi ulnaris on MRI. HAND 2015 S.J. Janssen, T. Teunis, T.G. Guitton, D. Ring. Do Surgeons Treat Their Patients Like They Would Treat Themselves? Clin Orthop Relat Res 2015 S.J. Janssen, J. Molleman, T.G. Guitton, D. Ring. What Middle Phalanx Base Fracture Characteristics are Most Reliable and Useful for Surgical Decision-making? Clin Orthop Relat Res 2015 S.T. Meijer, S.J. Janssen, T. Drijkoningen, D. Ring. Factors Associated with Unplanned Reoperation in Perilunate Dislocations and Fracture Dislocations. J Wrist Surgery 2015 S.J. Janssen, A.S. van der Heijden, M. van Dijke, J.E. Ready, K.A. Raskin, M.L. Ferrone, F.J. Hornicek, J.H. Schwab Marshall Urist Young Investigator Award: Prognostication in 273

276 Report Of Scholarship Patients With Long Bone Metastases: Does a Boosting Algorithm Improve Survival Estimates? Clin Orthop Relat Res 2015 J.T.P. Kortlever, S.J. Janssen, M.M.G. van Berckel, D. Ring, A.M. Vranceanu. What Is the Most Useful Questionnaire for Measurement of Coping Strategies in Response to Nociception? Clin Orthop Relat Res 2015 S.J. Janssen, T. Teunis, T.G. Guitton, D. Ring, J.H. Herndon. Orthopaedic Surgeons Views on Strategies for Improving Patient Safety. J Bone Joint Surg Am 2015 S.J. Janssen, T. Teunis, E. van Dijk, M.L. Ferrone, J.H. Shin, F. Hornicek, J.H. Schwab. Validation of the Spine Oncology Study Group Outcomes Questionnaire to assess quality of life in patients with metastatic spine disease. The Spine Journal 2015 W.F. van Leeuwen, S.J. Janssen, D.P. ter Meulen, D. Ring. What Is the Radiographic Prevalence of Incidental Kienbock Disease? Clin Orthop Relat Res 2015 D.P. ter Meulen, S.J. Janssen, M.G.J. Hageman, D.C. Ring. Quantitative three-dimensional computed tomography analysis of glenoid fracture patterns according to the AO/OTA classification. J Shoulder Elbow Surg 2015 S.J. Janssen, M. van Dijke, S.A. Lozano-Calderon, J.E. Ready, K.A. Raskin, M.L. Ferrone, F.J. Hornicek, J.H. Schwab. Complications after surgery for metastatic humeral lesions. J Shoulder Elbow Surg 2015 P. Osler, M.A. Bredella, K.A. Hess, S.J. Janssen, C.J. Park, Y.L. Chen, T.F. Delaney, F.J. Hornicek, J.H. Schwab. Sacral Insufficiency Fractures are Common After High-dose Radiation for Sacral Chordomas Treated With or Without Surgery. Clin Orthop Relat Res 2015 B. Lubberts, S.J. Janssen, J.J. Mellema, D. Ring. Quantitative 3-dimensional computed tomography analysis of olecranon fractures. J Shoulder Elbow Surg 2015 S.J.E. Becker, Y. Braun, S.J. Janssen, V. Neuhaus, D. Ring, C.S. Mudgal. Early Patient Satisfaction With Different Treatment Pathways for Trigger Finger and Thumb. J Hand Microsurg 2015 M.E. Menendez, S.J. Janssen, D. Ring. Electronic health record-based triggers to detect adverse events after outpatient orthopaedic surgery. BMJ Quality & Safety 2015 T. Teunis, S.J. Janssen, T.G. Guitton, D. Ring, R. Parisien. Do Orthopaedic Surgeons Acknowledge Uncertainty? Clin Orthop Relat Res 2016 S.J. Janssen, H.H. Hermanussen, T.G. Guitton, M.P.J. van den Bekerom, D.F.P. van Deurzen, D. Ring. Greater Tuberosity Fractures: Does Fracture Assessment and Treatment Recommendation Vary Based on Imaging Modality? Clin Orthop Relat Res 2016 W.F. van Leeuwen, Q.M.J. van der Vliet, S.J. Janssen, M. Heng, D. Ring, A.M. Vranceanu. Does perceived injustice correlate with pain intensity and disability in orthopaedic trauma patients? Injury 2016 S.J. Janssen, J.T.P. Kortlever, J.E. Ready, K.A. Raskin, M.L. Ferrone, F.J. Hornicek, J.H. Schwab. Complications After Surgery for Proximal Femoral Metastasis: A Retrospective Study of 417 Patients. J Am Acad Orthop Surg 2016 N.R. Paulino Pereira, R.B. Beks, S.J. Janssen, M.B. Harris, F.J. Hornicek, M.L. Ferrone, J.H. Schwab. Are allogeneic blood transfusions associated with decreased survival after surgical treatment for spinal metastases? The Spine Journal 2016 O.D. van Wulfften Palthe, V. Neuhaus, S.J. Janssen, T.G. Guitton, D. Ring, Among Musculoskeletal Surgeons, Job Dissatisfaction Is Associated With Burn Out. Clin Orthop Relat Res

277 Report Of Scholarship W.F. van Leeuwen, S.J. Janssen, D. Ring. Radiographic Progression of Kienböck Disease: Radial Shortening Versus No Surgery. J Hand Surg Am 2016 W.F. van Leeuwen, S.J. Janssen, D. Ring, N. Chen. Incidental magnetic resonance imaging signal changes in the extensor carpi radialis brevis origin are more common with age. J Shoulder Elbow Surg 2016 D.W.G. Langerhuizen, S.J. Janssen, Q.M.J. van der Vliet, K.A. Raskin, M.L. Ferrone, F.J. Hornicek, J.H. Schwab, S.A. Lozano-Calderon. Metastasectomy, Intralesional Resection, or Stabilization Only in the Treatment of Bone Metastases From Renal Cell Carcinoma. Journal of Surgical Oncology 2016 S.J. Janssen, T. Teunis, F.J. Hornicek, C.N. van Dijk, J.A. Bramer, J.H. Schwab. Outcome After Fixation of Metastatic Proximal Femoral Fractures: A Systematic Review of 40 Studies. Journal of Surgical Oncology 2016 J.T.P. Kortlever, S.J. Janssen, J. Molleman, M.G.J.S. Hageman, D. Ring. Discrete Pathophysiology is Uncommon in Patients with Nonspecific Arm Pain. Arch Bone Jt Sur 2015 S. Hartveldt, S.J. Janssen, K.B. Wood, T.D. Cha, J.H. Schwab, C.M. Bono, L.G. Jenis. Is There An Association of Epidural Corticosteroid Injection with Postoperative Surgical Site Infection after Surgery for Lumbar Degenerative Spine Disease? SPINE 2016 S.J. Janssen, E.A.J. van Rein, N.R. Paulino Pereira, K.A. Raskin, M.L. Ferrone, F.J. Hornicek, S.A. Lozano-Calderon, and J.H. Schwab. The Discrepancy between Patient and Clinician Reported Function in Extremity Bone Metastases. Sarcoma 2016 S.J. Janssen, N.R. Paulino Pereira, K.A. Raskin, M.L. Ferrone, F.J. Hornicek, C.N. van Dijk, S.A. Lozano-Calderon, J.H. Schwab. A Comparison of Questionnaires for Assessing Physical Function in Patients With Lower Extremity Bone Metastases. Journal of Surgical Oncology 2016 Q.M.J. van der Vliet, N.R. Paulino Pereira, S.J. Janssen, F.J. Hornicek, M.L. Ferrone, J.A.M. Bramer, C.N. van Dijk, J.H. Schwab. What Factors are Associated With Quality Of Life, Pain Interference, Anxiety, and Depression in Patients With Metastatic Bone Disease? Clin Orthop Relat Res 2016 M. van Dijke, S.J. Janssen, T.D. Cha, K.B. Wood, L.F. Borges, M.B. Harris, C.M. Bono, J.H. Schwab. Comparison of Decompression With and Without Fusion for Patients With Synovial Facet Cysts. Clin Spine Surg 2016 Lans, S.J. Janssen, D. Ring. Off-Hour Surgery Among Orthopedic Subspecialties at an Urban, Quaternary-Care, Level 1 Trauma Center. J Hand Surg Am 2016 N.R. Paulino Pereira, S.J. Janssen, E. van Dijk, M.B. Harris, F.J. Hornicek, M.L. Ferrone, J.H. Schwab. Development of a Prognostic Survival Algorithm for Patients with Metastatic Spine Disease. J Bone Joint Surg Am 2016 D. Zhang, M. Tarabochia, S.J. Janssen, D. Ring, N. Chen. Risk of Subluxation or Dislocation After Operative Treatment of Terrible Triad Injuries. J Orthop Trauma W.F. van Leeuwen, S.J. Janssen, T.G. Guitton, N. Chen, D. Ring. Interobserver Agreement in Diagnosing Early-Stage Kienböck Disease on Radiographs and Magnetic Resonance Imaging. HAND 2016 N.R. Paulino Pereira, D.W. Langerhuizen, S.J. Janssen, F.J. Hornicek, M.L. Ferrone, M.B. Harris, J.H. Schwab. Are Perioperative Allogeneic Blood Transfusions Associated With 90- Days Infection After Operative Treatment for Bone Metastases? Journal of Surgical Oncology

278 Report Of Scholarship O.D.R. van Wulfften Palthe, S.J. Janssen, J.S. Wunder, P.C. Ferguson, G. Wei, P.S. Rose, M.J. Yaszemski, F.H. Sim, P.J. Boland, J.H. Healey, F.J. Hornicek, J.H. Schwab. What questionnaires to use when measuring quality of life in sacral tumor patients: the updated sacral tumor survey. SPINE 2016 Lubberts, J.J. Mellema, S.J. Janssen, D. Ring. Fracture line distribution of olecranon fractures. Arch Orthop Trauma Surg 2016 A.M. Vranceanu, R.B. Beks, T.G. Guitton, S.J. Janssen, D. Ring. How do orthopedic surgeons address psychological aspects of illness? Arch Bone Jt Sur 2017 N.R. Paulino Pereira, L. Mclaughlin, S.J. Janssen, C.N. van Dijk, J.A.M. Bramer, I. Laufer, M.H. Bilsky, J.H. Schwab. The SORG nomogram accurately predicts 3- and 12-months survival for operable spine metastatic disease: External validation. Journal of Surgical Oncology 2017 N.R. Paulino Pereira, S.J. Janssen, K.A. Raskin, F.J. Hornicek, M.L. ferrone, J.H. Shin, J.A.M. Bramer, C.N. van Dijk, J.H. Schwab. Most efficient questionnaires to measure quality of life, physical function, and pain in patients with metastatic spine disease: a cross-sectional prospective survey study. The Spine Journal 2017 N.C. Theyskens, N.R. Paulino Pereira, S.J. Janssen, C.M. Bono, J.H. Schwab, T.D. Cha. The prevalence of spinal epidural lipomatosis on magnetic resonance imaging. The Spine Journal 2017 J.H. Schwab, S.J. Janssen, N.R. Paulino Pereira, Y.L.E. Chen, J.C. Wain, T.F. Delaney, F.J. Hornicek. Quality of life after resection of a chordoma of the mobile spine. The Bone & Joint Journal 2017 G.T.T. Helmerhorst, T. Teunis, S.J. Janssen, D. Ring. An epidemic of the use, misuse and overdose of opioids and deaths due to overdose, in the United States and Canada. The Bone & Joint Journal 2017 S.J. Janssen, N.R. Paulino Pereira, T.A. Meijs, M.A. Bredella, M.L. Ferrone, C.N. van Dijk, J.A.M. Bramer, S.A. Lozano-Calderon, J.H. Schwab. Predicting pathological fracture in femoral metastases using a clinical CT scan based algorithm: A case-control study. Journal of Orthopaedic Science 2017 S.J. Janssen, J.A.M. Bramer, T.G. Guitton, F.J. Hornicek, J.H. Schwab. Management of metastatic humeral fractures: Variations according to orthopedic subspecialty, tumor characteristics. Orthopaedics & Traumatology: Surgery & Research 2017 T.J. Crijns, S.J. Janssen, J.T. Davis, D. Ring, H.B. Sanchez. Reliability of proximal femur fractures: does clinical experience matter? Injury 2018 N. Stoop, B.A.T.D. van der Gronde, S.J. Janssen, M.T. Kuntz, D. Ring, N.C. Chen. Incidental Flexor Carpi Radialis Tendinopathy on Magnetic Resonance Imaging. HAND 2018 S.J. Janssen, P.T. Ogink, J.H. Schwab. The Prevalence of Incidental and Symptomatic Lumbar Synovial Facet Cysts. Clin Spine Surg 2018 Zhang, M. Tarabochia, S.J. Janssen, D. Ring, N. Chen. Acute compartment syndrome in patients undergoing fasciotomy of the forearm and the leg. Int Orthop 2018 J. van der Vis, S.J. Janssen, R. Haverlag, M.P.J. van den Bekerom. Functional outcome in patients who underwent distal biceps tendon repair. Arch Orthop Trauma Surg 2018 Q.C.B.S. Thio, A. Goudriaan, S.J. Janssen, N.R. Paulino Pereira, D.M. Sciubba, R.P. Rosovksy, J.H. Schwab. Prognostic role of neutrophil-to-lymphocyte ratio and platelet- tolymphocyte ratio in patients with bone metastases. Br J Cancer

279 Report Of Scholarship O.Q. Groot, P.T. Ogink, S.J. Janssen, N.R. Paulino Pereira, S.A. Lozano-Calderon, K. Raskin, F. Hornicek, J.H. Schwab. High Risk of Venous Thromboembolism After Surgery for Long Bone Metastases: A Retrospective Study of 682 Patients. Clin Orthop Relat Res 2018 PUBLICATIONS, NON-PEER REVIEWED S.J. Janssen. Het beoordelen van de intra-articulaire basis midphalanx fractuur. Het Verband 2017 PODIUM PRESENTATIONS (PRESENTER) S.J. Janssen, P. Jayakumar, D.P. ter Meulen, D. van Deurzen, D. Ring - Quantitative 3-Dimensional computed tomography modelling of isolated greater tuberosity fractures of the proximal humerus. May, 2014, Smith day, MGH, Boston, Massachusetts, USA B. Lubberts, S.J. Janssen, J. Mellema, D. Ring - Quantitative 3-dimensional computed tomography analysis of olecranon fractures. May, 2014, Smith day, MGH, Boston, Massachusetts, USA S.J. Janssen, B. Lubberts, J. Mellema, D. Ring - Fracture morphology of olecranon fractures: Quantitative 3D CT analysis. June, 2014, Harvard Orthopaedic Trauma Research Day, MGH, Boston, Massachusetts, USA S.J. Janssen, J.T.P. Kortlever, J.E. Ready, K.A. Raskin, M.L. Ferrone, F.J. Hornicek, J.H. Schwab Risk factors for complications after surgery for metastatic femur fractures. October, 2014, Musculoskeletal Tumor Society, Houston, Texas, USA S.J. Janssen, J.T.P. Kortlever, J.E. Ready, K.A. Raskin, S. Lozano-Calderon, M.L. Ferrone, F.J. Hornicek, J.H. Schwab Risk factors for complications after surgery for metastatic proximal femoral disease. March, 2015, American Academy of Orthopaedic Surgeons, Las Vegas, Nevada, USA S.J. Janssen, M. van Dijke, J.E. Ready, S. Lozano-Calderon, K.A. Raskin, M.L. Ferrone, F.J. Hornicek, J.H. Schwab Risk factors for complications after surgery for metastatic humerus disease. March, 2015, American Academy of Orthopaedic Surgeons, Las Vegas, Nevada, USA S.J. Janssen, T. Teunis, T.G. Guitton, D. Ring. Do Surgeons Treat Their Patients Like They Would Treat Themselves? May, 2015, Smith day, MGH, Boston, Massachusetts, USA B. Lubberts, S.J. Janssen, J. Molleman, T.G. Guitton, D. Ring. What Middle Phalanx Base Fracture Characteristics are Most Reliable and Useful for Surgical Decision-making? May, 2015, Smith day, MGH, Boston, Massachusetts, USA B. Lubberts, S.J. Janssen, T. Teunis, T.G. Guitton, D. Ring. Do Surgeons Treat Their Patients Like They Would Treat Themselves? June, 2015, Harvard Orthopaedic Trauma Research Day, MGH, Boston, Massachusetts, USA Mohamadi, A.R. Kachooei, A. Moradi, S.J. Janssen, D. Ring. The Influence of Dominant Limb Involvement on DASH and QuickDASH. May, 2015, Smith day, MGH, Boston, Massachusetts, USA. 277

280 Report Of Scholarship P.A.D. van Dijk, J.T.P. Kortlever, S.J. Janssen, J. Molleman, M.G.J.S. Hageman, D. Ring. Discrete pathophysiology is uncommon in patients with nonspecific arm pain. May, 2015, Smith day, MGH, Boston, Massachusetts, USA S.J. Janssen, A.S. van der Heijden, M. van Dijke, J.E. Ready, K.A. Raskin, M.L. Ferrone, F.J. Hornicek, J.H. Schwab Marshall Urist Young Investigator Award: Prognostication in Patients With Long Bone Metastases: Does a Boosting Algorithm Improve Survival Estimates? July, 2015, Association of Bone and Joint Surgeons, Eugene, Oregon, USA S.J. Janssen, Y. Braun, J.E. Ready, K.A. Raskin, M.L. Ferrone, F.J. Hornicek, J.H. Schwab. Are Allogeneic Blood Transfusions Associated With Decreased Survival After Surgery for Longbone Metastatic Fractures? July, 2015, Association of Bone and Joint Surgeons, Eugene, Oregon, USA S.J. Janssen, N.R. Paulino Pereira, K. Raskin, M. Ferrone, F. Hornicek, S. Lozano-Calderon, J.H. Schwab. Assessing Physical Function in Patients With Lower Extremity Bone Metastases: Which questionnaire is most useful? October, 2015, Musculoskeletal Tumor Society, Orlando, Florida, USA N.R. Paulino Pereira, S.J. Janssen, E. van Dijk, M.B. Harris, F.J. Hornicek, M. Ferrone, J.H. Schwab, Prognostication Of Survival In Patients Who Underwent Operative Treatment Of Metastatic Spine Disease. October, 2015, Musculoskeletal Tumor Society, Orlando, Florida, USA W. Van Leeuwen, S.J. Janssen, T. Guitton, N. Chen, D. Ring. Interobserver agreement in diagnosing early-stage Kienböck disease. May, 2016, Smith day, MGH, Boston, Massachusetts, USA W. Van Leeuwen, S.J. Janssen, D. Ring. Radiographic Progression of Kienböck Disease: Radial Shortening vs. No Surgery. May, 2016, Smith day, MGH, Boston, Massachusetts, USA S.J. Janssen, Jos Bramer, Thierry Guitton, Francis Hornicek, Joseph Schwab. Variation in Management of Metastatic Humeral Fractures. May, 2016, European Musculoskeletal Oncology Society, La Baule, France L.G. Jenis, S. Hartveldt, S.J. Janssen, K.B. Wood, T.D. Cha, J.H. Schwab, C.M. Bono. Is there an Association Between Epidural Steroid Injection and Postoperative Infection After Surgery for Lumbar Degenerative Spine Disease? April, 2016, Lumbar Spine Research Society Chicago, Illinois, USA O.D.R. van Wulfften Palthe, S.J. Janssen, J.S. Wunder, P.C. Ferguson, G. Wei, P.J. Boland, P.S. Rose, M.J. Yaszemski, F.H. Sim, F.J. Hornicek, J.H. Healey, J.H. Schwab, Sacral Tumor Study Group. What Questionnaires To Use When Measuring Quality Of Life In Sacral Tumor Patients? The Updated Sacral Tumor Survey. October, 2016, North American Spine Society, Boston, Massachusetts, USA N.R. Paulino Pereira, R.B. Beks, S.J. Janssen, M.B. Harris, F.J. Hornicek, M.L. Ferrone, J.H. Schwab. Are allogeneic blood transfusions associated with decreased survival after surgical treatment for spinal metastases? October, 2016, North American Spine Society, Boston, Massachusetts, USA N.R. Paulino Pereira, S.J. Janssen, E. Van Dijk, M.B. Harris, F.J. Hornicek, M.L. Ferrone, J.H. Schwab. Development of a Prognostic Survival Algorithm for Patients with Metastatic Spine Disease. October, 2016, North American Spine Society, Boston, Massachusetts, USA B.M. Verbeek, S.J. Janssen, J.A.M. Bramer, T.G. Guitton, F.J. Hornicek, J.H. Schwab. Variation in Management of Metastatic Humeral Fractures: Orthopedic Oncologist versus Trauma Surgeons. October, 2016, Musculoskeletal Tumor Society, Detroit, Michigan, USA 278

281 Report Of Scholarship N.R. Paulino Pereira, D. Langerhuizen, S.J. Janssen, F.J. Hornicek, M.L. Ferrone, M.B. Harris, J.H. Schwab. Are Perioperative Allogeneic Blood Transfusions Associated With 90-days Infection After Operative Treatment For Bone Metastases? October, 2016, Musculoskeletal Tumor Society, Detroit, Michigan, USA Q. Van der Vliet, N.R. Paulino Pereira, S.J. Janssen, F.J. Hornicek, M.L. Ferrone, J.A.M. Bramer, C.N. van Dijk, J.H. Schwab. Quality Of Life, Pain Interference, Anxiety, And Depression In Patients With Metastatic Bone Disease. October, 2016, Musculoskeletal Tumor Society, Detroit, Michigan, USA L.G. Jenis, S. Hartveldt, S.J. Janssen, K.B. Wood, T.D. Cha, J.H. Schwab, C.M. Bono. Is there an Association Between Epidural Steroid Injection and Postoperative Infection After Surgery for Lumbar Degenerative Spine Disease? October, 2016, North American Spine Society, Boston, Massachusetts, USA N.C. Theyskens, N.R. Paulino Pereira, S.J. Janssen, C.M. Bono, J.H. Schwab, T.D. Cha. The Prevalence of Spinal Epidural Lipomatosis on Magnetic Resonance Imaging. October, 2016, North American Spine Society, Boston, Massachusetts, USA A. Lans, S.J. Janssen, D. Ring. Off-Hour Surgery Among Orthopedic Subspecialties at an Urban, Quaternary-Care, Level 1 Trauma Center. January, 2017, Surgical resident trauma conference, Soestduinen, The Netherlands N.R. Paulino Pereira, L. Mclaughlin, S.J. Janssen, C.N. van Dijk, J.A.M. Bramer, I. Laufer, M.H. Bilsky, J.H. Schwab. External Validation of a Nomogram That Predicts 1, 3 and 12 Months Survival for Operable Spine Metastatic Disease. April, 2017, European Musculoskeletal Oncology Society, Budapest, Hungary J.H. Schwab, S.J. Janssen, N.R. Paulino Pereira, Y.L.E. Chen, J.C. Wain, T.F. Delaney, F.J. Hornicek. Quality of Life after Resection of Mobile Spine Chordoma. April, 2017, European Musculoskeletal Oncology Society, Budapest, Hungary D. Zhang, S.J. Janssen, B.E. Earp, P. Blazar. Predictors of Future Contralateral Carpal Tunnel Release. April, 2018, Smith Day, MGH, Boston, Massachusetts, USA S.J. Janssen, D.W.G. Langerhuizen, J.H. Schwab, J.A.M. Bramer. Outcome After Endoprosthetic Reconstruction Of Proximal Femoral Tumors: A Systematic Review. May, 2018, European Musculoskeletal Oncology Society, Amsterdam, The Netherlands POSTER PRESENTATIONS (PRESENTER) S.J. Janssen, D.P. ter Meulen, M.G.J.S. Hageman, B. Earp, D. Ring - Quantitative 3-dimensional computed tomography analyses of intra-articular fractures of the base of the middle phalanx. April, 2014, Scientific Advisory Committee conference, MGH, Boston, Massachusetts, USA S.J. Janssen, T. Teunis, F.J. Hornicek, J.H. Schwab Spine Oncology Study Group Outcomes Questionnaire Revised. May, 2014, Research career development conference, MGH, Boston, Massachusetts, USA S.J. Janssen, A. van der Heijden, M. van Dijke, J. Ready, K. Raskin, M. Ferrone, F. Hornicek, J.H. Schwab - Survival prognostication in patients with metastatic long-bone fractures: a machine learning and classic scoring algorithm to estimate survival. October, 2014, MGH Clinical Research Day, MGH, Boston, Massachusetts, USA 279

282 Report Of Scholarship S.J. Janssen, Y. Braun, J. Ready, K. Raskin, M. Ferrone, F. Hornicek, J.H. Schwab - Association of Perioperative Allogeneic Blood Transfusion with Survival in Patients with Metastatic Long-Bone Fractures. October, 2014, MGH Clinical Research Day, MGH, Boston, Massachusetts, USA C. Park, P. Osler, M. Bredella, K. Hess, S.J. Janssen, Y. Chen, T. Delaney, F. Hornicek, J.H. Schwab - Sacral Insufficiency Fractures after High-Dose Radiation with and without Partial Sacrectomy: A Twenty Year Retrospective. October, 2014, MGH Clinical Research Day, MGH, Boston, Massachusetts, USA J.J. Mellema, S.J. Janssen, T.G. Guitton, D. Ring - Quantitative Three-Dimensional Computed Tomography Measurements of Coronoid Fractures. October, 2014, MGH Clinical Research Day, MGH, Boston, Massachusetts, USA S.J. Janssen, T. Teunis, F.J. Hornicek, J.H. Schwab Spine Oncology Study Group Outcomes Questionnaire Revised. October, 2014, Musculoskeletal Tumor Society, Houston, Texas, USA S.J. Janssen, Y. Braun, J. Ready, K. Raskin, M. Ferrone, F. Hornicek, J.H. Schwab - Association of Perioperative Allogeneic Blood Transfusion with Survival in Patients with Metastatic Long-Bone Fractures. April, 2015, Scientific Advisory Committee conference, MGH, Boston, Massachusetts, USA S.J. Janssen, N.R. Paulino Pereira, T. Meijs, K. Raskin, M. Ferrone, F. Hornicek, S. Lozano- Calderon, J.H. Schwab. Structural Analysis of Metastatic Femoral Lesions Using Clinical Computed Tomography Scans. October, 2015, Musculoskeletal Tumor Society, Orlando, Florida, USA N.R. Paulino Pereira, N. Stoop, S. Hartveldt, S.J. Janssen, F.J. Hornicek, J.H. Schwab. Functional and oncological outcome after chondrosarcoma resection of the mobile spine. October, 2015, Musculoskeletal Tumor Society, Orlando, Florida, USA O. van Wulfften Palthe, S.J. Janssen, M. van Dijke, T. Cha, K. Wood, L. Borges, M. Harris, C. Bono, J.H. Schwab. Comparison of decompression with and without fusion for patients with synovial facet cysts: a retrospective study of 314 consecutive patients. October 2015, North American Spine Society, Chicago, USA D. Langerhuizen, S.J. Janssen, Q. van der Vliet, K. Raskin, M. Ferrone, F.J. Hornicek, J.H. Schwab, S. Lozano-Calderon. Metastasectomy versus Intralesional Resection in the Treatment of Bone Metastases from Renal Cell Carcinoma. October, 2015, MGH Clinical Research Day, MGH, Boston, Massachusetts, USA N.R. Paulino Pereira, R.B. Beks, D. Langerhuizen, S.J. Janssen, F.J. Hornicek, M. Ferrone, J.H. Schwab. Are allogeneic blood transfusions associated with decreased survival after surgical treatment for spinal metastases? October, 2015, MGH Clinical Research Day, MGH, Boston, Massachusetts, USA S.J. Janssen, E. van Rein, N.R. Paulino Pereira, K. Raskin, M. Ferrone, F.J. Hornicek, S. Lozano-Calderon, J.H. Schwab. Assessing Physical Function in Patients with Lower Extremity Bone Metastases: Which questionnaire is most useful? October, 2015, MGH Clinical Research Day, MGH, Boston, Massachusetts, USA E. van Rein, S.J. Janssen, K. Raskin, M. Ferrone, F.J. Hornicek, S. Lozano-Calderon, J.H. Schwab. Is There a Discrepancy Between Patient and Clinician Perceived Functional Outcome in Patients with Extremity Bone Metastases? October, 2015, MGH Clinical Research Day, MGH, Boston, Massachusetts, USA Kachooei, A. Moradi, S.J. Janssen, D. Ring. The Influence of Dominant Limb Involvement on Dash And QuickDash. September, 2015, ASSH, Seattle, USA 280

283 Report Of Scholarship S.J. Janssen, N.R. Paulino Pereira, K. Raskin, M. Ferrone, F.J. Hornicek, S. Lozano-Calderon, J.H. Schwab. Assessing Physical Function in Patients with Lower Extremity Bone Metastases. March, 2016, AAOS, Orlando, USA S.J. Janssen, Teun Teunis, Francis Hornicek, Niek van Dijk, Jos Bramer, Joseph Schwab. Outcome After Fixation of Metastatic Proximal Femoral Fractures: A Systematic Review. May, 2016, European Musculoskeletal Oncology Society, La Baule, France W.F. van Leeuwen, S.J. Janssen, D. Ring, N. Chen. Incidental magnetic resonance imaging signal changes in the extensor carpi radialis brevis origin are more common with age. May, 2016, Massachusetts General Hospital Orthopaedic Research Retreat, Boston, USA S. Hartveldt, S.J. Janssen, K.B. Wood, T.D. Cha, J.H. Schwab, C.M. Bono, L.G. Jenis. Is There An Association of Epidural Corticosteroid Injection with Postoperative Surgical Site Infection after Surgery for Lumbar Degenerative Spine Disease? April, 2016, Amsterdam Medical School Convention, Amsterdam, The Netherlands J.H. Schwab, S.J. Janssen, N.R. Paulino Pereira, Y.L. Chen, J. wain, T. Delaney, F.J. Hornicek. Quality of Life after Resection of Mobile Spine Chordoma. October, 2016, Musculoskeletal Tumor Society, Detroit, Michigan, USA N.R. Paulino Pereira, S.J. Janssen, K.A. Raskin, F.J. Hornicek, M.L. Ferrone, J.H. Shin, J.A.M. Bramer, C.N. van Dijk, J.H. Schwab. Most Efficient Questionnaires To Measure Quality Of Life, Physical Function, and Pain in Patients With Metastatic Spine Disease. October, 2016, Musculoskeletal Tumor Society, Detroit, Michigan, USA Y. Braun, O.D. van Wulfften Palthe, S.J. Janssen, B.M. Verbeek, J.H. Schwab. Intraobserver Agreement in Measuring the Neural Foramina Using a 360-Degree Ultrasound: A Human Cadaver Study. October, 2016, North American Spine Society, Boston, Massachusetts, USA S.J. Janssen, H.H. Hermanussen, T.G. Guitton, M.P.J. van den Bekerom, D.F.P. van Deurzen, D. Ring. Greater Tuberosity Fractures: Does Fracture Assessment and Treatment Recommendation Vary Based on Imaging Modality? November, 2016, Traumadagen, Amsterdam, Nederland S.J. Janssen, N.R. Paulino Pereira, Q.C.B.S. Thio, K.A. Raskin, J.A.M. Bramer, S.A. Lozano- Calderon, J.H. Schwab. Physical Function and Pain Intensity in Patients With Metastatic Bone Disease. May, 2018, European Musculoskeletal Oncology Society, Amsterdam, The Netherlands N.R. Paulino Pereira, S.J. Janssen, N. Stoop, S. Hartveldt, F. Hornicek, J.H. Schwab. Physical Function and Pain Intensity in Patients With Metastatic Bone Disease. May, 2018, European Musculoskeletal Oncology Society, Amsterdam, The Netherlands 281

284

285 About The Author Stein Jasper Janssen was born on May 26, 1987 in Utrecht, The Netherlands. After graduating high school (Sint Bonifatius College in Utrecht) in 2005, he started medical school at Utrecht University. During medical school he participated in several extracurricular organizations and did medical internships in India and Peru. In 2012, he did an elective rotation in orthopaedic surgery at the University Medical Center in Utrecht and started his first research project (under supervision of dr. J.J. Verlaan). After graduating from medical school in 2013, he did a six month research fellowship at the Hand & Upper extremity department of the Massachusetts General Hospital Harvard Medical School in Boston (under supervision of Professor David Ring). Inspired by these clinical research activities he pursued a two year research fellowship at the Orthopaedic Oncology and Spine department of the Massachusetts General Hospital (under supervision of Associate Professor Joseph Schwab). This work has resulted in numerous presentations at international conferences, peer-reviewed publications, and this PhD thesis. In 2015, he was awarded with the Marshall Urist Young Investigator Award by the Association of Bone and Joint Surgeons (founders of Clinical Orthopaedics and Related Research). He received several grants to support his research activities, including one from the Dutch Cancer Society. In July 2016 he started his residency in orthopaedic surgery (Academic Medical Center Amsterdam, under supervision of prof. dr. G.M.M.J. Kerkhoffs) at the department of general surgery of the Onze Lieve Vrouwe Gasthuis in Amsterdam. Currently, he is working at the department of orthopaedic surgery of the Amphia Ziekenhuis in Breda. 283