Management of endometrial abnormalities in postmenopausal women, an individualized approach Breijer, M.C.

Transcription

1 UvA-DARE (Digital Academic Repository) Management of endometrial abnormalities in postmenopausal women, an individualized approach Breijer, M.C. Link to publication Citation for published version (APA): Breijer, M. C. (2013). Management of endometrial abnormalities in postmenopausal women, an individualized approach General rights It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons). Disclaimer/Complaints regulations If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible. UvA-DARE is a service provided by the library of the University of Amsterdam ( Download date: 30 Sep 2018

2 Management of endometrial abnormalities in postmenopausal women, an individualized approach Maria Cornelia Breijer

3 Management of endometrial abnormalities in postmenopausal women, an individualized approach Maria Cornelia Breijer

4 Financial support for printing of this thesis was provided by: Vrouwenkliniek Academisch Medisch Centrum Astellas Pharma B.V. Olympus Nederland B.V. Johnson & Johnson Medical B.V. Chipsoft B.V. Sorg-Saem B.V. Astraia Software Benelux Bayer Healthcare BMA B.V. (Mosos) Abbott B.V. Albert Schweitzer Ziekenhuis Medical Dynamics Ton en Hetty Breijer Corrie Breijer - Achterberg Cover photo: Layout: Printing: Peter Dekker Nikki Vermeulen, Ridderprint BV, Ridderkerk, the Netherlands Ridderprint BV, Ridderkerk, the Netherlands ISBN: All rights reseved. No part of this thesis may be reproduced, stored in a retirieval system of any nature, or transmitted in any form or by any means, without prior written permission of the author.

5 Management of endometrial abnormalities in postmenopausal women, an individualized approach ACADEMISCH PROEFSCHRIFT ter verkrijging van de graad van doctor aan de Universiteit van Amsterdam op gezag van de Rector Magnificus prof. dr. D.C. van den Boom ten overstaan van een door het college voor promoties ingestelde commissie, in het openbaar te verdedigen in de Agnietenkapel op dinsdag 25 juni 2013, te 10:00 uur door Maria Cornelia Breijer geboren te Roosendaal en Nispen

6 Promotores: Co-promotores: Prof. dr. B.W.J. Mol Prof. dr. R.H.M. Verheijen Dr. A. Timmermans Dr. B.C. Opmeer Overige leden: Prof. dr. J. Stoker Prof. dr. M.J. van de Vijver Prof. dr. M.P.M. Burger Dr. W.M. Ankum Prof. dr. C.W. Burger Prof. dr. R.F.P.M. Kruitwagen Faculteit der Geneeskunde

7 Contents Chapter 1 General introduction 7 Chapter 2 Prediction models in women with postmenopausal bleeding: 17 a systematic review Chapter 3 Diagnostic strategies for endometrial cancer in women with 35 postmenopausal bleeding: cost-effectiveness of individualized strategies Chapter 4 External validation of a mathematical model to estimate the 49 probability of endometrial cancer in women with postmenopausal bleeding Chapter 5 Factors attributing to the failure of endometrial sampling in 63 women with postmenopausal bleeding Chapter 6 Comparing two diagnostic strategies based on the predicted 77 probability of a failed endometrial biopsy in women with postmenopausal bleeding: a cost-minimization analysis Chapter 7 A structured assessment to decrease the amount of inconclusive 85 endometrial biopsies in women with postmenopausal bleeding Chapter 8 The capacity of endometrial thickness measurement to diagnose 95 endometrial carcinoma in asymptomatic postmenopausal women: a systematic review and meta-analysis Chapter 9 Summary and General discussion 117 Nederlandse samenvatting 131 Curriculum Vitae 135 Dankwoord 137 PhD Portfolio 141

8

9 1 General introduction Partially adapted from: Diagnostic strategies for postmenopausal bleeding. M.C. Breijer, A. Timmermans, H.C. van Doorn, B.W.J. Mol, B.C. Opmeer. Obstetrics and Gynecology International 2010; 2010: Diagnostic evaluation of the endometrium in postmenopausal bleeding: an evidence-based approach. N. Van Hanegem, M.C. Breijer, K.S. Khan, T.J. Clark, M.P. Burger, B.W.J. Mol, A. Timmermans. Maturitas Feb;68(2):

10

11 General introduction 1 Postmenopausal bleeding (PMB), defined as blood loss occurring at least 12 months after menopause, is a common complaint in general gynecological practice. The prevalence of PMB is approximately 10% immediately after menopause. 1 Postmenopausal bleeding (PMB) signals endometrial carcinoma in around 10% of cases 2, 3 or less serious conditions, such as benign endometrial polyps, in a further 20 40%. 4-6 Endometrial carcinoma is the most common gynecologic malignancy and 95% of women with endometrial carcinoma present with PMB. 5, 7 Unlike ovarian cancer, endometrial cancer often presents at an early stage, when there is a possibility of curative treatment by hysterectomy (and bilateral salpingooophorectomy); therefore early, accurate and timely diagnosis is important and any PMB needs further investigation. There are five different diagnostic tests available to investigate women with PMB: transvaginal sonographic (TVS) measurement of endometrial thickness, saline infused sonography (SIS), outpatient endometrial biopsy, hysteroscopy and dilatation and curettage (D&C). D&C was traditionally the method of choice for investigating women with PMB. In approximately 60% of the D&C procedures however, less than half of the uterine cavity is evaluated by the curettage. 8 Another drawback of D&C is that this procedure is performed under general anesthesia in an inpatient setting. Endometrial biopsy and hysteroscopy have now almost completely replaced D&C. Transvaginal sonography The main goal of TVS is to exclude endometrial cancer. There are four meta-analyses published on this subject Almost every guideline refers to the meta-analysis performed by Smith-Bindman et al. in It used traditional statistical methods to combine the data from 35 published studies regarding the use of TVS in the evaluation of women with PMB. Using the reported data from each study, 2 2 tables were constructed of endometrial thickness measured by TVS (above or below a threshold) against the presence or absence of endometrial carcinoma. With a cut-off value of 5 mm, the sensitivity for detecting endometrial carcinoma was 96%, and the specificity 61%. This combination of sensitivity and specificity reduces a pre-test probability of 10% for endometrial cancer to a post-test probability (for a negative test) of 1%. Thus, based on the post-test probability of 1%, conservative management is recommended to women with an endometrial thickness of 5 mm. The three other meta-analyses reached different conclusions. Tabor et al. 11 conducted a meta-analysis of nine studies. They included studies only if the corresponding author was able to supply original data. For each included study, the median endometrial thickness per center was calculated and multiples of the median were used to pool data. They chose not to use a cut-off value, because there were statistically significant differences in endometrial thickness between centers, which may reflect differences in the populations studied or in the method of measuring endometrial thickness by TVS. In this 9

12 Chapter 1 study, a sensitivity of 96% and a specificity of 50% were found. These values give a post-test probability for a negative test of about 1% with a pre-test probability of carcinoma of 10%. These results are comparable to those of Smith-Bindman et al., but the authors disagreed on the interpretation of the results. The conclusion of Tabor et al. was that a 4% falsenegative rate is not acceptable and therefore the use of TVS in the evaluation of PMB is not recommended prior to invasive testing. Gupta et al. 9 performed a systematic quantitative review in which they focused on study quality assessment. None of the nine studies that used a cut-off for endometrial thickness of 4 mm were of good quality. Only four studies (out of 21) used a 5 mm cut-off, but these employed the best quality criteria. Pooling of the results of these four studies resulted in a negative likelihood ratio of This likelihood ratio implies that a patient with a negative test result (endometrial thickness 5 mm) and pre-test probability of 10% would have a posttest probability of 2.5%. Their conclusion was that TVS can be used to rule out endometrial hyperplasia or carcinoma using an endometrial thickness cut-off of 5 mm. With respect to meta-analysis of randomized trials, individual patient data are considered to be superior to meta-analysis of the literature. 13 The use of individual patient data instead of published summary data gives less optimistic but more accurate conclusions. In diagnostic reviews the same might apply. Timmermans et al. 12 tried to overcome this limitation using a meta-analytic approach in which individual patient data from a series of original studies were combined. This study showed that in previous studies and meta-analyses, the diagnostic accuracy of TVS had been overestimated. Timmermans et al. found a lower diagnostic accuracy for TVS than was reported previously: a sensitivity of 95% and a specificity of 47% at a cut-off of 4 mm, giving a post-test probability for a negative test of 1.2%. At a cut-off of 3 mm, they found a sensitivity of 98% and a specificity of 35%, giving a likelihood ratio for a negative test result of Thus, a cut-off level of 3 mm reduced a pre-test probability of 10% to a 0.7% post-test probability. The conclusion of this article was that the use of TVS measurement of endometrial thickness remains justified, but with a recommended cut-off level of 3 mm. Invasive endometrial assessment methods Endometrial biopsy Dijkhuizen et al. 14 performed a meta-analysis comparing different minimally invasive endometrial biopsy devices. In women with PMB, endometrial sampling with both the Pipelle device (Pipelle de Cornier, Paris, France) and the Vabra device (Berkeley Medevices, Inc., Richmond, CA, USA) are very sensitive techniques for the detection of endometrial carcinoma, with detection rates of 99.6% and 97.1% respectively. 14 However, in a substantial number of cases outpatient endometrial sampling fails because of technical problems (12-23%) or because of an insufficient amount of tissue for a histopathological diagnosis 10

13 General introduction 1 (16-68%) In 6% of these failures, women are subsequently diagnosed with both malignant and premalignant disorders of the endometrial tissue and cannot be reassured without further testing. 10 Hysteroscopy Compared with other methods, hysteroscopy offers the possibility of visualizing macroscopic or focal abnormalities and taking directed biopsies. 15, 16 With the development of smaller diameter hysteroscopic systems and the introduction of a vaginoscopic approach to hysteroscopy (without the use of a speculum or tenaculum), patient acceptance has improved considerably and hysteroscopy nowadays can be performed in an outpatient setting without the use of anaesthesia Clark et al. 20 performed a systematic quantitative review in which they focussed on the diagnostic accuracy of hysteroscopy in diagnosing endometrial cancer or hyperplasia. Postmenopausal women represented 29% of the populations studied. Only two studies concerning postmenopausal women were rated as high quality. Pooled data from these studies showed a post-test probability of a positive test of 71.8% (95% CI %). Sensitivity and specificity were not separately described for pre- and postmenopausal women, but the overall sensitivity and specificity were 86.4% and 99.2% respectively. The authors concluded that when the uterine cavity is adequately visualized, hysteroscopy is highly accurate and clinically useful in the diagnosis of endometrial cancer. However, its high accuracy relates to diagnosing cancer rather than its exclusion. Another systematic review and meta-analysis of diagnostic hysteroscopy was performed by van Dongen et al. 21 This article focused on studies on the use of hysteroscopy in the diagnosis of intrauterine abnormalities rather than endometrial cancer per se, because Clark et al. had already shown in their meta-analysis that diagnostic hysteroscopy is accurate in the diagnosis of endometrial cancer. 20 In this review five studies of postmenopausal women with homogeneous data were included. The pooled sensitivity and specificity in the assessment of uterine cavity abnormalities were 96% (95% CI 93 99%) and 90% (95% CI 83 95%) respectively. With a pre-test probability of uterine cavity abnormalities of 61.0% (the prevalence in this group), they found a post-test probability for a positive test of 93% (95% CI 88 95%). The conclusion was that this meta-analysis gives strong evidence that diagnostic hysteroscopy is accurate in the diagnosis of intrauterine abnormalities. International guidelines The published national and international guidelines describe different diagnostic pathways in the diagnostic work-up of women with postmenopausal bleeding. 7, All guidelines advise referral to a gynecologist for examination, cervical cytology and TVS or endometrial biopsy. In all guidelines except for the United States (US) guideline, the first step is TVS, based on the high sensitivity and non-invasive character of the procedure. Different 11

14 Chapter 1 guidelines use different cut-off values of endometrial thickness, varying from 3 mm to 5 7, 10, 22, 24 mm. These cut-off points are mostly based on the meta-analysis by Smith-Bindman, but also on Swedish literature, 23 and the review by Gupta et al. 9 The most important issue is what probability of endometrial carcinoma is deemed acceptable after a negative test. The 7, 23, 25 US guideline recommends either TVS or endometrial biopsy. The European guidelines advice endometrial sampling only when the endometrial thickness is above the cut-off value, possibly together with a SIS to distinguish between diffuse and focal pathology. Where the endometrium is thin, the guidelines recommend conservative management. Only the Scottish guideline recommends further investigation if the clinician, the patient or both are not reassured. The exact sequence of investigation will depend upon clinical judgment, local resources, local expertise and patient preference. 25 Prediction models for endometrial carcinoma in women with postmenopausal bleeding In women with PMB there is considerable variability in endometrial thickness and the likelihood of endometrial carcinoma. 26 Individual patient characteristics, including age, time since menopause, obesity, hypertension, diabetes mellitus and reproductive factors, are associated with a higher risk of endometrial carcinoma While the probability of PMB decreases with increasing age, 27 the probability of endometrial cancer in women with PMB increases significantly with increasing age. The probability rises from 1% in women younger than 50 years of age to 24% in women older than 80 years. 33 In clinical practice, tests are commonly combined in diagnostic sequences and disease probabilities are usually estimated in a hierarchical manner: first combining information from history and examination, followed by additional information obtained from diagnostic tests. The post-test probability is not only dependent on test characteristics, but also on the pretest probability, which is altered by patient characteristics. However, current diagnostic policy in women with PMB is not based on these patient-specific risk factors, but only on 7, one fixed cut-off point for endometrial thickness. Several articles have studied this subject and developed models to estimate the individual chance of endometrial carcinoma in women presenting with PMB. For a successful implementation, a prediction model should be validated externally in an independent population. None of the prediction models developed are externally validated. Diagnostic strategies In Figure 1. a flowchart with possible diagnostic pathways is presented. Two areas in this flowchart require further research: (1) Modeling to calculate the individual probability of endometrial cancer based on patient characteristics and the implementation of such a model in daily practice and (2) the diagnostic approach to benign endometrial abnormalities after a malignancy has been ruled out. 12

15 General introduction 1 Patient with postmenopausal bleeding Patient characteristics prediction model Based on predicted probability Transvaginal Sonography Expectant management Endometrial thickness 4mm Endometrial thickness > 4mm Endometrial Sampling Recurrent Bleeding Expectant management Recurrent bleeding (Pre-) Malignancy ruled out Insufficient endometrial sample (Pre-) Malignancy Expectant management Hysteroscopy Recurrent Bleeding Figure 1. Possible diagnostic strategies for postmenopausal bleeding. The areas surrounded by a dotted square require further research. Asymptomatic postmenopausal women In contrast with the guidelines on the management of women with PMB, clinicians are faced with uncertainty when endometrial thickness is measured for asymptomatic postmenopausal women. Symptom free women sometimes undergo transvaginal ultrasound for other indications than postmenopausal bleeding such as prolapse or abdominal complaints. Inevitably, the endometrium is visualised and a thickened endometrium may then be observed. It is not known how to manage such patients in whom a thick endometrium is observed incidentally. AIMS OF THE THESIS 1. To evaluate different diagnostic strategies for women with postmenopausal bleeding including strategies based on individual probabilities. 2. To study the value of endometrial thickness measurement in asymptomatic postmenopausal women. 13

16 Chapter 1 Outline of the thesis There are several articles on the development of prediction models to estimate the individual chance for endometrial cancer in women presenting with PMB. Chapter 2 describes a review to provide an overview of the existing prediction models for endometrial carcinoma in women with PMB. Chapter 3 describes a decision analytic model to investigate the cost-effectiveness of three diagnostic strategies based on the individual probabilities of women presenting with PMB. In Chapter 4 the results of an external validation study of two previously developed models on women with PMB are presented. A study on which patient and doctor related factors contribute to the failure of outpatient endometrial biopsies and a multivariable prediction model on the failure of outpatient endometrial biopsies are presented in Chapter 5. In Chapter 6 an analysis to investigate whether the multivariable prediction model described in chapter 5 could reduce the costs of diagnostic testing in women with PMB through a costminimization analysis is described. In Chapter 7 the results of a study to determine whether structured assessment of endometrial biopsy specimens with strict criteria decreases the amount of inconclusive samples are presented. Chapter 8 describes the results of a systematic review on the endometrial thickness, the prevalence of endometrial malignancy and the accuracy of endometrial thickness measurement in asymptomatic postmenopausal women. 14

17 General introduction 1 REFEREnCES 1. Astrup K, Olivarius NF. Frequency of spontaneously occurring postmenopausal bleeding in the general population. Acta Obstet Gynecol Scand 2004;83(2): Dijkhuizen FP, Brolmann HA, Potters AE, Bongers MY, Heinz AP. The accuracy of transvaginal ultrasonography in the diagnosis of endometrial abnormalities. Obstet Gynecol 1996;87(3): Emanuel M, Verdel M, Wamsteker K, Lammes F. An audit of true prevalence of intrauterine pathology: the hysteroscopic fi ndings, controled for pati ent selecti on in 1202 pati ents with abnormal uterine bleeding. Gynaecol Endosc 1995;4: Brenner PF. Differential diagnosis of abnormal uterine bleeding. Am J Obstet Gynecol 1996;175(3 Pt 2): Epstein E, Ramirez A, Skoog L, Valentin L. Dilatation and curettage fails to detect most focal lesions in the uterine cavity in women with postmenopausal bleeding. Acta Obstet Gynecol Scand 2001;80(12): Timmermans A, Veersema S. Office hysteroscopy in women with postmenopausal bleeding: see and treat of endometrial polyps using a Duckbill Polyp Snare. Gynecol Surg 2004;1(3): NVOG (Dutch Society of Obstetrics and Gynaecology). NVOG-richtlijn Abnormaal vaginaal bloedverlies in de menopauze [In Dutch]. NVOG Guideline: Abnormal vaginal bleeding during menopause Stock RJ, Kanbour A. Prehysterectomy curettage. Obstet Gynecol 1975;45(5): Gupta JK, Chien PF, Voit D, Clark TJ, Khan KS. Ultrasonographic endometrial thickness for diagnosing endometrial pathology in women with postmenopausal bleeding: a meta-analysis. Acta Obstet Gynecol Scand 2002;81(9): Smith-Bindman R, Kerlikowske K, Feldstein VA et al. Endovaginal ultrasound to exclude endometrial cancer and other endometrial abnormalities. JAMA 1998;280(17): Tabor A, Watt HC, Wald NJ. Endometrial thickness as a test for endometrial cancer in women with postmenopausal vaginal bleeding. Obstet Gynecol 2002;99(4): Timmermans A, Opmeer BC, Khan KS et al. Endometrial thickness measurement for detecting endometrial cancer in women with postmenopausal bleeding: a systematic review and meta-analysis. Obstet Gynecol 2010;116(1): Stewart LA, Parmar MK. Meta-analysis of the literature or of individual patient data: is there a difference? Lancet 1993;341(8842): Dijkhuizen FP, Mol BW, Brolmann HA, Heintz AP. The accuracy of endometrial sampling in the diagnosis of patients with endometrial carcinoma and hyperplasia: a meta-analysis. Cancer 2000;89(8): Gimpelson RJ, Rappold HO. A comparative study between panoramic hysteroscopy with directed biopsies and dilatation and curettage. A review of 276 cases. Am J Obstet Gynecol 1988;158(3 Pt 1): Loffer FD. Hysteroscopy with selective endometrial sampling compared with D&C for abnormal uterine bleeding: the value of a negative hysteroscopic view. Obstet Gynecol 1989;73(1): Bettocchi S, Selvaggi L. A vaginoscopic approach to reduce the pain of office hysteroscopy. J Am Assoc Gynecol Laparosc 1997;4(2): Bettocchi S, Nappi L, Ceci O, Selvaggi L. What does diagnostic hysteroscopy mean today? The role of the new techniques. Curr Opin Obstet Gynecol 2003;15(4): Cooper NA, Smith P, Khan KS, Clark TJ. Vaginoscopic approach to outpatient hysteroscopy: a systematic review of the effect on pain. BJOG 2010;117(5): Clark TJ, Voit D, Gupta JK, Hyde C, Song F, Khan KS. Accuracy of hysteroscopy in the diagnosis of endometrial cancer and hyperplasia: a systematic quantitative review. JAMA 2002;288(13): Van Dongen H, De Kroon CD, Jacobi CE, Trimbos JB, Jansen FW. Diagnostic hysteroscopy in abnormal uterine bleeding: a systematic review and meta-analysis. BJOG 2007;114(6): American College of Obstetricians and Gynecologists. ACOG Committ ee Opinion No. 426: The role of transvaginal ultrasonography in the evaluati on of postmenopausal bleeding. Obstet Gynecol 2009;113(2 Pt 1): Epstein E. Management of postmenopausal bleeding in Sweden: a need for increased use of hydrosonography and hysteroscopy. Acta Obstet Gynecol Scand 2004;83(1):

18 Chapter Goldstein RB, Bree RL, Benson CB et al. Evaluation of the woman with postmenopausal bleeding: Society of Radiologists in Ultrasound-Sponsored Consensus Conference statement. J Ultrasound Med 2001;20(10): Scottish Intercollegiate Guidelines Network. Investigation of postmenopausal bleeding. Scottish Intercollegiate Guidelines Network, Royal College of Physicians van Doorn LC, Dijkhuizen FP, Kruitwagen RF, Heintz AP, Kooi GS, Mol BW. Accuracy of transvaginal ultrasonography in diabetic or obese women with postmenopausal bleeding. Obstet Gynecol 2004;104(3): Anderson KE, Anderson E, Mink PJ et al. Diabetes and endometrial cancer in the Iowa women s health study. Cancer Epidemiol Biomarkers Prev 2001;10(6): Gull B, Karlsson B, Milsom I, Granberg S. Factors associated with endometrial thickness and uterine size in a random sample of postmenopausal women. Am J Obstet Gynecol 2001;185(2): McPherson CP, Sellers TA, Pott er JD, Bosti ck RM, Folsom AR. Reproducti ve factors and risk of endometrial cancer. The Iowa Women s Health Study. Am J Epidemiol 1996;143(12): Reeves GK, Pirie K, Beral V, Green J, Spencer E, Bull D. Cancer incidence and mortality in relation to body mass index in the Million Women Study: cohort study. BMJ 2007;335(7630): Weiderpass E, Persson I, Adami HO, Magnusson C, Lindgren A, Baron JA. Body size in different periods of life, diabetes mellitus, hypertension, and risk of postmenopausal endometrial cancer (Sweden). Cancer Causes Control 2000;11(2): Xu WH, Xiang YB, Ruan ZX et al. Menstrual and reproducti ve factors and endometrial cancer risk: Results from a populati on-based case-control study in urban Shanghai. Int J Cancer 2004;108(4): Gredmark T, Kvint S, Havel G, Matt sson LA. Histopathological fi ndings in women with postmenopausal bleeding. Br J Obstet Gynaecol 1995;102(2):

19 2 Prediction models in women with postmenopausal bleeding: a systematic review N. van Hanegem M.C. Breijer B.C. Opmeer B.W.J. Mol A. Timmermans Women's Health (2012) 8(3),

20 Chapter 2 Abstract Postmenopausal bleeding is associated with an elevated risk of having endometrial cancer. The aim of this review is to give an overview of existing prediction models on endometrial cancer in women with postmenopausal bleeding. In a systematic search of the literature, we identified nine prognostic studies, of which we assessed the quality, the different phases of development and their performance. From these data, we identified the most important predictor variables. None of the detected models completed external validation or impact analysis. Models including power Doppler showed best performance in internal validation, but Doppler in general gynecological practice is not easy accessible. We can conclude that we have indications that the first step in the approach of women with postmenopausal bleeding should be to distinguish between women with low risk versus high risk of having endometrial carcinoma and the next step would be to refer high-risk patients for further (invasive) testing. 18

21 Prediction models in women with postmenopausal bleeding: a systematic review InTRODuCTIOn Endometrial carcinoma is the most common gynecologic malignancy. Approximately 95% of women with endometrial carcinoma present with postmenopausal bleeding (PMB). 1, 2 PMB signals endometrial carcinoma, which is present in about 10% of cases, 3, 4 or less serious 3, 5-7 conditions, such as benign endometrial polyps or endometrial atrophy. To reduce invasive procedures in women with PMB, measurement of the endometrial thickness is used to stratify women into low versus high risk of having endometrial carcinoma. Measurement of endometrial thickness has shown to be accurate in excluding endometrial cancer, although the risk of endometrial carcinoma with a negative test is still % depending on the cutoff point used. 8, 9 In women with PMB there is considerable variability in endometrial thickness and the likelihood of endometrial carcinoma. 10 Individual patient characteristics, including age, time since menopause, obesity, hypertension, diabetes mellitus and reproductive factors, are associated with a higher risk of endometrial carcinoma While the probability of postmenopausal bleeding decreases with increasing age, 17 the probability of endometrial cancer in women with PMB increases significantly with increasing age. The probability rises from 1% in women younger than 50 years to 24% in women older than 80 years. 18 In clinical practice, tests are commonly combined in diagnostic sequences and disease probabilities are usually estimated in a hierarchical manner: first combining information from history and examination, followed by additional information obtained from diagnostic tests. The post-test probability is not only dependent on test characteristics but also on the pre-test probability, which is altered by patient s characteristics. However, current diagnostic policy in women with PMB is not based on these patient specific risk factors, but only on one 2, fixed cutoff point for endometrial thickness. Clinical doctors want to identify women with a high risk for endometrial cancer when presenting with PMB. Several articles have studied this subject and developed models to estimate the individual chance of endometrial carcinoma in women presenting with PMB. The purpose of this review is to give an overview of the existing prediction models for endometrial carcinoma in women with PMB, to assess their quality and to identify important predictor variables. 2 METHODS Study identification We performed a computerized MEDLINE and EMBASE search to identify all studies on prediction models in women with postmenopausal bleeding published from inception to June The search was limited to human studies, no restrictions were held concerning publication year or language. We included articles reporting on multivariable models 19

22 Chapter 2 predicting endometrial cancer in women with PMB. We checked references cited in the selected articles for further relevant prediction models not identified by the electronic searches. We used all known synonyms for the terms postmenopausal bleeding and endometrial cancer and we used a search-filter for prediction models. 22 The search strategy can be found in appendix 1. Study selection This review focused on articles, which report on a prediction model for endometrial carcinoma in women with PMB. In this review, a prediction model was defined as a multivariable model that expresses the chance of endometrial carcinoma as a function of two or more predictor variables. PMB was defined as vaginal bleeding after more than one year of amenorrhea after the age of 40 or persistent (>3 months) unscheduled bleeding on hormone replacement therapy (HRT). Two independently working reviewers (NvH and MB) selected the articles, by assessing titles and abstracts. If there were any doubts about eligibility after reading title and abstract, we read the full text version to make sure no articles were missed. In case of disagreement the article was included for full text reading and assessed by a third reviewer (AT). Study quality assessment A framework for quality assessment was developed based on the recommendations of Hayden et al. 23 and on a quality assessment framework for prediction models in subfertile women to predict the chance of pregnancy. 24 The framework was divided into four sections: study participation, predictor variables, outcome measurement and analysis. Each item in the different sections was scored with yes, no or unclear. Predictor variables All predictor variables were collected for each prediction model. The predictor variables are the potential predictors, which were tested, both during model development and in the final model. The original articles selected multiple variables or risk factors, which are thought to be associated with an increased risk of endometrial cancer. These variables have been tested in the original articles for univariate association and, if sufficiently contributing to predictive accuracy in multivariable regression analysis, combined to construct a clinical prediction model. We collected all different predictor variables from the original articles, together with their significance, to identify the most important predictor variables for endometrial cancer. The most important predictor variables had been considered as statistically significant input variables in three or more studies or were considered statistically significant in two studies and had not been tested in other studies. 20

23 Prediction models in women with postmenopausal bleeding: a systematic review Model development assessment The development of a prediction model consists of three phases: model derivation, model validation and impact analysis. 25 In the first phase, model derivation, predictor variables are identified by logistic regression. Model validation, the second phase, consists of an internal and external validation phase. 24 In internally validated models, the performance of the model is tested in the same data set in which the model was developed, or in a group of subsequent patients within the same center. In external validation, the goal is to demonstrate generalizability and reproducibility in patients different from the patients used for derivation of the original model. Therefore, the prediction model is evaluated on new data collected from an appropriate patient population in a different center. 26 The final phase of model development is impact analysis, in which prediction models are tested for their ability to change clinician s decisions and to change patient outcomes. 27 All prediction models identified in this review are classified into the different phases of model development. We sent an to all authors of the identified articles to investigate if their models are undergoing external validation and are not published yet. 2 Model performance Performance measures (calibration, discrimination and clinical usefulness) and the range of probabilities given by the different prediction models were recorded. Calibration refers to the agreement between observed probabilities and predicted probabilities for groups of patients; this is usually reported as a calibration plot or a Hosmer-Lemeshow statistic (test for goodness-of-fit ). 28 Discrimination is commonly reported as the c-statistic (concordance), also referred to as the Area Under the receiver-operating characteristic Curve (AUC). It measures the ability of a prediction model in separating patients with endometrial cancer and patients without endometrial cancer. An AUC of 0.5 describes a non-informative test, whereas an AUC of 1.0 represents a test that discriminates perfectly between presence and absence of a disease. 29 Clinical usefulness measures how close a prediction for an individual patient is to her actual outcome. This is mostly reported as accuracy (percentage of patients correctly classified), sensitivity or specificity, positive or negative predictive value (PPV or NPV) or likelihood ratios (LR) of a prediction model. 30 As we are interested in identifying a group of patients with a high risk for endometrial cancer, we are most interested in a high sensitivity, high NPV and a low negative LR. RESulTS Study identification and selection Of 754 articles identified by the MEDLINE and EMBASE search, a total of nine articles met the inclusion criteria of our review We identified another three articles by scanning the 21

24 Chapter 2 reference lists of included articles 40-42, however none of these matched our inclusion criteria after reading the abstract and full text version of these articles (Fig 1). 754 Potentially relevant studies identified and screened for retrieval from electronic search 732 studies excluded based on title/abstract 22 studies retrevied for more detailed evaluation Three papers retreived from cross references 16 studies excluded: Seven model based on one variable Five diagnosis, not prognosis One review Three pre- and postmenopausal women included Nine prognostic articles included, after consensus by two independent reviewers Figure 1. Study selection diagram Study characteristics Study characteristics are shown in Table 1. Of the nine selected articles on prediction models for women with PMB, five articles described the development of one model and four articles described two or more different prediction models. In the nine selected articles, four models were based primarily on patient characteristics, 31, 34, 36, 38 four prediction models were based on a combination of patient characteristics and gray-scale transvaginal sonography (TVS) findings, 31, 33, 36 two prediction models were based on a combination of patient characteristics, hysteroscopy and/or gray-scale TVS findings, 31 two prediction models were based on TVS findings only 37, 39 and three models used Doppler TVS findings as 35, 37, 38 a predictor variable. Patient selection and inclusion criteria were not the same in all articles. All nine articles included women with PMB, but three of these articles studied a population of women with a 35, 37, 38 high risk profile for endometrial cancer, based on an endometrial thickness of 5mm. 22

25 Prediction models in women with postmenopausal bleeding: a systematic review Table 1. Study characteristics of included articles Outcome Prediction model Study design Age (years) n malignant(%) Patients Exclusion criteria n First author, year 2 models: both combination patient history, TVS and Endometrial carcinoma (19%) Mean 66 Prospective CH Doppler artefacts, incorrect processing PMP bleeding, ET 5mm Epstein 2002 Doppler 1 model: TVS only Endometrial carcinoma Endometrial carcinoma Prospective CH Prospective CH PMP bleeding HRT/Tamoxifen use (30%) (median 68) 4 models: patient history, patient history + US, patient history + hyst, patient history (4.6%) (median 54) PMP bleeding, unscheduled bleeding on HRT Randelzhofer 2002 Bachmann TVS + hyst 1 model: patient history + TVS 2 models: patient characteristics and 1 combined model with patient characteristics and TVS Many models: TVS and Doppler CH* Endometrial carcinoma Prospective Endometrial CH carcinoma or premalignancy PMP bleeding HRT use 95 9 (9.5%) (mean 60) (10.3%) (mean 62) PMP bleeding History of hysterectomy, HRT use, recurrent bleeding, TVS not possible/not performed Bruchim 2004 Opmeer 2006 Endometrial carcinoma (25%) * Prospective CH Fluid in cavity, absence of power Doppler signals, large myomas, no histologic diagnosis PMP bleeding, ET 4.5mm Opolskiene model: patient history + TVS 1 model: patient history only Endometrial carcinoma Endometrial carcinoma Endometrial carcinoma Prospective CH PMP bleeding Asymptomatic women (median 59) (6%) Prospective CH (24%) (median Prospective 67) CH PMP bleeding History of hysterectomy, asymptomatic women PMP bleeding, ET Absence of processed 4.5mm ultrasound images or reliable histologic diagnosis Burbos 2010 Burbos 2011 Opolskiene models: 1 based on patient history, 1 patient history + TVS, 2 patient history + TVS + Doppler PMP = postmenopausal * not well described ET = endometrial thicknes CH = cohort study Hyst = hysteroscopy (pos/neg) TVS = transvaginal sonography HRT = hormone replacement therapy 2 23

26 Chapter 2 Study quality The results of the quality assessment are reported in Figure 2. Overall, study quality was good. The quality of the description of the setting and study period was rated as moderate; this was not described in three out of nine articles. Three articles included all women with postmenopausal bleeding, but performed histology only in patients with an increased endometrial thickness. All three articles explained that no further investigations were performed in women with an endometrial thickness less than 5 mm, because evidence 33, 34, 36 suggests a very low probability of cancer below this threshold. Figure 2. Quality of included studies Predictor variables The nine included articles investigated 27 different possible prediction variables (Table 2). Age was tested in all nine articles, turned out to be statistically significant in multivariable analysis in six articles and was used in the prediction model in six articles. Endometrial thickness was tested in eight articles, statistically significant in multivariable analysis in eight articles and used in eight prediction models. The most important predictor variables in patient history were: age, body mass index (BMI), diabetes, frequency of bleeding, use of anticoagulants and HRT. Endometrial thickness, endometrial morphology and endometrial border were identified as significant gray-scale TVS variables. In the three articles studying the use of Doppler for predicting endometrial cancer, endometrial color score and vascularity index were identified as the most important predictor variables. 24

27 Prediction models in women with postmenopausal bleeding: a systematic review Table 2. Predictor variables evaluated and used in the prediction models. 2 Epstein 2002 Randelzhofer 2002 Bachmann 2003 Bruchim 2004 Opmeer 2006 Opolskiene 2007 Burbos 2010 Burbos 2011 Opolskiene 2011 Patient history variables Age BMI Diabetes Frequency of bleeding 1 1 Amount of bleeding 3 3 HRT use Anticoagulant use 1 1 Nulliparity 1 3 Time since menopause 3 2 VAS-score 1 Examiner 3 Hypertension Thyroid dysfunction 3 3 History of malignancy Ultrasound variables gray-scale Endometrial thickness Heterogeneous echogenicity/ morphology Endometrial border 1 1 Endometrial fluid in cavity 3 Endometrial area 2 Ultrasound variables Doppler Vascularized area 2 1 Endometrial color score 1 1 Irregular branching 1 Vascularity index 1 1 Mean intensity of pixels in endometrial area 2 Mean intensity of pixels in vascularized area 3 Hysteroscopy variables Suspicious hysteroscopy findings 1 1. statistically significant in multivariate analysis and included in model 2. statistically significant in univariate analysis and not included in model 3. not statistically significant and not included in prediction model 25

28 Chapter 2 Phases of model development All articles selected in this review addressed the first phase of developing a prediction model: model derivation. 24 Of the nine articles on predicting endometrial cancer in women with PMB, eight had been internally validated but none of these models passed the external validation phase. We asked all six research-groups, which developed the nine different prediction models if their models are undergoing external validation and we received 37, 38 response from all six research-groups. The two prediction models of Opolskiene et al, are undergoing temporal validation (internal validation in a newly recruited patient group) and external validation in an international multicenter study by Valentin et al. No results are available yet, since they are still recruiting patients for these studies. The two prediction models developed by Burbos et al. 33, 34 were recently used in an article to compare the performance in internal validation of these models. 43 This group is working on external validation. Finally, we can report that the prediction model of Opmeer et al. 36 is currently being externally validated in two cohorts: one cohort in three different hospitals in the Netherlands and one in Skåne University Hospital Malmö in collaboration with the group of Valentin et al, but this external validation is not published yet. There were no impact analysis studies, i.e. studies that showed that the prediction model indeed improved patient outcome or was cost-effective in clinical practice. Performance of the prediction models The performance of the eight articles that were internally validated their models 31, is presented in Table 3. Calibration was described in one article. 39 The estimated probability of malignancy and the observed proportion of patients with endometrial carcinoma are mentioned in Randelzhofer et al. 39 However, calibration is generally reported as a calibration plot. None of the studies reported on calibration in a calibration plot. Discrimination was studied in seven out of eight articles by calculating an AUC. The AUC varied from 0.66 to 0.92 for different prediction models, with the highest AUC for a model combining Doppler and gray-scale TVS. 37 In all internally validated studies clinical usefulness is described, with the highest sensitivity and the lowest negative LR for a combined model with patient characteristics, gray-scale TVS and Doppler. 38 The highest NPV found for a model was for a model, which combined patient history, endometrial thickness and histology in a sequential strategy. 36 The performance of the four models using only patient characteristics showed a high sensitivity or high NPV in two models 36, 38 and a low LR for a negative outcome in one model. 38 All three studies in which Doppler was studied as a predictor variable, reported this information 35, 37, 38 to contribute to the prediction of endometrial carcinoma in women with PMB. Endometrial thickness was used as a variable in eight prediction models and seven found that incorporating endometrial thickness may improve diagnostic accuracy of a model. 26

29 Prediction models in women with postmenopausal bleeding: a systematic review Table 3. Evaluation of model development and model performance Specification of model performance at internal validation Prediction model Discrimination # Calibration Clinical usefulness Result as reported in paper Development phase First author, year Power Doppler can contribute to diagnosis of endometrial cancer. Sens 0.75 Spec 0.96 Sens 0.88 Spec Subjective prob of malignancy (PH, TVS, Doppler) Objective prob of malignancy (PH, US, Doppler) Epstein 2002 Internal validation - - The combined assessment of ET and endometrial morphology may improve diagnostic accuracy. Sens 0.97 Spec 0.62 NPV 0.98 Accuracy 0,72 - Estimated prob 2.8% vs. 2.1% real malignant ET+endometrial structure+myometrial border (cutoff point ET >10 mm) Internal validation Randelzhofer Not much increased value in testing with ultrasound, if hysteroscopy was already performed PH PH+ET PH+ hysteroscopy PH+ET+ hysteroscopy Internal validation Bachmann 2003 ET+time since menopause Time since menopause and ET can define when invasive testing is needed. Model derivation Bruchim 2004 Compared with US only, efficiency gain is reflected in increased AUC and reduced number of procedures, with PH+US in a sequential strategy. - NPV Efficiency +0.6 NPV Efficiency NPV Efficiency PH PH+TVS if prob>4% PH+TVS, histology if prob >4% Internal validation Opmeer 2007 A model including ET and heterogeneous echogenicity of the endometrium was best in predicting endometrial carcinoma, with Doppler diagnostic performance improved marginally ET, echogenicity (and Doppler) 0.91 / 0.92*** - Sens 0.93 / 0.87*** Spec 0.79 / 0.83 LR- 0.1 / 0.2 Internal validation Opolskiene 2007* Fair accuracy in separating women without cancer from women with cancer / 0.66** - Sens 0.82** Spec 0.50 Accuracy 0.52 LR DEFAB: diabetes, ET, frequency of bleeding, age and BMI (DEFAB 3) Internal validation Burbos 2010 Reasonable discriminatory ability / 0.66** - Sens 0.80** Spec 0.51 Accuracy 0.53 LR FAD 31: frequency of bleeding, age, diabetes, BMI cut-off 31 (cut-off FAD 31 4) Internal validation Burbos 2011 Fairly good in excluding endometrial cancer when power Doppler is added. Sens 0.89 LR Sens 0.84 LR Sens 0.70 LR Sens 0.89 LR PH PH+ET PH+ET+Doppler(VAS) PH+ET+Doppler(VI) Internal validation Opolskiene * As many models are described we selected the model with the best performance. # Discrimination is reported as area under the curve (AUC) of the ROC-curve. ** Cut-off point DEFAB 3 ## Cut-off point FAD 31 4 *** Including Doppler Prob = probability PH = patient history ET = endometrial thickness VAS = color intensity power Doppler VI = vascularity index NPV = negative predictive value 2 27

30 Chapter 2 Discussion We systematically reviewed existing prediction models for endometrial carcinoma in women with PMB and to identify the most important predictor variables. We found nine studies reporting on the development of prediction models for endometrial carcinoma in women with PMB. Eight of these studies described at least one aspect of internal validation and until now, none of the prediction models have been externally validated. The different predictor variables can roughly be divided into four subjects: patient characteristics, gray-scale ultrasound variables, Doppler ultrasound variables and hysteroscopy variables. Most prediction models used a combination of these subjects to predict the chance of endometrial carcinoma. We chose to limit our list of most important predictor variables to those which had been considered as statistically significant input variables in three or more studies and to those which were significant input variables in two studies and had not been tested in other studies. By doing this, we identified the most important variables, without missing possible important variables, which have not yet been extensively studied. Using these limits we identified 11 important input variables for predicting endometrial cancer in women with PMB (Table 2). Almost all articles reported performance in terms of discrimination and/or clinical usefulness, whereas calibration was reported only incidentally. In this study, we identified 31, five articles describing a prediction model with good discrimination (AUC of >0.8). Because only one study described data on calibration, there is insufficient data available to draw conclusions on calibration. Two studies showed best performance regarding discrimination and clinical usefulness: Opolskiene et al 2011 and Opmeer et al In the model by Opolskiene et al 2011, a combination of patient characteristics, gray-scale TVS and Doppler was used. They concluded that their model excludes endometrial cancer reasonably well when power Doppler is added. Furthermore, in all three studies that used Doppler, Doppler was found to contribute to the prediction of endometrial carcinoma in women with PMB. 35, 37, 38 Based on this we could conclude that the best model in predicting endometrial cancer is a model, which uses a combination of patient characteristics, endometrial thickness and power Doppler. However, power Doppler cannot be used in all patients. All three Doppler-models excluded patients based on different reasons: Doppler artifacts, incorrect processing of TVS image, fluid in the cavity and absence of Doppler signals or large myomas. Another limitation in the use of power Doppler is that these studies do not give information on the interobserver variability and learning curve in measuring Doppler variables. For application of results found in Doppler studies it is important to use the same ultrasound system, as the color content of a power Doppler scan depends heavily on Doppler sensitivity. 38 Although the performance of the models using Doppler seems reasonable, a model using patient characteristics and endometrial thickness might be more useful in daily practice. In a 28

31 Prediction models in women with postmenopausal bleeding: a systematic review health care system with general practitioners referring patients with a high risk of malignant disease to a specialist, the best model would be a model that can distinguish women with a high risk of endometrial cancer from women with a low risk based on patient characteristics only. Such a model would also be useful in situations where TVS is not directly available. Only women with a high risk could be referred for TVS or to the gynecologist for a further evaluation and women with a low risk could be reassured and referred only at recurrent bleeding. Based on this review we couldn t identify a model with a good performance in internal validation based on patient characteristics only. However, two of four models based on patient characteristics showed good performance in clinical usefulness with a high sensitivity, a high NPV and/or a low LR for a negative outcome. 36, 38 Based on these results we can conclude that although these models do not show a high AUC, they could be useful in clinical practice. These models were found to discriminate women with a high risk for endometrial cancer from women with low risk and to select women for further (invasive) testing. The conclusions above are based on reported model performance based on internal validation only. To implement a prediction model into clinical practice, external validation is essential. McGinn et al. describe three reasons. 25 A prediction model may reflect associations between given predictors and outcomes that are primarily due to chance. Secondly, the predictor variables used in a model may be idiosyncratic to that specific population, which suggests that the prediction model may fail in a new setting. And thirdly, clinicians may fail to implement the model comprehensively or accurately in their clinical practice. The result would be that a model succeeds in theory, but fails in practice. For a successful implementation, a model should be validated both internally and externally and finally go through the phase of impact analysis in the same population in which a model is derived. As none of the prediction models have completed the phase of external validation, they cannot be used in clinical practice yet. When evaluating these prediction models by external validation or finally in impact analysis, one should keep in mind that these models were developed in different patient populations. The target population in which a model is derived should be the same as the population in which a model is tested or clinically used. Selecting a high-risk population (for example a population with an ET 5mm) will result in a different performance and possibly in the selection of different predictor variables compared to an unselected population of women with PMB. Furthermore, in an unselected population there could be implicate selection dependent of a population within a general practice or a population within a gynecological practice or differences in health systems in different countries. Different populations have different prevalence of endometrial cancer, which could be an explanation for the differences found in the performance of the models. 2 29

32 Chapter 2 A consensus has not been found in systematic reviews or in international guidelines regarding the best sequence of diagnostic procedures for women with PMB. 9 Considering the performance of the existing prediction models, we can conclude that we have indications that the first step in the approach of women with PMB should be to distinguish between women with low versus increased risk of having endometrial carcinoma and the next step would be to refer patients for TVS or further invasive testing. Future perspective The prediction models that have been developed for women with postmenopausal bleeding showed good performance but have only reached the phase of internal validation. Future research should focus on external validation and impact analysis of these prediction models. We hope that these will confirm their prognostic abilities, so that in the next few years, prediction models can be implemented in general gynecological practice. Based on this review, we conclude that clinical prediction models show promising results, but further external validation is required as well as impact analysis to maximize diagnostic accuracy of the models at an acceptable patient burden and for acceptable health care costs. 30

33 Prediction models in women with postmenopausal bleeding: a systematic review REFEREnCES 1. Brenner PF. Differential diagnosis of abnormal uterine bleeding. Am J Obstet Gynecol 1996;175(3 Pt 2): NVOG (Dutch Society of Obstetrics and Gynaecology). NVOG-richtlijn Abnormaal vaginaal bloedverlies in de menopauze [In Dutch]. NVOG Guideline: Abnormal vaginal bleeding during menopause Dijkhuizen FP, Brolmann HA, Potters AE, Bongers MY, Heinz AP. The accuracy of transvaginal ultrasonography in the diagnosis of endometrial abnormalities. Obstet Gynecol 1996;87(3): Emanuel M, Verdel M, Wamsteker K, Lammes F. An audit of true prevalence of intrauterine pathology: the hysteroscopic findings, controled for patient selection in 1202 patients with abnormal uterine bleeding. Gynaecol Endosc 1995;4: Epstein E, Ramirez A, Skoog L, Valentin L. Dilatation and curettage fails to detect most focal lesions in the uterine cavity in women with postmenopausal bleeding. Acta Obstet Gynecol Scand 2001;80(12): Ferrazzi E, Torri V, Trio D, Zannoni E, Filiberto S, Dordoni D. Sonographic endometrial thickness: a useful test to predict atrophy in patients with postmenopausal bleeding. An Italian multicenter study. Ultrasound Obstet Gynecol 1996;7(5): Timmermans A, Veersema S. Ambulatory transcervical resection of polyps with the Duckbill polyp snare: a modality for treatment of endometrial polyps. J Minim Invasive Gynecol 2005;12(1): Breijer MC, Timmermans A, Van Doorn HC, Mol BW, Opmeer BC. Diagnostic strategies for postmenopausal bleeding. Obstet Gynecol Int 2010;2010: Van Hanegem N, Breijer MC, Khan KS et al. Diagnostic evaluation of the endometrium in postmenopausal bleeding: an evidence-based approach. Maturitas 2011;68(2): van Doorn LC, Dijkhuizen FP, Kruitwagen RF, Heintz AP, Kooi GS, Mol BW. Accuracy of transvaginal ultrasonography in diabetic or obese women with postmenopausal bleeding. Obstet Gynecol 2004;104(3): Anderson KE, Anderson E, Mink PJ et al. Diabetes and endometrial cancer in the Iowa women s health study. Cancer Epidemiol Biomarkers Prev 2001;10(6): Gull B, Karlsson B, Milsom I, Granberg S. Factors associated with endometrial thickness and uterine size in a random sample of postmenopausal women. Am J Obstet Gynecol 2001;185(2): McPherson CP, Sellers TA, Potter JD, Bostick RM, Folsom AR. Reproductive factors and risk of endometrial cancer. The Iowa Women s Health Study. Am J Epidemiol 1996;143(12): Reeves GK, Pirie K, Beral V, Green J, Spencer E, Bull D. Cancer incidence and mortality in relation to body mass index in the Million Women Study: cohort study. BMJ 2007;335(7630): Weiderpass E, Persson I, Adami HO, Magnusson C, Lindgren A, Baron JA. Body size in different periods of life, diabetes mellitus, hypertension, and risk of postmenopausal endometrial cancer (Sweden). Cancer Causes Control 2000;11(2): Xu WH, Xiang YB, Ruan ZX et al. Menstrual and reproductive factors and endometrial cancer risk: Results from a population-based case-control study in urban Shanghai. Int J Cancer 2004;108(4): Astrup K, Olivarius NF. Frequency of spontaneously occurring postmenopausal bleeding in the general population. Acta Obstet Gynecol Scand 2004;83(2): Gredmark T, Kvint S, Havel G, Mattsson LA. Histopathological findings in women with postmenopausal bleeding. Br J Obstet Gynaecol 1995;102(2): American College of Obstetricians and Gynecologists. ACOG Committee Opinion No. 426: The role of transvaginal ultrasonography in the evaluation of postmenopausal bleeding. Obstet Gynecol 2009;113(2 Pt 1): Epstein E. Management of postmenopausal bleeding in Sweden: a need for increased use of hydrosonography and hysteroscopy. Acta Obstet Gynecol Scand 2004;83(1): Goldstein RB, Bree RL, Benson CB et al. Evaluation of the woman with postmenopausal bleeding: Society of Radiologists in Ultrasound-Sponsored Consensus Conference statement. J Ultrasound Med 2001;20(10): Ingui BJ, Rogers MA. Searching for clinical prediction rules in MEDLINE. J Am Med Inform Assoc 2001;8(4):

34 Chapter Hayden JA, Cote P, Bombardier C. Evaluation of the quality of prognosis studies in systematic reviews. Ann Intern Med 2006;144(6): Leushuis E, van der Steeg JW, Steures P et al. Prediction models in reproductive medicine: a critical appraisal. Hum Reprod Update 2009;15(5): McGinn TG, Guyatt GH, Wyer PC, Naylor CD, Stiell IG, Richardson WS. Users guides to the medical literature: XXII: how to use articles about clinical decision rules. Evidence-Based Medicine Working Group. JAMA 2000;284(1): Altman DG, Royston P. What do we mean by validating a prognostic model? Stat Med 2000;19(4): Reilly BM, Evans AT. Translating clinical research into clinical practice: impact of using prediction rules to make decisions. Ann Intern Med 2006;144(3): Steyerberg E.W. Calibration. Clinical Prediction Models - A Practical Approach to Development, Validation and Updating. Germany: Springer; 2009: Steyerberg E.W. Discriminative ability. Clinical Prediction Models - A Practical Approach to Development, Validation and Updating. Germany: Springer; 2009: Vergouwe Y. Validation of clinical prediction models: theory and applications in testicular germ cell cancer. Erasmus University Rotterdam, the Netherlands; Bachmann LM, ter RG, Clark TJ, Gupta JK, Khan KS. Probability analysis for diagnosis of endometrial hyperplasia and cancer in postmenopausal bleeding: an approach for a rational diagnostic workup. Acta Obstet Gynecol Scand 2003;82(6): Bruchim I, Biron-Shental T, Altaras MM et al. Combination of endometrial thickness and time since menopause in predicting endometrial cancer in women with postmenopausal bleeding. J Clin Ultrasound 2004;32(5): Burbos N, Musonda P, Giarenis I et al. Predicting the risk of endometrial cancer in postmenopausal women presenting with vaginal bleeding: the Norwich DEFAB risk assessment tool. Br J Cancer 2010;102(8): Burbos N, Musonda P, Duncan TJ, Crocker SG, Morris EP, Nieto JJ. Estimating the risk of endometrial cancer in symptomatic postmenopausal women: a novel clinical prediction model based on patients characteristics. Int J Gynecol Cancer 2011;21(3): Epstein E, Skoog L, Isberg PE et al. An algorithm including results of gray-scale and power Doppler ultrasound examination to predict endometrial malignancy in women with postmenopausal bleeding. Ultrasound Obstet Gynecol 2002;20(4): Opmeer BC, Van Doorn HC, Heintz AP, Burger CW, Bossuyt PM, Mol BW. Improving the existing diagnostic strategy by accounting for characteristics of the women in the diagnostic work up for postmenopausal bleeding. BJOG 2007;114(1): Opolskiene G, Sladkevicius P, Valentin L. Ultrasound assessment of endometrial morphology and vascularity to predict endometrial malignancy in women with postmenopausal bleeding and sonographic endometrial thickness >or= 4.5 mm. Ultrasound Obstet Gynecol 2007;30(3): Opolskiene G, Sladkevicius P, Valentin L. Prediction of endometrial malignancy in women with postmenopausal bleeding and sonographic endometrial thickness >/= 4.5 mm. Ultrasound Obstet Gynecol 2011;37(2): Randelzhofer B, Prompeler HJ, Sauerbrei W, Madjar H, Emons G. Value of sonomorphological criteria of the endometrium in women with postmenopausal bleeding: a multivariate analysis. Ultrasound Obstet Gynecol 2002;19(1): Feldman S, Cook EF, Harlow BL, Berkowitz RS. Predicting endometrial cancer among older women who present with abnormal vaginal bleeding. Gynecol Oncol 1995;56(3): Khan KS, Bachmann LM, ter RG. Systematic reviews with individual patient data meta-analysis to evaluate diagnostic tests. Eur J Obstet Gynecol Reprod Biol 2003;108(2): Weber AM, Belinson JL, Piedmonte MR. Risk factors for endometrial hyperplasia and cancer among women with abnormal bleeding. Obstet Gynecol 1999;93(4): Musonda P, Burbos N, Duncan TJ, Crocker SG, Morris EP, Nieto JJ. Comparing the performance of two clinical models in estimating the risk of endometrial cancer in symptomatic postmenopausal women. Eur J Obstet Gynecol Reprod Biol 2011;159(2):

35 Prediction models in women with postmenopausal bleeding: a systematic review APPEnDIX Search strategy MEDLINE Searches #1. postmenopause [mesh] #2. postmenopau* [tw] #3. post-menopau* [tw] #4. #1 OR #2 OR #3 #5. hemorrhage [tw] #6. bleed* [tw] #7. hemorrhag* [tw] #8. haemorrhag* [tw] #9. blood loss* [tw] #10. #5 OR #6 OR #7 OR #8 OR #9 #11. endometrial neoplasms [mesh] #12. endometrial neoplasm* [tw] #13. endometrial carcinoma* [tw] #14. endometrial cancer* [tw] #15. endometrial malignanc* [tw] #16. endometrial tumo* [tw] #17. corpus uteri cancer* [tw] #18. #11 OR #12 OR #13 OR #14 OR #15 OR #16 OR #17 #19. endometrial hyperplasia [mesh] #20. endometrial hyperplasia* [tw] #21. #19 OR #20 #22. #18 OR #21 #23. predict* [tiab] #24. clinical* [tiab] #25. outcome* [tiab] #26. risk* [tiab] #27. #23 OR #24 OR #25 OR #26 #28. #4 AND #10 AND #22 AND #27 Search strategy EMBASE Searches #1. postmenopause/ #2. (postmenopau* or post-menopau*).tw. #3. (after menopaus* or after the menopaus* or following menopaus* or following the menopaus*).tw. 4. or/1-3 #5. exp bleeding/ #6. (bleed* or hemorrhag* or haemorrhag* or blood loss*).tw. #7. or/5-6 #8. endometrial neoplasms/ #9. endometrial neoplasm* or endometrial carcinoma* or endometrial cancer* or endometrial malignanc* or endome trial tumo* or corpus uteri cancer*).tw #10. or/8-9 #11. endometrial hyperplasia/ #12. endometrial hyperplasia.tw. #13. or/11-12 # or 13 #15. (predict* or clinical* or outcome* or risk*).ti,ab. #16. 4 and 7 and 14 and

36

37 3 Diagnostic strategies for endometrial cancer in women with postmenopausal bleeding: costs-effectiveness of individualized strategies M.C. Breijer H.C. Van Doorn T.J. Clark K.S. Khan A. Timmermans B.W.J. Mol B.C. Opmeer Eur J Obstet Gynecol Reprod Biol Jul;163(1):91-6.

38 Chapter 3 Abstract Objective: To evaluate the cost-effectiveness of diagnostic strategies incorporating the diagnostic value of patient characteristics for endometrial carcinoma using prediction models. Study design: A decision analytic model was created to compare four diagnostic strategies for women with postmenopausal bleeding: The main outcome measures were 5-year survival, costs, and cost-effectiveness of three model based strategies compared to the strategy reflecting current practice. Results: A strategy selecting women for endometrial biopsy based on their history only, dominated all other strategies (more effective, less costs). In a clinical scenario where transvaginal sonography (TVS) was assumed to be an integral part of the consultation without additional costs, a strategy selecting high risk women for measurement of endometrial thickness became the most cost-effective strategy. Conclusions: Strategies taking into account the individual probability based on a prognostic model, are less costly than the currently applied strategy for a similar effectiveness. The most cost-effective strategy depends on the clinical setting; in areas where TVS is performed by the consulting gynecologist without extra costs, selective TVS based on history is the most cost-effective strategy. When TVS is not readily available and therefore incurs extra costs, a risk selection based on patient characteristics is most cost-effective. 36

39 Diagnostic strategies for endometrial cancer in women with postmenopausal bleeding InTRODuCTIOn The aim of the diagnostic workup in women presenting with postmenopausal bleeding (PMB) is to rule out endometrial carcinoma without over investigation. 1 Current guidelines recommend to start with TVS of the endometrium, and to perform histological analysis using office endometrial biopsy (EB) devices in case of endometrial thickness of 5 mm or more. 2-5 There is considerable variability in the endometrial thickness and the likelihood of endometrial carcinoma across women. This variability has been associated with individual patient characteristics including age, time since menopause, obesity, hypertension, diabetes mellitus, and reproductive factors. The probability of endometrial carcinoma in women with PMB rises from 1% in women younger than 50 years to 23.8% in women older than 80 years and the incidence of malignancy is, regardless of age, higher in women with PMB and obesity (18%) or diabetes (21%) as compared to women without one of these risk factors In 2006 we evaluated the combined diagnostic accuracy and costs of various diagnostic strategies based on combinations of commonly used diagnostic procedures. These analyses indicated that the currently recommended guideline for PMB, i.e. start with TVS, and further evaluate women at risk with EB (endometrial thickness cut-off 5 mm), was the most costeffective approach. However, this analysis did not take into account aforementioned clinical characteristics. 13 Subsequently, our group developed two multivariable prediction models to estimate the risk of endometrial carcinoma in patients with PMB taking into account their clinical characteristics. 14 These two models can be used in three different diagnostic strategies: 1. The patient characteristics strategy, i.e. probability estimates based on characteristics of the women and histological analysis in case the probability of (pre)malignancy exceeded 4%. In this strategy, TVS is not performed. 2. The sequential strategy, i.e. probability estimates based on characteristics of the women, with TVS in case the probability for cancer exceeded 4% and subsequent histological analysis when the endometrial thickness exceeds 4mm. 3. The integrated strategy, i.e. TVS in all women, with a probability estimate based on both characteristics of the women and TVS results, completed by histological analysis when the probability of cancer exceeded 4%. 14 The objective of this study was to evaluate the cost-effectiveness of the individualized strategies in the diagnostic work-up for patients with PMB, using a decision analytic approach. 3 37

40 Chapter 3 Materials and Methods Decision analytic model We used a decision analytic approach to evaluate the outcomes of four diagnostic strategies for PMB. Model assumptions and input parameters were based on the earlier reported model based study by Clark et al. 13 however, we used 5-year survival instead of duration of survival as of the primary health outcome. Three new diagnostic strategies were constructed using the previously developed multivariable models based on patient characteristics. 14 Data to construct these patient characteristics models were obtained from a prospective cohort study of 614 women presenting with PMB in one university hospital and seven teaching hospitals in the Netherlands. 10 The four diagnostic strategies based on combinations of TVS, EB and/or diagnostic information from patient characteristics are presented in Table 1. Table 1. Evaluated Diagnostic strategies Strategy Abbreviation Description I TVS 5mm A strategy selecting women for EB with TVS using a 5 mm cut-off for endometrial thickness (reference strategy) II History Only A strategy selecting women for EB based on the patient characteristics model ( History Only ) III History and TVS A strategy selecting women for EB based on the patient characteristics and TVS model ( History and TVS ) IV Selective Testing A strategy with selective TVS in high risk women based on the patient characteristics model ( Selective Testing ); i.e. TVS was performed in patients with a high risk of endometrial cancer based on their history, subsequently in case of an increased endometrial thickness an EB was performed TVS, transvaginal sonography. EB, endometrial biopsy. Decision Analysis TreeAge software (DATA, TreeAge Software Inc., Williamstown, MA, USA) was used to specify the model, while SAS 9.1 software (SAS Institute Inc., Cary, NC, USA) was used to estimate additional input parameters. Model input; probabilities Model assumptions and input parameters to populate the decision tree are summarized in Table 2. Parameters used in the main analyses as well as ranges for sensitivity analyses are indicated. To identify women at risk of endometrial cancer based on clinical information in the newly developed strategies, we used a 4% threshold for both the History Only model and the History and TVS model. Compared to the currently applied strategy of TVS and EB in case of increased endometrial thickness (cut-off 5 mm), this risk threshold results in a similar detection rate of endometrial cancer, with less false positive test results in the subsequent workup

41 Diagnostic strategies for endometrial cancer in women with postmenopausal bleeding Table 2. Estimated base-case values and ranges for model input parameters variable Base case probability Sensitivity analysis Source Failure rates EB 0.12 (95%CI: 0.09 to 0.15) DC TVS 0.00 (95%CI: 0.00 to 0.02) DC EB after successful TVS 0.12 (95%CI: 0.09 to 0.15) DC EB after History Only model 0.12 (95%CI: 0.05 to 0.13) BO EB after History + TVS model 0.12 (95%CI: 0.09 to 0.15) DC Complication rates Outpatient diagnostic procedures - - DC D&C DC True Positive rates EB 0.94 (95%CI: 0.84 to 0.99) DC TVS 5 mm 0.97 (95%CI: 0.94 to 0.98) DC D&C 0.96 (95%CI: 0.82 to 1.00) DC History Only model 0.99 (95%CI: 0.97 to 1.00) BO History + TVS model 0.99 (95%CI: 0.96 to 1.00) BO Conditional true positive rates EB if TVS positive 0.94 (95%CI: 0.94 to 0.95) DC EB if History Only positive 0.94 (95%CI: 0.93 to 0.97) BO EB if History + TVS positive 0.94 (95%CI: 0.94 to 0.95) BO TVS 5 mm if EB negative 0.97 (95%CI: 0.80 to 0.99) DC False positive rates EB 0.01 (95%CI: 0.00 to 0.02) DC TVS 5 mm 0.45 (95%CI: 0.43 to 0.47) DC D&C 0.01 (95%CI: 0.00 to 0.03) DC History Only model 0.80 (95%CI: 0.72 to 0.87) BO History + TVS model 0.41 (95%CI: 0.39 to 0.44) BO Surgical stage at hysterectomy (FIGO) Probability of stage I (first presentation) 0.70 (95%CI: 0.60 to 0.80) DC Probability of stage II-IV (first presentation) 0.30 (95%CI: 0.20 to 0.40) DC Probability of stage I (representation) 0.65 (95%CI: 0.40 to 0.70) DC Probability of stage II-IV (representation) 0.35 (95%CI: 0.30 to 0.60) DC 5-years survival rates Stage I 0.87 Stage II-IV 0.60 EB, endometrial biopsy. TVS, transvaginal sonography. D&C, dilatation and curettage. FIGO, International Federation of Obstetrics and Gynecology. DC, Clark BO, Opmeer

42 Chapter 3 Model input; costs Costs evaluated in the model were generated by diagnostic procedures to detect or rule out carcinoma, and by treatment in case of positive findings. We used the cost-estimates that have been reported by Clark et al. 13 Outcomes; 5-year-survival In the model, the primary measure of effectiveness was incorporated in terms of 5-year survival. We differentiated between expected survival of women where malignancy is immediately detected and treated appropriately and women where the first diagnostic procedure fails to detect carcinoma, after which survival decreases due to the delayed diagnosis. The extent to which survival decreases is based on an estimated increase in disease stage from local (FIGO stage I) to advanced (FIGO stages II-IV) 15 endometrial cancer and associated 5-year mortality rates (Table 2). Outcomes; cost-effectiveness Cost-effectiveness can be expressed in incremental costs-effectiveness ratios (ICERs) reflecting the costs needed to obtain one additional unit of health benefit (i.e. one woman extra to survive at 5-years). ICERs were estimated to evaluate the relative cost-effectiveness of each strategy as compared to the reference strategy. In case a strategy is more effective and less costly than the reference case, the ICER has a negative sign (less than 0). However, in case of a strategy that is more costly and less effective, the sign of the ICER also becomes negative. It is therefore important for the interpretation of the ICER to inspect the values of incremental effectiveness and incremental costs. Sensitivity analyses To evaluate robustness of our findings and ascertain how uncertainty arising from underlying assumptions influence results, sensitivity analyses were conducted. We evaluated the uncertainty in the estimated ICERs assuming three different scenarios: A. prevalence of endometrial carcinoma of 10%, B. prevalence of endometrial carcinoma of 5% and C. since TVS is increasingly being performed by the consulting gynecologist we evaluated a scenario where TVS is free of charge. Nevertheless, TVS will require resources (physician time, maintenance of machine) and incur costs whether the physician is paid or not. To investigate at what costs for TVS the cost-effectiveness of the History Only strategy exceeds the cost-effectiveness of the Selective Testing strategy, we performed a threshold analysis for this clinical scenario. The analyses were performed for the three new developed strategies compared to the reference strategy. After altering the above mentioned parameters matching the clinical 40

43 Diagnostic strategies for endometrial cancer in women with postmenopausal bleeding scenario, uncertainty in the estimated ICERs was determined with bootstrap analyses(monte Carlo simulation), in which all other parameter estimates are systematically varied using multivariate random sampling from a distribution as indicated by the lower and upper limits (Table 2). The estimates of the incremental costs and incremental effectiveness for 1000 repeated analyses with randomly varying model parameters are plotted in a cost-effectiveness plane. Diagnostic strategies are represented according to their effectiveness (x-axis) and cost (y-axis) relative to those of the reference strategy. Ideally, strategies fall within the lower right quadrant, because these strategies are more effective and less costly than the reference strategy, while strategies in the upper right quadrant are still considered to be cost-effective if the ICER does not exceed some defined acceptable cost-effectiveness ratio RESulTS Table 3 shows the expected costs, effectiveness and ICER of the four diagnostic strategies for the most probable assumptions. Table 3. Summary of model results for effectiveness (5-yr survival), costs (of diagnosis and treatment, in euros per patient) and incremental cost-effectiveness ratio (euros per extra patient surviving 5 years) based on most probable assumptions (prevalence EC 0.10) Strategy ICER Effectiveness Costs Reference TVS 5mm TVS 5 mm D (History Only) History Only , History & TVS D (History Only) Selective Testing D (History Only) D (History Only), strategy dominated by History Only (i.e. more costs and less effective). ICER, incremental costeffectiveness ratio. TVS, transvaginal sonography. Effectiveness The estimated proportion of women surviving after five years varied between for the TVS 5 mm strategy and for the History Only strategy. As compared to the currently recommended strategy of measuring endometrial thickness with TVS using a cutoff of 5 mm, the incremental effectiveness of the model based strategies was small: in every 10,000 women, approximately two more women were expected to survive after 5 years with model based strategies History Only and History and TVS, and one in each 10,000 women with the Selective Testing strategy. 41

44 Chapter 3 Costs Costs of the currently recommended strategy based upon initial investigation with TVS were 330 per patient presenting with PMB. For diagnostic strategies using patient characteristics model based probability estimates of endometrial cancer, costs varied between 278 for the History Only strategy and 325 for the History and TVS strategy. As compared to the currently recommended strategy of measuring endometrial thickness with TVS, using a cut-off of 5 mm, the incremental costs were generally in favor of the model-based strategies, ranging from 5 less per patient for the History and TVS strategy to 52 less per patient for the History Only strategy. Cost-effectiveness Compared to the currently applied strategy, the model based strategies were all costeffective. The History Only strategy dominates all other strategies because it was the most effective and less costly strategy. In view of both the incremental costs and the incremental effectiveness results, the strategies have limited benefits in terms of survival, yet result in substantial reduction of costs for each additional women surviving after 5 years. Sensitivity analyses The results of the sensitivity analyses are shown in Figure 1. The results for the History Only and the History and TVS strategies, clearly indicate that for the scenarios with an endometrial cancer prevalence of 10% and 5%, most of the analyses with varying model input parameters are found within the lower right quadrant, and only incidental analyses show a decrease in effectiveness. For the Selective Testing strategy, the sensitivity analyses indicated less uncertainty about the expected reduction in costs, but the average expected reduction was substantially lower ( 5 for each woman) as compared to the strategy based on the History Only model. An analysis performed with model parameters adapted to a scenario with an endometrial cancer prevalence of 10% but without additional costs for TVS showed that the cost reduction per patient was unchanged for the History and TVS strategy, but was substantially reduced ( 18) for the Selective Testing strategy. For the History Only strategy, costs per patient were increased (ca 4). In this clinical scenario, the Selective Testing strategy was most cost-effective (Table 4). A threshold analysis showed that if costs for TVS performed by the consulting gynecologist exceeded 30 the History Only strategy was the most costeffective. 42

45 Diagnostic strategies for endometrial cancer in women with postmenopausal bleeding 3 A. Prevalence endometrial carcinoma B. Prevalence endometrial carcinoma C. Prevalence endometrial carcinoma 0.10, costs of TVS 0. Figure 1. Incremental cost-effectiveness plane, three strategies compared to reference strategy for different clinical scenarios. Results of sensitivity analyses for cost-effectiveness, for three strategies using probability-estimates based on the statistical models, as compared to the currently recommend strategy of TVS measurement of endometrial thickness, with a cut-off at 5 mm. 43

46 Chapter 3 Table 4. ICER clinical scenario C (prevalence EC 0.10; costs TVS 0), three strategies vs TVS 5 mm Strategy ICER ΔEffectiveness ΔCosts Reference TVS 5 mm History Only , History & TVS , Selective Testing , ICER, incremental cost-effectiveness ratio. TVS, transvaginal sonography. Comment Statement of principal findings In this study, we found that incorporating the individual probability of cancer based on patient characteristics in the diagnostic work up for patients with PMB can increase the cost-effectiveness. In addition to previously studied strategies by Clark et al. 13, three new individualized diagnostic strategies were introduced. Using this additional information related to women s individual risk profiles improves efficiency of the diagnostic work-up for patients with PMB by reducing the number of diagnostic procedures (and costs) and increasing the sensitivity without an increase in the number of invasive procedures (and costs). Multivariate sensitivity analyses indicated that with varying model input parameters, all three strategies remained more cost-effective. In a scenario without additional costs for TVS, analyses showed that in this case, the History Only strategy generated more costs compared to the currently applied TVS with 5 mm cut-off. The other two strategies were still cost-effective but there was a smaller expected decrease in costs. The most cost-effective diagnostic strategy therefore depends on the clinical setting. In areas where TVS is performed by the consulting gynecologist with minimal additional cost, the Selective Testing strategy will be most cost-effective. When TVS is not readily available and its use therefore generates extra costs, a more specified risk selection based on patient characteristics ( History Only ) strategy will be most cost-effective. This study concerns patients referred to the gynecologist with PMB and not on screening for endometrial carcinoma; the model was designed accordingly. In other settings, e.g. countries were screening for endometrial carcinoma is performed by a general practitioner, costs will be different. This was not evaluated in our current study. Strengths and limitations The strength of this study is the incorporation of patient characteristics in the diagnostic work-up. Evaluating test accuracy in isolation from patient history and clinical characteristics 44

47 Diagnostic strategies for endometrial cancer in women with postmenopausal bleeding tends to underestimate the potential accuracy of the test in daily practice. By accounting for these characteristics, an individualized pre-test probability of disease can be estimated, thus reflecting clinical practice. Several multivariable models were developed to estimate the risk of endometrial carcinoma in patients with PMB. 14, The model developed by our group, evaluated all clinical characteristics known to influence the risk of endometrial carcinoma. 6-8,10-12 None of the models was previously evaluated for its potential to improve cost-effectiveness in the clinical work-up for PMB. 3 A limitation of this study is the model based approach. As for all model based approaches the validity of the model is dependent on the model parameters. We used the same model assumptions that were previously used in the cost-effectiveness analyses of initial diagnostic strategies. For the three new strategies incorporating the patient characteristics model, we calculated the conditional estimates of test failure and accuracy using our prospective collected data. The model previously described by Clark et al was designed to reflect the situation in the United Kingdom. Data for failure rates and estimates of diagnostic accuracy were obtained from systematic quantitative reviews and are applicable to other populations; cost estimates however will only be partially comparable to cost figures in the other countries. The objective of this study however, was to explore whether incorporating patient characteristics would increase the efficiency of initial diagnostic strategies for PMB. For comparability reasons, we therefore extended the existing model without changing the assumptions. In contrast to all other parameters, we changed the outcome into 5 year survival. Patients with endometrial carcinoma, if diagnosed and treated at an early stage, have a high 5 year survival rate 24, 25. Although life expectancy is usually considered the standard outcome measure in decision analysis, we considered 5 years survival more relevant for this particular clinical problem. Future research Future research should focus on validation and reproducibility of the models. To show the generalizability of the patient characteristics models, they should be validated in external 19, 26 datasets, e.g. based on systematic reviews using individual patient data meta-analysis. This external validation should address the discriminative performance and calibration of the models. If external validation proves the diagnostic accuracy of the models, estimation of cancer risk before subsequent diagnostic testing by a patient characteristics model should be incorporated in the clinical guidelines. The incremental effectiveness and cost reduction per patient are low, but on a population level the differences in costs and effects demonstrated with this decision analytic model are expected to substantially affect health care costs 27 45

48 Chapter 3 since PMB is a common clinical problem with high resource use. 28 To implement the model in clinical practice, because of the minimal effect per patient, it is important to make the model easily available for practitioners for example by making a web based application or designing a simplified risk score. Conclusions In conclusion, our study demonstrated the potential to increase the cost-effectiveness by integrating individual patient characteristics in the diagnostic workup for PMB. The most cost-effective strategy depends on the prevalence of endometrial cancer as well as whether TVS is readily available in the clinical setting. In all strategies, the use of patient characteristics reduces costs without compromising the effectiveness of the strategy. 46

49 Diagnostic strategies for endometrial cancer in women with postmenopausal bleeding REFEREnCES 1. Gupta JK, Wilson S, Desai P, Hau C. How should we investigate women with postmenopausal bleeding? Acta Obstet Gynecol Scand 1996;75(5): Epstein E. Management of postmenopausal bleeding in Sweden: a need for increased use of hydrosonography and hysteroscopy. Acta Obstet Gynecol Scand 2004;83(1): Goldstein RB, Bree RL, Benson CB et al. Evaluation of the woman with postmenopausal bleeding: Society of Radiologists in Ultrasound-Sponsored Consensus Conference statement. J Ultrasound Med 2001;20(10): NVOG (Dutch Society of Obstetrics and Gynaecology). NVOG-richtlijn Abnormaal vaginaal bloedverlies in de menopauze [In Dutch]. NVOG Guideline: Abnormal vaginal bleeding during menopause Scottish Intercollegiate Guidelines Network. Investigation of postmenopausal bleeding. Scottish Intercollegiate Guidelines Network, Royal College of Physicians Anderson KE, Anderson E, Mink PJ et al. Diabetes and endometrial cancer in the Iowa women s health study. Cancer Epidemiol Biomarkers Prev 2001;10(6): Gull B, Karlsson B, Milsom I, Granberg S. Factors associated with endometrial thickness and uterine size in a random sample of postmenopausal women. Am J Obstet Gynecol 2001;185(2): McPherson CP, Sellers TA, Potter JD, Bostick RM, Folsom AR. Reproductive factors and risk of endometrial cancer. The Iowa Women s Health Study. Am J Epidemiol 1996;143(12): Tabor A, Watt HC, Wald NJ. Endometrial thickness as a test for endometrial cancer in women with postmenopausal vaginal bleeding. Obstet Gynecol 2002;99(4): van Doorn LC, Dijkhuizen FP, Kruitwagen RF, Heintz AP, Kooi GS, Mol BW. Accuracy of transvaginal ultrasonography in diabetic or obese women with postmenopausal bleeding. Obstet Gynecol 2004;104(3): Weiderpass E, Persson I, Adami HO, Magnusson C, Lindgren A, Baron JA. Body size in different periods of life, diabetes mellitus, hypertension, and risk of postmenopausal endometrial cancer (Sweden). Cancer Causes Control 2000;11(2): Xu WH, Xiang YB, Ruan ZX et al. Menstrual and reproductive factors and endometrial cancer risk: Results from a population-based case-control study in urban Shanghai. Int J Cancer 2004;108(4): Clark TJ, Barton PM, Coomarasamy A, Gupta JK, Khan KS. Investigating postmenopausal bleeding for endometrial cancer: cost-effectiveness of initial diagnostic strategies. BJOG 2006;113(5): Opmeer BC, van Doorn HC, Heintz AP, Burger CW, Bossuyt PM, Mol BW. Improving the existing diagnostic strategy by accounting for characteristics of the women in the diagnostic work up for postmenopausal bleeding. BJOG 2007;114(1): Pecorelli S. Revised FIGO staging for carcinoma of the vulva, cervix, and endometrium. Int J Gynaecol Obstet 2009;105(2): Black WC. The CE plane: a graphic representation of cost-effectiveness. Med Decis Making 1990;10(3): Bachmann LM, ter RG, Clark TJ, Gupta JK, Khan KS. Probability analysis for diagnosis of endometrial hyperplasia and cancer in postmenopausal bleeding: an approach for a rational diagnostic workup. Acta Obstet Gynecol Scand 2003;82(6): Bruchim I, Biron-Shental T, Altaras MM et al. Combination of endometrial thickness and time since menopause in predicting endometrial cancer in women with postmenopausal bleeding. J Clin Ultrasound 2004;32(5): Khan KS, Bachmann LM, ter RG. Systematic reviews with individual patient data meta-analysis to evaluate diagnostic tests. Eur J Obstet Gynecol Reprod Biol 2003;108(2): Clark TJ, Voit D, Gupta JK, Hyde C, Song F, Khan KS. Accuracy of hysteroscopy in the diagnosis of endometrial cancer and hyperplasia: a systematic quantitative review. JAMA 2002;288(13): Clark TJ, Mann CH, Shah N, Khan KS, Song F, Gupta JK. Accuracy of outpatient endometrial biopsy in the diagnosis of endometrial cancer: a systematic quantitative review. BJOG 2002;109(3): Gupta JK, Chien PF, Voit D, Clark TJ, Khan KS. Ultrasonographic endometrial thickness for diagnosing endometrial pathology in women with postmenopausal bleeding: a meta-analysis. Acta Obstet Gynecol Scand 2002;81(9):

50 Chapter Smith-Bindman R, Kerlikowske K, Feldstein VA et al. Endovaginal ultrasound to exclude endometrial cancer and other endometrial abnormalities. JAMA 1998;280(17): Aoki Y, Watanabe M, Amikura T et al. Adjuvant chemotherapy as treatment of high-risk stage I and II endometrial cancer. Gynecol Oncol 2004;94(2): Kumar VJ, Nin CY, Kuei LY, Tan KH, Yeo R, Lam PY. Survival and disease relapse in surgical stage I endometrioid adenocarcinoma of the uterus after adjuvant vaginal vault brachytherapy. Int J Gynecol Cancer 2010;20(4): Broeze KA, Opmeer BC, Bachmann LM et al. Individual patient data meta-analysis of diagnostic and prognostic studies in obstetrics, gynaecology and reproductive medicine. BMC Med Res Methodol 2009;9: Lilford RJ, Pauker SG, Braunholtz DA, Chard J. Decision analysis and the implementation of research findings. BMJ 1998;317(7155): Spencer CP, Whitehead MI. Endometrial assessment re-visited. Br J Obstet Gynaecol 1999;106(7):

51 4 External validation of a mathematical model to estimate the probability of endometrial cancer in women with postmenopausal bleeding M.C. Breijer N. van Hanegem B.C. Opmeer H.C. van Doorn L. Valentin T. van den Bosch R.H.M. Verheijen B.W.J. Mol A. Timmermans Submitted

52 Chapter 4 Abstract Objective: To externally validate the previously developed patient characteristics and patient characteristics and transvaginal sonography (TVS) models estimating the probability of endometrial cancer in women with postmenopausal bleeding (PMB). Design: External validation study. Setting: Two independent datasets of consecutive women not using hormone replacement therapy with a first episode of PMB. Population: One dataset was prospectively collected in three general hospitals in the Netherlands including 559 women, whereas the other dataset was prospectively collected in a university hospital in Sweden including 433 women. Methods: We applied the two models on the two validation databases. The performance of the models was assessed by examining discrimination and calibration. Main Outcome Measures: Discrimination is presented with a receiver operator characteristic (ROC) curve and an area under the ROC curve (AUC). Calibration is presented graphically with calibration plots. Results: The AUC for the patients characteristics -model was 0.71 (95% CI 0.65 to 0.76) in the Dutch database and 0.69 (95% CI 0.64 to 0.73) in the Swedish database. The AUC for the patient characteristics and TVS -model was 0.89 (95% CI 0.86 to 0.92) and 0.89 (95% CI 0.86 to 0.91) in the Dutch and Swedish database, respectively. Conclusions: Although both patient characteristics and TVS models maintained their diagnostic performance in two independent validation databases, they did not seem to offer advantages over endometrial thickness measurement alone. The patient characteristics model is able to select women with a low risk of endometrial cancer to be reassured without further testing, thus allowing its use in a setting without TVS. 50

53 External validation of a mathematical model to estimate the probability of endometrial cancer InTRODuCTIOn Postmenopausal bleeding (PMB) is a common complaint in postmenopausal women who present in both primary and secondary care. Immediately after menopause, PMB occurs in 10% of women. 1 The main objective of the diagnostic work-up for women presenting with PMB is to rule out endometrial cancer. In women with PMB, the probability of endometrial cancer rises from 1% in women younger than 50 years to 24% in women older than 80 years and, regardless of age; the risk of malignancy is higher in women with obesity (18%) or diabetes (21%) than in women without these risk factors (8.0%). 2-8 In the 1990s endometrial thickness measurement with transvaginal ultrasonography (TVS) was introduced as a test to rule out endometrial cancer. 9 However, there is debate on which cut-off value for endometrial thickness should be used when deciding if endometrial sampling needs to be performed or not. 10 A post-test probability of endometrial cancer of 1% seems the worldwide-accepted threshold for patient reassurance The post-test probability depends not only on test characteristics but also on the pre-test probability, which depends on patient s characteristics. In women with PMB, characteristics that define the pre-test probability of endometrial cancer are: age, time since menopause, body mass index (BMI), hypertension, diabetes mellitus, and parity. 2-7, 16 Current guidelines are mainly based on sonographic endometrial thickness only and do not systematically take these 11-14, 17 additional characteristics into account. In 2007 we reported on the development of two multivariable prediction models to estimate the probability of endometrial cancer in women with PMB. Details on the development of the two prediction models can be found in a previous publication. 18 In short, data on 540 women with PMB, not using hormone replacement therapy (HRT), were included in a multivariable regression model. Patient characteristics that satisfied the criteria for inclusion in the model were age, BMI, diabetes, parity and the use of anticoagulants. Two models were developed: (1) the patient characteristics -model, including the five characteristics listed above and (2) the patient characteristics and TVS -model, including the five patient characteristics and endometrial thickness as measured by TVS. The area under the receiver operating characteristic curve (AUC) for the patient characteristics model in the development database was 0.76 (95% CI ) and for the patient characteristics and TVS model 0.90 (95% CI ). 18 For a successful implementation, a prediction model should be validated externally in an independent population. 19 The aim of the present study was to externally validate the diagnostic performance of the existing two models in two independent prospectively collected datasets of women with PMB and to compare these models to the accuracy of endometrial thickness measurement alone. 4 51

54 Chapter 4 Methods The development of the multivariable models We developed two multivariable logistic regression models, which we described in the article by Opmeer et al. 18 The first model only contains characteristics from the history of the women and is referred to as the patient characteristics model. The second model, an extension of the first model also includes endometrial thickness as measured with TVS and is referred to as the patient characteristics and TVS model. Categorical variables with subdivisions (e.g. type and management of diabetes) were dichotomised (e.g. diabetes: yes/ no). Since we reported previously that the accuracy of endometrial thickness measurement was different in obese and non-obese women and in women with diabetes and those without, 5 differences in diagnostic performance across sub-groups were evaluated through interaction terms. Further statistical details are provided in the original article. 18 Study population used for external validation For the present study, we used two prospectively collected databases: I. Dutch database: This database includes all women presenting with postmenopausal bleeding at the TweeSteden hospital Tilburg, the Maxima Medical Centre Veldhoven and the St. Antonius hospital Nieuwegein in the Netherlands between January 2009 and April Menopause was defined as at least one year of amenorrhea. No age criterion was used, if there were doubts about the postmenopausal status this was confirmed by hormone level testing. The following patient characteristics were recorded: age, years since menopause, BMI, parity, HRT use, hypertension, diabetes, use of anticoagulants and endometrial thickness as measured by TVS. If double endometrial thickness was > 4.0 mm endometrial biopsy using the Pipelle (Labaratoire CCD, Paris, France) was performed. In case of a failed endometrial biopsy, hysteroscopy with directed biopsy was performed. II. Swedish database: This database includes all women presenting with PMB at the Skåne University Hospital Malmö postmenopausal bleeding clinic between November 2002 and June Menopause was defined as at least one year of amenorrhea after the age of 40. The following patient characteristics were recorded: age, age at menopause, weight, height, parity, HRT use, hypertension, diabetes, use of anticoagulants and endometrial thickness as measured by TVS. If double endometrial thickness was 4.5 mm, saline infused sonography (SIS) was performed. If there were no focal lesions in the uterine cavity at SIS, an endometrial sample using the Endorette (Medscan AB, Malmö, Sweden) was taken. If there were focal lesions at SIS or if SIS failed, hysteroscopic resection was performed

55 External validation of a mathematical model to estimate the probability of endometrial cancer All women in both databases were instructed to contact the hospital if recurrence of bleeding occurred. Follow-up in the Dutch database was based on data collected from case notes. For the purpose of this study, all women with an endometrial thickness below the cut-off value (without endometrial biopsy and without recurrent bleeding) were considered negative for endometrial cancer. In case of recurrent bleeding hysteroscopy was performed. In one patient endometrial cancer was diagnosed after recurrent postmenopausal bleeding during follow-up. The median follow-up time in this database was 26 months (range 18 to 43 months). In the Swedish database, all women with an endometrial thickness < 4.4 mm (and therefore without a histological diagnosis of the endometrium) were matched with the regional cancer register to ascertain if any of these women were diagnosed with endometrial cancer after inclusion in the study. Women were excluded from further analysis if the endometrial thickness was not measurable. In line with the classification in the development database, definite disease state was determined as benign in women whose histology specimen showed atrophy, benign polyps, simple hyperplasia or proliferative endometrium. Premalignancy, defined as any form of hyperplasia with atypia, and malignancy in the histology specimen were combined in the analysis in the diagnostic group endometrial malignancy since both diagnostic categories warrant further treatment. 4 Statistical Analysis Missing values Generally, dropping cases with missing values (complete case analysis) yields biased results, and the discriminative ability of a multivariable model is reduced when a case with missing values is dropped from the analysis. 21 In multiple imputation, each missing value is imputed several times. The variation among the imputations reflects the uncertainty with which the missing values can be predicted from the observed ones. After combining the results, the overall estimates and standard errors reflect missing data uncertainty In our validation study, we performed multiple imputations for missing variables, with separate imputation rounds for each of the two databases. External validation We applied the two models to the women in the Dutch and Swedish databases. We assessed the performance of the models by examining the calibration (agreement between predicted risks and observed frequencies of endometrial cancer) and discrimination (the ability of the models to distinguish between women with and without endometrial cancer). To assess calibration, we plotted the predicted probabilities of endometrial cancer and the observed 53

56 Chapter 4 proportion of endometrial cancer by deciles of the predicted probabilities in a calibration plot , 27 Calibration is considered perfect if the intercept is 0 and the calibration slope is 1. We assessed discrimination by calculating the area under the receiver operator characteristic curve (AUC). To compare the performance of the two models with the performance of the currently applied strategy of measuring endometrial thickness ( TVS-only ), we calculated the AUC for TVS-only. Results The two databases available for external validation consisted of 559 Dutch and 433 Swedish women with PMB not using HRT. Table 1 shows the characteristics of women in the two databases and the percentage of missing data per database. Age, time since menopause, anticoagulants use, body mass index (BMI), endometrial thickness and the frequency of endometrial cancer differed significantly between the two validation populations, women in the Swedish database being older, having lower BMI, thicker endometrium, a higher percentage of endometrial cancer and more women used anticoagulant therapy. Table 1. Patient characteristics and missing values in validation databases. Swedish database Missing Dutch database Missing P N Age (years) / / <0.01# DM 66 (15.2) 0 72 (12.9) 1 (0.2) 0.29 HT 176 (40.6) (35.1) AC 88 (31.9) 157 (13.2) 94 (16.8) 0 <0.01 BMI (kg/m 2 ) / (5.5) / (43.4) <0.01# TMP (years)* 16 (5-26) 6 (1.4) 5 (2-14) 124 (22.2) <0.01 Nulliparity 45 (10.6) 8 (1.8) 60 (13.5) 114 (20.4) 0.19 ET (mm)* 6.0 ( ) NA 5.7 ( ) NA 0.02 Endometrial (pre-) malignancy Atypical hyperplasia Endometrial cancer 65 (15.0) 3 (0.69) 62 (14.3) 0 57 (10.2) 7 (1.3) 50 (8.9) DM, diabetes mellitus; HT, hypertension; AC, anticoagulants; BMI, body mass index; TMP, time since menopause; ET, endometrial thickness; NA, not applicable. Results presented are n (%), mean +/-SD or median and (interquartile range) * not normally distributed, values presented as median and (interquartile range) # Independent samples T-test Chi-square Independent samples Mann-Whitney U Test 54

57 External validation of a mathematical model to estimate the probability of endometrial cancer Calibration Calibration plots for the patient characteristics model and the patient characteristics and TVS model in the Dutch and Swedish database are presented in Figure 1. In the Dutch database, the calibration slope is better for the patient characteristics and TVS model compared to the patient characteristics model. In patients with a high risk of endometrial carcinoma predicted probabilities are slightly overestimated when using the patient characteristics model. Because this is only in the high-risk region, the model is capable of selecting women with a low risk, who can be reassured without further invasive testing. Actual Probability Dxy C (ROC) R2 D U Q Brier Intercept Slope Emax S:z S:p Ideal Logistic calibration Nonparametric Grouped observations Actual Probability Dxy C (ROC) R2 D U Q Brier Intercept Slope Emax S:z S:p Ideal Logistic calibration Nonparametric Grouped observations Predicted Probability A. Patient characteristics -model Dutch validation database Predicted Probability B. Patient characteristics and TVS -model Dutch validation database. Actual Probability Dxy C (ROC) R2 D U Q Brier Intercept Slope Emax S:z S:p Ideal Logistic calibration Nonparametric Grouped observations Actual Probability Dxy C (ROC) R2 D U Q Brier Intercept Slope Emax S:z S:p Ideal Logistic calibration Nonparametric Grouped observations Predicted Probability C. Patient characteristics -model Swedish validation database Predicted Probability D. Patient characteristics and TVS -model Swedish validation database. Figure 1. Calibration plots Patient characteristics only -model and Patient characteristics and TVS - model. 55

58 Chapter 4 In the Swedish database, the calibration slope performed better for the patient characteristics model compared to the patient characteristics and TVS model. For the patient characteristics model, predicted probabilities were close to the observed frequency in all patients, low- or high risk. The patient characteristics and TVS model underestimated the probability of endometrial cancer over almost the whole range of probabilities except for very low probabilities. 1 0,8 0,6 sensitivity 0,4 0, ,2 0,4 0,6 0,8 1 1-specificity A. Dutch validation database --- Patient characteristics -model, AUC: 0.71 (95% CI: 0.65 to 0.76) Patient characteristics and TVS - model, AUC: 0.89 (95% CI: 0.86 to 0.92) Endometrial thickness AUC: 0.87 (95% CI: 0.83 to 0.90) 1 0,8 0,6 sensitivity 0,4 0, ,2 0,4 0,6 0,8 1 1-specificity B. Swedish validation database --- Patient characteristics - model, AUC: 0.69 (95% CI: 0.64 to 0.73) Patient characteristics and TVS - model, AUC: 0.89 (95% CI: 0.86 to 0.91) Endometrial thickness AUC: 0.90 (95% CI: 0.88 to 0.93) Figure 2. ROC curves Patient characteristics only -model and Patient characteristics and TVS -model. 56

59 External validation of a mathematical model to estimate the probability of endometrial cancer Discrimination Figure 2 shows ROC curves for the two models and for TVS alone in the two validation datasets. In both the Dutch and the Swedish databases, discrimination as reflected by the AUC for the patient characteristics and TVS model (respectively 0.89 (95% CI ) and 0.89 (95% CI ) was higher compared to the AUC for the patients characteristics - model (0.71 (95% CI ) in the Dutch database and 0.69 (95% CI ) in the Swedish database). For TVS-only, the AUC was 0.87 (95% CI ) in the Dutch Database and 0.90 (95% CI ) in the Swedish database. Comparison with current practice In the original publication on model development, model-based strategies were compared to the currently applied strategy of endometrial thickness measurement with a cut-off value of five mm, to reflect clinical practice in the Netherlands at that time. 18 Because the AUC s in the patient characteristics model and in the patient characteristics and TVS model were calculated with the predicted probability as a continuous variable, we re-calculated the AUC for TVS-only with endometrial thickness as a continuous variable instead of a dichotomized variable (with a fixed cut-off). The AUC for TVS-only was 0.87 (95% CI ) in the Dutch and 0.90 (95% CI ) in the Swedish database. The curves (figure 2) are almost identical to the curves of the patient characteristics and TVS model. The patient characteristics model has a much lower AUC than TVS alone in both databases. Nevertheless, this patient characteristics model could be used in situations where an ultrasound is not (immediately) available in the physicians office. 4 DISCuSSIOn We assessed the external validity and generalizability of two previously developed clinical prediction models 18 to estimate the probability of endometrial cancer in women presenting with PMB in two independent prospective cohorts. In this external validation study, we demonstrated that the diagnostic performance of the models was similar to the diagnostic performance of the original model in internal validation. The diagnostic performance of the patient characteristics and TVS model however, is comparable to endometrial thickness measurement by TVS. The patient characteristics model is able to select women with a low risk of endometrial cancer, who can be reassured without further invasive testing. In the original publication, a predicted risk lower than 4% was equal to a negative predictive value of 99%. 18 This means that in the original model, calibration was not optimal in the low risk range; the predicted risk was higher than the observed proportion of endometrial cancer. In clinical practice, a woman can be reassured without TVS and without further testing if she is 57

60 Chapter 4 aged 50 years or younger and has up to one additional risk factor. In a primary care setting or in a health care setting where TVS cannot be performed immediately and has to be ordered separately and be performed by a radiologist, this model is able to select women that can be reassured without TVS and without further testing. Different calibration was observed for the models in the two different databases. In the Swedish database the predicted probability of having endometrial carcinoma is underestimated in the patient characteristics and TVS model, except in women with a very low risk for having endometrial cancer. This difference may be due to a different composition of the two validation databases. Women in the Swedish database had lower BMI, were older, had thicker endometrium and more women had endometrial cancer. Strengths of our current study are the external validation of the model using data from a different region within the Netherlands as well as data from another European country. The diagnostic performance of the models is comparable to the diagnostic performance in the development database, although patient groups are significantly different. Another strength is that we performed a comparison with current clinical practice: measurement of endometrial thickness in all women with PMB. The use of the model, adding patient characteristics to endometrial thickness, does not seem to improve the efficiency of the diagnostic workup over TVS alone. With a statistical approach, we hoped to improve the diagnostic work-up for women with PMB by individualizing the strategy. 28 In the publication on model development we concluded that accounting for the characteristics of the women could increase the efficiency. 18 With a more clinical approach in this validation study, we find no added value of the patient characteristics and TVS model over the use of TVS-only. One of the limitations of this study is missing data. Multiple imputation was used to handle these missing data. Multiple imputation, even with a relatively large amount of missing data, gives a more precise and valid measure of association for variables with missing values than complete case analysis. 21, 22 Another limitation is the fact that in the two validation databases a different cut-off value for endometrial thickness was used. In the Dutch database, there was one patient with an endometrial thickness between 4.0 and 4.5 mm diagnosed with endometrial cancer. All women with an endometrial thickness 4.4 mm from the Swedish database were linked to the national cancer registry. None of these women were diagnosed with endometrial cancer during follow-up. Therefore we think that the different cut-offs used in two databases did not influence our results. Several prediction models have been published to estimate the risk of endometrial cancer in women with postmenopausal bleeding. 29 Opolskiene et al developed a prediction model for women with an endometrial thickness 4.5 mm focusing on the combination of clinical and ultrasound characteristics: endometrial thickness and power Doppler examination of the endometrium. 20 Burbos et al. developed two models based on patient characteristics 58

61 External validation of a mathematical model to estimate the probability of endometrial cancer 18, 30, 31 with and without endometrial thickness, similar to the models we published in None of these models were externally validated. 29 External validation, assessing the validity and generalizability of a model is an essential step before a model can be implemented in practice. 25, 32 Our study is the first to describe external validation of a model estimating the risk of endometrial carcinoma in women with PMB. Conclusions This study shows that after external validation in two independent datasets, the existing multivariable models maintain their diagnostic performance and are able to distinguish between women with low or high risk of endometrial cancer in women with postmenopausal bleeding not using HRT. The patient characteristic and TVS model however, offers no diagnostic advantage over the measurement of endometrial thickness alone. The patient characteristics model is able to select women with a low risk of endometrial cancer, who can be reassured without further testing. This is especially useful in a setting where TVS is not (directly) available. 4 59

62 Chapter 4 References 1. Astrup K, Olivarius NF. Frequency of spontaneously occurring postmenopausal bleeding in the general population. Acta Obstet Gynecol Scand 2004;83(2): Anderson KE, Anderson E, Mink PJ et al. Diabetes and endometrial cancer in the Iowa women s health study. Cancer Epidemiol Biomarkers Prev 2001;10(6): Gull B, Karlsson B, Milsom I, Granberg S. Factors associated with endometrial thickness and uterine size in a random sample of postmenopausal women. Am J Obstet Gynecol 2001;185(2): McPherson CP, Sellers TA, Potter JD, Bostick RM, Folsom AR. Reproductive factors and risk of endometrial cancer. The Iowa Women s Health Study. Am J Epidemiol 1996;143(12): van Doorn LC, Dijkhuizen FP, Kruitwagen RF, Heintz AP, Kooi GS, Mol BW. Accuracy of transvaginal ultrasonography in diabetic or obese women with postmenopausal bleeding. Obstet Gynecol 2004;104(3): Weiderpass E, Persson I, Adami HO, Magnusson C, Lindgren A, Baron JA. Body size in different periods of life, diabetes mellitus, hypertension, and risk of postmenopausal endometrial cancer (Sweden). Cancer Causes Control 2000;11(2): Xu WH, Xiang YB, Ruan ZX et al. Menstrual and reproductive factors and endometrial cancer risk: Results from a population-based case-control study in urban Shanghai. Int J Cancer 2004;108(4): Gredmark T, Kvint S, Havel G, Mattsson LA. Histopathological findings in women with postmenopausal bleeding. Br J Obstet Gynaecol 1995;102(2): Smith-Bindman R, Kerlikowske K, Feldstein VA et al. Endovaginal ultrasound to exclude endometrial cancer and other endometrial abnormalities. JAMA 1998;280(17): Timmermans A, Opmeer BC, Khan KS et al. Endometrial thickness measurement for detecting endometrial cancer in women with postmenopausal bleeding: a systematic review and meta-analysis. Obstet Gynecol 2010;116(1): Epstein E. Management of postmenopausal bleeding in Sweden: a need for increased use of hydrosonography and hysteroscopy. Acta Obstet Gynecol Scand 2004;83(1): Goldstein RB, Bree RL, Benson CB et al. Evaluation of the woman with postmenopausal bleeding: Society of Radiologists in Ultrasound-Sponsored Consensus Conference statement. J Ultrasound Med 2001;20(10): NVOG (Dutch Society of Obstetrics and Gynaecology). NVOG-richtlijn Abnormaal vaginaal bloedverlies in de menopauze [In Dutch]. NVOG Guideline: Abnormal vaginal bleeding during menopause Scottish Intercollegiate Guidelines Network. Investigation of postmenopausal bleeding. Scottish Intercollegiate Guidelines Network, Royal College of Physicians Van Hanegem N, Breijer MC, Khan KS et al. Diagnostic evaluation of the endometrium in postmenopausal bleeding: an evidence-based approach. Maturitas 2011;68(2): Reeves GK, Pirie K, Beral V, Green J, Spencer E, Bull D. Cancer incidence and mortality in relation to body mass index in the Million Women Study: cohort study. BMJ 2007;335(7630): American College of Obstetricians and Gynecologists. ACOG Committee Opinion No. 426: The role of transvaginal ultrasonography in the evaluation of postmenopausal bleeding. Obstet Gynecol 2009;113(2 Pt 1): Opmeer BC, van Doorn HC, Heintz AP, Burger CW, Bossuyt PM, Mol BW. Improving the existing diagnostic strategy by accounting for characteristics of the women in the diagnostic work up for postmenopausal bleeding. BJOG 2007;114(1): McGinn TG, Guyatt GH, Wyer PC, Naylor CD, Stiell IG, Richardson WS. Users guides to the medical literature: XXII: how to use articles about clinical decision rules. Evidence-Based Medicine Working Group. JAMA 2000;284(1): Opolskiene G, Sladkevicius P, Valentin L. Prediction of endometrial malignancy in women with postmenopausal bleeding and sonographic endometrial thickness >/= 4.5 mm. Ultrasound Obstet Gynecol 2011;37(2): Janssen KJ, Donders AR, Harrell FE, Jr. et al. Missing covariate data in medical research: to impute is better than to ignore. J Clin Epidemiol 2010;63(7):

63 External validation of a mathematical model to estimate the probability of endometrial cancer 22. Clark TG, Altman DG. Developing a prognostic model in the presence of missing data: an ovarian cancer case study. J Clin Epidemiol 2003;56(1): Schafer JL. Multiple imputation: a primer. Stat Methods Med Res 1999;8(1): van BS, Boshuizen HC, Knook DL. Multiple imputation of missing blood pressure covariates in survival analysis. Stat Med 1999;18(6): Steyerberg E. Clinical prediction models. A practical approach to development, validation, and updating. New York: Springer; Janssen KJ, Moons KG, Kalkman CJ, Grobbee DE, Vergouwe Y. Updating methods improved the performance of a clinical prediction model in new patients. J Clin Epidemiol 2008;61(1): Steyerberg EW. Assessing the performance of prediction models: a framework for traditional and novel measures Khan KS, Bachmann LM, Ter RG. Systematic reviews with individual patient data meta-analysis to evaluate diagnostic tests. Eur J Obstet Gynecol Reprod Biol 2003;108(2): Van Hanegem N, Breijer MC, Opmeer BC, Mol BW, Timmermans A. Prediction models in women with postmenopausal bleeding: a systematic review. Womens Health (Lond Engl ) 2012;8(3): Burbos N, Musonda P, Giarenis I et al. Predicting the risk of endometrial cancer in postmenopausal women presenting with vaginal bleeding: the Norwich DEFAB risk assessment tool. Br J Cancer 2010;102(8): Burbos N, Musonda P, Duncan TJ, Crocker SG, Morris EP, Nieto JJ. Estimating the risk of endometrial cancer in symptomatic postmenopausal women: a novel clinical prediction model based on patients characteristics. Int J Gynecol Cancer 2011;21(3): Justice AC, Covinsky KE, Berlin JA. Assessing the generalizability of prognostic information. Ann Intern Med 1999;130(6):

64

65 5 Factors attributing to the failure of endometrial sampling in women with postmenopausal bleeding N.C.M. Visser M.C. Breijer M.C. Herman R.L.M. Bekkers S. Veersema B.C. Opmeer B.W.J. Mol A. Timmermans J.M.A. Pijnenborg Submitted

66 Chapter 5 Abstract Objective: To determine which doctor and patient related factors affect failure of outpatient endometrial sampling in women with postmenopausal bleeding, and to develop a multivariable prediction model to select women with a high probability of a failed sampling. Design: Prospective multicenter cohort study. Setting: Three teaching hospitals in the Netherlands Population: Women presenting with postmenopausal bleeding with an indication for endometrial sampling. Methods: Multivariable logistic regression was performed to evaluate the impact of doctor s training level and patient s characteristics on failure of sampling. Main outcome measures: Failure of endometrial sampling, classified as technical failure or insufficient tissue for histological diagnosis. Results: In 74 (20.8%) of the 356 included women sampling technically failed, and in 84 (29.8%) the amount of tissue was insufficient for histological diagnosis. Nulliparity (OR 3.8, 95% CI ) and advanced age (OR 1.03 per year, 95% CI ) were associated with technical failure. Advanced age was associated with insufficient tissue sampling (OR 1.04 per year, 95% CI ), and increased endometrial thickness decreased the chance of insufficient sampling (OR 0.3, 95% CI ). The prediction model for total failure had an area under the receiver operating characteristics (ROC) curve of 0.64 (95% CI ). Conclusions: In women with postmenopausal bleeding, failure rate of endometrial sampling is relatively high, and is associated with nulliparity, advanced age and endometrial thickness. A multivariable prediction model for total failure based on patient characteristics has a moderate capacity to discriminate between women with a high or a low risk of failure. 64

67 Factors attributing to the failure of endometrial sampling in women with postmenopausal bleeding InTRODuCTIOn Postmenopausal bleeding accounts for approximately five percent of gynecological consultations. 1 Guidelines for women presenting with postmenopausal bleeding advocate measurement of the endometrial thickness by transvaginal ultrasound. 2, 3 In women with an endometrial thickness > 4 mm, endometrial sampling is warranted to rule out a malignancy of the endometrium. 2-5 Dilatation and curettage (D&C) was traditionally the method of choice to obtain endometrial tissue for the histologic investigation in women with postmenopausal bleeding. Nowadays, outpatient endometrial sampling is recommended to rule out endometrial carcinoma in women with an increased endometrial thickness, and postmenopausal bleeding. 6 This outpatient endometrial sampling has a sensitivity of 99.6% for excluding endometrial carcinoma in women presenting with postmenopausal bleeding. 7 It has been demonstrated that outpatient endometrial sampling after transvaginal ultrasound is a more cost-effective strategy than immediate hysteroscopy. 8, 9 However, in a substantial number of cases outpatient endometrial sampling fails because of technical problems (12-23%) or because of an insufficient amount of tissue for a histopathological diagnosis (16-68%) In 6% of these failures, women are subsequently diagnosed with both malignant and premalignant disorders of the endometrial tissue. 11 Previous studies investigated factors that could influence failure of endometrial sampling, i.e.: nulliparity, age, endometrial thickness and training level Yet, these studied populations also included premenopausal women and results were mainly based on univariate analyses. Furthermore, since most studies did not focus primarily on failure rate, the results of failed endometrial sampling might be underestimated. In order to investigate which doctor and patient related factors contribute to failure of outpatient endometrial sampling in postmenopausal women, the current study was designed. The secondary aim of the current study was to develop a multivariable prediction model in order to select women with a high probability of a failed outpatient endometrial sampling. 5 MATERIAl AnD METHODS From January 2009 to April 2011 all women presenting with postmenopausal bleeding at the TweeSteden hospital, Tilburg, the Máxima Medical Center, Veldhoven and the St. Antonius hospital, Nieuwegein in The Netherlands were included in a prospective study. Postmenopausal bleeding was defined as one-time or recurrent bleeding after a period of at least 12 months of amenorrhea. According to the guideline of the Dutch Society for Obstetrics and Gynecology, 3 the diagnostic work-up consisted of a double-layer measurement of the endometrium by 65

68 Chapter 5 transvaginal ultrasound. Women with a first episode of postmenopausal bleeding and an endometrial thickness 4 mm were reassured and instructed to contact the gynecologist in case of recurrent postmenopausal bleeding. When the endometrial thickness was > 4 mm, not measurable, or when the episode of postmenopausal bleeding was recurrent, outpatient endometrial sampling was performed. The following patient characteristics were recorded: age, body mass index (BMI), parity, years since menopause, use of hormone therapy, presence of hypertension, diabetes, use of anticoagulants, and double-layer endometrial thickness in the midsagittal plane measured by transvaginal ultrasound. The training level of doctors who performed the outpatient endometrial sampling was categorized as resident or gynecologist. If the procedure performed by the resident failed, endometrial sampling was not repeated by a gynecologist. There was no standardized sampling method or recommendations on using a tenaculum or entering the uterine cavity more than once. Failure of endometrial sampling was classified as either an endometrial sampling that could not be performed (A: technical failure) or as an endometrial sampling in which the procedure was performed, but the amount of tissue was insufficient for a reliable histopathological diagnosis (B: insufficient sampling). Inconclusive samples were reviewed by one pathologist with special interest in gynecology. All outpatient endometrial samplings were performed using a Pipelle sampling device (Pipelle de Cornier, Paris, France). In case of technical failure or insufficient sample for histopathological diagnosis, outpatient endometrial sampling was repeated, hysteroscopic evaluation of the endometrium was performed, or dilatation and curettage was performed. Follow-up data were collected from the patient charts. Diagnosis was based on a final histological examination by a pathologist with special interest in gynecology. No ethical approval was needed for this observational study according to the Code of Conduct for the use of data in Health Research (Dutch federation of Biomedical Scientific Societies, www. federa.org). Statistical analysis Statistical analysis was performed with the PASW statistics 18.0 package (SPSS, Inc., Somers, NY, USA). Factors that could be predictive for failure of endometrial sampling were evaluated in univariate analysis. Categorical data were analyzed with a Chi-square test. Continuous variables were tested for normal distribution, and if not a Mann-Whitney U test for univariate analysis was used. Endometrial thickness was categorized into not-measurable, 8 mm, 8 12 mm and > 12 mm, based on cubic splines. To evaluate the independent effect of the tested variables, a multivariable logistic regression analysis with backward stepwise selection was performed. Variables were included in the backward selection procedure if the p-value in the univariate analysis was <0.30 (p

69 Factors attributing to the failure of endometrial sampling in women with postmenopausal bleeding to stay in the model). A multivariable prediction model was developed to predict failure of endometrial sampling by combining technical failure (A) and insufficient sampling (B) into total failure. Both discrimination and calibration were assessed for the model. Subsequently, a receiver operating characteristics (ROC) curve was made to assess the discriminative value of the model. Calibration was assessed to measure the agreement between predicted and observed outcomes. The extent of over- or underestimation relative to the observed and predicted rate was explored graphically using calibration plots. 15 The predicted probability was compared to the observed proportion of failed samples in subgroups divided by deciles of the predicted probability. Multiple imputation was used in order to obtain more precise and valid measures of possible association for variables with missing values. 16 To account for differences between imputation sets, we performed variable selection in each imputed dataset separately. The variables for the final multivariable analysis were selected using the majority method, meaning that variables were selected in at least three of five imputed datasets. A pooled analysis was performed with the selected variables. Variables were checked for collinearity. A p-value of <0.05 was considered to indicate statistical significance. 5 RESulTS A total of 675 women presented with postmenopausal bleeding. Endometrial thickness was 4 mm in 239 women and in 80 patients no endometrial sampling was attempted. In the remaining 356 women, an attempt to perform an endometrial sampling was undertaken. These women were included for analysis. The median age was 60 years (range years), the median BMI was 29 kg/m 2 (range kg/m 2 ) and the median time since menopause was 10 years (range 1-49 years). 46 women (12.9%) were nulliparous. Endometrial cancer was diagnosed in 57 (16%) of the women. All patient characteristics are summarized in Table 1. Overall, outpatient endometrial sampling was successful with a histological diagnosis in 198 of the 356 women (55.6%) (Figure 1). Systematic revision of all inconclusive samples by a pathologist specialized in gynecology did not yield different results. In 32 (9.0%) women endometrial thickness could not be measured. A separate analysis eliminating patients in which endometrial thickness could not be measured did not change the results (data not shown). In 74 (20.8%) of 356 women sampling technically failed. In 16 (21.6%) of these 74 women an endometrial carcinoma was diagnosed and in one (1.4%) woman atypical hyperplasia. Additional investigations were omitted in five (6.8%) women, and hysteroscopy failed in three (4.7%) women due to cervical stenosis or pain. 67

70 Chapter 5 Table 1. Patient characteristics of the 356 postmenopausal women Characteristics of the women Age (years) 60 (41-92) BMI (kg/m 2 ) 29 (18-76) Years since menopause (years) 10 (1-49) Training level resident 108 (30.3) gynecologist 248 (69.7) Endometrial thickness not measurable 32 (9.0) 8 mm 162 (45.5) 8-12 mm 69 (19.4) > 12 mm 93 (26.1) Parity nulliparous 46 (12.9) multiparous 310 (87.1) Hormone therapy yes 19 (5.3) no 337 (94.7) Hypertension yes 143 (40.2) no 213 (59.8) Diabetes Mellitus yes 55 (15.4) no 301 (84.6) Anticoagulants use yes 73 (20.5) no 283 (79.5) Endometrial cancer yes 57 (16.0) no 299 (84.0) Values are presented as median (range), number (percent). The results of univariate and multivariable analysis are presented in Table 2. In univariate analysis the percentage of technical failure was significantly higher in nulliparous women compared to multiparous women (41.3 vs. 17.7%, odds ratio [OR] 3.2, 95% confidence interval [CI] ). Advanced age, years since menopause, and presence of hypertension were both significantly associated with technical failure. The doctor s training level was not significantly associated with technical failure. As can be expected, age and years since menopause were highly correlated (Spearman s r 0.86), therefore years since menopause was excluded from multivariable analysis. 68

71 Factors attributing to the failure of endometrial sampling in women with postmenopausal bleeding Office endometrial sampling (n=356) Uterine cavity (n=282, 79.2%) Endometrial carcinoma (n=30, 10.6%) Atypical hyperplasia (n=8, 2.8%) Polyp (n=8, 2.8%) Benign (n=152, 53.9%) Insufficient sample (n=84, 29.8%) Hysteroscopy (n=3) Hysterectomy (n=27) Hysteroscopy (n=3) Hysterectomy (n=5) Hysteroscopy (n=3) No further procedures (n=5) Technical failure (n=74, 20.8%) Hysteroscopy (n=64) No further procedures (n=5) Hysteroscopy (n=53) No further procedures (n=99) Other diagnostic procedure (n=5) Hysteroscopy (n=41) No further procedures (n=38) Endometrial carcinoma: 3 Endometrial carcinoma: 27 Benign: 2 Endometrial carcinoma: 1 Polyp: 5 Polyp: 3 Atypical hyperplasia: 4 Endometrial carcinoma: 1 Atrophy: 6 Polyp: 28 Benign: 15 Endometrial carcinoma: 3 Atypical hyperplasia: 1 Benign: 97 Cervical polyp: 1 Endometrial carcinoma: 1 Cervical polyp: Cervical carcinoma: 1 No diagnosis: 5 Atrophy: 9 Polyp: 18 Benign: 19 Atypical hyperplasia: 1 Endometrial carcinoma: 15 No diagnosis: 2 Other diagnostic procedure (n=5) Polyp: 1 Endometrial carcinoma: 1 Cervical polyp: 2 Atrophy: 5 Polyp: 6 No diagnosis: 27 Atypical hyperplasia: 1 Endometrial carcinoma: 5 Cervical polyp: Atrophy: 5 Polyp: 6 No diagnosis: 27 Cervical carcinoma: 1 Sigmoid carcinoma: Cervical carcinoma: 2 Figure 1. Flow chart of diagnostic work-up and final diagnosis in 356 women with postmenopausal bleeding 5 69

72 Chapter 5 In the multivariable analysis only nulliparity (OR 3.8, 95% CI ) and advanced age (OR 1.03 per year, 95% CI ) were independently associated with technical failure. Table 2. Results of univariate and multivariable analysis of factors attributing to technical failure Technical failure Yes, n =74 (20.8%) No, n=282 (79.2%) Univariate analysis Multivariable analysis OR p Adjusted OR p Age (years) 65 (45-90) 60 (41-92) ( ) 0.02 BMI (kg/m 2 ) 29 (18-50) 29 (18-76) Years since menopause 14 (1-48) 9 (1-49) (years) Training level resident 16 (14.8) 92 (85.2) 0.57 ( ) ( ) 0.15 gynecologist 58 (23.4) 190 (76.6) Endometrial thickness not measurable 10 (31.3) 22 (68.8) reference * mm 28 (17.3) 134 (82.7) 0.46 ( ) mm 16 (23.2) 53 (76.8) 0.66 ( ) 0.39 > 12 mm 20 (21.5) 73 (78.5) 0.60 ( ) 0.27 Parity nulliparous 19 (41.3) 27 (58.7) 3.24 ( ) ( ) multiparous 55 (17.7) 255 (82.3) Hormone therapy yes 4 (21.1) 15 (78.9) 1.00 ( ) no 70 (20.8) 267 (79.2) Hypertension yes 38 (26.6) 105 (73.4) 1.79 ( ) ( ) 0.14 no 36 (16.9) 177 (83.1) Diabetes Mellitus yes 16 (29.1) 39 (70.9) 1.70 ( ) no 58 (19.3) 243 (80.7) Anticoagulants use yes 21 (28.8) 52 (71.2) 1.75 ( ) no 53 (18.7) 230 (81.3) Values are presented as median (range), number (percent). Two-sample Wilcoxon rank sum test (Mann-Whitney test) Χ 2 test * Overall p-value In 84 (29.8%) of 282 women, the amount of tissue obtained was insufficient for a histopathological diagnosis. In 6 (7.1%) of these an endometrial carcinoma was diagnosed and in one (1.2%) woman atypical hyperplasia. In 38 (45.2%) women with an insufficient sample no further histological evaluation of the endometrium was performed. In one patient an endometrial carcinoma was diagnosed after recurrent postmenopausal bleeding 70

73 Factors attributing to the failure of endometrial sampling in women with postmenopausal bleeding during a median follow-up time of 26 months (range months). Women in whom no additional examination was performed were significantly less obese (p = 0.023), had a thinner endometrium (p = 0.001), and less diabetes (p = 0.037) when compared to women with additional histological examination. Results of univariate and multivariable analysis are presented in Table 3. Table 3. Results of univariate and multivariable analysis of factors attributing to insufficient sampling Technical failure Yes, n =84 (29.8%) No, n=198 (70.2%) Univariate analysis Multivariable analysis OR p Adjusted OR p Age (years) 63 (42-88) 59 (41-92) ( ) 0.01 BMI (kg/m 2 ) 28 (18-52) 29 (18-76) Years since menopause 13 (1-39) 7 (1-49) (years) Training level resident 32 (34.8) 60 (65.2) 1.42 ( ) ( ) 0.11 gynecologist 52 (27.4) 138 (72.6) Endometrial thickness not measurable 10 (45.5) 12 (54.5) reference 0.09* reference 0.03* 8 mm 45 (33.6) 89 (66.4) 0.61 ( ) ( ) mm 14 (26.4) 39 (73.6) 0.43 ( ) ( ) 0.10 > 12 mm 15 (20.5) 58 (79.5) 0.31 ( ) ( ) 0.01 Parity nulliparous 9 (33.3) 18 (66.7) 1.20 ( ) multiparous 75 (29.4) 180 (70.6) Hormone therapy yes 2 (13.3) 13 (86.7) 0.41 ( ) ( ) 0.24 no 82 (30.7) 185 (69.3) Hypertension yes 37 (35.2) 68 (64.8) 1.51 ( ) no 47 (26.6) 130 (73.4) Diabetes Mellitus yes 13 (33.3) 26 (66.7) 1.22 ( ) no 71 (29.2) 172 (70.8) Anticoagulants use yes 20 (38.5) 32 (61.5) 1.59 ( ) no 64 (27.8) 166 (72.2) Values are presented as median (range), number (percent). Two-sample Wilcoxon rank sum test (Mann-Whitney test) Χ 2 test * Overall p-value 5 71

74 Chapter 5 Advanced age was associated with insufficient sampling (p = 0.009). Endometrial thickness > 12 mm was negatively associated with insufficient sampling (OR 0.3, 95% CI ). Training level was not significantly associated with insufficient sampling. Again, because age and years since menopause were highly correlated (Spearman s r 0.79), years since menopause was excluded in multivariable analysis. Advanced age was independently associated with insufficient sampling (OR 1.04 per year, 95% CI ) and endometrial thickness > 12 mm independently decreased the chance of insufficient sampling (OR 0.3, 95% CI ). Overall, outpatient endometrial sampling failed in 158 out of 356 women (44.4%). Based on the outcomes of the univariate analyses, factors that satisfied the criteria for inclusion in the model were age (OR 1.04 per year, 95% CI ), endometrial thickness ( 8 mm OR 0.7, 95% CI ; 8-12 mm OR 0.5, 95% CI ; >12 mm OR 0.4, 95% CI ), nulliparity (OR 2.3, 95% CI ) and presence of hypertension (OR 1.5, 95% CI ). The results of the multivariable logistic regression analysis are presented in Table 4. Table 4. Multivariable logistic regression model for the prediction of failed endometrial sampling in women with postmenopausal bleeding Predictor OR (95% CI) Age 1.04 ( ) Endometrial thickness not measurable reference category 8 mm 0.66 ( ) 8 12 mm 0.52 ( ) > 12 mm 0.39 ( ) Nulliparity 2.28 ( ) Hypertension 1.53 ( ) Figure 2 shows the receiver operating characteristics (ROC) curve for the model. The area under the ROC curve was 0.64 (95% CI ), indicating a moderate capacity to discriminate between women in whom endometrial sampling will fail and women in whom it will succeed. Figure 3 shows the calibration plot. Predicted probabilities ranged from 18% to 88%. There was a good agreement between the predicted probabilities and the observed proportion of failed endometrial samples. 72

75 Factors attributing to the failure of endometrial sampling in women with postmenopausal bleeding Figure 2. Receiver operating characteristic (ROC) curve for multivariable model to predict failure of endometrial sampling. The ROC curve was constructed by changing the cut-off point of the probability of failure. 5 Figure 3. Internal validation of the prediction model. Calibration plot comparing the observed proportion of failed endometrial sampling versus the predicted probability of failed endometrial sampling using the multivariable prediction model. 73

76 Chapter 5 Discussion In this prospective multicenter study we observed successful endometrial sampling in just over half of the women presenting with postmenopausal bleeding. Sampling failed for technical reasons in 21% of the women, whereas in 30% the amount of tissue obtained was insufficient for a histopathological diagnosis, resulting in a total failure of endometrial sampling of 44%. Nulliparity and advanced age were associated with failure and an endometrial thickness > 12 mm decreased the chance of failure. Training level was not associated with technical failure or insufficient sampling. A failed sampling does not rule out a malignancy of the endometrium. The strength of this study is its multicenter prospective character, which enabled us to recruit a large number of women. One of the limitations of this study is missing data. Multiple imputation was used to handle these missing data and obtain more precise and valid measures of association for variables with missing values. 16 The percentage of technical failure is in accordance with previous studies on outpatient endometrial sampling in postmenopausal women (12-23%) The rate of samples with insufficient material for histological diagnosis varies widely in the literature, from 16% to 68% The wide variance of insufficient sampling rates may result from the study design. Single center studies 12, 13, 17 10, 11 showed higher insufficient rates than multicenter studies. Our multicenter design led to results that are more representative for daily practice. Based on a small population, Gordon et al. suggested that the chance of insufficient sampling is related to operator experience. 13 In the current study no significant association was observed between training level and technical failure or insufficient sampling. A malignancy of the endometrium was diagnosed in 16 (21.6%) out of 74 women with a technical failed endometrial sampling and in 6 (7.1%) out of 84 women with an insufficient sampling with subsequent diagnostic evaluation. In a study performed by Van Doorn et al. an endometrial carcinoma (n=3) or atypical hyperplasia (n=1) was diagnosed in 6% (4/66) of the women with insufficient sampling. 11 Other investigators quoted uterine malignancies in % of the women with insufficient sampling, but this was in small series Despite this knowledge and the recommendations in the national guideline 3 no further effort was made to obtain material for a final diagnosis in 7% of women with a technical failure and 45% of women with an insufficient sample. In conclusion, nulliparity, advanced age and endometrial thickness are variables independently associated with failure of outpatient endometrial sampling in women with postmenopausal bleeding. A multivariable prediction model for total failure based on specific patient characteristics has a moderate capacity to discriminate between women with a high or a low risk of failure. As endometrial sampling resulted in a histological diagnosis in 56%, we think it is still advised in the primary work-up for women with postmenopausal bleeding. A failed sampling is an indication for further histological examination since serious 74

77 Factors attributing to the failure of endometrial sampling in women with postmenopausal bleeding underlying pathology may be present. The high failure rate of endometrial sampling (44%) may have consequences for the cost-effectiveness of this technique. Further research is required to externally validate the prediction model and to investigate if cost-effectiveness of the diagnostic work-up could be improved by applying the prediction model and selecting women with a high probability of a failed sampling to be referred for immediate diagnostic outpatient hysteroscopy without an attempted outpatient endometrial sampling. 5 75

78 Chapter 5 References 1. Moodley M, Roberts C. Clinical pathway for the evaluation of postmenopausal bleeding with an emphasis on endometrial cancer detection. J Obstet Gynaecol 2004;24(7): Epstein E. Management of postmenopausal bleeding in Sweden: a need for increased use of hydrosonography and hysteroscopy. Acta Obstet Gynecol Scand 2004;83(1): NVOG (Dutch Society of Obstetrics and Gynaecology). NVOG-richtlijn Abnormaal vaginaal bloedverlies in de menopauze [In Dutch]. NVOG Guideline: Abnormal vaginal bleeding during menopause Gupta JK, Chien PF, Voit D, Clark TJ, Khan KS. Ultrasonographic endometrial thickness for diagnosing endometrial pathology in women with postmenopausal bleeding: a meta-analysis. Acta Obstet Gynecol Scand 2002;81(9): Smith-Bindman R, Kerlikowske K, Feldstein VA et al. Endovaginal ultrasound to exclude endometrial cancer and other endometrial abnormalities. JAMA 1998;280(17): Van Hanegem N, Breijer MC, Khan KS, Clark TJ, Burger MP, Mol BWJ, Timmermans A. Diagnostic evaluation of the endometrium in postmenopausal bleeding: an evidence-based approach.maturitas. 2011;68(2): Dijkhuizen FP, Brolmann HA, Potters AE, Bongers MY, Heinz AP. The accuracy of transvaginal ultrasonography in the diagnosis of endometrial abnormalities. Obstet Gynecol 1996;87(3): Clark TJ, Barton PM, Coomarasamy A, Gupta JK, Khan KS. Investigating postmenopausal bleeding for endometrial cancer: cost-effectiveness of initial diagnostic strategies. BJOG 2006;113(5): Feldman S, Berkowitz RS, Tosteson AN. Cost-effectiveness of strategies to evaluate postmenopausal bleeding. Obstet Gynecol 1993;81(6): Clark TJ, Mann CH, Shah N, Khan KS, Song F, Gupta JK. Accuracy of outpatient endometrial biopsy in the diagnosis of endometrial cancer: a systematic quantitative review. BJOG 2002;109(3): van Doorn HC, Opmeer BC, Burger CW, Duk MJ, Kooi GS, Mol BW. Inadequate office endometrial sample requires further evaluation in women with postmenopausal bleeding and abnormal ultrasound results. Int J Gynaecol Obstet 2007;99(2): Williams AR, Brechin S, Porter AJ, Warner P, Critchley HO. Factors affecting adequacy of Pipelle and Tao Brush endometrial sampling. BJOG 2008;115(8): Gordon SJ, Westgate J. The incidence and management of failed Pipelle sampling in a general outpatient clinic. Aust N Z J Obstet Gynaecol 1999;39(1): Bakour SH, Khan KS, Gupta JK. Controlled analysis of factors associated with insufficient sample on outpatient endometrial biopsy. BJOG 2000;107(10): Steyerberg E. Clinical prediction models. A practical approach to development, validation, and updating Donders AR, van der Heijden GJ, Stijnen T, Moons KG. Review: a gentle introduction to imputation of missing values. J Clin Epidemiol 2006;59(10): Batool T, Reginald PW, Hughes JH. Outpatient pipelle endometrial biopsy in the investigation of postmenopausal bleeding. J Obstet Gynaecol 1997;17(4): De Silva BY, Stewart K, Steven JD, Sathanandan M. Transvaginal ultrasound measurement of endometrial thickness and endometrial pipelle sampling as an alternative diagnostic procedure to hysteroscopy and dilatation and curettage in the management of post-menopausal bleeding. J Obstet Gynaecol 1997;17(4): Farrell T, Jones N, Owen P, Baird A. The significance of an insufficient Pipelle sample in the investigation of post-menopausal bleeding. Acta Obstet Gynecol Scand 1999;78(9):

79 6 Comparing two diagnostic strategies based on the predicted probability of a failed endometrial biopsy in women with postmenopausal bleeding: a costminimization analysis M.C. Breijer N. van Hanegem N.C.M. Visser R.H.M. Verheijen B.W.J. Mol J.M.A. Pijnenborg B.C. Opmeer A. Timmermans Submitted

80 Chapter 6 Abstract Objective: To evaluate whether a model to predict a failed endometrial biopsy in women with postmenopausal bleeding and an endometrial thickness > 4.0 mm can reduce costs for the same accuracy through a cost-minimization analysis. Study design: A decision analytic model was designed to compare two diagnostic strategies for women with postmenopausal bleeding: (I) Attempt office endometrial biopsy and perform outpatient hysteroscopy after failed biopsy and (II) Endometrial biopsy or direct referral to outpatient hysteroscopy based on the predicted probability of a failed endometrial biopsy. Robustness of assumptions regarding costs was evaluated in sensitivity analyses. Results: At different cut-offs for the predicted probability of failure of an endometrial biopsy, strategy (I) was generally less expensive than strategy (II). The costs for strategy (I) were always 460; the costs for strategy (II) varied between 457 and 475. At a 65% cut-off, a possible saving of 3 per woman could be achieved. Conclusions: Individualizing the decision to perform an endometrial biopsy or hysteroscopy in women presenting with postmenopausal bleeding based on patient characteristics is unlikely to increase the efficiency of the diagnostic work up. 78

81 Comparing two diagnostic strategies based on the predicted probability of a failed endometrial biopsy InTRODuCTIOn Postmenopausal bleeding (PMB) is the most common presenting symptom of endometrial cancer and warrants further investigation. 1 Since the 1990s endometrial thickness measured by transvaginal ultrasound was introduced to select women for further invasive diagnostic testing to detect or rule out endometrial cancer. 2-4 Although the optimal endometrial thickness cut-off for women with PMB still remains questionable, at present most guidelines advise an endometrial thickness cut-off of 4 mm or 5 mm to select patients for further histological verification. 1, 5-9 Outpatient endometrial biopsy is the least invasive technique to obtain material for histological assessment. Pipelle endometrial biopsy (Pipelle de Cornier, Paris, France) is the most accurate endometrial sampling device to detect endometrial carcinoma and endometrial hyperplasia in patients with PMB. 10 Furthermore, a strategy with endometrial biopsy after endometrial thickness measurement is the most cost-effective diagnostic strategy for patients with PMB. 11 Although endometrial biopsy is the most accurate and frequently used diagnostic procedure it has some major drawbacks in clinical practice. In 12-21% of cases, endometrial sampling fails due to technical reasons and in 7-68% of cases the amount of tissue obtained is insufficient for a reliable histological diagnosis Because an endometrial (pre-) malignancy is present in 6-23% of the women with a failed endometrial biopsy, these patients cannot be 13, 14 reassured without further more invasive investigations. In a previous publication we described a multivariable prediction model to predict the probability of a failed endometrial biopsy in women with PMB. The purpose of the current study was to evaluate whether this model has the potential to reduce costs for the same accuracy of diagnostic testing in women with PMB through a cost-minimization analysis. 6 METHODS We performed a cost-minimization analysis using a decision analytic approach. The aim was to evaluate whether individualizing the decision to perform a diagnostic hysteroscopy rather than an endometrial biopsy in women with a first episode of PMB can decrease the costs of the diagnostic work-up. This decision can be based on the probability of a failed biopsy, estimated with a clinical prediction model based on patient characteristics. In a previous publication, we described the development of a model to predict the risk of a failed endometrial biopsy in women presenting with PMB and an endometrial thickness of more than 4.0 mm. Details on model development are presented in the original paper. 14 In short, data on 356 women with PMB were included in a multivariable regression analysis. Characteristics satisfying the criteria for inclusion in the model were time since menopause, hypertension, endometrial thickness (categorized) and nulliparity. Endometrial biopsy failed in 44.4% (95% CI: %) of the women (158/356). The discriminative capacity 79

82 Chapter 6 of the model was assessed with the area under the receiver operator characteristic curve and was 0.66 (95% CI ). The calibration of the model was good e.g. there was high agreement between the predicted probabilities and the observed proportion of failed endometrial biopsies. Cost minimization analysis The cost minimization analysis compared two diagnostic strategies: (I) Attempt office endometrial biopsy in all women and perform outpatient hysteroscopy after failed biopsy and (II) Decision for endometrial biopsy or direct referral to outpatient hysteroscopy based on the model based probability of a failed biopsy. The diagnostic strategies are represented in Figure 1. The sensitivity and specificity associated with different failure rate cut-offs were calculated in the prospective cohort. As the individualized strategy may become cost saving at a certain cut-off, an analysis was performed to identify this cut-off. Patient with postmenopausal bleeding Strategy I Visit to outpatient clinic Strategy II Predict probability of a failed endometrial biopsy Endometrial biopsy Probability < cut-off Endometrial biopsy succeeded Endometrial biopsy failed Probability cut-off Outpatient hysteroscopy with biopsy Outpatient hysteroscopy with biopsy Histological diagnosis Histological diagnosis Histological diagnosis Figure 1. Flowchart representing the two diagnostic strategies. 80

83 Comparing two diagnostic strategies based on the predicted probability of a failed endometrial biopsy Assumptions Outpatient hysteroscopy with biopsy was assumed to be the gold standard and had a 100% success rate. Furthermore we assumed that in all women diagnostic investigations are continued until adequate material for a histological diagnosis is obtained thus, all women in whom endometrial biopsy failed are subsequently referred to outpatient hysteroscopy with biopsy. With both strategies material for diagnosis will be obtained in all patients. Costs The analysis was conducted from a health care provider s perspective and included direct medical costs in euro s (Table 1). Costs were estimated using Dutch reference unit costs 17 and local cost calculations (Academic Medical Center Amsterdam). Costs for outpatient hysteroscopy were estimated costs per procedure and included costs for maintenance, disinfection and sterilization. Costs for an outpatient clinic visit included the specialist fee, costs for the assisting personnel and overhead costs. Table 1. Model input: direct medical costs. variable Cost ( ) Range ( ) Outpatient appointment (first visit) Outpatient appointment (follow up visit) Outpatient appointment (hysteroscopy clinic visit) Hysteroscopy including endometrial biopsy and pathologists fee, maintenance and sterilization Endometrial biopsy during outpatient visit including pathologists fee Statistical analyses Analyses were performed using TreeAgePro 2008 (TreeAge Software Inc, Williamstown, Mass). In addition to the base-case analysis, evaluating the results for the most likely estimates of model parameters, one-way sensitivity analyses were carried out to explore the robustness of the results for uncertainty in these parameters, including model accuracy, incidence of failure, and cost estimates. RESulTS In this economic analysis, expected costs for the two diagnostic strategies were compared. As endometrial biopsy failed in 44% of the patients in the cohort study, this percentage was used for strategy I. In the prospective cohort, the predicted probability of failure varied between 22 and 73%. The sensitivity and specificity of a 50, 55, 60, 65 and 70% predicted failure cut-off is reported in Table 2. In the base-case analysis, expected costs for the two 81

84 Chapter 6 strategies were comparable. Only at a 65% predicted failure cut-off, strategy II appeared to be less costly than strategy I. At this threshold, strategy II would result in a saving of 3 per patient presenting with PMB (Table 2). Table 2. Accuracy, uncertainty and impact of cut-off values for predicted probability of a failed endometrial biopsy in women with postmenopausal bleeding and endometrial thickness > 4.0 mm Cut-off failure probability (%) Sensitivity Specificity Costs Strategy I ( ) Costs Strategy II ( ) ( ) 0.86 ( ) ( ) 0.92 ( ) ( ) 0.96 ( ) ( ) 0.99 ( ) ( ) 0.99 ( ) Expected difference (strategy II strategy I) ( ) All sensitivity analyses performed for cost variables to explore robustness of the results for uncertainties in cost-estimates, showed results consistent with the base case analysis. For a cut-off value of 65%, the minimal expected cost-difference between strategy I and strategy II was 1 and the maximal expected cost-difference was 5 (Table 3). A sensitivity analysis varying the percentage of failure showed that strategy I is always cost-effective if the failure rate of endometrial biopsies is 25% or less. Table 3. Sensitivity analysis: minimum and maximum expected costs and cost-difference for 65% cut off value. Variable Strategy Strategy II Difference Minimum ( ) Maximum ( ) Minimum ( ) Maximum ( ) Minimum ( ) Outpatient appointment (first visit) Outpatient appointment (follow up visit) Outpatient appointment (hysteroscopy clinic visit) Hysteroscopy Endometrial biopsy Maximum ( ) 82

85 Comparing two diagnostic strategies based on the predicted probability of a failed endometrial biopsy DISCuSSIOn In this study we performed a cost-minimization analysis to investigate the potential for cost savings of selecting women with a first episode of PMB for endometrial biopsy or hysteroscopy based on the predicted probability of a failed biopsy. At a cut-off of 65% predicted failure rate, there is possibly a very small saving per patient presenting with PMB. The strength of this analysis is the new approach to the problem of a failed endometrial biopsy in women with PMB. Increasingly, clinical prediction models are developed for clinical practice. Before a model can be implemented in clinical practice, the external validity and impact on cost-effectiveness has to be evaluated. A limitation of our study is the model based approach. As for all model based studies, the validity of the model depends on the model input parameters and assumptions. We assumed that a hysteroscopy, as golden standard diagnostic procedure, would always succeed and lead to a diagnosis. We also assumed that an outpatient endometrial biopsy if successful, would lead to a reliable diagnosis. Another possible limitation is the costs used in our analyses. Because precise economic data are not available, we used the best available data that could be acquired from national and local sources. 17 The work-up in the model was conform recommendations in the national guideline on the work-up for women with PMB. 7 Clark and colleagues investigated the cost-effectiveness of initial diagnostic strategies for women with PMB. Their conclusion was that depending on the prevalence of endometrial malignancy, an endometrial biopsy after endometrial thickness measurement with a cut-off 4 or 5 mm was the most cost-effective strategy. 11 In this analysis however, the probability of a failed endometrial biopsy was 12% and in contrast to our current study, this probability was independent of patient characteristics. Because the implementation of a prediction model in clinical practice is time consuming, the model should be easy applicable and the benefits in terms of cost-effectiveness should be substantial. In view of the results of our study, the potential value of using the prediction model for a failed endometrial biopsy in women presenting with PMB appears to be limited and not contributing to a more efficient use of health care resources. Based on the current study, we recommend outpatient endometrial biopsy for all women presenting with a first episode of PMB and an endometrial thickness more than 4.0 mm. Patients in whom endometrial biopsy fails have 6-23% risk of an endometrial (pre-)malignancy and cannot be reassured without further testing. 13, 14 In case of a failed endometrial biopsy, a hysteroscopy with guided biopsy should be performed. Future research should, to increase the efficiency of the diagnostic work-up for women with PMB, focus on decreasing the amount of failed endometrial biopsies. 6 83

86 Chapter 6 References 1. Goldstein RB, Bree RL, Benson CB et al. Evaluation of the woman with postmenopausal bleeding: Society of Radiologists in Ultrasound-Sponsored Consensus Conference statement. J Ultrasound Med 2001;20(10): Gupta JK, Wilson S, Desai P, Hau C. How should we investigate women with postmenopausal bleeding? Acta Obstet Gynecol Scand 1996;75(5): Smith-Bindman R, Kerlikowske K, Feldstein VA et al. Endovaginal ultrasound to exclude endometrial cancer and other endometrial abnormalities. JAMA 1998;280(17): Tabor A, Watt HC, Wald NJ. Endometrial thickness as a test for endometrial cancer in women with postmenopausal vaginal bleeding. Obstet Gynecol 2002;99(4): American College of Obstetricians and Gynecologists. ACOG Committee Opinion No. 426: The role of transvaginal ultrasonography in the evaluation of postmenopausal bleeding. Obstet Gynecol 2009;113(2 Pt 1): Epstein E. Management of postmenopausal bleeding in Sweden: a need for increased use of hydrosonography and hysteroscopy. Acta Obstet Gynecol Scand 2004;83(1): NVOG (Dutch Society of Obstetrics and Gynaecology). NVOG-richtlijn Abnormaal vaginaal bloedverlies in de menopauze [In Dutch]. NVOG Guideline: Abnormal vaginal bleeding during menopause Scottish Intercollegiate Guidelines Network. Investigation of postmenopausal bleeding. Scottish Intercollegiate Guidelines Network, Royal College of Physicians Timmermans A, Opmeer BC, Khan KS et al. Endometrial thickness measurement for detecting endometrial cancer in women with postmenopausal bleeding: a systematic review and meta-analysis. Obstet Gynecol 2010;116(1): Dijkhuizen FP, Mol BW, Brolmann HA, Heintz AP. The accuracy of endometrial sampling in the diagnosis of patients with endometrial carcinoma and hyperplasia: a meta-analysis. Cancer 2000;89(8): Clark TJ, Barton PM, Coomarasamy A, Gupta JK, Khan KS. Investigating postmenopausal bleeding for endometrial cancer: cost-effectiveness of initial diagnostic strategies. BJOG 2006;113(5): Clark TJ, Mann CH, Shah N, Khan KS, Song F, Gupta JK. Accuracy of outpatient endometrial biopsy in the diagnosis of endometrial cancer: a systematic quantitative review. BJOG 2002;109(3): van Doorn HC, Opmeer BC, Burger CW, Duk MJ, Kooi GS, Mol BW. Inadequate office endometrial sample requires further evaluation in women with postmenopausal bleeding and abnormal ultrasound results. Int J Gynaecol Obstet 2007;99(2): Visser NCM, Breijer MC, Herman MC et al. Factors attributing to the failure of endometrial sampling in women with postmenopausal bleeding: a prospective multicenter cohort study Gordon SJ, Westgate J. The incidence and management of failed Pipelle sampling in a general outpatient clinic. Aust N Z J Obstet Gynaecol 1999;39(1): Williams AR, Brechin S, Porter AJ, Warner P, Critchley HO. Factors affecting adequacy of Pipelle and Tao Brush endometrial sampling. BJOG 2008;115(8): Dutch Healtcare Authority (NZa)

87 7 A structured assessment to decrease the amount of inconclusive endometrial biopsies in women with postmenopausal bleeding M.C. Breijer N.C.M. Visser N. van Hanegem A.A. van der Wurff B.C. Opmeer H.C. van Doorn B.W.J. Mol J.M.A. Pijnenborg A. Timmermans Submitted

88 Chapter 7 Abstract Objective: To determine whether structured assessment of outpatient endometrial biopsies decreases the number of inconclusive samples. Design: Retrospective cohort study. Setting: Single hospital pathology laboratory. Population: Endometrial biopsy samples of 66 women with postmenopausal bleeding, collected during the normal diagnostic work-up and assessed as insufficient for reliable histological diagnosis. Methods: Endometrial biopsy samples were requested from the pathology laboratories. The retrieved samples were systematically reassessed by a single pathologist specialized in gynecology. Main outcome measure: Disagreement between initial assessment and conclusion after structured reassessment. Results: We retrieved 36 of 66 endometrial biopsy samples from six different pathology laboratories. Structured reassessment of the retreived samples by one pathologist specialized in gynecology did not change the conclusion in 35 of the 36 samples. This sample contained a large amount of endometrial tissue and the diagnosis at reassessment was endometrial hyperplasia without atypia. All other samples contained insufficient material for a reliable diagnosis. Conclusion: Although there are no guidelines on standardized pathological assessment of endometrial biopsy samples, a structured reassessment of endometrial biopsies samples that were classified as inconclusive due to insufficient material did not change the conclusion except for one. Thus, we can conclude that although it might be helpful for pathologists to have diagnostic criteria for adequacy and/or inadequacy of an endometrial biopsy sample, the gain in efficiency is likely to be small. 86

89 A structured assessment to decrease the amount of inconclusive endometrial biopsies InTRODuCTIOn Patients with postmenopausal bleeding (PMB) are at risk for endometrial carcinoma and therefore PMB warrants further investigation. Histological endometrial assessment is indicated when a patient presents with PMB and an increased endometrial thickness on transvaginal sonography (TVS). 1-3 Outpatient endometrial biopsy is the least invasive technique to obtain material for histological assessment. Endometrial biopsies have a very high sensitivity in postmenopausal women (95%). 4 Furthermore, performing an endometrial biopsy in women with PMB and an increased endometrial thickness is the most cost-effective strategy. 5 Yet, 7-68% of outpatient endometrial biopsy samples are inconclusive because the amount of tissue obtained is insufficient for a reliable histopathological diagnosis. 6-9 In such cases a more invasive hysteroscopy or dilatation and curettage (D&C) is necessary in order to rule out endometrial carcinoma or atypical hyperplasia which is present in 6% of these women. 8 The high failure rate due to inconclusive endometrial biopsies might affect the cost-effectiveness of the diagnostic work-up. Among pathologists considerable disagreement exists about what constitutes an adequate endometrial biopsy sample. A questionnaire sent to all members of the British Association of Gynecological Pathologists and participants in the National Gynecological Pathology External Quality Assessment Scheme in the UK, revealed that 88.5% of the respondents would think criteria for adequacy and/or inadequacy would be useful. 10 At present, there are no such guidelines on standardized pathological assessment of endometrial biopsy samples. This results in a diagnosis that is influenced by subjectivity, with a high inter-pathologist variability. The objective of the current study is to determine whether structured assessment of outpatient endometrial biopsy samples with strict criteria decreases the amount of inconclusive samples due to insufficient material. 7 MATERIAl AnD METHODS Patient material was used from a previously published study on women with PMB. 8 Data on all women presenting with PMB in seven teaching hospitals and one university hospital in the Netherlands between January 2001 and June 2003 were prospectively collected to evaluate the diagnostic work-up in women with PMB in the Netherlands. 8 In 403 women with an endometrial thickness more than 4 mm or not measurable, an outpatient endometrial biopsy was performed. In this study there was an inconclusive endometrial biopsy sample due to insufficient tissue in 66 of the 403 women. All 66 women were subsequently evaluated with hysteroscopy or D&C and in 4 women (6%) an endometrial (pre-) malignancy was diagnosed. The 66 cases with an inconclusive endometrial biopsy in the prospective cohort were used for the current study (Figure 1). 87

90 Chapter 7 Women with postmenopausal bleeding and endometrial thickness > 4 mm or not measurable (N = 516) Endometrial biopsy not performed (N = 113) Endometrial biopsy failed (N = 93) Insufficient material for reliable diagnosis (N = 66) Diagnostic sample (N = 244) Final diagnosis after further testing: Benign histology (N = 50) Endometrial (pre-) malignancy (N = 4) Other malignancy (N = 2, bladder, breast) No diagnosis (N = 10) Figure 1. Flow-chart of women with postmenopausal bleeding included in the original prospective cohort study. Table 1. Items for assessment of the adequacy of an endometrial biopsy sample Items Estimated amount of material: Quality of material No material < 0.5 cm cm cm 2 > 2.0 cm 2 No endometrial tissue Only superficial endometrial tissue A large amount of endometrial tissue The endometrial biopsy samples were requested from the pathology laboratories of the eight hospitals. Based on two reports on the assessment of endometrial biopsy samples, we proposed items for a structured assessment (Table 1). 10, 11 The material was systematically scored on the items estimated amount of material and quality of material. The endometrial biopsy samples were systematically reassessed by one pathologist specialized in gynecology (AW). The pathologist knew that it was a reassessment of previously non-diagnostic endometrial biopsy samples but was blinded for the final diagnosis in the original cohort study. The study was reviewed and approved by the institutional review board. 88

91 A structured assessment to decrease the amount of inconclusive endometrial biopsies Statistical analysis Statistical analysis was performed with the PASW statistics 18.0 package (SPSS, Inc., Somers, NY, USA). Conditional on the distribution, continuous variables were presented as mean +/- standard deviation or as median and interquartile range, and differences between groups tested with an independent samples T-test or a Mann-Whitney U test. Categorical data were presented as N (%) and groups compared using a Chi-square test. RESulTS We were able to retrieve endometrial biopsy samples from 36 (55%) women from six different pathology laboratories. Pathology laboratories from two hospitals did not respond to our request to sent material. Pathology laboratories that did not respond to the initial request were contacted by phone once. Afterwards, no further efforts were made to obtain the samples. In endometrial biopsy samples from 29 (80.5%) of the 36 women, no endometrial tissue was found. The pathologist described just slime, blood or superficial columnar epithelium. In endometrial biopsy samples from six (16.7%) women, superficial endometrium was found, with only one very small strip of endometrial stroma and therefore these samples were also inconclusive. In the endometrial biopsy sample from one (2.8%) woman there was a large amount of endometrium and the diagnosis endometrial hyperplasia without atypia was established on this material. The reassessment for all 36 women with the macroscopic description is reported in Table 2. There were no differences in the characteristics of women with a reassessed sample when compared to women without a reassessment (Table 3). Samples of all four women with an endometrial (pre-) malignancy at subsequent testing in the original cohort study were retrieved for reassessment. 7 Table 3. Characteristics of women with a non-diagnostic sample, retrieved samples versus not retrieved samples. Characteristic Material retrieved (N = 36) Material not retrieved (N = 30) P-value Age (years) a / / P = 0.39 Body mass index (kg/m2) a / / P = 0.53 Endometrial thickness (mm) b 7.5 ( ) 7.0 ( ) P = 0.42 Hormone therapy use n (%) c 7 (19.4%) 7 (23.3%) P = 0.89 Nulliparity n (%) c 2 (5.6%) 3 (10%) P = 0.69 Endometrial (pre-) malignancy n (%) c 4 (11.1%) 0 P = 0.03 a Independent samples T-test, mean +/- SD b Mann Whitney U test, median IQR c Chi-Square 89

92 Chapter 7 Table 2. Reassessment of endometrial samples. Patient Macroscopic Amount (cm2) Quality Diagnosis 1 5ml slime < 0.5 No endometrial tissue No diagnosis 2 Tissue 15 x 3 mm < 0.5 No endometrial tissue No diagnosis 3 0.5ml tissue < 0.5 No endometrial tissue No diagnosis 4 Just slime and blood < 0.5 No endometrial tissue No diagnosis 5 1ml tissue < 0.5 No endometrial tissue No diagnosis 6 Just slime and blood < 0.5 No endometrial tissue No diagnosis 7 Some tissue 1 to 4 mm < 0.5 Superficial endometrial tissue No diagnosis 8 Some white tissue < 0.5 No endometrial tissue No diagnosis 9 Red/brown slimy tissue < 0.5 No endometrial tissue No diagnosis 10 Very little slimy tissue < 0.5 No endometrial tissue No diagnosis 11 Very little grey/red material < 0.5 No endometrial tissue No diagnosis 12 Some grey/red tissue fragments maximum 3cm > 2.0 Large amount of endometrial tissue Hyperplasia without atypia 13 Little grey/brown material < 0.5 No endometrial tissue No diagnosis 14 Hardly any material < 0.5 No endometrial tissue No diagnosis 15 Very little grey/brown material < 0.5 No endometrial tissue No diagnosis 16 1ml slime < 0.5 No endometrial tissue No diagnosis ml slime < 0.5 No endometrial tissue No diagnosis 18 < 0.5ml slime < 0.5 No endometrial tissue No diagnosis 19 Some tissue fragments < 0.5mm < 0.5 No endometrial tissue No diagnosis ml slimy bloody material < 0.5 No endometrial tissue No diagnosis 21 15ml slimy material < 0.5 No endometrial tissue No diagnosis 22 < 0.25ml slime < 0.5 No endometrial tissue No diagnosis 23 Some slimy fragments < 0.5 Superficial endometrial tissue No diagnosis ml slimy bloody material < 0.5 Superficial endometrial tissue No diagnosis ml material < 0.5 Superficial endometrial tissue No diagnosis 26 Some very small fragments < 0.5 No endometrial tissue No diagnosis ml slimy material < 0.5 No endometrial tissue No diagnosis 28 < 0.5ml slimy material < 0.5 No endometrial tissue No diagnosis 29 Three slimy fragments 1-2mm < 0.5 Superficial endometrial tissue No diagnosis ml slimy material < 0.5 No endometrial tissue No diagnosis 31 < 0.25ml slime < 0.5 No endometrial tissue No diagnosis ml slime < 0.5 No endometrial tissue No diagnosis 33 < 0.2ml material < 0.5 No endometrial tissue No diagnosis 34 Just slime < 0.5 No endometrial tissue No diagnosis ml material < 0.5 Superficial endometrial tissue No diagnosis 36 Slimy material < 0.5 No endometrial tissue No diagnosis 90

93 A structured assessment to decrease the amount of inconclusive endometrial biopsies DISCuSSIOn Although there are no guidelines on standardized pathological assessment of endometrial biopsy samples, a structured reassessment of endometrial biopsy samples that were classified as inconclusive due to insufficient material did not change the conclusion except for one. To our knowledge, there are no previous reports on the attempt to increase the diagnostic efficiency of outpatient endometrial biopsy by structured assessment. Because in women with an endometrial thickness > 4 mm or not measurable and an inconclusive endometrial biopsy, an endometrial (pre-) malignancy is found in 6%, these women cannot be reassured without further more invasive diagnostics. 8 The strength of our study is that all endometrial biopsy samples were assessed by one pathologist specialized in gynecology. Furthermore, the pathologist was blinded for the definite diagnosis. A limitation of our study is that we retrieved 55% of the requested endometrial biopsy samples from the pathology laboratories. Despite repeated requests, two hospitals did not respond. One of those hospitals had 21 women with an inconclusive sample in the original cohort study. Although we had only 55% of the requested endometrial biopsy samples available, it is unlikely that a complete response would have changed our conclusions. First, only one of the 36 available endometrial biopsy samples was considered diagnostic at reassessment within this selection and second, endometrial biopsy samples of all four patients that were subsequently diagnosed with endometrial (pre-) malignancies were present. Characteristics of the women we did not receive material from were comparable to the characteristics of women with reassessed samples. Another limitation is the fact that the pathologist performing the reassessment knew that the endometrial biopsy samples were initially assessed as inconclusive, which potentially leads to a biased interpretation at reassessment. Performing a prospective cohort study where all endometrial biopsy samples are assessed with structured criteria during a certain period could solve this. The amount of insufficient samples could be compared to the amount of insufficient samples in a previous period without structured pathology assessment. In 2005 Phillips et al reported on a questionnaire among 61 pathologists in the UK. Most respondents felt that it would be useful if criteria for adequacy and/or inadequacy were proposed. 10 Similar criteria are already used for the evaluation of specimen adequacy in fine needle aspiration of thyroid nodules. 12 In the cost-effectiveness analysis by Clark et al, the failure rate due to inconclusive endometrial biopsy samples was 12% (95% CI ) based on a systematic review. 5 The high failure rate of the endometrial biopsies might have consequences for the cost-effectiveness of the technique. Reviewing hospital protocols revealed that in none of the hospitals there was a protocol on standardized sampling methods available, let alone recommendations on using a tenaculum or entering the uterine cavity more than once, or the use of analgesia in painful procedures. 7 91

94 Chapter 7 Our findings suggest that although it might be helpful for pathologists to have diagnostic criteria for adequacy and/or inadequacy of an endometrial biopsy sample, the gain in efficiency is likely to be small. We therefore think that to increase the effectiveness of outpatient endometrial biopsies, effort has to be made to obtain as much material as possible to minimize the failure of endometrial biopsies due to insufficient material. Further research should focus on the best way to achieve this. 92

95 A structured assessment to decrease the amount of inconclusive endometrial biopsies REFEREnCES 1. American College of Obstetricians and Gynecologists. ACOG Committee Opinion No. 426: The role of transvaginal ultrasonography in the evaluation of postmenopausal bleeding. Obstet Gynecol 2009;113(2 Pt 1): NVOG (Dutch Society of Obstetrics and Gynaecology). NVOG-richtlijn Abnormaal vaginaal bloedverlies in de menopauze [In Dutch]. NVOG Guideline: Abnormal vaginal bleeding during menopause Scottish Intercollegiate Guidelines Network. Investigation of postmenopausal bleeding. Scottish Intercollegiate Guidelines Network, Royal College of Physicians Dijkhuizen FP, Brolmann HA, Potters AE, Bongers MY, Heinz AP. The accuracy of transvaginal ultrasonography in the diagnosis of endometrial abnormalities. Obstet Gynecol 1996;87(3): Clark TJ, Barton PM, Coomarasamy A, Gupta JK, Khan KS. Investigating postmenopausal bleeding for endometrial cancer: cost-effectiveness of initial diagnostic strategies. BJOG 2006;113(5): Clark TJ, Mann CH, Shah N, Khan KS, Song F, Gupta JK. Accuracy of outpatient endometrial biopsy in the diagnosis of endometrial cancer: a systematic quantitative review. BJOG 2002;109(3): Gordon SJ, Westgate J. The incidence and management of failed Pipelle sampling in a general outpatient clinic. Aust N Z J Obstet Gynaecol 1999;39(1): van Doorn HC, Opmeer BC, Burger CW, Duk MJ, Kooi GS, Mol BW. Inadequate office endometrial sample requires further evaluation in women with postmenopausal bleeding and abnormal ultrasound results. Int J Gynaecol Obstet 2007;99(2): Williams AR, Brechin S, Porter AJ, Warner P, Critchley HO. Factors affecting adequacy of Pipelle and Tao Brush endometrial sampling. BJOG 2008;115(8): Phillips V, McCluggage WG. Results of a questionnaire regarding criteria for adequacy of endometrial biopsies. J Clin Pathol 2005;58(4): McCluggage WG. My approach to the interpretation of endometrial biopsies and curettings. J Clin Pathol 2006;59(8): Guidelines of the Papanicolaou Society of Cytopathology for the examination of fine-needle aspiration specimens from thyroid nodules. The Papanicolaou Society of Cytopathology Task Force on Standards of Practice. Diagn Cytopathol 1996;15(1):

96

97 8 The capacity of endometrial thickness measurement to diagnose endometrial carcinoma in asymptomatic postmenopausal women: a systematic review and meta analysis M.C. Breijer J.A.H. Peeters B.C. Opmeer T.J. Justin Clark R.H.M. Verheijen B.W.J. Mol A. Timmermans Ultrasound Obstet Gynecol ; 40:

98 Chapter 8 Abstract Objectives: Measurement of endometrial thickness is an important tool in the assessment of women with postmenopausal bleeding. The place of endometrial thickness measurement by ultrasound in asymptomatic women is unclear. The aims of this study were to address: (I) the normal endometrial thickness measured with ultrasonography, (II) the prevalence of serious endometrial pathology and (III) the sensitivity and specificity of endometrial thickness measurement by transvaginal ultrasonography (TVS) for diagnosing premalignant and malignant endometrial disease in asymptomatic postmenopausal women. Methods: A MEDLINE and EMBASE search (from inception to January 2011) was performed. Articles reporting on endometrial thickness measurement in the diagnosis of endometrial carcinoma and atypical hyperplasia in asymptomatic postmenopausal women without HRT were selected. Endometrial thickness and the prevalence of endometrial (pre-)malignancies were recorded. If possible, 2x2 tables were extracted. Results: We included 32 studies reporting on 11,100 women to answer our three objectives: (I) The estimated mean endometrial thickness was 2.9 mm (95% CI ). (II) The pooled estimated prevalence of endometrial carcinoma and atypical endometrial hyperplasia were 0.62% (95% CI ), and 0.59% (95% CI ), respectively. (III) Summary estimates for sensitivity and specificity of TVS endometrial thickness measurement were 0.83 (95% CI ) and 0.72 (95% CI ) for 5 mm cut-off and 0.33 (95% CI ) and 0.94 (95% CI ) for 6 mm cut-off. Conclusions: The results from this review do not justify the use of endometrial thickness as a screening test for endometrial carcinoma and atypical hyperplasia in asymptomatic postmenopausal women not using HRT. 96

99 Endometrial thickness measurement in asymptomatic postmenopausal women InTRODuCTIOn Endometrial carcinoma is the most common malignancy of the female genital tract in the developed countries and presents itself with postmenopausal bleeding in more than 95% of cases. 1, 2 In patients with postmenopausal bleeding, sonographic measurement of endometrial thickness (ET) is, the first test to determine whether further investigations are needed to rule out malignancy. 3 Guidelines recommend a cut-off value of 4 or 5 mm by transvaginal ultrasonography (TVS), below which endometrial cancer is unlikely. 4-7 When the endometrial thickness is below this cut-off, the probability of endometrial carcinoma is below 1%. 3 In contrast with the clear guidelines on the management of women with postmenopausal bleeding, clinicians are faced with uncertainty when endometrial thickness is measured in asymptomatic postmenopausal women. Symptom free women may undergo TVS for other indications, such as vaginal prolapse or abdominal complaints. Inevitably, the endometrium is then visualized and incidentally a thickened endometrium may be observed. It is unknown how to manage such an incidentally observed thick endometrium. Based on a decision analysis in a theoretical cohort, Smith-Bindman et al concluded that in asymptomatic postmenopausal women with an ET of 11 mm or thicker an endometrial biopsy should be performed. 8 They argued that women with an endometrial thickness above this threshold have a malignancy risk of 6.7% which is comparable to the risk in patients with postmenopausal bleeding and an ET larger than 5 mm (7.3%), the latter being the widely accepted threshold to perform a biopsy in symptomatic patients. Apart from this debate on the accuracy of endometrial thickness measurement in asymptomatic postmenopausal women, its potential value also depends on the prevalence of the disease searched for, i.e. endometrial thickness and or its precursors. Since in asymptomatic women the prevalence of endometrial carcinoma is lower than in symptomatic women, the cut-off of endometrial thickness for abnormality in these women should be higher. To address the above mentioned dilemmas, we reviewed the literature on asymptomatic postmenopausal women not using hormone replacement therapy (HRT). The aims of this review were: to address in asymptomatic postmenopausal women: (1) normal endometrial thickness measured with ultrasonography, (2) the prevalence of serious endometrial pathology and (3) the sensitivity and specificity of endometrial thickness measurement by TVS for diagnosing premalignant and malignant endometrial disease. 8 METHODS Identification of studies We performed an electronic search in January 2011 in MEDLINE (from 1948) and EMBASE (from 1980) to identify articles reporting on endometrial thickness and/or endometrial 97

100 Chapter 8 carcinoma and hyperplasia in asymptomatic postmenopausal women. We used the following keywords: postmenopausal, asymptomatic, screening, endometrial, thickness, ultrasound, hyperplasia and carcinoma. The complete search syntax is reported in appendix 1. Language restrictions were not applied. Abstracts or articles written in languages other than English were read by a member of the team with sufficient knowledge of the language, if there was no team member available, the article was translated by a native speaker. References and cross references of selected studies were searched for articles not identified by the electronic searches. No review protocol was registered. Selection of studies and data abstraction Two independent reviewers (MB, JP) screened the electronic search results by reading titles and / or abstracts. Studies that were restricted to patients with postmenopausal bleeding or women using HRT or Tamoxifen were excluded. Titles and abstracts were assessed to identify eligible studies. Subsequently, these articles were evaluated in full text for each of the three objectives independently for the final study selection. Any disagreements were resolved by consensus. In case of persistent disagreement, the judgment of a third reviewer (AT) was decisive. If multiple publications reported analyses on the same dataset, only the largest study was included. If a dataset was split into different subgroups and the subgroups were reported separately in multiple publications, we combined the results of these publications. We used the quality assessment of diagnostic accuracy studies (QUADAS) 9 checklist to assess the methodological quality of included studies (Appendix 2). Endometrial thickness To evaluate normal endometrial thickness in asymptomatic postmenopausal, we searched for studies that (1) reported a mean endometrial thickness with a measure of variance and (2) described a standardized approach to measurement of endometrial thickness by TVS. Mean endometrial thickness as well as standard deviation, standard error or 95% confidence intervals were recorded for each included study. A pooled estimate of mean endometrial thickness was then calculated using inverse variance weighting in a random effects model. Prevalence of endometrial (pre-) malignancies To assess the prevalence of (pre-) malignant lesions of the endometrium in asymptomatic postmenopausal women, we included studies that reported on any form of endometrial verification on the total population of asymptomatic postmenopausal women without HRT. The endometrium could be assessed with histology (hysterectomy, dilatation and curettage (D&C), hysteroscopy with biopsy, endometrial biopsy) or with cytology. Studies that performed partial verification (histological verification only in a subgroup based on a 98

101 Endometrial thickness measurement in asymptomatic postmenopausal women previous test e.g. ET above a cut-off level or progesterone challenge test positive patients) were excluded. The method of verification, the number of women with atypical endometrial hyperplasia and the number of women with endometrial carcinoma was recorded for each selected study. The prevalence of endometrial carcinoma and atypical endometrial hyperplasia was then calculated for each study, and a weighted pooled estimate was derived. Diagnostic accuracy of endometrial thickness for endometrial (pre-) malignancy Finally, to estimate the diagnostic accuracy of endometrial thickness for (pre)malignancy of the endometrium, we selected studies that reported on both endometrial thickness measurement and endometrial histological verification in asymptomatic postmenopausal women. Information required to construct a two-by-two table for each reported endometrial thickness cut-off value was recorded for each selected study. Three different outcomes were considered: benign (including atrophy, endometrial polyps, and endometrial hyperplasia without atypia), atypical endometrial hyperplasia, and endometrial carcinoma. The data from the two-by-two tables were used to calculate sensitivity and specificity, as well as positive and negative predictive values for each study. If the study reported on multiple thresholds, we included two-by-two tables for all reported thresholds. Subsequently, summary point estimates for sensitivity and specificity were generated for each reported endometrial cut-off value using a bivariate random effects approach 10, for endometrial carcinoma, atypical endometrial hyperplasia, and these two diagnoses combined in one group. RESulTS Search strategy Our search resulted in 503 citations; another 31 studies were identified through reference search. There were 95 studies eligible for inclusion based on title and abstract. After assessment of the full text articles, 63 studies were discarded as there were no separate data for asymptomatic patients (N = 7), non HRT using patients or no clear information on HRT use (N = 29), postmenopausal patients (N = 3), being a review or editorial (N = 4), endometrial thickness reported in subgroups only (N = 4), no mean endometrial thickness or measure for variance reported (N = 12), or having multiple publications on the same dataset (N = 4) (Figure 1). As a result, a set of 32 relevant studies was available to answer our three questions: 10 studies could be used for question 1 (estimating normal endometrial thickness); 15 studies could be used for question 2 (prevalence of endometrial malignancy and pre-malignancy) and 20 studies could be used for question 3 (diagnostic accuracy of TVS for these endometrial diseases). 8 99

102 Chapter 8 Retrieved from searches: electronic search ( n = 503) cross-reference search ( n = 31) Excluded after reading: titles (n = 289) abstracts ( n = 150) Retrieved in full ( n = 95) Studies available to answer different questions ( n = 32) Excluded after reading article: No separate data for: Asymptomatic women ( n = 7) Women not using HRT ( n = 17) Postmenopausal women (n = 3) No information regarding HRT use (n = 12) Review / editorial (n = 4) Endometrial thickness reported for subgroups only ( n = 4) No mean endometrial thickness or measure for variance reported (n = 12) Multiple publications on the same dataset (n = 4) Studies included for endometrial thickness (n = 10) Studies included for histological verification (n = 15) Studies included for diagnostic accuracy (n = 20) Figure 1. Flow chart of studies included in the meta-analysis. HRT, hormone replacement therapy. Assessment of methodological quality Study quality was considered generally good when eight of the 14 QUADAS items were met by over 70% of the included studies (Figure 2). 100 % 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Patient spectrum Selection criteria Reference standard Time interval Partial verification Differential verification Incorporation Index text details Reference standard details Review bias (index) Review bias (reference) Clinical data Uninterpretable results Withdrawals Figure 2. Summary of evaluation of the 32 studies analyzed using the quality assessment of diagnostic accuracy studies (QUADAS) checklist. See appendix S2 for the 14 questions summarized as labels on the X-axis. 100

103 Endometrial thickness measurement in asymptomatic postmenopausal women Studies scored poor on the items regarding blinded interpretation of the reference test and description of withdrawals. In 5 of the 32 studies, almost the entire QUADAS checklist was scored as not applicable because e.g. no endometrial thickness measurement (index test) was reported or no endometrial verification (reference test) was reported. Absence of the index or reference test, automatically led to problems in completing the QUADAS checklist. For these five studies it was impossible to answer some questions of the QUADAS checklist with yes, no or unclear and these items were scored not applicable. Endometrial thickness We found ten studies that reported on endometrial thickness measurement with a measure of variance in asymptomatic women without HRT The 10 studies that were assessed for inclusion had been conducted in nine different countries. One study was published in German; the other nine were published in English. In total, these 10 studies reported on 3,049 women, with a median sample size of 207 (range 97 to 1182). Mean endometrial thickness in the 10 studies varied from 2.1 mm to 5.7 mm. The pooled estimate of the mean endometrial thickness was 3.2 mm (95% CI ). There was one outlier, with a mean endometrial thickness of 5.7 mm, compared to the other studies; Güven et al reported a thicker endometrium. 13 Mean BMI in this study was 29.5 kg/m 2. In contrast, mean BMI in the other included studies ranged between 22.4 and 25.6 kg/m 2. The purpose of the study by Güven et al was to correlate BMI to endometrial thickness, this potentially could have lead to inclusion bias. We therefore excluded this study from the meta-analysis for endometrial thickness. The remaining nine studies reported on 2952 women with a median sample size of 259 (Table 1). The pooled estimate of the mean endometrial thickness was 2.9 mm (95% CI ) (Figure 3). Statistical heterogeneity between studies (I 2 ) was 28%. Table 1. Characteristics of studies included in meta-analysis for mean endometrial thickness measured by transvaginal ultrasound in asymptomatic postmenopausal women without hormone replacement therapy. Endometrial thickness (mm) Reference Year Country n Mean SD Range Andolf Sweden Gull Sweden Kasraeian Iran Malinova Bulgaria NR Minagawa Japan Neele The Netherlands Osmers Germany Pirhonen Finland NR Warming Denmark NR standard error. NR, not reported 8 101

104 Chapter 8 First author Andolf 11 Gull 12 Kasraeian 14 Malinova 15 Minagawa 16 Neele 17 Osmers 18 Warming 20 Pirhonen 19 Summary Endometrial thickness (mm) Figure 3. Forest plot of the meta analysis for mean endometrial thickness. Prevalence of endometrial (pre-) malignancies In 28 studies, a histological diagnosis of the endometrium (endometrial verification) was obtained. We excluded 13 studies from further analysis, as they had only partial endometrial verification (verification in a subgroup of patients selected by a previous test e.g. endometrial thickness or progesterone challenge test). The 15 remaining studies were performed in nine different countries. 13, 14, Thirteen studies were published in English, one in Portuguese and one in Italian. Together, these studies described a total of 3,595 women. Characteristics of these studies are reported in Table 2. The median sample size was 145 (range 30 to 883). The prevalence of endometrial carcinoma varied between 0% and 2.1%, the prevalence of atypical endometrial hyperplasia varied between 0% and 3.5%, and the prevalence of combined (pre-) malignancy varied between 0% and 4.3%. The pooled estimated prevalence of endometrial carcinoma was 0.62% (95% CI ) of endometrial hyperplasia was 0.59% (95% CI ) and of combined (pre-) malignancies 1.21% (95% CI ). Diagnostic accuracy of endometrial thickness for endometrial (pre-) malignancy There were 20 studies that reported on the endometrial thickness cut-off value and histological or cytological endometrial verification , 23, 26, 29, 31, 32, These studies were performed in 13 different countries. Sixteen studies were published in English, one in Italian, one in Spanish, one in Portuguese, and one in German. 102

105 Endometrial thickness measurement in asymptomatic postmenopausal women Table 2. Characteristics of studies included in meta analysis for assessment of prevalence of endometrial (pre)malignancies in asymptomatic postmenopausal women not using hormone replacement therapy. Author Year Country n Verification method Endometrial carcinoma n Prevalence (%) n AEH Prevalence (%) Prevalence of (pre-) malignancy (%) Bortoletto Brazil 150 EBNS Buccoliero Italy 107 Hysterectomy, EBN, EES Cohen USA 60 EBP Elewa Egypt 30 Hyst + biopsy Gol Turkey 556 D&C Gouveia Brazil 47 EBP Güven Turkey 97 D&C Kasraeian Iran 259 Hyst + biopsy Langer USA 145 EBNS Macia Spain 130 EBS Marello Italy 328 Hyst + biopsy NR NR NR Martinez-Rubio Spain 369 EBP Paraskevaidis Greece 59 EBK Tsuda Japan 375 EES, EBNS NR NR NR Tsuda Japan 883 SC EC, endometrial carcinoma; AEH, atypical endometrial hyperplasia; NA, not applicable; NR, not reported. Endometrial verification methods: EB NS, endometrial biopsy not specified; EB N, endometrial biopsy, Novak curette; EES, endocyte endometrial sampler; EB P, endometrial biopsy, Pipelle; Hyst, hysteroscopy; D&C, dilatation and curettage; EB S, endometrial biopsy, Semm s canula; EB K, endometrial biopsy, Karman; SC, softcyto. In total, these 20 studies described 6,974 women with a median sample size of 209 (range 47 to 1,926). Study characteristics are reported in Table 3. In 13 studies partial verification occurred, whereas in the other seven studies endometrial verification was performed in all women. In 5,198 out of 6,974 women an endometrial sample was obtained. Endometrial carcinoma was found in 32 and atypical endometrial hyperplasia in 21 women. 8 Diagnosis of endometrial carcinoma Eight different endometrial thickness cut-off values were reported in the 20 studies. Sensitivity, specificity, negative predictive value and positive predictive value for the detection of endometrial carcinoma for each reported endometrial thickness cut-off value are reported in Table 4. Positive predictive values varied between 0 and 0.2 and negative predictive value, if estimable, was 1. For the reported endometrial thickness cut-offs 4, 5 and 6 mm we were able to calculate a summary sensitivity and specificity for the detection of endometrial carcinoma, the remaining five cut-offs were reported by one single study, for these cut-offs we were therefore unable to calculate summary sensitivity and specificity. 103

106 Chapter 8 Table 3. Characteristics of studies included for estimation of sensitivity and specificity of endometrial thickness as measured by transvaginal ultrasound with regard to premalignant or malignant endometrium in asymptomatic postmenopausal women not using hormone replacement therapy. Author Year Country n Verification method Patients verified (n) Patients < cut off Patients > cut off Cut-off Total (n) With EC (n) With AEH (n) Total (n) With EC (n) With AEH (n) Andolf Sweden 300 D&C 11 5 mm 289 NR NR Cohen USA 60 EBP 60 5 mm Exacoustos Italy 910 Hyst + biopsy 83 8 mm 827 NR NR 83 3 NR Fernandez Spain 209 Hyst + biopsy mm NR NR NR NR Fleischer USA 1926 EBNS mm Gouveia Brazil 47 EBP 47 5 mm 28 0 NR 19 1 NR Gull Sweden 361 Hyst + D&C 18 8 mm 343 NR NR Güven Turkey 97 D&C 97 5 mm Kasraeian Iran 259 Hyst + biopsy mm Malinova Bulgaria 130 D&C 30 6 mm 95 NR NR 35 0 NR Marello Italy 328 Hyst + biopsy mm NR NR Minagawa Japan 146 SC 5 5 mm 141 NR NR Paraskevaidis Greece 59 EBK 59 9 mm Pardo Israel 85 Hyst + biopsy 85 7 mm NR NR NR Psyllaki Greece 850 Hyst + D&C mm 701 NR NR Ribeiro Brazil 399 Hyst + biopsy mm NR NR NR Schmidt Germany 209 Hyst + D&C mm NR NR NR Tsuda Japan 375 EES + EBNS mm NR NR 4 mm NR 63 0 NR 6 mm NR 30 0 NR 8 mm NR 23 0 NR 10 mm NR 13 0 NR Valadares Portugal 150 Hyst + biopsy mm NR NR NR NR Zacchi 42 * 1993 Italy 74 Hyst + biopsy 6 8 mm 68 NR NR *Single layer measurement, values reported in article are multiplied by two. The reported number is included in the cut-off, i.e. cut off 5mm means ET 5mm. EC, endometrial carcinoma; AEH, atypical endometrial hyperplasia;tvs, transvaginal ultrasonography; ET, endometrial thickness; NR, not reported; NA, not applicable Endometrial verification methods: D&C, dilatation and curettage; EB P, endometrial biopsy, Pipelle; Hyst, hysteroscopy; EB NS, endometrial biopsy not specified; SC, softcyto; EB K, endometrial biopsy, Karman; EES, endocyte endometrial sampler. 104

107 Endometrial thickness measurement in asymptomatic postmenopausal women Table 4. Sensitivity, specificity, positive and negative predictive value of endometrial thickness as measured by transvaginal ultrasound with regard to endometrial carcinoma. Studies are grouped by reported cut-off value. Study Year TP FP Fn Tn Sensitivity Specificity PPv npv Endometrial thickness cut-off 3 mm Tsuda ( ) 0.71 ( ) Endometrial thickness cut-off 4 mm Marello ( ) 0.61 ( ) 0 1 Ribeiro NR NR NE NE NE Tsuda ( ) 0.83 ( ) 0 1 Valadares NR NR NE NE 0.01 NE Endometrial thickness cut-off 5 mm Andolf NR NR NE NE 0 NE Cohen NE 0.63 ( ) 0 1 Fernandez NR NR NE NE 0.02 NE Gouveia ( ) 0.61 ( ) Guven NE 0.77 ( ) 0 1 Kasraeian ( ) 0.84 ( ) Minagawa NR NR NE NE 0.2 NE Psyllaki NR NR NE NE 0 NE Endometrial thickness cut-off 6 mm Fleischer ( ) 0.95 ( ) Malinova NR NR NE NE 0 NE Schmidt NR NR NE NE 0.04 NE Tsuda ( ) 0.92 ( ) 0 1 Endometrial thickness cut-off 7 mm Pardo NR NR NE NE 0.04 NE Endometrial thickness cut-off 8 mm Exacoustos NR NR NE NE 0.04 NE Gull NR NR NE NE 0 NE Tsuda ( ) 0.94 ( ) 0 1 Zacchi NR NR NE NE 0 NE Endometrial thickness cut-off 9 mm Paraskevaidis ( ) 0.67 ( ) Endometrial thickness cut-off 10 mm Tsuda ( ) 0.97 ( ) TP, true positive; FP, false positive; FN, false negative; TN, true negative; PPV, positive predictive value; NPV, negative predictive value; NR, not reported; NE, not estimable. 105

108 Chapter 8 The summary estimate of the sensitivity for the different cut-offs varied between 0.00 and The summary estimate of the specificity for the different cut-offs varied between 0.72 and 0.93 (Table 5). For the summary estimates of sensitivity and specificity as well as for the reported sensitivity and specificity, the 95% confidence intervals were very wide, indicating the high uncertainty surrounding these estimates. Table 5. Summary estimates of sensitivity and specificity with regard to endometrial carcinoma for different transvaginal ultrasound derived endometrial thickness cutoff values. Threshold mm No of studies No of women sens (95% CI) spec (95% CI) ( ) 0.71 ( ) ( ) 0.73 ( ) ( ) 0.72 ( ) ( ) 0.94 ( ) NE NE ( ) 0.94 ( ) ( ) 0.67 ( ) ( ) 0.96 ( ) 95% CI, 95% confidence interval; NE, not possible to estimate Diagnosis of atypical endometrial hyperplasia Six different endometrial thickness cut-off values were reported by 13 studies. The other seven studies did not report on an endometrial thickness cut-off for atypical endometrial hyperplasia. Sensitivity, specificity, negative predictive value and positive predictive value for the detection of atypical endometrial hyperplasia for each reported endometrial thickness cut-off value are reported in Table 6. There were no cut-off values for which multiple studies reported sufficient data to calculate both sensitivity and specificity. Therefore, we were unable to calculate summary estimates of sensitivity and specificity for the different cut-off values for atypical endometrial hyperplasia. With regard to atypical endometrial hyperplasia, the range of sensitivity was and the range of specificity was Diagnosis of combined (pre-) malignancy For this combined analysis, studies were included if they reported on both endometrial carcinoma and atypical endometrial hyperplasia. All 20 studies reported on endometrial carcinoma and 13 studies reported on atypical endometrial hyperplasia as well. Six different endometrial cut-off values were reported. Sensitivity, specificity, negative predictive value and positive predictive value for the detection of endometrial (pre-) malignancy for each reported endometrial thickness cut-off value are reported in Table 7. As for atypical endometrial hyperplasia, there were no cut-off values for which multiple studies reported 106

109 Endometrial thickness measurement in asymptomatic postmenopausal women sufficient data to calculate both sensitivity and specificity. Therefore, we were unable to calculate summary estimates of sensitivity and specificity for the different cut-off values for the combined outcome of (pre-) malignancy. With regard to the combined diagnosis of premalignant and malignant endometrium, the range of sensitivity was and the range of specificity was Table 6. Sensitivity, specificity, positive and negative predictive value of endometrial thickness as measured by transvaginal ultrasound with regard to atypical endometrial hyperplasia. Studies are grouped by reported cut-off value. Study Year TP FP Fn Tn Sensitivity Specificity PPv npv Endometrial thickness cut-off 4 mm Ribeiro NR NR NE NE NE Endometrial thickness cut-off 5 mm Andolf NR NR NE NE 0 NE Cohen NE 0.63 ( ) 0 1 Guven NE 0.77 ( ) 0 1 Kasraeian ( ) 0.85 ( ) Minagawa NR NR NE NE 0 NE Psyllaki NR NR NE NE NE Endometrial thickness cut-off 6 mm Fleischer ( ) 0.95 ( ) 0 1 Schmidt NR NR NE NE 0.02 NE Endometrial thickness cut-off 7 mm Pardo NR NR NE NE 0 NE Endometrial thickness cut-off 8 mm Gull NR NR NE NE 0 NE Zacchi NR NR NE NE 0 NE Endometrial thickness cut-off 9 mm Paraskevaidis ( ) 0.67 ( ) TP, true positive; FP, false positive; FN, false negative; TN, true negative; PPV, positive predictive value; NPV, negative predictive value; NR, not reported; NE, not estimable

110 Chapter 8 Table 7. Sensitivity, specificity, positive and negative predictive value of endometrial thickness as measured by transvaginal ultrasound with regard to endometrial malignancy /premalignancy. Studies are grouped by reported cut-off value. Study Year TP FP FN TN Sensitivity Specificity PPV NPV Endometrial thickness cut-off 4 mm Ribeiro NR NR NE NE NE Endometrial thickness cut-off 5 mm Andolf NR NR NE NE 0 NE Cohen NE 0.63 ( ) 0 1 Guven NE 0.77 ( ) 0 1 Kasraeian ( ) 0.86 ( ) Minagawa NR NR NE NE 0.2 NE Psyllaki NR NR NE NE NE Endometrial thickness cut-off 6 mm Fleischer ( ) 0.95 ( ) Schmidt NR NR NE NE 0.06 NE Endometrial thickness cut-off 7 mm Pardo NR NR NE NE 0.04 NE Endometrial thickness cut-off 8 mm Gull NR NR NE NE 0 NE Zacchi NR NR NE NE 0 NE Endometrial thickness cut-off 9 mm Paraskevaidis ( ) 0.68 ( ) TP, true positive; FP, false positive; FN, false negative; TN, true negative; PPV, positive predictive value; NPV, negative predictive value; NR, not reported; NE, not estimable. Discussion Our review shows that in a population of postmenopausal women without PMB and not using HRT, the mean endometrial thickness is 2.9 mm and the prevalence of endometrial carcinoma and atypical endometrial hyperplasia is 0.62% and 0.59% respectively. Positive predictive values for the three outcomes (endometrial carcinoma, atypical endometrial hyperplasia and both combined) for all reported endometrial thickness cut-offs were between 0 and 0.2.The NPV of TVS was between 0.98 and 1.0, at all endometrial thickness cut-offs and for all three disease outcomes. However, the utility of a negative test in an asymptomatic postmenopausal population is limited because the absolute risk of disease is already low, as demonstrated in this review (prevalence of endometrial carcinoma 0.62% and atypical endometrial hyperplasia 0.59%). This contrasts to symptomatic postmenopausal women where the pre-testing risk of endometrial cancer or atypical hyperplasia varies between 5 and 20%. 43, 44 Thus, TVS is only of value in postmenopausal women with vaginal bleeding 108

111 Endometrial thickness measurement in asymptomatic postmenopausal women (PMB) because a clinically substantial reduction in disease probability may be achieved by the use of TVS. This reduction is typically from around 10% to below 1% for endometrial carcinoma, 3 a probability threshold where firstly, the majority of clinicians recommend reassurance and no need for further evaluation of the endometrium, and secondly, a post-tvs probability is demonstrated to be equivalent to the prevalence in asymptomatic postmenopausal female population. 4-7 The strength of our analysis is the complete overview of data combining endometrial thickness, endometrial carcinoma and atypical endometrial hyperplasia in asymptomatic postmenopausal women without HRT. We describe mean endometrial thickness and the prevalence of endometrial (pre-) malignancies in these women. Furthermore, we assessed the diagnostic accuracy of endometrial thickness measurements in this population. Efforts were made to identify all available publications on this subject and we used the most appropriate technique to summarize the sensitivity and specificity, to come to better estimates than the formerly applied summary receiver operating characteristic curve (sroc) technique for meta-regression in diagnostic meta-analysis. 10, To our knowledge, there is no previously published review or meta-analysis on this subject. A limitation of our study is that despite the thousands of women included in our analysis, the estimates of sensitivity and specificity are very imprecise, especially the estimates of sensitivity. Another limitation could be a bias because of the quality of the included studies. We tried to minimize this bias by performing quality assessment and applying strict criteria for inclusion of studies in the meta-analysis. In a decision analysis performed by Smith-Bindman et al. 8, an endometrial thickness cut-off of 11 mm for an incidentally measured increased endometrial thickness in an asymptomatic woman was proposed. In this decision analysis the risk of malignancy in a woman below the threshold is extremely low and the risk of malignancy above the threshold varied between 2.2% and 9.3%. In contrast to this analysis, which was a decision analysis in a theoretical cohort, we analyzed observational data. Unfortunately, we had insufficient data available from the published studies to calculate an optimal threshold for endometrial thickness based on the sensitivity and specificity reported in the different studies. Because of the low prevalence of the disease, the 95% CI for the summary estimates of sensitivity are very wide indicating a high inaccuracy. The use of TVS is not limited to women with PMB. The portability and improved resolution of TVS has contributed to the ubiquity of the test in routine gynecological practice. Postmenopausal women undergo TVS for a variety of gynecological indications (e.g. pelvic pain, suspicion of a pelvic mass, uterine prolapse). During TVS for such nonbleeding indications, images of the endometrium are frequently obtained and a thickened endometrium may be observed. This situation of an apparently incidental finding of an abnormal endometrium will be familiar to all practicing ultrasonographists. Faced with such 8 109

112 Chapter 8 a TVS finding, it is difficult for the physician to decide on the right management and usually results in a decision to undertake further, more invasive testing with endometrial sampling and / or hysteroscopy in keeping with current PMB pathways. Therefore, the findings of this review, describing normative values for endometrial thickness, determining serious disease prevalence and estimating diagnostic accuracy at various TVS thresholds, in this nonbleeding postmenopausal population are clinically important. Our review has shown that the average TVS derived endometrial thickness is 2.9 mm. However, the significance of an endometrial thickness beyond 4 mm is not the same as for a symptomatic PMB population and extrapolating protocols from PMB to an asymptomatic population is not justifiable in view of the poor performance of TVS in detecting serious endometrial disease at all cut-offs and the low overall disease prevalence. Because the prevalence of the target disease in an unselected postmenopausal population without bleeding symptoms and without HRT is very low, and endometrial thickness measurement in this population cannot achieve a sufficiently high sensitivity to provide additional reassurance to women with a negative test nor achieve a sufficiently high specificity to justify further invasive testing in women with a positive test, endometrial thickness measurement has no value in this population. Furthermore, there is no evidence that patients in whom endometrial cancer was discovered while being asymptomatic have a prognostic advantage over postmenopausal endometrial cancer patients who visited their gynecologists immediately after bleeding had occurred. 48 Thus, the results from this systematic review do not justify the use of endometrial thickness as a screening test for endometrial carcinoma and atypical endometrial hyperplasia in any asymptomatic postmenopausal woman without HRT. Hence, the need for further diagnostic evaluation of the endometrium should be made by the clinician on an individual patient basis taking into account clinical signs (e.g. abnormal findings at physical examination, pelvic pain, distension, urinary and bowel complaints etc.), risk factors for endometrial disease (e.g. abnormal BMI, medical co-morbidities, family history etc.) and patient preference

113 Endometrial thickness measurement in asymptomatic postmenopausal women REFEREnCES 1. Brenner PF. Differential diagnosis of abnormal uterine bleeding. Am J Obstet Gynecol 1996;175(3 Pt 2): Parkin DM, Bray F, Ferlay J, Pisani P. Estimating the world cancer burden: Globocan Int J Cancer 2001;94(2): Smith-Bindman R, Kerlikowske K, Feldstein VA et al. Endovaginal ultrasound to exclude endometrial cancer and other endometrial abnormalities. JAMA 1998;280(17): Epstein E. Management of postmenopausal bleeding in Sweden: a need for increased use of hydrosonography and hysteroscopy. Acta Obstet Gynecol Scand 2004;83(1): Goldstein RB, Bree RL, Benson CB et al. Evaluation of the woman with postmenopausal bleeding: Society of Radiologists in Ultrasound-Sponsored Consensus Conference statement. J Ultrasound Med 2001;20(10): NVOG (Dutch Society of Obstetrics and Gynaecology). NVOG-richtlijn Abnormaal vaginaal bloedverlies in de menopauze [In Dutch]. NVOG Guideline: Abnormal vaginal bleeding during menopause Scottish Intercollegiate Guidelines Network. Investigation of postmenopausal bleeding. Scottish Intercollegiate Guidelines Network, Royal College of Physicians Smith-Bindman R, Weiss E, Feldstein V. How thick is too thick? When endometrial thickness should prompt biopsy in postmenopausal women without vaginal bleeding. Ultrasound Obstet Gynecol 2004;24(5): Whiting P, Rutjes AW, Reitsma JB, Bossuyt PM, Kleijnen J. The development of QUADAS: a tool for the quality assessment of studies of diagnostic accuracy included in systematic reviews. BMC Med Res Methodol 2003;3: Reitsma JB, Glas AS, Rutjes AW, Scholten RJ, Bossuyt PM, Zwinderman AH. Bivariate analysis of sensitivity and specificity produces informative summary measures in diagnostic reviews. J Clin Epidemiol 2005;58(10): Andolf E, Dahlander K, Aspenberg P. Ultrasonic thickness of the endometrium correlated to body weight in asymptomatic postmenopausal women. Obstet Gynecol 1993;82(6): Gull B, Karlsson B, Milsom I, Wikland M, Granberg S. Transvaginal sonography of the endometrium in a representative sample of postmenopausal women. Ultrasound Obstet Gynecol 1996;7(5): Guven MA, Pata O, Bakaris S, Kafkasli A, Mgoyi L. Postmenopausal endometrial cancer screening: is there a correlation between transvaginal sonographic measurement of endometrial thickness and body mass index? Eur J Gynaecol Oncol 2004;25(3): Kasraeian M, Asadi N, Ghaffarpasand F, Karimi AA. Value of transvaginal ultrasonography in endometrial evaluation of non-bleeding postmenopausal women. Climacteric 2011;13(6): Malinova M, Pehlivanov B. Transvaginal sonography and progesterone challenge for identifying endometrial pathology in postmenopausal women. Int J Gynaecol Obstet 1996;52(1): Minagawa Y, Sato S, Ito M, Onohara Y, Nakamoto S, Kigawa J. Transvaginal ultrasonography and endometrial cytology as a diagnostic schema for endometrial cancer. Gynecol Obstet Invest 2005;59(3): Neele SJ, Marchien van BW, van der Mooren MJ, Kessel H, Netelenbos JC, Kenemans P. Ultrasound assessment of the endometrium in healthy, asymptomatic early post-menopausal women: saline infusion sonohysterography versus transvaginal ultrasound. Ultrasound Obstet Gynecol 2000;16(3): Osmers R, Volksen M, Rath W, Teichmann A, Kuhn W. [Vaginosonographic measurement of the postmenopausal endometrium in the early detection of endometrial cancer]. Geburtshilfe Frauenheilkd 1989;49(3): Pirhonen JP, Vuento MH, Makinen JI, Salmi TA. Long-term effects of hormone replacement therapy on the uterus and on uterine circulation. Am J Obstet Gynecol 1993;168(2): Warming L, Ravn P, Skouby S, Christiansen C. Measurement precision and normal range of endometrial thickness in a postmenopausal population by transvaginal ultrasound. Ultrasound Obstet Gynecol 2002;20(5): Bortoletto CC, Baracat EC, Goncalves WJ, Lima GR, Stavale JN. Transvaginal ultrasonography and the progestogen challenge test in postmenopausal endometrial evaluation. Int J Gynaecol Obstet 1997;58(3): Buccoliero AM, Caldarella A, Noci I et al. [Thin-layer cytology in endometrial diagnosis]. Pathologica 2003;95(4):

114 Chapter Cohen MA, Sauer MV, Keltz M, Lindheim SR. Utilizing routine sonohysterography to detect intrauterine pathology before initiating hormone replacement therapy. Menopause 1999;6(1): Elewa AMA. Correlation of vaginal ultrasound and hysteroscopy with endometrial histopathology in postmenopausal women. Middle East Fertility Society Journal 2001;6(1): Gol K, Saracoglu F, Ekici A, Sahin I. Endometrial patterns and endocrinologic characteristics of asymptomatic menopausal women. Gynecol Endocrinol 2001;15(1): Gouveia DA, Bahamondes L, Aldrighi JM, Tamanaha S, Ribeiro AL, Aoki T. [Prevalence of endometrial injury in asymptomatic obese women]. Rev Assoc Med Bras 2007;53(4): Langer RD, Pierce JJ, O Hanlan KA et al. Transvaginal ultrasonography compared with endometrial biopsy for the detection of endometrial disease. Postmenopausal Estrogen/Progestin Interventions Trial. N Engl J Med 1997;337(25): Macia M, Novo A, Ces J, Gonzalez M, Quintana S, Codesido J. Progesterone challenge test for the assessment of endometrial pathology in asymptomatic menopausal women. Int J Gynaecol Obstet 1993;40(2): Marello F, Bettocchi S, Greco P et al. Hysteroscopic evaluation of menopausal patients with sonographically atrophic endometrium. J Am Assoc Gynecol Laparosc 2000;7(2): Martinez-Rubio MP, Alcazar JL. Ultrasonographic and pathological endometrial findings in asymptomatic postmenopausal women taking antihypertensive drugs. Maturitas 2003;46(1): Paraskevaidis E, Papadimitriou D, Kalantaridou SN et al. Screening transvaginal uterine ultrasonography for identifying endometrial pathology in postmenopausal women. Anticancer Res 2002;22(4): Tsuda H, Kawabata M, Yamamoto K, Inoue T, Umesaki N. Prospective study to compare endometrial cytology and transvaginal ultrasonography for identification of endometrial malignancies. Gynecol Oncol 1997;65(3): Tsuda H, Nakamura H, Inoue T, Kawamura N, Adachi K, Bandera CA. Transvaginal ultrasonography of the endometrium in postmenopausal Japanese women. Gynecol Obstet Invest 2005;60(4): Exacoustos C, Chiaretti M, Minghetti MC, Bianchi L, Arduini D, Romanini C. Endometrial Evaluation in Asymptomatic Postmenopausal Women by Transvaginal Sonography and Color Flow Doppler. J Am Assoc Gynecol Laparosc 1996;3(4, Supplement):S Fernandez Parra JGP. Endometrial alterations on ultrasonography in asymptomatic postmenopausal women. Progresos en Obstetricia y Ginecologia 2008;51(7): Fleischer AC, Wheeler JE, Lindsay I et al. An assessment of the value of ultrasonographic screening for endometrial disease in postmenopausal women without symptoms. Am J Obstet Gynecol 2001;184(2): Pardo J, Aschkenazi S, Kaplan B, Orvieto R, Nitke S, Ben-Refael Z. Abnormal sonographic endometrial findings in asymptomatic postmenopausal women: possible role of antihypertensive drugs. Menopause 1998;5(4): Psyllaki AN, I. Transvaginal sonographic prognostic value to detect endometrial pathology in postmenopausal asymptomatic women without hormone replacement therapy. Archives of Gynecology and Obstetrics 2010;Conference(Deutsche Gesellschaft fur Gynakologie und Geburtshilfe, DGGG):Munich-S Ribeiro CT, Rosa-E-Silva JC, Silva-de-Sa MF et al. Hysteroscopy as a standard procedure for assessing endometrial lesions among postmenopausal women. Sao Paulo Med J 2007;125(6): Schmidt TR. The role of hysteroscopy in the management of asymptomatic postmenopausal patients with suspicious ultrasound findings of the uterine endometrium - Correlation with sonographic and histologic findings. Geburtshilfe und Frauenheilkunde 1999;59(4): Valadares S.Coutinho. Ambulatory hysteroscopy results post-menopause: Comparative study between patients with and without metrorrhagia. Gynecological Surgery 2005;2(4): Zacchi V, Zini R, Canino A. [Transvaginal sonography as a screening method for the identification of patients at risk of postmenopausal endometrial pathology]. Minerva Ginecol 1993;45(7-8): Dijkhuizen FP, Brolmann HA, Potters AE, Bongers MY, Heinz AP. The accuracy of transvaginal ultrasonography in the diagnosis of endometrial abnormalities. Obstet Gynecol 1996;87(3): Emanuel M, Verdel M, Wamsteker K, Lammes F. An audit of true prevalence of intrauterine pathology: the hysteroscopic findings, controled for patient selection in 1202 patients with abnormal uterine bleeding. Gynaecol Endosc 1995;4:

115 Endometrial thickness measurement in asymptomatic postmenopausal women 45. Deeks JJ. Systematic reviews in health care: Systematic reviews of evaluations of diagnostic and screening tests. BMJ 2001;323(7305): Leeflang MM, Deeks JJ, Gatsonis C, Bossuyt PM. Systematic reviews of diagnostic test accuracy. Ann Intern Med 2008;149(12): Shapiro DE. Issues in combining independent estimates of the sensitivity and specificity of a diagnostic test. Acad Radiol 1995;2 Suppl 1:S37-S Gerber B, Krause A, Muller H et al. Ultrasonographic detection of asymptomatic endometrial cancer in postmenopausal patients offers no prognostic advantage over symptomatic disease discovered by uterine bleeding. Eur J Cancer 2001;37(1): Anderson KE, Anderson E, Mink PJ et al. Diabetes and endometrial cancer in the Iowa women s health study. Cancer Epidemiol Biomarkers Prev 2001;10(6): McPherson CP, Sellers TA, Potter JD, Bostick RM, Folsom AR. Reproductive factors and risk of endometrial cancer. The Iowa Women s Health Study. Am J Epidemiol 1996;143(12): Timmermans A, Opmeer BC, Veersema S, Mol BW. Patients preferences in the evaluation of postmenopausal bleeding. BJOG 2007;114(9): van Doorn LC, Dijkhuizen FP, Kruitwagen RF, Heintz AP, Kooi GS, Mol BW. Accuracy of transvaginal ultrasonography in diabetic or obese women with postmenopausal bleeding. Obstet Gynecol 2004;104(3): Weiderpass E, Persson I, Adami HO, Magnusson C, Lindgren A, Baron JA. Body size in different periods of life, diabetes mellitus, hypertension, and risk of postmenopausal endometrial cancer (Sweden). Cancer Causes Control 2000;11(2): Xu WH, Xiang YB, Ruan ZX et al. Menstrual and reproductive factors and endometrial cancer risk: Results from a population-based case-control study in urban Shanghai. Int J Cancer 2004;108(4):

116 Chapter 8 APPENDIX Appendix 1. Search Strategy Medline #1. Postmenopause [MeSH] #2. Postmenopau* [tiab] #3. Post-menopau* [tiab] #4. #1 OR #2 OR #3 #5. Asymptomatic [tiab] #6. Screening [tiab] #7. #5 OR #6 #8. #4 AND #7 #9. Ultrasonography [MeSH] #10. Utrasound* [tiab] #11. Ultrasonograph* [tiab] #12. Sonograph* [tiab] #13. Echograph* [tiab] #14. Ultrasonic imaging* [tiab] #15. #9 OR #10 OR #11 OR #12 OR #13 OR #14 #16. Endometrium [MeSH] #17. Endometrial thickness [tiab] #18. #16 OR #17 #19. #15 AND #18 #20. #19 AND #8 #21. Endometrial neoplasms [MeSH] #22. Endometrial neoplasm* [tiab] #23. Endometrial carcinoma* [tiab] #24. Endometrial cancer* [tiab] #25. Endometrial malignanc* [tiab] #26. Endometrial tumo* [tiab] #27. #21 OR #22 OR #23 OR #24 OR #25 OR #26 #28. Endometrial hyperplasia [MeSH] #29. Endometrial hyperplasia* [tiab] #30. #28 OR # 29 #31. #27 OR #30 #32. # 8 AND # 31 #33. # 20 OR # 32 Embase #1. Postmenopausal #2. Postmenopau$.ti,ab #3. Post-menopau$.ti,ab #4. #1 OR #2 OR #3 #5. Asymptomatic.ti,ab #6. Screening.ti,ab #7. #5 OR #6 #8. #4 AND #7 #9. Ultrasound$.ti,ab #10. Ultrasonograph$.ti,ab #11. Sonograph$.ti,ab #12. Echograph$.ti,ab #13. Ultrasonic imaging$.ti,ab #14. #9 OR #10 OR #11 OR #12 OR #13 #15. Endometri$.ti,ab #16. Endometrial thickness.ti,ab #17. #15 OR #16 #18. # 14 AND #17 #19. #18 AND #8 #20. Endometrial carcinoma$.ti,ab #21. Endometrial neoplasm$.ti,ab #22. Endometrial cancer$.ti,ab #23. Endometrial malignanc$.ti,ab #24. Endometrial tumo$.ti,ab #25. #20 OR #21 OR #22 OR #23 OR #24 #26. Endometrial hyperplasia$.ti,ab #27. #25 OR #26 #28. #8 AND #27 #29. #19 OR #28 114

117 Endometrial thickness measurement in asymptomatic postmenopausal women Appendix 2. QUADAS checklist Item Question Answer Yes no Unclear 1 Was the spectrum of patients representative of the patients who will receive the test in practice? 2 Were selection criteria clearly described? 3 Is the reference standard likely to correctly classify the target condition? 4 Is the time period between reference standard and index test short enough to be reasonably sure that the target condition did not change between the two tests? 5 Did the whole sample or a random selection of the sample, receive verification using a reference standard of diagnosis? 6 Did patients receive the same reference standard regardless of the index test result? 7 Was the reference standard independent of the index test? 8 Was the execution of the index test described in sufficient detail to permit replication of the test? 9 Was the execution of the reference standard described in sufficient detail to permit its replication? 10 Were the index test results interpreted without knowledge of the results of the reference standard? 11 Were the reference standard results interpreted without knowledge of the results of the index test? 12 Were the same clinical data available when test results were interpreted as would be available when the test is used in practice? 13 Were uninterpretable / intermediate test results reported? 14 Were withdrawals from the study explained? 8 115

118

119 9 Summary and General discussion

120

121 Summary and General Discussion SuMMARY The goal of this thesis is to evaluate different diagnostic strategies for women with postmenopausal bleeding (PMB), including strategies based on individual probabilities and to study the value of endometrial thickness measurement in asymptomatic postmenopausal women. Chapter 2 presents a systematic review on existing prediction models for endometrial cancer in women with PMB. We identified nine studies reporting on the development and validation of prediction models in women with PMB. From the data we identified the most important predictor variables, which could be roughly divided into four subjects: patient characteristics, gray-scale transvaginal sonography (TVS) variables, Doppler TVS variables and hysteroscopy variables. Most prediction models used a combination of these subjects to predict the chance of endometrial cancer. Eight of these studies described at least one aspect of internal validation and none of the prediction models was externally validated. Models including power Doppler showed the best performance in internal validation, but based on the difficult use of Doppler in general gynecological practice, we concluded that the best models up to present are prediction models combining patient characteristics with endometrial thickness. In Chapter 3 we evaluated the cost-effectiveness of strategies incorporating the diagnostic value of patient characteristics for endometrial carcinoma using prediction models with a decision analytic approach. We designed a decision analytic model to compare four diagnostic strategies for women with PMB. (1) The patient characteristics strategy, i.e. probability estimates based on characteristics of the women and histological analysis in case the probability of (pre)malignancy exceeded 4%. In this strategy, TVS is not performed. (2) The sequential strategy, i.e. probability estimates based on characteristics of the women, with TVS in case the probability for cancer exceeded 4% and subsequent histological analysis when the endometrial thickness exceeds 4mm. (3) The integrated strategy, i.e. TVS in all women, with a probability estimate based on both characteristics of the women and TVS results, completed by histological analysis when the probability of cancer exceeded 4%. The three model based strategies were compared to a strategy with endometrial thickness measurement reflecting current practice. Strategy (1), selecting women for endometrial biopsy based on their history only, dominated all other strategies (more effective, less costs). In a clinical scenario where TVS was assumed to be an integral part of the consultation without additional costs, a strategy selecting high-risk women for TVS (the sequential strategy) became the most cost-effective strategy

122 Chapter 9 For a successful implementation, a prediction model should be validated externally in an independent population. Chapter 4 describes the external validation of two previously developed models predicting endometrial cancer in women with PMB: the patient characteristics model and the patient characteristics and TVS model. We applied the models on two independent databases collected in the Netherlands and Sweden to assess the diagnostic performance of the models. In both the Dutch and the Swedish databases discrimination for the patient characteristics and TVS model showed an AUC of 0.89 (95% CI /92), which was higher compared to the AUC for the patients characteristics - model, 0.71 (95% CI ) and 0.69 (95% CI ) respectively. The calculated AUC for TVS-only was 0.87 in the Dutch Database (95% CI ) and 0.90 in the Swedish database (95% CI ). This study shows that after external validation in two independent datasets, the existing multivariable models maintain their diagnostic performance and are able to distinguish between women with low or high risk of endometrial cancer in women with PMB. The patient characteristic and TVS model however, offers no diagnostic advantage over the measurement of endometrial thickness alone. The patient characteristics model is able to select women with a low risk of endometrial cancer, who can be reassured without further testing. This is especially useful in a setting where TVS is not (directly) available. In a substantial number of cases, endometrial sampling fails because of technical problems or because of an insufficient amount of tissue for a histological diagnosis. In Chapter 5 we investigated which doctor and patient related factors contribute to failure of outpatient endometrial sampling in women with PMB and an endometrial thickness > 4 mm. Nulliparity and advanced age were associated with failure and an endometrial thickness > 12 mm decreased the chance of failure. Training level was not associated with technical failure or insufficient sampling. A prediction model for total failure was subsequently designed. The AUC of the model was 0.64 (95% CI ) indicating a moderate capacity to discriminate between women with a high or a low risk of failure. Chapter 6 evaluates the use of the prediction model for a failed endometrial biopsy to reduce costs for the same accuracy of diagnostic testing in women with PMB through a cost-minimization analysis. If the probability of failure exceeds 65% it is less costly to refer the patient immediately for outpatient hysteroscopy with biopsy, instead of an attempt for outpatient endometrial biopsy. However this is only a cost-reduction of three euro s per patient, so we concluded that individualizing the decision to perform endometrial biopsy or hysteroscopy in women presenting with PMB based on patient characteristics is unlikely to increase the efficiency of the diagnostic work up. 120

123 Summary and General Discussion No guidelines on standardized assessment of endometrial biopsy specimens are available, which results in a diagnosis that is influenced by subjectivity, with a high inter-pathologist variability. The objective of Chapter 7 was to determine whether structured assessment of outpatient endometrial biopsy specimens with strict criteria decreases the amount of inconclusive samples due to insufficient material. We requested 66 and retrieved 36 endometrial samples from eight different hospitals that were collected during the normal diagnostic work-up and assessed as insufficient for reliable histological diagnosis. Structured reassessment of the retrieved samples by one pathologist specialized in gynecology did not change the conclusion in 35 of the 36 samples. Only one sample contained a large amount of endometrial tissue and the final diagnosis was endometrial hyperplasia without atypia. All other samples contained insufficient material for a reliable diagnosis. Our findings suggest that although it might be helpful for pathologists to have diagnostic criteria for adequacy and/or inadequacy of an endometrial biopsy sample, the gain in efficiency is likely to be small. We therefore think that to increase the effectiveness of outpatient endometrial biopsies, effort has to be made to obtain as much material as possible to minimize the failure of endometrial biopsies due to insufficient material. Measurement of endometrial thickness is an important tool in the assessment of women with PMB. The relevance of endometrial thickness measurement by ultrasound in asymptomatic women is unclear. Chapter 8 describes a systematic review and meta-analysis to address: (I) the normal endometrial thickness measured by TVS, (II) the prevalence of endometrial pathology and (III) the sensitivity and specificity of endometrial thickness measurement by TVS for diagnosing (pre-) malignant endometrial disease in asymptomatic postmenopausal women. We included 32 studies reporting on 11,100 women to answer our three objectives: (I) The estimated mean endometrial thickness was 2.9 mm (95% CI ). (II) The pooled estimated prevalence of endometrial carcinoma and atypical endometrial hyperplasia were 0.62% (95% CI ), and 0.59% (95% CI ), respectively. (III) Summary estimates for sensitivity and specificity of TVS endometrial thickness measurement were 0.83 (95% CI ) and 0.72 (95% CI ) for 5 mm cut-off and 0.33 (95% CI ) and 0.94 (95% CI ) for 6 mm cut-off. The results from this systematic review do not justify the use of endometrial thickness as a screening test for endometrial carcinoma and atypical endometrial hyperplasia in asymptomatic postmenopausal women not using HRT

124

125 Summary and General Discussion general DISCuSSIOn The studies presented in this thesis focus on an individualized approach for women with PMB. We assessed the cost-effectiveness of an individualized approach and externally validated a prediction model for endometrial cancer in women with PMB. Furthermore, we studied the value of endometrial thickness measurement in asymptomatic postmenopausal women. Below, we discuss the implications of the main findings in this thesis for clinical practice and future research. Cost-effectiveness and external validation of prediction models for endometrial cancer in women with postmenopausal bleeding. In the current diagnostic work-up for PMB, an endometrial thickness cut-off value of 4 mm is used to determine whether a patient needs further invasive testing for endometrial carcinoma. This cut-off value of 4 mm advised by Dutch guidelines is based on a metaanalysis performed by Smith-Bindman et al. 1 The determination of the cut-off value for the Netherlands was based on a 10% pre-test probability, the risk for all women with PMB regardless of other risk factors. 2 While the probability of PMB decreases with increasing age, 3 the probability of endometrial cancer increases significantly with increasing age. 4 Depending on other risk factors including age, time since menopause, obesity, hypertension, diabetes mellitus and reproductive factors, the risk of endometrial cancer varies between 1% and 24%. 3, 5-10 However, in currently used guidelines individual patient characteristics are not taken into account. Previous systematic reviews showed that TVS and hysteroscopy both have high individual accuracy in predicting endometrial (pre) malignancy. 11, 12 In these studies the diagnostic tests are evaluated for accuracy against a reference standard, independent of their clinical context or individual patient characteristics. Estimates of diagnostic accuracy derived in this way ignore information that may have been acquired earlier in the diagnostic process. Such an approach can lead to erroneous inferences and may artificially increase the value of diagnostic tests. 13 Inclusion of individual patient characteristics could result in a more individualized and possibly more accurate and efficient work up strategy. In this strategy, a very high a priori chance of endometrial carcinoma warrants further histological testing, whereas women with a very low prior chance might be reassured even without TVS. Our group developed two multivariate models to calculate the individual risk of endometrial carcinoma in women with PMB, which showed good accuracy in internal validation. 14 In this thesis, these previously developed patient characteristics and patient characteristics and TVS models, 14 were externally validated in two independent databases. Both models maintained their diagnostic performance in two independent databases. We compared the patient characteristics and TVS model with endometrial thickness measurement alone. This led to new insights on the applicability of the model. The study presented in Chapter 3 showed that the patient characteristics and TVS model is slightly more cost-effective 9 123

126 Chapter 9 than TVS alone but the diagnostic performance is comparable. We believe that in clinical practice, due to the complexity of the prediction model, TVS alone outweighs this small increase in cost-effectiveness. The patient characteristics model, to be used without an ultrasound, has a much lower diagnostic accuracy than TVS alone. Nevertheless, this patient characteristics model could be used to reassure women without further (invasive) testing in situations where an ultrasound is not immediately available in the physicians office, for example in primary care or in a health system were additional TVS has to be ordered separately and be performed by a radiologist. In clinical practice, a woman can be reassured without TVS and without further testing if she is aged 50 years or younger and has up to one additional risk factor. Efficiency of minimally invasive endometrial biopsy. In case of an endometrial thickness of more than 4 mm, further histological assessment with an endometrial sampling device is advised: the Dutch guideline advises a Pipelle endometrial sampling device. 2 The Pipelle endometrial sampling device has the highest sensitivity in postmenopausal women (99.6%), compared to other sampling devices. 15 A major drawback of this technique is the high failure rate due to technical problems (12-23%) or because of an insufficient amount of material for a reliable diagnosis (16-68%) We hypothesized that less experienced professionals would have a higher endometrial biopsy failure rate. The multivariable analysis presented in Chapter 5 showed that only patient characteristics contributed to the failure of outpatient endometrial biopsies and thus, inexperience was not a risk factor for failure. With a moderate accuracy, a prediction model based on patient characteristics can predict the failure of an endometrial biopsy. Individualizing the decision to perform an endometrial biopsy of hysteroscopy in women presenting with PMB based on patient characteristics is unlikely to increase the efficiency of the diagnostic work up. In our study, the failure rate of endometrial biopsies due to insufficient material was 30%. At subsequent testing, we found a (pre-) malignancy in 7% of these women. In a study performed by Van Doorn et al. an endometrial (pre-) malignancy was diagnosed in 6% of the women with an insufficient sample. 18 Therefore, women with postmenopausal bleeding and a failed endometrial biopsy cannot be reassured without further testing. Pathologists feel that it would be useful if criteria for adequacy and/or inadequacy of an endometrial biopsy were proposed. 20 We evaluated such criteria by applying them on samples that were assessed as insufficient in Chapter 7. This did not change the conclusion in all but one of the samples. Therefore, we concluded that the gain in efficiency of such criteria is likely to be small. In order to increase the effectiveness of outpatient endometrial biopsies, effort has to be made to obtain as much material as possible to minimize the failure of endometrial biopsies due to insufficient diagnosis. Further research should focus on the best way to achieve this improvement. 124

127 Summary and General Discussion Asymptomatic postmenopausal women Also women not suffering from PMB sometimes undergo TVS, when visiting the gynecologist for other indications such as prolapse or abdominal complaints. Inevitably, the endometrium is visualized by TVS and a thickened endometrium may then be observed. The prevalence of endometrial cancer and endometrial hyperplasia in postmenopausal women without bleeding symptoms and without HRT is very low. Endometrial thickness measurement in this population cannot achieve a sufficiently high sensitivity to provide additional reassurance to women with a negative test nor achieve a sufficiently high specificity to justify further invasive testing in women with a positive test. Therefore, endometrial thickness measurement has no value in women without PMB. The clinician should decide on an individual patient basis for further diagnostic evaluation. Current diagnostic work-up When a woman presents with PMB at a general practitioner, he or she will refer her to a gynecologist to exclude the presence of endometrial carcinoma. Current guidelines are not clear and unambiguous about the diagnostic pathway. Based on the available evidence, we think that measurement of endometrial thickness should be the first step in the diagnostic pathway. If endometrial thickness is more than 4 mm, further invasive testing, by endometrial biopsy, is warranted. Pathology results can be: 1. (pre) malignancy; 2. benign; or 3. insufficient. If endometrial biopsy shows an insufficient sample, we confirmed in Chapter 7 that further testing by hysteroscopy is indicated because of an increased risk of endometrial carcinoma. If endometrial biopsy shows a benign result, the Dutch guideline is not clear about further diagnostic work up. The individual doctor can decide for expectant management or further invasive testing by hysteroscopy. To answer this clinical question, we designed the POMPOEN trial (effectiveness of saline-infused sonography and hysteroscopy in the work-up of women with PMB), which is currently still including patients. In this trial, women with PMB, an endometrial thickness > 4 mm and a benign endometrial biopsy are randomized to either expectant management or uterine cavity assessment (and if necessary treatment) with saline infused sonography (SIS) and hysteroscopy. Primary outcome of this study is recurrence of PMB after twelve months. If hysteroscopy for benign pathology is proven to be effective, we need to reconsider the added value of office endometrial biopsy in the work-up for women with PMB. Until the POMPOEN trial is finished, we would like to propose a diagnostic work-up presented in Figure

128 Chapter 9 Risk factors: - Diabetes Mellitus - Nulliparity - BMI > 26 kg/m2 - No anticoagulant use Woman with postmenopausal bleeding 50 years with up to one additional risk factor > 50 years No Transvaginal ultrasound Yes Endometrial thickness 4mm Endometrial thickness > 4 mm Reassure without further testing Endometrial biopsy Recurrent bleeding Failed biopsy (Pre) malignancy ruled out (Pre) malignancy Hysteroscopy Recurrent bleeding Treatment Figure 1. Future research The POMPOEN trial will answer the question whether or not treatment of benign endometrial pathology is effective in women with an endometrial thickness > 4 mm. The diagnostic flow chart shown in Figure 1, should be updated according to the results of this trial, and implemented in guidelines on women with PMB. Before changing current guidelines, it is important to know the implementation in clinical practice of the guideline at this moment. 126

129 Summary and General Discussion The Dutch guideline only recommends further testing when SIS or office-hysteroscopy is possible. However, most resections of endometrial polyps are still performed under general anesthesia. 21 By studying the variety in clinical practice of TVS, endometrial biopsy, SIS and hysteroscopy throughout our country insight is obtained in the implementation of current guidelines. Furthermore, the impact of reassurance of women with a low risk of cancer without further research needs to be evaluated after implementation in clinical practice

130 Chapter 9 References 1. Smith-Bindman R, Kerlikowske K, Feldstein VA et al. Endovaginal ultrasound to exclude endometrial cancer and other endometrial abnormalities. JAMA 1998;280(17): NVOG (Dutch Society of Obstetrics and Gynaecology). NVOG-richtlijn Abnormaal vaginaal bloedverlies in de menopauze [In Dutch]. NVOG Guideline: Abnormal vaginal bleeding during menopause Anderson KE, Anderson E, Mink PJ et al. Diabetes and endometrial cancer in the Iowa women s health study. Cancer Epidemiol Biomarkers Prev 2001;10(6): Gredmark T, Kvint S, Havel G, Mattsson LA. Histopathological findings in women with postmenopausal bleeding. Br J Obstet Gynaecol 1995;102(2): Gull B, Karlsson B, Milsom I, Granberg S. Factors associated with endometrial thickness and uterine size in a random sample of postmenopausal women. Am J Obstet Gynecol 2001;185(2): McPherson CP, Sellers TA, Potter JD, Bostick RM, Folsom AR. Reproductive factors and risk of endometrial cancer. The Iowa Women s Health Study. Am J Epidemiol 1996;143(12): Reeves GK, Pirie K, Beral V, Green J, Spencer E, Bull D. Cancer incidence and mortality in relation to body mass index in the Million Women Study: cohort study. BMJ 2007;335(7630): van Doorn LC, Dijkhuizen FP, Kruitwagen RF, Heintz AP, Kooi GS, Mol BW. Accuracy of transvaginal ultrasonography in diabetic or obese women with postmenopausal bleeding. Obstet Gynecol 2004;104(3): Weiderpass E, Persson I, Adami HO, Magnusson C, Lindgren A, Baron JA. Body size in different periods of life, diabetes mellitus, hypertension, and risk of postmenopausal endometrial cancer (Sweden). Cancer Causes Control 2000;11(2): Xu WH, Xiang YB, Ruan ZX et al. Menstrual and reproductive factors and endometrial cancer risk: Results from a population-based case-control study in urban Shanghai. Int J Cancer 2004;108(4): Clark TJ, Voit D, Gupta JK, Hyde C, Song F, Khan KS. Accuracy of hysteroscopy in the diagnosis of endometrial cancer and hyperplasia: a systematic quantitative review. JAMA 2002;288(13): Gupta JK, Chien PF, Voit D, Clark TJ, Khan KS. Ultrasonographic endometrial thickness for diagnosing endometrial pathology in women with postmenopausal bleeding: a meta-analysis. Acta Obstet Gynecol Scand 2002;81(9): Bachmann LM, Ter RG, Clark TJ, Gupta JK, Khan KS. Probability analysis for diagnosis of endometrial hyperplasia and cancer in postmenopausal bleeding: an approach for a rational diagnostic workup. Acta Obstet Gynecol Scand 2003;82(6): Opmeer BC, van Doorn HC, Heintz AP, Burger CW, Bossuyt PM, Mol BW. Improving the existing diagnostic strategy by accounting for characteristics of the women in the diagnostic work up for postmenopausal bleeding. BJOG 2007;114(1): Dijkhuizen FP, Mol BW, Brolmann HA, Heintz AP. The accuracy of endometrial sampling in the diagnosis of patients with endometrial carcinoma and hyperplasia: a meta-analysis. Cancer 2000;89(8): Clark TJ, Mann CH, Shah N, Khan KS, Song F, Gupta JK. Accuracy of outpatient endometrial biopsy in the diagnosis of endometrial cancer: a systematic quantitative review. BJOG 2002;109(3): Gordon SJ, Westgate J. The incidence and management of failed Pipelle sampling in a general outpatient clinic. Aust N Z J Obstet Gynaecol 1999;39(1): van Doorn HC, Opmeer BC, Burger CW, Duk MJ, Kooi GS, Mol BW. Inadequate office endometrial sample requires further evaluation in women with postmenopausal bleeding and abnormal ultrasound results. Int J Gynaecol Obstet 2007;99(2): Williams AR, Brechin S, Porter AJ, Warner P, Critchley HO. Factors affecting adequacy of Pipelle and Tao Brush endometrial sampling. BJOG 2008;115(8): Phillips V, McCluggage WG. Results of a questionnaire regarding criteria for adequacy of endometrial biopsies. J Clin Pathol 2005;58(4): van Dijk LJ, Breijer MC, Veersema S, Mol BW, Timmermans A. Current practice in the removal of benign endometrial polyps: a Dutch survey. Gynecol Surg 2012;9(2):

131 Nederlandse samenvatting Curriculum Vitae Dankwoord PhD Portfolio

132

133 Nederlandse samenvatting Nederlandse samenvatting Het doel van dit proefschrift is het evalueren van verschillende diagnostische strategieën voor vrouwen met postmenopauzaal bloedverlies (PMB), waaronder strategieën gebaseerd op individuele kansen en om de waarde van endometriumdikte meting bij asymptomatische postmenopauzale vrouwen te onderzoeken. In Hoofdstuk 2 presenteren we een systematische review over de bestaande predictiemodellen voor endometriumcarcinoom bij vrouwen met PMB. We hebben negen studies geïdentificeerd waarin de ontwikkeling en validatie van predictiemodellen in vrouwen met PMB wordt beschreven. Vanuit de data hebben we de belangrijkste voorspellers geïdentificeerd, deze kunnen opgedeeld worden in vier onderwerpen: Patiëntkarakteristieken, variabelen van transvaginale echo (TVE) met grijswaarden, variabelen van TVE met Doppler en hysteroscopie variabelen. De meeste predictiemodellen gebruikten een combinatie van deze onderwerpen voor het voorspellen van de kans op endometriumcarcinoom. Acht studies beschreven minstens één aspect van interne validatie, geen van de modellen werd extern gevalideerd. Modellen met daarin Doppler variabelen laten de beste voorspellende waarde zien in interne validatie maar vanwege de moeilijkheden van het gebruik van Doppler in de dagelijkse gynaecologische praktijk, concluderen we dat de beste modellen tot nu predictiemodellen zijn waarin patiëntkarakteristieken worden gecombineerd met endometriumdikte. In Hoofdstuk 3 evalueren we de kosteneffectiviteit van strategieën met daarin de diagnostische waarde van patiëntkarakteristieken in de vorm van predictiemodellen voor endometriumcarcinoom door middel van een besliskundige aanpak. We hebben een beslismodel ontwikkeld waarin vier diagnostische strategieën voor vrouwen met PMB worden vergeleken: (1) de patiëntkarakteristieken strategie, waarin de kans op endometriumcarcinoom wordt geschat op basis van patiëntkarakteristieken en histologische verificatie wordt verricht in het geval van een kans hoger dan 4%. In deze strategie wordt geen endometriumdikte meting verricht. (2) De sequentiële strategie, met een schatting van de kans op endometriumcarcinoom op basis van patiëntkarakteristieken en een endometriumdikte meting bij een kans meer dan 4%. Histologische verificatie wordt vervolgens verricht bij een endometriumdikte van meer dan 4 mm. (3) De geïntegreerde strategie, waarin bij alle vrouwen een endometriumdikte wordt gemeten en een kansschatting wordt gemaakt op basis van de endometriumdikte en patiëntkarakteristieken gecombineerd. Hierna wordt een histologische verificatie verricht indien de kans groter is dan 4%. De drie strategieën gebaseerd op individuele kansen werden vergeleken met een strategie waarin alleen endometriumdikte wordt gemeten, in overeenstemming met de huidige praktijk. Strategie (1), het selecteren van patiënten voor histologische verificatie 131

134 Nederlandse samenvatting op basis van alleen patiëntkarakteristieken, domineert alle andere strategieën (effectiever en goedkoper). In een klinisch scenario waarbij wordt aangenomen dat de TVE integraal onderdeel uitmaakt van het consult en geen extra kosten met zich meebrengt, wordt een strategie waarbij patiënten met een hoog risico worden geselecteerd voor TVE (de sequentiële strategie) het meest kosten effectief. Voor een succesvolle implementatie, moet een predictiemodel extern worden gevalideerd in een onafhankelijke populatie. Hoofdstuk 4 beschrijft de externe validatie van twee eerder ontwikkelde predictiemodellen voor endometriumcarcinoom bij vrouwen met PMB: het patiëntkarakteristieken model en het patiëntkarakteristieken en TVE model. De modellen werden toegepast op twee onafhankelijke databases verzameld in Nederland en Zweden. In zowel de Nederlandse als de Zweedse database had het patiëntkarakteristieken en TVE model een AUC van 0.89 (95% CI /92). Deze was hoger dan de AUC van het patiëntkarakteristieken model welke respectievelijk 0.71 (95% CI ) en 0.69 (95% CI ) was. De AUC voor alleen TVE was 0.87 in de Nederlandse database (95% CI ) en 0.90 in de Zweedse database (95% CI ). Deze studie laat zien dat na externe validatie in twee onafhankelijke databases, de bestaande multivariabele modellen hun diagnostische waarde houden en in staat zijn te differentiëren tussen vrouwen met een hoog en een laag risico op endometriumcarcinoom bij vrouwen met PMB. Het patiëntkarakteristieken en TVS model heeft echter geen meerwaarde boven het meten van endometriumdikte alleen. Het patiëntkarakteristieken model is in staat vrouwen met een laag risico op endometriumdikte te selecteren, deze vrouwen kunnen worden gerustgesteld zonder verder onderzoek. Dit is voornamelijk bruikbaar in een setting waarin TVE niet (direct) beschikbaar is. Minimaal invasieve endometriumbiopsie mislukt in een substantieel aantal gevallen door technische problemen of omdat er onvoldoende materiaal voor een histologische diagnose wordt verkregen. In Hoofdstuk 5 onderzoeken we welke dokter en patiënt gerelateerde factoren bijdragen aan het mislukken van een poliklinisch endometriumbiopt bij vrouwen met PMB en een endometriumdikte > 4 mm. Nullipariteit en hogere leeftijd waren geassocieerd met het mislukken van het biopt en een endometriumdikte > 12 mm verminderde de kans op mislukken. Opleidingsniveau van de dokter was niet geassocieerd met technisch falen of te weinig materiaal. Een predictiemodel voor het mislukken van een endometriumbiopt werd vervolgens ontwikkeld. De AUC van het model was 0.64 (95% CI ) wat duidt op een matige capaciteit om te differentiëren tussen patiënten met een hoge en een lage kans op een mislukt biopt. 132

135 Nederlandse samenvatting Hoofdstuk 6 evalueert het gebruik van het predictiemodel voor een mislukt endometriumbiopt bij vrouwen met PMB door middel van een kostenminimalisatie analyse. Wanneer de voorspelde kans op mislukken meer dan 62% is, is het goedkoper om een patiënte direct voor hysteroscopie met biopsie te verwijzen in plaats van een poging tot het doen van een minimaal invasief endometriumbiopt. Dit geeft echter een kostenreductie van drie euro per patiënt en daarom concluderen we dat het onwaarschijnlijk is dat het individualiseren van de beslissing om een minimaal invasief endometriumbiopt of een hysteroscopie met biopt te doen bij patiënten met PMB de efficiëntie van de work-up gaat verbeteren. Er zijn geen richtlijnen voor een gestandaardiseerde beoordeling van endometriumbiopten wat resulteert in een diagnose die is beïnvloed door subjectiviteit en veel variatie tussen de verschillende pathologen. Het doel van Hoofdstuk 7 was onderzoeken of gestructureerde beoordeling van endometriumbiopten met behulp van strikte criteria het aantal niet conclusieve biopten als gevolg van te weinig materiaal kan verminderen. We hebben 66 endometriumbiopten uit acht verschillende ziekenhuizen opgevraagd waarvan er 36 werden verkregen. Deze biopten werden verzameld tijdens de normale diagnostische work-up voor vrouwen met PMB en waren beoordeeld als onvoldoende materiaal voor betrouwbare diagnose. Gestructureerde herbeoordeling van de verkregen biopten door een patholoog gespecialiseerd in gynaecologie veranderde de conclusie niet in 35 van de 36 biopten. Eén biopt bevatte een ruime hoeveelheid endometriumweefsel en de conclusie na herbeoordeling was endometrium hyperplasie zonder atypie. Alle andere biopten bevatten onvoldoende materiaal voor betrouwbare diagnose. Onze bevindingen suggereren dat, hoewel het voor pathologen behulpzaam zou zijn criteria te hebben voor het beoordelen van de adequaatheid van een endometriumbiopt, de te verwachten winst in efficiëntie hiervan klein is. We vinden daarom dat, om de efficiëntie van minimaal invasieve endometriumbiopten te vergroten, er moeite moet worden gedaan om zoveel mogelijk weefsel te verkrijgen bij de afname om zo de kans op een mislukt biopt door te weinig materiaal te verkleinen. Meting van endometriumdikte is een belangrijk instrument bij het onderzoeken van patiënten met PMB. De relevantie van een echoscopisch gemeten endometriumdikte bij asymptomatische postmenopauzale vrouwen is onbekend. Hoofdstuk 8 beschrijft een systematische review en meta-analyse met drie hoofdvragen: (I) de normale endometriumdikte gemeten met TVE, (II) de prevalentie van maligne en premaligne endometriumpathologie en (III) de sensitiviteit en specificiteit van endometriumdikte meting voor het diagnosticeren van maligne en premaligne endometriumpathologie bij asymptomatische postmenopauzale vrouwen. We hebben 32 studies geïncludeerd waarin 11,100 vrouwen werden beschreven om deze drie vragen te beantwoorden: (I) de geschatte 133

136 Nederlandse samenvatting gemiddelde endometriumdikte was 2.9 mm (95% CI ). De gepoolde prevalentie van endometriumcarcinoom en atypische hyperplasie was respectievelijk 0.62% (95% CI ) en 0.59% (95% CI ). Gewogen schattingen voor de sensitiviteit en specificiteit van endometriumdikte meting waren 0.83 (95% CI ) en 0.72 (95% CI ) voor afkapwaarde van 5 mm en 0.33 (95% CI ) en 0.94 (95% CI ) voor een afkapwaarde van 6 mm. De resultaten van deze systematische review rechtvaardigen niet het gebruik van endometriumdikte meting als een screeningstest voor endometriumcarcinoom en atypische hyperplasie bij asymptomatische postmenopauzale vrouwen zonder hormoonvervangende therapie. 134

137 Curriculum Vitae Curriculum Vitae Maria Cornelia (Merel) Breijer werd op 6 december 1982 geboren in Roosendaal. Ze groeide op in Grijpskerke. Vanaf haar 16 e woonde ze in Zeist waar ze in 2001 haar VWO diploma behaalde aan de Stichtse Vrije School in Zeist. Datzelfde jaar startte ze met haar studie geneeskunde aan de Erasmus Universiteit in Rotterdam. In 2007 startte ze als ANIOS in het TweeSteden Ziekenhuis in Tilburg op de afdeling gynaecologie en verloskunde. Na anderhalf jaar begon ze voor de helft van haar tijd als onderzoeker in het AMC, het eerste jaar op de afdeling klinische epidemiologie en later op de afdeling gynaecologie en verloskunde. Vanaf oktober 2011 is ze in opleiding tot gynaecoloog in het Albert Schweitzer Ziekenhuis Dordrecht (opleider Mw. dr. G.S. Kooi). Vanaf oktober 2013 zal ze de opleiding voortzetten in het Erasmus MC Rotterdam (opleider Mw. dr. M.J. ten Kate-Booij). Merel woont samen met Peter in Utrecht. 135

138

139 Dankwoord Dankwoord Onderzoek doe je niet alleen. Ik wil iedereen die mij in de afgelopen jaren op enige manier heeft bijgestaan en de mensen die mij de kans hebben gegeven dit onderzoek te doen hiervoor bedanken. Een aantal personen wil ik in het bijzonder noemen. Allereerst de patiënten die hebben deelgenomen aan de POMPOEN studie. Hoewel de studie zelf dit proefschrift niet heeft gehaald, is de bijbehorende database de basis van een heel aantal hoofdstukken in dit proefschrift. Mijn promotoren Ben Willem Mol en René Verheijen en mijn co-promotoren Anne Timmermans en Brent Opmeer. Ben Willem, buiten de reguliere wegen om heb je me de kans gegeven dit onderzoek te doen. Bedankt voor je vertrouwen en heldere blik. Je inschattingsfout wat betreft de beste club van Nederland vergeef ik je. Veel geluk met je nieuwe avontuur. René, iets later ben je betrokken geraakt bij mijn onderzoek. Je hebt ervoor gezorgd dat we het zicht op de kliniek niet zijn kwijtgeraakt. Bedankt voor het snelle en relevante commentaar op de manuscripten. Anne, vanaf mijn eerste dag als arts-assistent in Tilburg was je mijn collega, jij was toen nog bezig met het afronden van je promotie. Ik heb je ooit gevraagd hoe het mogelijk was om te promoveren naast de opleiding en dan ook nog een sociaal leven te hebben. Inmiddels weet ik het antwoord. Bedankt voor alles wat ik van je heb geleerd, zowel in de kliniek als in het onderzoek. Brent, in het begin van mijn onderzoek had ik een plekje op de KEBB, bijna recht tegenover je kamer. Eigenlijk alles wat ik in de afgelopen jaren heb geleerd over statistiek, epidemiologie en kosteneffectiviteit heb ik van jou geleerd. Het was altijd mogelijk om even samen ergens naar te kijken als dat nodig was. Bedankt daarvoor. De leden van de promotiecommissie, prof. dr. J. Stoker, prof. dr. M.J. van de Vijver, prof. dr. M.P.M. Burger, dr. W.M. Ankum, prof. dr. C.W. Burger en prof. dr. R.F.P.M. Kruitwagen wil ik bedanken voor het lezen en beoordelen van mijn manuscript en dat zij zitting hebben willen nemen in mijn promotiecommissie. Onderzoekscollega s van de KEBB en de gynaecologie. Bedankt voor de gezellige praatjes en lunch, onmisbare hulp bij eigenlijk alles wat met onderzoek doen te maken heeft en voor de 137

140 Dankwoord Utrechtse AMC-ers borrels op vrijdagmiddag. Al was ik er altijd op onregelmatige dagen en tijden als buitenbeentje, ik heb me erg welkom gevoeld. Ik zal het AMC gaan missen. Onderzoekers en researchmedewerkers van het consortium: ik bewaar goede herinneringen aan het kastelenweekend in Frankrijk. Het is fijn om bekenden tegen te komen overal waar ik kom. De gynaecologen en onderzoekers van het benigne gynaecologie consortium wil ik met name bedanken voor de samenwerking, nu en in de toekomst, de POMPOEN inclusie vordert gestaag maar gaat zeker zijn einde bereiken. Nu de rest van de studies nog. Lennie, wat fijn dat jij ook als onderzoeker bij de POMPOEN betrokken bent, we gaan het samen afmaken! Heel erg bedankt voor je hulp de afgelopen tijd. Maya, Zelda, Cynthia en Cornelieke: bedankt voor jullie ondersteuning en enthousiasme en natuurlijk jullie gastvrijheid na mijn verhuizing naar H4. Ingrid en Marjan, tijdens mijn onderzoek heb ik veel aan jullie hulp gehad, daar wil ik jullie voor bedanken. Tessa, bedankt voor de ondersteuning vanuit het UMC. Oud-collega s van het TweeSteden Ziekenhuis (en met collega s bedoel ik natuurlijk ook de verpleegkundigen, verloskundigen, assistentes en secretaresses). Bij jullie heb ik mijn eerste stapjes als arts-assistent gezet en jullie hebben me de kans gegeven om onderzoek te combineren met de kliniek. Zonder jullie was dit proefschrift er zeker niet geweest. Het voelde bijna als een tweede thuis en nog steeds noem ik wel eens het verkeerde ziekenhuis als ik de telefoon opneem. Bedankt voor alles. Collega s van het Albert Schweitzer Ziekenhuis. In 2011 mocht ik bij jullie beginnen met de opleiding. Bedankt voor het warme welkom. De afdeling is ontzettend opleidingsgericht en ik voel me bevoorrecht dat ik hier mag worden opgeleid. Het is niet altijd makkelijk geweest om de kliniek met het onderzoek te combineren maar dat heeft zeker niet aan jullie gelegen. Bedankt voor de flexibiliteit en ondersteuning. Lieve Tamar en Liesbeth, ik heb een tijd gedacht dat ik gynaecologie zo leuk vond omdat ik het samen met jullie deed. Vanaf het begin was duidelijk dat als ik ooit zou promoveren jullie naast me zouden staan. Bedankt voor de fijne vriendschap. Lieve familie en schoonfamilie, dank voor jullie oprechte interesse in mijn onderzoek en 138

141 Dankwoord begrip als ik ergens weer eens niet bij kan zijn vanwege mijn werk. Ik ben bang dat ik hierin geen beterschap kan beloven. Mama, je straalt een onvoorwaardelijk vertrouwen in ons uit en ik weet dat je trots bent op ons allemaal. Ik vind het jammer dat André er vandaag niet bij kan zijn, al zou hij waarschijnlijk al weken van tevoren niet meer slapen van de zenuwen. We missen hem. Papa en Hetty, bedankt voor de nodige afleiding tijdens feestjes en feestdagen. Jullie zijn een geweldige gastvrouw en gastheer. Jullie hebben geen makkelijke tijd achter de rug en het einde daarvan is misschien ook nog niet in zicht. Nu papa is gestopt met werken hebben jullie meer tijd om samen door te brengen, geniet er van. Lieve broers en zussen, jullie zijn er in alle soorten en maten en jullie zijn allemaal op je eigen manier belangrijk voor me. Bedankt daarvoor. Lieve Peter, ik wil nog heel lang samen genieten. Ik hou van je. Merel 139

142

143 PhD Portfolio PhD Portfolio Name PhD student: M.C. Breijer PhD period: September 2009 June 2013 Name PhD supervisor: Prof. dr. B.W.J. Mol Prof. dr. R.H.M. Verheijen 1. PhD training General courses - AMC World of Science - Evidence based medicine - Principles of research in medicine (NIHES summer programme) - Expert management of medical literature - BROK: Basic course in Legislation and Organisation for clinical Researchers - Practical Biostatistics Specific courses - Systematic reviews, Dutch Cochrane Centre - Clinical decision analysis (NIHES summer programme) - Prognostic research (MSC epidemiology UMC Utrecht) - Methods for Evaluating Medical tests and biomarkers (pre-conference course) Seminars, workshops and master classes - Weekly seminar department of clinical epidemiology, biostatistics and bioinformatics - Weekly scientific meeting PhD students Obstetrics and Gynaecology Year Workload (Hours/ECTS)

144 PhD Portfolio Presentations - Patiëntkarakteristieken bij patiënten met postmenopauzaal bloedverlies: een modelmatige aanpak. Poster presentation, NVOG Gynaecongres. - Validation and refinement of a model to predict risk of endometrial cancer in patients with postmenopausal bleeding. Oral presentation, Methods for Evaluating Medical Tests and Biomarkers. - Endometrial thickness and the prevalence of endometrial carcinoma in asymptomatic postmenopausal patients, a systematic review of the literature. Oral presentation, 20th World Congress on Ultrasound in Obstetrics and Gynecology. - Estimating probability of cancer in patients with postmenopausal bleeding prior to ultrasonography. Validation and refinement of a prognostic model. Oral presentation, 20th World Congress on Ultrasound in Obstetrics and Gynecology. - Diagnostiek naar endometriumcarcinoom in patiënten met postmenopauzaal bloedverlies, kosten effectiviteit van strategieën gebaseerd op individuele patiëntkenmerken. Poster presentation, NVOG Gynaecongres. (Inter)national conferences - NVOG gynaecongres 10 en 11 juni 2009, de Galgewaard Utrecht. - Methods for evaluating medical tests and biomarkers 1th and 2nd July 2010, Medical School, University of Birmingham - NVOG gynaecongres 12 en 13 november 2009, Hotel en Congrescentrum Papendal, Arnhem. - 20th World Congress on Ultrasound in Obstetrics and Gynecology, 10 to 14 October 2010, Prague Congress Center, Prague. - NVOG gynaecongres 11 en 12 november 2010, Hotel en Congrescentrum Papendal, Arnhem. - NVOG gynaecongres 10 en 11 november 2011, Hotel en Congrescentrum Papendal, Arnhem. - NVOG gynaecongres 10 en 11 mei 2012, World Forum, Den Haag. Year Workload (Hours/ECTS) Other - Journal Club

145 PhD Portfolio 2. Teaching Workload Year (Hours/ECTS) Supervising - Nicole Visser, Afstudeeronderzoek Geneeskunde Other - Lecture Women s visiting gynaecological club - Review for European Journal of Obstetrics & Gynecology and Reproductive Biology - Review for Gynecologic and Obstetric Investigation

146

147