Predicting Positive Margins in Resection of Cutaneous Melanoma of the Head and Neck

Transcription

1 The Laryngoscope VC 2013 The American Laryngological, Rhinological and Otological Society, Inc. Predicting Positive Margins in Resection of Cutaneous Melanoma of the Head and Neck J. Jared Christophel, MD, MPH; Andrew K. Johnson, BS; Timothy L. McMurry, PhD; Stephen S. Park, MD; Paul A. Levine, MD Objectives/Hypothesis: Head and neck melanoma surgeons must achieve negative margins before performing margin compromising reconstructions such as a local flap closure. This often necessitates staged operations, including further margin resection. Peripheral sampling is often used before definitive resection to help guide the extent of the resection. If melanoma margin status could be predicted based on lesion characteristics, the surgeon could be more confident in performing definitive closure immediately after resection of some lesions or confident in the need to take larger margins in predictably extensive lesions. Study Design: Retrospective review and logistic regression analysis. Methods: Institutional review board approval was obtained. Out of 637 patients treated for head and neck melanoma by the Department of Otolaryngology Head and Neck Surgery in the last 10 years, 409 patients had primary resection with available histopathologic margin status used as the outcome variable. Predictor variables of demographics, lesion size, pathologic subtype, location on face, and depth of invasion were collected. Results: Histopathologic margin status could be predicted by age but not by the other predictor variables. Conclusions: In this large series of head and neck melanomas excised using National Comprehensive Cancer Network recommended margins, histopathologic margin status could be predicted based on age but not on lesion characteristics. This finding is surprising given the published data showing that melanoma in situ has a higher rate of positive margin compared to subtypes of invasive melanoma. It reinforces the need for delaying reconstruction until margins are clear or performing reconstruction at a time of resection that does not compromise the ability to resect margins further (e.g., skin graft). Key Words: Melanoma, head and neck, cutaneous melanoma, margins, reconstruction. Level of Evidence: 2c. Laryngoscope, 123: , 2013 INTRODUCTION The incidence of cutaneous melanoma has been increasing over the past three decades and was most recently reported at 20.8 cases per 100,000 people. 1 Of these lesions, 25% to 35% present on the head and neck region as the primary site. 2 Although various immunotherapy modalities have been developed in the past decade, the gold standard treatment for all primary cutaneous melanomas continues to be surgical. 3 5 The current National Comprehensive Cancer Network (NCCN) surgical margin guidelines for initial definitive excision of cutaneous melanoma are shown in Table I. 3 From the Department of Otolaryngology Head and Neck Surgery (J.J.C., S.S.P., P.A.L.) and the Department of Health Evaluation Sciences (J.J.C., T.L.M.), University of Virginia Health System, Charlottesville, the University of Virginia School of Medicine (A.K.J.), Charlottesville, Virginia, U.S.A. Editor s Note: This Manuscript was accepted for publication September 24, Presented at the Triological Society Annual Meeting at COSM, San Diego, California, U.S.A., April 18 22, The authors have no funding, financial relationships, or conflicts of interest to disclose. Send correspondence to J. Jared Christophel, MD, Department of Otolaryngology Head and Neck Surgery, University of Virginia Health System, PO Box , Charlottesville, VA jjc3y@virginia.edu DOI: /lary Although these margins are often readily achieved on the trunk and closed primarily, similar resections on the head and neck are more difficult due to anatomic constraints. Fear of disfigurement or loss of function of critical structures may lead the surgeon to take more narrow margins, resulting in incomplete excision. 6 Involved margins may leave residual tumor cells in the wound and incur a higher likelihood of locoregional recurrence or metastasis. 7 Common practice has evolved to delay reconstruction of the excisional defect until negative margins are confirmed with histopathologic staining. This requires a second surgery and maintenance of a large wound in the interim. In addition, the 0.5-cm margin recommended for melanoma in situ (MIS) leaves 14% to 50% of patients with positive margins requiring re-excision Some centers have used peripheral sampling techniques for MIS or large melanomas to ensure clear margins before definitive excision, and others have used the Mohs technique. 9,11,12 Even with peripheral sampling, many surgeons will still delay reconstruction (meaning 3 surgeries) until final pathology review of the primary specimen because finding a melanoma within an MIS (up to 12%) or a deeper melanoma than originally biopsied may require wider excision or nodal biopsies. 6,10 Although more widely accepted, the delayed reconstruction approach incurs significant morbidity in patients, 683

2 TABLE I. Principles of Surgical Margins for Wide Excision of Primary Melanoma. Tumor Thickness, mm Recommended Clinical Margins, cm In situ >4 2.0 National Comprehensive Cancer Network Melanoma Guidelines and some centers utilize immediate reconstruction in selected cases. 13,14 There are minimal survival data available for this technique, although the positive margin rate and local recurrence rate in these selected cases remains low. 15 The objective of our study was to define a set of predictive factors for a positive histopathologic margin for melanoma of the head and neck when following NCCN surgical margin guidelines. This information could help surgeons determine which reconstructions could possibly be performed immediately. MATERIALS AND METHODS Study Design Retrospective review and logistic regression analysis. Approval was obtained from the University of Virginia Institutional Review Board for Health Sciences Research. All patients who underwent surgery for cutaneous melanoma at the University of Virginia Department of Otolaryngology Head and Neck Surgery (UVA OTO-HNS) between February 2001 and June 2011 were included in a data set and analyzed for factors predictive of positive histopathologic margin after primary resection. Patients. A total of 637 patients were identified from the Clinical Data Repository at the University of Virginia who had a cutaneous melanoma resected by surgeons at the UVA OTO- HNS. After we excluded patients for whom the pathology report was unavailable, primary definitive excision was performed elsewhere, or cutaneous excision was performed for reason other than primary resection (recurrence or cutaneous metastasis), 412 patients remained. Surgical Resection Surgical margins depended on the depth of invasion (DOI), with margins taken according to NCCN guidelines, which have remained constant over the time period studied (Table I). 3,16 20 Surgical margin was recorded from operative reports when available. In a small subset of patients (n ¼ 8), the recommended margins could not be achieved because of anatomic constraints (e.g., nasal ala or eyelid). Statistical Analysis Hypothesis. We hypothesized that positive histopathologic margin is predictable based on patient and lesion characteristics. Predictor variables. Variables hypothesized to correlate with histopathologic margin status included age, pathologic subtype, location on face, DOI, and lesion diameter. These data were collected from patient charts and recorded as missing if unavailable. Outcome variable. Histopathologic margin was used as the dependent variable and was determined from the pathology report. A histopathologic margin was considered positive if tumor was present <1.0 mm from the edge of the specimen. Logistic regression. The data consisted of 412 subjects, three of whom were missing pathologic margin data, leaving 409 subjects for analysis. The data were randomly split into a training data set of 300 and a test set of 109. The variables considered as possible predictors of positive margin were an indicator for MIS, DOI, an indicator for tumor on the scalp, log size of lesion, and age on the date of surgery. The predictors were individually screened for significance on the training data set using logistic regression, except that the MIS indicator and DOI were combined for the initial screening. R version (R Foundation for Statistical Computing, Vienna, Austria. ISBN ; was used to perform the logistic regression and formulate the receiver operating characteristic (ROC) curve. RESULTS Descriptive Statistics and Frequencies Table II summarizes the data set by margin status stratified according to the nominal variables of melanoma subtype and location on the head and neck. Table III summarizes margin status by the scale variables of DOI, lesion size, and patient age. Table IV summarizes margin status by the ordinal variables of tumor stage and clinically relevant lesion size. MIS was the most common type of melanoma resected, followed by lentiginous melanoma. The cheek is the most common location treated followed by ear and then scalp. T1 is the most common invasive tumor stage treated, with frequency decreasing as tumor stage increases. TABLE II. Margin Status by Nominal Variables (Subtype and Subsite). Frequency (þ) Margin (%) Subtype MIS (12) Lentiginous (16) SS 66 6 (9) Nodular 17 1 (6) Amelanotic 3 2 (67) NOS 92 4 (4) Spindle 21 5 (24) Subsite Nose 39 6 (15) Eyelid 12 2 (17) Cheek (14) Ear 79 5 (6) Neck 52 5 (10) Forehead 27 5 (19) Lips 2 0 (0) Temple 25 3 (12) Scalp 67 7 (10) MIS ¼ melanoma in situ; NOS ¼ not otherwise specified; SS ¼ superficial spreading. 684

3 TABLE III. Margin Status by Scale Variables (Depth of Invasion and Age). Mean DOI, mm 1.89 * 1.89 Size, mm Age, yr *Includes only invasive melanoma subtypes (melanoma in situ excluded). DOI ¼ depth of invasion; SD ¼ standard deviation. SD These descriptive characteristics of the data set were important to confirm, showing that the data comprise a representative sample similar to what is treated elsewhere in terms of tumor type, tumor size, and patient demographics. 10,13 TABLE IV. Margin Status by Ordinal Variables (Tumor Stage, Lesion Size). Frequency (þ) Margin (%) Tumor stage Tis (12) T (13) T (10) T (12) T (9) Missing 4 1 (25) Lesion size, mm (17) (12) (17) > (11) Missing (3) Fig. 1. Receiver operating curve (ROC) for logistical model fit to entire data set. Predicting Positive Histopathologic Margin Of the individual predictors listed here that were screened on the training data set, only age was statistically significant (P ¼.0166 on the training set). We then attempted to assess how well a model relying on age is able to predict a positive margin. Using a logistic model containing only age, the area under the ROC curve was on the training data set, and when tested on the held-out test data set, gave an area under the ROC curve of 0.555, suggesting that age has a positive but limited predictive power. A final logistic model was then refit on the entire data set, resulting in the model: Ln odds of positive margin ¼ þ [Age at Surgery], where the age coefficient in the final model is significant at P ¼ Using this model on the entire data set, the area under the ROC curve was (Fig. 1). To illustrate the changing likelihood of positive margin with age, we split the data into approximate quartiles by age and looked at the percentage positive histopathologic margins in each quartile. (Note: one patient s age was missing, so this record was excluded.) The percentages of positive margins are given in Table V. DISCUSSION This data set provided a large number of patients for an outcomes study. An infrequent outcome variable requires a large data set to reliably study multiple predictor variables. The outcome variable positive histopathologic margin after resection of cutaneous melanoma has been reported to be between 6% and 24% in other studies. 10,13 The overall positive histopathologic margin rate in this study was 11.7% (48 of 412) for all tumors treated. A formal power analysis was not performed because in the case of multiple logistic regression, the power depends strongly on the relationship between individual predictors, which is not possible to meaningfully estimate in advance. There is a statistical rule of thumb that recommends 10 events per predictor of interest. 21 With almost 50 outcomes of interest, it allowed us to reliably investigate five predictor variables. We further attempted to ensure reliable results by first randomly splitting the data into training and test sets, estimating effects on the training set only, and then confirming the results on the previously untouched data; this is standard practice. TABLE V. Margin Status by Age. Age, yr No. of Patients No. Positive Margins Percent Positive þ

4 The predictor variables were chosen based on a combination of previously published reports and clinical suspicion. It has been noted that MIS has an unusually high rate of positive histopathologic margin compared to invasive subtypes of melanoma. 9,10,22 Based on the high published rate of positive margins after resection of MIS and our clinical experience, we chose to look at the various melanoma subtypes including MIS as a predictor of positive margins. It was interesting to note that MIS failed to show a higher rate of positivity than invasive subtypes. As noted in the surgical technique section, planned margins for MIS resection were 5 mm, and based on published reports with this margin, the expected margin positivity rate would be 14% to 50% The 12% positive margin rate in 114 MIS patients is distinctly less than is often posited when taking 5-mm margins. The low positivity rate in the MIS group could be due to the slightly smaller lesion size in this sample ( mm in greatest diameter) compared to the >2-cm lesions in other studies. 10 However, that explanation assumes a higher positivity rate with larger lesions, which this study failed to prove. In addition, many of the studies quoting higher positivity rates (or larger than 5-mm margins needed for clearance) utilized Mohs technique to assess the entire margin periphery. 9,22,23 Margins for the lesions in this study were assessed by a combination of transverse vertical sectioning and en face margins, depending on the pathologist. Transverse vertical sections may not assess the entire margin periphery, resulting in an elevated false-negative rate. 24 To assess whether this is the case, we will need to examine longterm survival data including local recurrence from this data set. Regardless, using standard pathologic techniques, these data confirm the NCCN guidelines for 0.5- cm margins in resecting MIS. It has also been noted that DOI correlates with positive histopathologic margin when the same size margin is taken. 10,13 This clinical tenet has been the foundation for the various guidelines regarding surgical margins in resecting melanoma; the deeper the invasion, the larger the margin needed. The appropriate surgical margin for melanomas of various depths of invasion is an important clinical topic with a great deal of evidence behind the recommendations. 3,7,17,25 30 Those recommendations were inherently incorporated into this study as NCCN margin guidelines were followed at resection. Therefore, investigating whether DOI correlated with positive margin was essentially controlled for by taking the recommended larger margins for deeper tumors. The relatively constant margin positivity rate with increasing DOI and lack of DOI as a statistically significant predictor are therefore expected when all DOIs are studied in this data set. This result helps to verify the current NCCN recommendations. To study whether DOI correlates with margin positivity, each level of surgical margin would have to be studied separately (i.e., by evaluating margin positivity within intermediate-depth lesions excised with a 1- vs. 2-cm margin). This study does not have enough patients to perform such a risk adjustment. It has also been noted that lesion diameter correlates with positive margin. 10 Our data set did not reveal an increase in lesion diameter correlating with positive margin, although again this assessment would be even stronger if analysis could be risk adjusted by margin of resection. Clinical experience led us to study subsite of the head and neck as a possible predictor of positive margin, as scalp and forehead lesions seemed to have a higher rate of positive margin and need for re-excision. However, analysis of our data showed no statistical difference between subsites in terms of positive margin rate when analyzed using logistic regression. As a methods check, a simple contingency table was made to study the site with the highest positivity rate (forehead, 19%) compared to all the rest, and a Fischer exact test performed. We still failed to show a statistically significant difference, confirming the findings of the logistic regression that subsite does not predict positive margin after resection. Of the predictor variables studied, the least relevant variable from a clinical experience standpoint was patient age. However, the logistic regression analysis proved this to be the only significant predictor of margin positivity. The notable aspect of this conclusion is not an extremely high margin positivity rate in patients older than 75 years (15.6%þ), but the relatively low positivity rate in patients 55 and younger (4.6%). This rate of 4.6% in the youngest quartile of the data set is lower than the lowest rate cited by groups performing immediate reconstruction (6%). 13 One possible explanation for the finding that age is a statistically significant predictor of margin positivity is its role as a confounder for an unstudied variable. If age were studied alone, one would question if any noted correlation was possibly confounded by increasing tumor burden with age (increasing DOI or tumor diameter). However, those are controlled for by including them in the logistic regression. This can also be studied by looking at descriptive data by age as shown in Table VI. Note similarity of melanoma, except for a slightly larger diameter melanoma in the older patients (P ¼.05). Recall that tumor diameter, when tested independently, did not correlate with positive margin status. Something possibly not controlled for, yet inherently related to age, is the concept of multifocal disease; there is a higher likelihood of adjacent sun-damaged skin containing melanoma as the patient ages. Controlling for multifocal disease would be difficult unless it was routinely specified on the pathology report. Our pathology reports do not routinely note whether marginal disease was from a separate focus or extension of the same tumor. Therefore, we are left to view age as either a surrogate of an unknown confounder such as multifocal disease or a small but significant predictor of margin positivity. The data set had similar rates of overall positive margins when compared to the literature and similar distribution of tumor stages, tumor depths, and tumor size, indicating that the statistical conclusions from this study apply to the larger population. One aberrant characteristic of the data was the relatively low proportion of 686

5 TABLE VI. Dataset Description by Age. Age, yr No. of Patients DOI, mm* SD Size, mm SD Subtype No. (%) Location % < MIS 80 (26) Cheek 0.25 Invasive 226 (74) Ear 0.20 Scalp þ MIS 33 (32) Cheek 0.30 Invasive 69 (68) Ear 0.19 Scalp 0.13 *Independent samples t test, P ¼.430. Independent samples t test, P ¼.049. v 2 test of proportions for MIS, P ¼.131. v 2 test of proportions for location P ¼.502. DOI ¼ depth of invasion; MIS ¼ melanoma in situ; SD ¼ standard deviation. tumors that were MIS; a sample statistic of 28% compared with a parameter of 45% in the population. 10 This likely represents a slight selection bias for more invasive melanomas referred to the OTO-HNS department, compared to MIS, which is more commonly resected by the dermatologists in our region. However, given the similar positive margin rates compared with invasive melanomas, a larger MIS cohort would not change the conclusion. Had there been a distinct difference in positive margin rates between MIS and invasive melanomas, the smaller MIS cohort might have led to a type I error. In summary, the unexpected findings from this outcomes project were 1) age as a statistically significant predictor of margin positivity, with patients younger than 55 having a very low positivity rate, 2) the lack of higher margin positivity in MIS lesions, 3) the lack of higher positivity with larger diameter lesions, 4) the lack of higher margin positivity with any specific location on the head and neck, and 5) the similar rates of margin positivity with increasing DOI for invasive melanomas, confirming the NCCN resection guidelines. There must be a balance in resection of head and neck cutaneous melanoma. To clear 99% of tumors histopathologically, the initial margins would be 0.9 cm for MIS and larger for invasive disease. 9 This practice would incur a significant increase in functional and cosmetic disability in patients who would have had clear margins with smaller resection. Therefore, until a definitive Mohs-type technique that can be incorporated with sentinel node biopsy is developed, a certain low rate of margin positivity is acceptable. CONCLUSION In a logistic regression model of melanoma margins, increasing patient age renders the margins more likely to be positive after initial definitive resection. Melanoma subtype, DOI, location on the face, and lesion diameter do not have a statistically significant bearing on the likelihood of margin positivity. Resection of cutaneous head and neck melanoma remains fraught with a small, constant, and relatively unpredictable positive margin rate. This should lend caution to the surgeon performing an immediate, margin-compromising reconstruction. The surgeon should also be hesitant to compromise surgical margin width in older patients and should give consideration to more aggressive margins in hopes of performing a single resection. These results also prompt consideration of future evaluation of survival data in elderly melanoma patients. Although they may have higher rates of positive margin at the initial resection, we assume they ultimately achieve negative pathologic margin status by further resection. However, if the tumors in the elderly are more aggressive, it would show in evaluation of stage-matched disease-specific survival. Positive margin has been shown to correlate with local recurrence, which infers decreased survival. 6,31 The next feasible study from this data set will examine survival analysis as predicted by distance to clear histopathologic margin. This could help to provide data for histopathologic margins needed to survive melanoma, as current recommendations merely state clear is all that is needed. Acknowledgment The authors thank Dane Barrett, MD, for assistance with the data collection. BIBLIOGRAPHY 1. National Cancer Institute. SEER Stat Fact Sheets: Melanoma of the skin. Available at: Accessed March 25, Gomez-Rivera F, Santillan A, McMurphey AB, et al. Sentinel node biopsy in patients with cutaneous melanoma of the head and neck: recurrence and survival study. Head Neck 2008;30: Coit DG, Andtbacka R, Anker CJ, et al. Melanoma. J Natl Compr Canc Netw 2012;10: Pollack LA, Li J, Berkowitz Z, et al. Melanoma survival in the United States, 1992 to J Am Acad Dermatol 2011;65:S78 S Blank CU, Hooijkaas AI, Haanen JB, Schumacher TN. Combination of targeted therapy and immunotherapy in melanoma. Cancer Immunol Immunother 2011;60: Erickson Foster J, Velasco JM, Hieken TJ. Adverse outcomes associated with noncompliance with melanoma treatment guidelines. Ann Surg Oncol 2008;15: Thomas JM, Newton-Bishop J, A Hern R, et al. Excision margins in highrisk malignant melanoma. N Engl J Med 2004;350: Agarwal-Antal N, Bowen GM, Gerwels JW. Histologic evaluation of lentigo maligna with permanent sections: implications regarding current guidelines J Am Acad Dermatol 2002;47: Kunishige JH, Brodland DG, Zitelli JA. Surgical margins for melanoma in situ. J Am Acad Dermatol 2012;66: Moller MG, Pappas-Politis E, Zager JS, et al. Surgical management of melanoma-in-situ using a staged marginal and central excision technique Ann Surg Oncol 2009;16:

6 11. Dengel L, Turza K, Noland MM, Patterson JW, Slingluff CLJr. Skin mapping with punch biopsies for defining margins in melanoma: when you don t know how far to go. Ann Surg Oncol 2008;15: Anderson KW, Baker SR, Lowe L, Su L, Johnson TM. Treatment of head and neck melanoma, lentigo maligna subtype: a practical surgical technique. Arch Facial Plast Surg 2001;3: Sullivan SR, Scott JR, Cole JK, et al. Head and neck malignant melanoma: margin status and immediate reconstruction. Ann Plast Surg 2009;62: Bogle M, Kelly P, Shenaq J, Friedman J, Evans GR. The role of soft tissue reconstruction after melanoma resection in the head and neck. Head Neck 2001;23: Sullivan SR, Liu DZ, Mathes DW, Isik FF. Head and neck malignant melanoma: local recurrence rate following wide local excision and immediate reconstruction. Ann Plast Surg 2012;68: Houghton A, Coit D, Bloomer W, et al. NCCN melanoma practice guidelines. National Comprehensive Cancer Network. Oncology (Williston Park) 1998;12: Lens MB, Dawes M, Goodacre T, Bishop JA. Excision margins in the treatment of primary cutaneous melanoma: a systematic review of randomized controlled trials comparing narrow vs wide excision. Arch Surg 2002;137: National Comprehensive Cancer Network Melanoma. Melanoma. Clinical practice guidelines in oncology J Natl Compr Canc Netw 2004;2: Houghton AN, Coit DG, Daud A, et al. Melanoma. J Natl Compr Canc Netw 2006;4: Coit DG, Andtbacka R, Bichakjian CK, et al. Melanoma. J Natl Compr Canc Netw 2009;7: Vittinghoff E, Glidden D, Shiboski S, McCulloch C, eds. Regression Methods in Biostatistics: Linear, Logistic, Survival, and Repeated Measures Models. New York, NY: Springer; Zitelli JA, Brown CD, Hanusa BH. Surgical margins for excision of primary cutaneous melanoma J Am Acad Dermatol 1997;37: Jejurikar SS, Borschel GH, Johnson TM, Lowe L, Brown DL. Immediate, optimal reconstruction of facial lentigo maligna and melanoma following total peripheral margin control. Plast Reconstr Surg 2007;120: Trotter MJ. Melanoma margin assessment. Surgical Pathology Clinics 2009;2: Wong JY, Sondak VK. Unanswered questions about margin recommendations for primary cutaneous melanoma. J Natl Compr Canc Netw 2012; 10: Veronesi U, Cascinelli N, Adamus J, et al. Thin stage I primary cutaneous malignant melanoma. Comparison of excision with margins of 1 or 3 cm. N Engl J Med 1988;318: Sladden MJ, Balch C, Barzilai DA, et al. Surgical excision margins for primary cutaneous melanoma. Cochrane Database Syst Rev 2009;(4): CD Balch CM, Soong SJ, Smith T, et al. Long-term results of a prospective surgical trial comparing 2 cm vs. 4 cm excision margins for 740 patients with 1-4 mm melanomas. Ann Surg Oncol 2001;8: Gillgren P, Drzewiecki KT, Niin M, et al. 2-cm versus 4-cm surgical excision margins for primary cutaneous melanoma thicker than 2 mm: a randomised, multicentre trial. Lancet 2011;378: Khayat D, Rixe O, Martin G, et al. Surgical margins in cutaneous melanoma (2 cm versus 5 cm for lesions measuring less than 2.1-mm thick). Cancer 2003;97: McKinnon JG, Starritt EC, Scolyer RA, McCarthy WH, Thompson JF. Histopathologic excision margin affects local recurrence rate: analysis of 2681 patients with melanomas < or ¼2 mm thick. Ann Surg 2005;241: