Clinical outcomes. NED indicates no evidence of disease; plus sign, positive.
Kaplan-Meier overall survival curve by American Joint Committee on Cancer staging.
Kaplan-Meier overall survival curve by American Joint Committee on Cancer T status.
Kaplan-Meier overall survival curve by margin status.
Shuman AG, Light E, Olsen SH, Pynnonen MA, Taylor JMG, Johnson TM, Bradford CR. Mucosal Melanoma of the Head and NeckPredictors of Prognosis. Arch Otolaryngol Head Neck Surg. 2011;137(4):331-337. doi:10.1001/archoto.2011.46
To identify significant clinical and pathological predictors of survival in mucosal melanoma of the head and neck.
Retrospective case series. We reviewed cases of mucosal melanoma of the head and neck from a prospectively collected database after institutional review board approval.
A single academic institution.
Fifty-two patients with mucosal melanoma of the head and neck.
With a median follow-up of 97 months, the median overall survival was 52 months, with a 5-year overall survival of 38%. The median disease-free survival was 15 months, with a 5-year disease-free survival of 22%. Younger age (P = .02), lower T status (P = .003), and lower American Joint Committee on Cancer stage (P < .001) were associated with better overall survival. Positive surgical margins predicted poorer overall survival (P = .01), but patients who required reexcision to achieve negative margins had outcomes that were not significantly different from those with initially negative surgical margins (P = .71). Sex, smoking history, and primary site did not affect disease-free or overall survival. Adjuvant radiotherapy and/or chemotherapy did not predict improved outcomes. Fewer mitoses (P = .02) and the absence of ulceration (P = .01) predicted improved overall survival.
Our experience confirms the utility of current staging systems in predicting outcomes of mucosal melanoma of the head and neck and stresses the importance of achieving negative surgical margins. Pathologically, fewer mitoses and the absence of ulceration predict better outcomes and should be reported as part of routine histological profiles of mucosal melanoma. Further studies are necessary to change the paradigm of care for this rare and deadly disease.
Mucosal melanoma is a rare disease that involves the mucosa of the upper aerodigestive tract and perineum and comprises less than 1% of all diagnosed melanomas.1 The disease has a poor prognosis and is not well understood as a distinct clinical entity owing to its rarity.2 As a result, recommended management is typically based on pooled data from case reports and smaller case series. Although recurrent and metastatic disease occur frequently, most cases present solely with primary site involvement.3 Treatment is typically surgical despite the absence of prospective and/or randomized trials assessing the efficacy of treatment strategies in the primary or adjuvant setting.4
Mucosal melanoma has a unique staging system because traditional aspects of cutaneous melanoma staging may not apply.5 Histological diagnosis can be challenging owing to its rarity and variable appearance.6 Standard histopathological predictors of poor prognosis that affect cutaneous melanoma staging, such as Breslow depth, ulceration, and mitoses, have not been shown to influence survival in mucosal melanoma.7 As a result, a salient management guideline does not exist for treatment solely based on clinical and pathological staging.
This study was designed to report the clinical characteristics of patients diagnosed as having mucosal melanoma of the head and neck, including survival, treatments received, and clinical and histopathological predictors of improved survival. We hypothesized that specific clinical and pathological factors might significantly affect survival and potentially alter the management of this challenging clinical entity.
This study reports a retrospective case series approved by the University of Michigan Medical School institutional review board. The patients were identified and the data were collected from a comprehensive, prospective, single-institution, melanoma clinicopathological database from 1992 through 2009. Additional queries were made of a separate University of Michigan electronic pathology database using the terms melanoma AND mucosa, mouth, oral cavity, lips, tongue, buccal, sinus, nose OR nasal to ensure that all eligible patients were captured. Patients with biopsy-confirmed primary melanoma of the mucosal surface of the upper aerodigestive mucosa were included; patients who had poorly differentiated tumors that were not conclusively melanoma, primary involvement of cutaneous skin (including the lip), metastatic lesions involving aerodigestive mucosa, and primary mucosal melanoma not involving the head and neck were excluded. The demographics and clinical course as documented in the melanoma database were confirmed and cross-referenced via the electronic medical record by manual medical record abstraction.
The clinical variables evaluated included age at diagnosis, smoking status, clinical presentation at the time of diagnosis, site of the tumor, presence of other diagnosed neoplasms, response to and type of initial therapy, and timing and location of recurrence. Tumor staging was reported according to the guidelines set by the American Joint Committee on Cancer (AJCC).8
All pathological data (tissue blocks, slides, and reports) were independently reviewed by a pathologist (S.H.O.) with specific expertise in melanoma; data points were omitted in the case of missing pathological material. Specific factors assessed included tumor cell morphologic features, depth and anatomic level of invasion, growth phase (vertical vs radial), number of mitoses per square millimeter (roughly 4 high-power fields), presence or absence of ulceration, host response (nonbrisk vs brisk), presence or absence of regression, necrosis, pseudopapillomatous morphologic features, angiolymphatic invasion, and neurotropism. Depth of invasion was measured as thickness, in millimeters, from the most superficial layer of the epithelium, ulcer base, or granular cell layer, where present, and then stratified into levels based on the relationship to the submucosal lamina propria. The latter microstaging classification was derived from Prasad et al,9 who defined level I as mucosal melanoma in situ (without invasion of the lamina propria or with only microinvasion), level II as invasion into the lamina propria only, and level III as invasion into deep tissue structures, such as bone, muscle, and cartilage. Brisk host response is defined as at least a moderate lymphocytic inflammatory infiltrate at the base of the tumor and/or infiltrating within the tumor; nonbrisk host response is defined as an absent or scant inflammatory component surrounding the tumor.
We calculated descriptive statistics for all variables. Frequencies and percentages are presented for categorical variables, and means (SDs) are presented for continuous variables. We analyzed associations using univariate Cox proportional hazards regression models to calculate P values and hazard ratios (HRs) for all the clinical and pathological variables under evaluation. For class variables, such as stage, overall Wald type III tests are reported within the text, and contrast tests across levels are reported in the tables.
We analyzed survival statistics, including median follow-up in months and median overall and disease-free survival at 2 and 5 years, using the Kaplan-Meier product-limit method. Disease-free survival was determined as the time from initial treatment to the patient's last visit with no evidence of recurrent disease. Overall survival time was determined using a combination of electronic medical records, the melanoma database, and the Social Security Death Index and was defined as the time from the initial pathological diagnosis to death or to last known contact.
A total of 52 patients were identified, with a mean (SD) age of 66 (13) years. Thirty-one (60%) were female. Thirteen (25%) had been diagnosed as having other cancers; 1 (2%), with prior cutaneous melanoma; and 5 (10%), with prior nonmelanomatous cutaneous malignant neoplasms. Twenty-two patients (42%) were former or current smokers. Sixteen (31%) presented with oral cavity lesions (8 of the lip; 5, gingiva; 2, hard palate; and 1, retromolar trigone), and 36 (69%) presented with sinonasal primary tumors (22 of the nasal cavity; 7, maxillary sinus; 6, ethmoid sinus; and 1, sphenoid sinus). All oral cavity tumors presented as discrete lesions. Of the 36 sinonasal tumors, 14 presented with nasal obstruction, 12 with bleeding/epistaxis, 4 with pain, and 6 with other symptoms.
Forty-six patients (88%) initially were treated surgically, of whom 42 (81% of the total) achieved complete remission. All underwent extirpation at the primary site; in 31, negative margins were achieved initially; in 13, reexcision was required to achieve negative margins; and in 2, margins remained positive despite resection (both were T4 ethmoid sinus tumors with positive margins at the skull base, 1 of which was also positive in the orbit). Five individuals (10%) underwent neck dissection for clinically evident regional metastatic disease. Ten (19%) underwent adjuvant external beam radiotherapy, and an additional 5 individuals (10%) underwent palliative radiotherapy to the primary site and/or cervical lymphatics. Three (6%) received adjuvant cytotoxic chemotherapy, and an additional 15 individuals (29%) underwent palliative chemotherapy for unresectable locoregional or distant recurrence.
Data summarizing clinical staging are provided in Table 1.
A flowchart summarizing patient outcomes is provided in Figure 1. As of the last follow-up, 17 patients were still alive. With a median follow-up of 97 months, median overall survival was 52 months (95% confidence interval [CI], 24-74 months) with 2- and 5-year overall survival of 64% and 38%, respectively. With a median follow-up of 82 months, median disease-free survival was 15 months (95% CI, 8-33 months) with 2- and 5-year disease-free survival of 43% and 22%, respectively.
Increased age was significantly associated with poorer overall and disease-free survival (P = .02 and P = .07, respectively). The AJCC stage and individual T, N, and M status were all significantly associated with overall survival (Wald type III test, P = .003, P = .003, P = .001, and P < .001, respectively) (Figure 2 and Figure 3). The AJCC stage and T, N, and M status also significantly predicted disease-free survival (Wald type III test, P < .001, P < .001, P = .003, and P < .001, respectively). Sex, primary site, and smoking status were not associated with overall or disease-free survival (Table 2).
Patients with negative surgical margins had significantly better overall survival than those who did not achieve negative margins (median, 56 vs 9 months; P = .01) (Figure 4). Of the patients with initially positive margins who underwent a second resection and achieved negative margins, there was no difference in overall or disease-free survival compared with the patients who had initially negative margins (overall, 81 vs 56 months [P = .71]; disease-free, 26 vs 16 months [P = .28]). The HR for margin status (6.3 [95% CI, 1.0-39.1]) remained marginally significant (P = .05) after controlling for T status. Surgical margins were extremely variable given the variations in extirpative technique and tumor sites, limiting analysis of this variable.
In 16 patients, tissue was unavailable for retrospective pathological evaluation. There was no evidence of a statistically significant difference between those with and without tissue available for analysis in overall or disease-free survival (overall survival HR, 1.8 [95% CI, 0.9-3.5; P = .11]; disease-free survival 1.7 [0.8-3.4; P = .15]). Among patients whose tissue was analyzed, the following histopathological findings were less common: ulceration (51%), tumor regression (9%), radial growth (10%), necrosis (38%), pseudopapillomatous growth (25%), angiolymphatic spread (19%), neurotropism (18%), and brisk host response (26%) (Table 3). The mean (SD) thickness (depth of invasion) was 6.2 (3.8) mm (range, 1.6-16.0 mm). There was a mean (SD) of 7.1 (7.2) mitoses/mm2 (range, 0-25.0 mitoses/mm2). Level of invasion based on the Prasad scale (available for 36 patients) was I in 3 patients (8%), II in 13 (36%), and III in 20 (56%); these levels did not predict overall or disease-free survival in our cohort.
An increased number of mitoses was associated with poorer overall and disease-free survival (P = .02 and P = .002, respectively). Ulceration was associated with poorer overall survival (HR, 3.71 [95% CI, 1.43-9.63; P = .01]), with a trend toward poorer disease-free survival that did not reach clinical significance (2.23 [0.93-5.35; P = .07]).
Our demographic data with regard to age, sex, site, and stage are fairly similar to those reported elsewhere.5 Mucosal melanoma of the head and neck occurs at similar rates among both sexes, is typically diagnosed after the fifth decade of life, and is more frequent in the sinonasal cavity compared with the oral cavity. Other aerodigestive mucosal surfaces are more rarely involved. Increased age predicts worse outcomes, consistent with prior reports.10 We did not identify other demographic factors that were significant predictors of outcome, nor was primary site associated with a significantly different prognosis.
Our data corroborates the utility of the AJCC mucosal melanoma staging system. Overall stage and T, N, and M status all predict overall and disease-free survival. Our data are consistent with recent analyses with similar findings.11,12 As we demonstrate, prognosis is extremely poor among patients presenting with advanced primary disease or regional and/or distant metastases.
Our median overall survival of 52 months and 5-year overall survival of 38% are consistent with the larger series published in the literature: Yanagi et al13 report 27% 5-year overall survival, and Patel et al3 report 35%. Our dismal outcomes, with particularly poor disease-free intervals despite the achievement of complete remission in 81% of patients, underscore the aggressiveness of the disease.
One of the difficulties in diagnosing mucosal melanoma relates to its clinical rarity and variable histological presentation.6 Morphologically, mucosal melanoma may mimic classic pigmented cutaneous melanomas or may have a widely variable and heterogeneous appearance, with desmoplastic, epithelioid, sarcomatoid, undifferentiated, or pleomorphic components, all with or without pigmentation. Immunohistochemistry can be invaluable in making an accurate diagnosis; in general, staining patterns for mucosal melanoma parallel those seen in cutaneous melanoma. Specifically, mucosal melanomas tend to yield positive findings for S-100 and vimentin, occasionally positive findings for HMB-45 and Melan-A, and negative findings for cytokeratin and epithelial membrane antigen (both of which suggest an epithelial neoplasm). In general, pathologists reviewing challenging tumors of the upper aerodigestive tract mucosa may benefit from using immunohistochemistry to help exclude or confirm a diagnosis of mucosal melanoma.
Historically, the histopathological variables that are considered part of the cutaneous melanoma profile do not apply to mucosal melanoma. Our study, to our knowledge for the first time, demonstrates a greater than 3-fold impact of the presence of ulceration with survival in head and neck mucosal melanoma. Higher mitotic indices have been suggested to correlate with poorer outcome in selected other series, and we corroborate these results.12,14 We did not identify any other histopathological predictors of outcome.
Case series have widely varied findings concerning the prognostic role of specific histological characteristics. Prasad et al7 reported that vascular invasion, polymorphous tumor cell population, and necrosis significantly affected overall and disease-free survival, whereas tumor thickness, level of invasion, ulceration, mitotic index, and neurotropism did not affect outcome. We could not confirm the utility of microstaging based on tumor depth of invasion as a predictive tool.9 Unlike recently published findings,12 pseudopapillary architecture did not correlate with outcomes in our series.
As a result of our data and the varied reports in the literature, we propose that pathological analysis of mucosal melanoma should involve the formulation of a histopathological template profile of the primary lesion, which is required for most cancers, including cutaneous melanoma, and is shown in the following tabulation.
Development of a structured primary lesion template for histopathological reports will facilitate more consistent and detailed information transfer between centers and publications to better understand clinicopathological correlation with outcomes for future studies. Specifically, ulceration and mitotic index, as well as several other variables, should be routinely reported.
Mucosal melanoma of the head and neck is primarily a surgical disease.15 One of our intriguing observations was the equivalent outcomes among patients treated surgically who required reexcision to achieve negative margins compared with those with initially negative margins, as well as the dramatically worse prognosis among patients for whom margins could not be cleared. This suggests the importance not only of achieving negative surgical margins but also of carefully assessing resectability preoperatively, specifically at the level of the skull base and dura. This observation is limited by the fact that larger unresectable tumors clearly have worse prognoses; this has been demonstrated in multiple series, including those that account for tumor volume.13 In addition, the retrospective nature of our series makes it difficult to conclude whether initial procedures that yielded positive margins were truly oncologic excisions rather than diagnostic and/or excisional biopsies; there also exists the possibility of a type II statistical error due to lack of power. Nevertheless, our data imply that aggressive surgical extirpation of positive margins is indicated when feasible, even in cases that require reoperation.
Given the poor prognosis regardless of treatment modality, the role of aggressive surgical extirpation must be critically weighed against expected surgical morbidity. Acceptable functional outcomes are often achievable—even with extensive skull base and oral cavity surgery—by judicious attention to preservation of cranial nerves, meticulous reconstruction of surgical defects, and avoidance of brain retraction during skull base exposure by using an endoscopic and/or subcranial approach. Moreover, in many cases, the consequences of uncontrolled tumor progression may outweigh expected postoperative morbidity, even when the possibility of locoregional or distant recurrence is accounted for, especially given the absence of efficacious nonsurgical treatment options. In all cases, patient input and preferences are instrumental in deciding on appropriate treatment modalities, and a frank and open preoperative discussion concerning oncologic outcomes and perioperative expectations is crucial.
With regard to the role of neck dissection, our practice has been surgical extirpation of clinically evident regional metastatic disease, without treatment of N0 disease. Our data demonstrate the uniformly poor prognosis of all node-positive patients. Some authors advocate elective neck dissection in a subset of patients with N0 disease.16 In general, although there are no conclusive data, in our experience few patients have recurrences in the neck without accompanying local or distant disease, thereby calling into question the role of surgical management of the neck in these cases. Sentinel node biopsy is an intriguing concept for mucosal melanoma of the head and neck and may merit further consideration.17
Our study was not powered or designed to study the role of adjuvant or palliative radiotherapy or chemotherapy; given these limitations, we did not identify a significant difference in outcome among the patients treated in the adjuvant or palliative setting. The data to date remain inconclusive concerning the role of radiotherapy; multiple series have reported on external beam radiotherapy as the initial and adjuvant treatment.18 Although a consensus is difficult to achieve from a small case series, Temam et al19 and Moreno et al12 suggest that adjuvant radiotherapy may improve locoregional control despite the absence of significant improvement in overall survival. Carbon ion radiotherapy has undergone trial as well, with acceptable rates of achievement of clinical remission and initial locoregional control, but recurrences were frequent and long-term survival was rare.19 Although the data on chemotherapy are generally poor, M. D. Anderson Cancer Center, Houston, Texas, has pioneered “biochemotherapy,” involving combined cytotoxic and immunomodulatory regimens, with a small series demonstrating optimistic results that merit further investigation.20 Encouraging studies of therapeutic targets based on specific gene mutations found in a subset of melanomas may also merit application to mucosal melanoma.21
Mucosal melanoma is a neoplasm with a nefarious course despite aggressive treatment. Given this fact, surgical procedures that yield significant functional morbidity must be weighed against the prognosis and the patient's own choices. Further strategies to stratify patients according to risk and more efficacious treatment modalities remain elusive. One intriguing study22 reports on the loss of heterozygosity in free DNA circulating in plasma as a method of identifying subclinical metastasis in this patient population, but clinical utility remains to be defined.
Our report constitutes one of the larger case series of mucosal melanoma of the head and neck, which facilitates valid statistical analyses that are less prone to sampling error and confounding variables. Our data are strengthened by long follow-up and the use of a prospective clinicopathological database to limit recall bias. The independent review by a pathologist with expertise in this clinical realm also fortifies the weight of the data.
As with any retrospective review, our conclusions and recommendations are limited by the absence of more definitive prospective data; however, this is unlikely to emerge given the rarity of the clinical condition. In addition, the small numbers of patients and inherent selection bias among those undergoing nonsurgical therapy and the variable courses of adjuvant and palliative chemotherapy and radiotherapy limit our ability to make conclusions concerning their role and efficacy in this disease and create the risk of type II errors in our analysis. Missing histopathological data points associated with the absence of a standardized template could not be obtained in some cases. Moreover, the number of variables studied and the ensuing dearth of statistical power limited the role for a multivariate analysis.
In conclusion, mucosal melanoma is an aggressive disease with a poor prognosis despite aggressive treatment. Our experience confirms the utility of current staging systems in predicting outcomes of mucosal melanoma of the head and neck. The data stress the importance of achieving negative surgical margins and suggest that adjuvant radiotherapy and/or chemotherapy may not yield improved survival. Pathologically, fewer mitoses and the absence of ulceration predict better outcomes. Further studies and incorporation of standard histological profiles will be necessary to change the paradigm of care for this rare and deadly disease.
Correspondence: Carol R. Bradford, MD, Department of Otolaryngology, University of Michigan Hospitals, 1904 Taubman Center, Ann Arbor, MI 48109 (email@example.com).
Submitted for Publication: November 4, 2010; final revision received December 18, 2010; accepted December 29, 2010.
Author Contributions: All the authors had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Shuman, Light, Olsen, Pynnonen, Taylor, Johnson, and Bradford. Acquisition of data: Shuman, Light, Olsen, Taylor, Johnson, and Bradford. Analysis and interpretation of data: Shuman, Light, Olsen, Pynnonen, Taylor, Johnson, and Bradford. Drafting of the manuscript: Shuman, Light, Olsen, Pynnonen, Taylor, Johnson, and Bradford. Critical revision of the manuscript for important intellectual content: Shuman, Light, Olsen, Pynnonen, Taylor, Johnson, and Bradford. Statistical analysis: Shuman, Light, Olsen, Taylor, Johnson, and Bradford. Administrative, technical, and material support: Shuman, Light, Olsen, Pynnonen, Taylor, Johnson, and Bradford. Study supervision: Shuman, Light, Olsen, Pynnonen, Taylor, Johnson, and Bradford.
Financial Disclosure: None reported.
Previous Presentation: This study was presented as a poster at the American Head and Neck Society Research Workshop on Biology, Prevention, and Treatment of Head and Neck Cancer; October 28-30, 2010; Arlington, Virginia.