The Role of Microsatellites as a Prognostic Factor in Primary Malignant Melanoma

Objective  To determine the impact of microsatellites as a prognostic factor in primary cutaneous melanoma.

Design  Retrospective cohort study.

Setting  Tertiary referral center.

Patients  A total of 504 patients with a history of primary melanoma observed for 2 years or having experienced a first relapse.

Main Outcome Measures  Overall survival (OS) and relapse-free survival (RFS).

Results  Forty-five patients had evidence of microsatellites in their primary melanoma. Presence of microsatellites significantly correlated with the presence of several other histologic high-risk factors such as tumor thickness, ulceration, Clark level, vascular factors, and mitotic rate. Univariate analysis demonstrated decreased RFS and OS in patients with microsatellites. Presence of microsatellites was associated with increased locoregional metastasis but not distant metastasis. In multivariate analysis, with the inclusion of 6 other clinical and histologic factors, presence of microsatellites was a significant predictor of RFS but not OS. Patients with clinical macrosatellites had a trend toward worsening OS compared with those with microsatellites.

Conclusions  The presence of microsatellites is intimately tied to other markers of melanoma aggressiveness. Microsatellites appear to predict locoregional relapse and RFS but neither distant metastasis nor OS. These results may have implications for patient care as well as the inclusion of microsatellites in stage III of the current classification.

In 1981, Day and colleagues¹ first used the term microscopic satellites to represent nests of melanoma cells separated from the main body of the tumor mass by a layer of collagen or subcutaneous fat. These cells were thought to be a microscopic representation of the tumor’s metastatic potential and as such a plausible marker of the tumor’s innate aggressiveness. Subsequent survival analyses have revealed decreased survival and increased probability of regional relapse among patients with evidence of microsatellites in their primary tumor.^2,3

Recently, the latest version of the American Joint Committee on Cancer (AJCC) staging classification for melanoma⁴ has up-staged microsatellites to the N classification of the TNM system. Specifically, patients with microsatellites are assigned to the pathologic N2c and clinical IIIB staging categories, prognostically placing them alongside patients with in-transit and clinical satellite metastases (macrosatellites). The inclusion of microsatellites in the same category as macrosatellites was partly based on the empirical assumption that microsatellites represent an early step in the development of all intralymphatic metastases. However, the exact relationship between microscopic satellitosis and the former entities has not been fully delineated.

To date, only a handful of studies have investigated the prognostic significance of microscopic satellites in primary melanoma.^1-3,5,6 In the present study, we examine the role of microsatellitosis as a predictive marker of outcome associated with melanoma.

This analysis was approved by the committee on human research, the University of California, San Francisco, institutional review board. The records of 525 patients from the University of California, San Francisco, Melanoma Center database were reviewed for the presence of microsatellites. Selection guidelines for this database have been described previously⁷ and include patients with (1) at least 2 years of follow-up or (2) having experienced a first relapse. The range of follow-up was 0.04 to 24.0 years, with a mean follow-up of 5.37 years. The cohort had a median tumor thickness of 1.55 mm. An updated database of our previously defined cohort was used for all analyses. Relapse was defined to include local dermal recurrence and in-transit, regional nodal, and distant metastasis.

Nine clinical factors were included in the database used to conduct this analysis: age, sex, location of primary melanoma, status of regional lymph nodes, site of first recurrence, site of distant metastasis, presence or absence of macrosatellites, date of relapse, and date of death. Nine histopathologic parameters were also evaluated for each patient by a single pathologist (R.W.S) based on evaluation of routine hematoxylin-eosin slides used to make the diagnosis of primary melanoma at the time of the patient’s initial visit to the University of California, San Francisco, Melanoma Center: tumor thickness, histogenetic subtype, Clark level, ulceration, vascular involvement, tumor vascularity, mitotic rate (defined as mitoses per mm²), regression, and microsatellites. The systematic approach to the histopathologic review of primary melanoma has been described previously.⁸

Data on 504 patients out of a total of 525 patients were sufficiently complete for analysis. A microsatellite was strictly defined as a discrete nest of tumor cells distinctly separated by a minimum of 0.5 mm (by ocular micrometer) from the main body (vertical growth phase) of the tumor by a layer of collagen or subcutaneous fat.^1,6 Tumor cells separated by tumor stroma or single cells (by hematoxylin-eosin or using special stains) were excluded from this definition. Microsatellites were not included in the Breslow measurement or Clark level. Care was taken to distinguish between microsatellites and local persistent disease, vascular involvement, local dermal recurrence, and in-transit disease. Macrosatellites were defined as clinically evident tumor nodules in the dermis or fat located between the primary tumor and the regional nodal basin and therefore encompassed both local recurrences and satellite or in-transit disease. Vascular involvement and tumor vascularity were scored as previously described.^7,9

All statistical analyses were performed as previously described.^7,9 Multivariate analysis of all study parameters was performed using the Cox regression model of proportional hazards. In the Cox regression analysis, the various prognostic factors were entered into the model according to data coding criteria previously used and described by the AJCC staging committee.⁴ Thus, tumor thickness was entered into the model in 4 strata (≤1.00 mm, 1.01-2.00 mm, 2.01-4.00 mm, and >4.00 mm); age was entered in 10-year increments; and Clark level was entered as a dichotomous variable (level IV/V vs II/III). The following factors were entered as dichotomous variables: sex, location (extremity vs axial and head and neck), ulceration, microsatellites, nodal status, vascular involvement (present or absent), and tumor vascularity (absent vs increased vascularity). Except where otherwise noted, the results of all statistical tests were reported based on 2-tailed probabilities.

Forty-five patients (9%) had evidence of microsatellites in their primary tumor in this cohort. Initially, the relationship between sex, age, and location of the primary tumor and microsatellites was explored. A significant association was observed between presence of microsatellites and location of the primary tumor, with most microsatellites occurring in the head and neck region. Fifty-two percent of patients with microsatellites (n = 23) had primary melanomas located in the head and neck region, 27% (n = 12) in the trunk, 18% (n = 8) in the lower extremity, and 2% (n = 1) in the upper extremity regions (P<.001). In contrast, patients with microsatellites were not significantly different in age or sex from control counterparts (data not shown).

The presence of microsatellites correlated with many other known histopathologic factors. Tumors with microsatellites were mostly of the nodular and superficial spreading melanoma subtypes (Table 1). Stratifying the database on thickness revealed increasing frequency of microsatellites with increasing thickness ranges (Table 1). In addition, tumors with microsatellites were more often associated with a high mitotic rate or a higher Clark level. Finally, presence of microsatellites also correlated with ulceration, tumor vascularity, and vascular involvement (Table 1).

Kaplan-Meier survival curves were compared in tumors with and without microsatellites. Both relapse-free survival (RFS) and overall survival (OS) were significantly lower in patients with microsatellites (P<.001, data not shown). Presence of microsatellites was also associated with reduced RFS and OS parameters by several other univariate analyses (Table 2). Intriguingly, microsatellites were associated with significantly increased rates of macrosatellite and lymph node metastasis but not distant metastasis. Finally, comparison of OS curves between patients with microsatellites and those with macrosatellites demonstrated a trend toward lower OS in the macrosatellite group, with 5-year OS of 40.2% and 22.4%, respectively (1-tailed P<.07).

To determine the independent impact of microsatellites on outcome, we performed multivariate analysis of OS and RFS considering the following parameters: sex, location, age, thickness, ulceration, Clark level, and microsatellites (Table 3). After adjusting for the aforementioned features, microsatellites provided independent prognostic information in predicting RFS but not OS. Finally, with the addition of 3 other prognostic factors (nodal status, vascular involvement, and tumor vascularity), tumor thickness emerged as the strongest predictor of OS by stepwise Cox regression analysis, followed by nodal status, vascular involvement, tumor vascularity, location, and age, all of which were also significantly and independently predictive of OS in this analysis (data not shown).

In this study, we analyzed the role of microsatellites in the outcome associated with melanoma. Our data demonstrate increased frequency of relapse and death and decreased RFS and OS rates in patients with microsatellites by univariate analysis. Microsatellites had a stronger impact on in-transit and nodal relapse than on distant metastasis. Significant correlations were observed between microsatellites and other powerful histopathologic risk factors. Patients with macrosatellites had a trend toward poorer 5-year survival than patients with microsatellites. Multivariate analysis by Cox regression revealed an independent role for microsatellites in predicting RFS but not OS.

To date, only a handful of studies have investigated the role of microsatellites in predicting outcome.^1-3,10,11 Day et al¹ showed a 36% 5-year RFS in patients with microscopic satellites, compared with an 89% RFS in patients without. A later study² by this group analyzing 20 clinical and histologic factors identified thickness, ulceration, and microsatellites as the combination of variables most predictive of regional nodal metastasis. Likewise, Leon and colleagues³ showed decreased RFS and OS in a matched cohort of 30 patients with microsatellites. The microsatellite group demonstrated a 37% 5-year OS compared with 65% in the nonmicrosatellite group. Cox regression analysis identified microscopic satellites, regression, ulceration, high mitotic rate, tumor-infiltrating lymphocytes, positive lymph node dissection, and location as independent predictors of survival when 12 factors were considered.

Despite these results, the role of microsatellites as an independent predictor of outcome remains unclear.^6,7,9 Some of the difficulty stems from different definitions and measurements of this histopathologic factor. Rates ranging from 6%¹² to 17%¹ have been observed among different cohorts, likely owing to differences in measurements of this high-risk feature and the cohort of patients included for analysis. Harrist et al² included in their analyses of microsatellites not only true dermal satellites but also tumor emboli within vascular spaces as well as tumor islands potentially contiguous with the main body of the tumor. Other groups, however, have included only true dermal satellites in the definition of microsatellites and considered other entities such as tumor within vasculature as a representation of vascular involvement.^6,7 This would potentially explain the higher frequency of microsatellites observed in the Day et al¹ cohort (17%) as well as the robust decrease in OS and RFS observed by these investigators in the microsatellite group. Furthermore, most of the earlier literature on microsatellites has not considered the relationship between this entity and more recently described pathologic markers such as vascular involvement and tumor vascularity. A stepwise Cox regression analysis revealed microsatellites to be less critical in predicting OS and RFS when vascular factors were also included in the model.

Recently, the AJCC has placed microsatellites in the N2c category of the TNM classification, thus placing it in the stage III category of disease.⁴ Microsatellites were included in stage III based on multiple survival analyses of patients with clinical satellites, local recurrence, and in-transit and nodal metastases and who all demonstrated similar 10-year survival rates of 20% to 40%.^13-15 The decision to include microsatellites in this same category has mainly been based on the idea that the development of microsatellites, clinical satellites, and in-transit and regional nodal metastases are all continuous biologic events and can be treated as 1 entity.^13,16-18 This reclassification will have significant prognostic and therapeutic consequences, especially in regard to adjuvant therapy and clinical trial eligibility.

However, using a database that recapitulated the prognostic impact of thickness and ulceration,¹⁹ we found that microsatellites do not have additional prognostic impact on OS when added to the factors recently analyzed by the AJCC staging committee.⁴ In addition, the frequency of distant relapse, a more powerful indicator of disease-specific survival, was not significantly different in patients with and without microsatellites. In this regard, a prior study of 284 patients showed thick tumors with and without microsatellites to have similar overall (locoregional and distant) recurrence rates.⁶ Further stratification of relapse into local and distant relapse showed a significantly higher local but not distant recurrence rate in the microsatellite group. By comparison, more powerful prognostic factors in our database such as vascular involvement, tumor vascularity, and ulceration have been shown to significantly predict both regional and distant metastasis.^7,9 Furthermore, there was a trend toward decreased OS among patients with macrosatellites compared with those with microsatellites, suggesting that these represent 2 distinct entities.

While the inclusion of ulceration in the latest staging classification was based on its independent impact on survival by Cox regression,^4,13,20,21 the inclusion of microsatellites did not undergo similar validation. A multivariate analysis of our data using the same factors as the AJCC⁴ identified tumor thickness, age, location, and ulceration as the strongest predictors of OS. By contrast, presence of microsatellites was predictive of RFS but not OS according to this analysis. Since microsatellites did not have the same prognostic power in determining OS as did other prognostic factors, its automatic inclusion in stage III disease deserves further scrutiny. The observation that microsatellites occurred more frequently in the context of all the other high-risk features considered in this study suggests that by the time microsatellites develop, numerous other high-risk events have already taken place, further reducing its impact in multivariate analysis. Finally, the finding of a trend toward reduced OS in the subgroup of patients with macrosatellites may provide further evidence that microsatellitosis and clinical satellitosis represent potentially distinct points along the tumor progression pathway of primary melanoma.

In summary, microsatellites seem to impart their influence on outcome through their correlation with locoregional as opposed to distant relapse. Thus, they represent yet another factor that refines the prognosis of primary melanoma, given a range of tumor thickness. In this regard, a previous study²² showed microsatellites to be 1 of 4 factors (along with mitotic index, vascular invasion, and ulceration) that increase the risk of identifying a positive sentinel lymph node. In view of these findings, we recommend that microsatellites be placed alongside ulceration (and vascular invasion) as a second determinant (ie, in the b subclassification) of the T staging category.

Correspondence: Mohammed Kashani-Sabet, MD, Melanoma Center, University of California, San Francisco Cancer Center, 1600 Divisadero St, Second Floor, San Francisco, CA 94115 (kashanim@derm.ucsf.edu).

Accepted for Publication: January 11, 2005.

Funding/Support: This study was supported by an educational grant from the Skin Disease Research Foundation and by the Mark Shafir Fund. Dr Kashani-Sabet is supported by the Herschel and Diana Zackheim Endowment Fund in Cutaneous Oncology.

Acknowledgment: We thank Rosie Casella for preparing the manuscript.

Financial Disclosure: None.