See related editorial on pages 1403-5, this issue.
Sentinel lymph node (SLN) biopsy generally is recommended for patients who have melanoma with a Breslow thickness ≥1 mm. Most patients with melanoma between 1 mm and 2 mm thick have tumor-negative SLNs and an excellent long-term prognosis. The objective of the current study was to evaluate prognostic factors in this subset of patients and determine whether all such patients require SLN biopsy.
Patients with melanoma between 1 mm and 2 mm in Breslow thickness were evaluated from a prospective multi-institutional study of SLN biopsy for melanoma. Disease-free survival (DFS) and overall survival (OS) were evaluated by Kaplan-Meier analysis to compare patients with melanoma that measured from 1.0 mm to 1.59 mm (Group A) versus patients with melanoma that measured from ≥1.6 mm to 2.0 mm thick (Group B). Univariate and multivariate analyses were performed to evaluate factors predictive of tumor-positive SLN status, DFS, and OS.
The current analysis included 1110 patients with a median follow-up of 69 months. SLN status was tumor-positive in 133 of 1110 patients (12%) including 66 of 762 patients (8.7%) in Group A and 67 of 348 patients (19.3%) in Group B (P < .0001). On multivariate analysis, age, Breslow thickness, and lymphovascular invasion were independently predictive of a tumor-positive SLN (P < .05). DFS (P < .0001) and OS (P = .0001) were significantly better for Group A than for Group B. When tumor thickness was treated as either a continuous variable (P < 0.0001) or a categorical variable (P < .0001), it was significantly predictive of DFS and OS. On multivariate analysis, Breslow thickness, age, ulceration, histologic subtype, regression, Clark level, and SLN status were significant factors predicting DFS; and Breslow thickness, age, primary tumor location, sex, ulceration, and SLN status were significant factors predicting OS (P < .05). A subgroup of patients who had tumors <1.6 mm in Breslow thickness, had no lymphovascular invasion, and were aged ≥59 years had a low risk (5%) of tumor-positive SLN.
Regional lymph node status remains the most important prognostic indicator in early stage melanoma.1-3 Sentinel lymph node (SLN) biopsy has emerged as an accurate method for evaluating the draining lymph node basin, allowing for the generation of valuable prognostic information. In addition, early evaluation of the regional lymph node basin with SLN biopsy can identify clinically occult lymph node metastases, allowing early therapeutic lymph node dissection and guiding further therapy with minimal morbidity.3-7 Left in situ, clinically silent nests of metastatic disease eventually will progress to clinically palpable lymph node disease if they are not removed,7, 8 thus adding to the value of the SLN technique.
Several studies have investigated the clinicopathologic factors associated with SLN metastasis. Wagner et al9 demonstrated that mitotic index and Breslow thickness are significant factors for predicting SLN metastasis. The presence of multiple lymph node basins,10 patient age,11 and sex12 also are associated with the risk of lymph node metastasis. Other tumor-related factors associated with lymph node metastasis include tumor thickness, ulceration, regression, mitotic rate, Clark level of invasion, histopathologic subtype of melanoma, and primary tumor site.3, 13-19
Tumor thickness is a powerful predictor of patient outcome and is closely associated with the development of metastasis in regional lymph nodes.2, 9 The risk of SLN metastasis increases as melanoma thickness increases. For melanomas that measure <1 mm in thickness, the risk of having a positive SLN is <5%.3 This risk increases to >30% for melanomas that measure >4 mm in thickness.2, 20, 21 Although SLN biopsy generally is recommended for all patients who have tumor thickness >1 mm, the majority of these patients will have a negative SLN biopsy. In fact, only 20% of patients who have intermediate thickness melanoma will have tumor-positive lymph nodes.2, 3, 22
Although SLN biopsy remains a low-risk procedure with little morbidity and is highly accurate as a staging procedure, attempts to define a cohort at low risk for lymph node metastasis might improve clinicians' ability to counsel patients regarding risks and benefits of the procedure. Because tumor thickness has an impact on the risk of lymph node metastasis, we sought to determine whether a population of patients with relatively thin melanoma might have a very low risk of lymph node metastasis and, thus, might derive little benefit from SLN biopsy. On the basis of the published data, we hypothesized that patients who had cutaneous melanoma measuring 1 mm to 2 mm in thickness might have a very low risk of lymph node metastases and that associated clinicopathologic factors would predict a tumor-negative SLN. The objective of the current study was to determine whether clinicopathologic factors might predict a subset of patients with cutaneous melanoma measuring 1 mm to 2 mm in thickness that has a minimal risk of SLN metastasis.
MATERIALS AND METHODS
The Sunbelt Melanoma Trial is a prospective, randomized trial that involves 79 centers in North America. Institutional review boards of each participating institution evaluated and approved this study. Inclusion criteria included ages 18 to 70 years, cutaneous melanoma ≥1 mm in Breslow thickness, and no clinical evidence of regional or distant metastasis. The current report represents a post hoc analysis of the study data. An independent data safety and monitoring committee supervised the study.
After obtaining informed consent, patients underwent wide local excision of the primary melanoma and SLN biopsy, as described previously.11 Because 1 of the objectives of the Sunbelt Melanoma Trial was to determine the role of molecular staging for melanoma, a portion of each SLN (defined as 25% of the lymph node or a 2-mm3 portion of the lymph node, whichever was smaller) was snap frozen on dry ice or liquid nitrogen and stored at −80°C until it was shipped on dry ice. If more than 1 SLN was identified, then each SLN was processed identically. The remaining SLN tissue was processed by hematoxylin and eosin staining at multiple levels, including ≥5 sections per block and 2 additional random sections for S-100 immunohistochemistry. The a priori definition of a tumor-positive SLN was any SLN that contained histologic evidence of metastatic melanoma cells on either hematoxylin and eosin-stained tissue or S-100 immunohistochemistry. A central pathology review committee evaluated the first 10 cases from each participating institution and evaluated all SLNs that contained metastases. Then, patients with SLNs that contained metastatic melanoma underwent completion lymph node dissection. Patients who had a single tumor-positive SLN were randomized to receive either adjuvant interferon or observation, whereas patients who had >1 tumor-positive SLN or those with extracapsular extension subsequently received adjuvant interferon therapy.
The subgroup of patients with melanoma between 1 mm and 2 mm in Breslow thickness was evaluated in this analysis. For the purposes of separating these patients into comparable groups, tumor thickness was used as a stratifying variable. Tumor thickness initially was evaluated as a continuous variable. Partitioning tree software was used to analyze tumor thickness and suggested dividing patients into 2 groups: those with melanoma from 1 mm to 1.59 mm thick (Group A) and those with melanoma from 1.60 mm to 2.0 mm thick (Group B). Clinicopathologic variables such as age, sex, ulceration, and lymphovascular invasion (defined as histologic evidence of melanoma cells invading either dermal lymphatics or blood vessels), were collected and analyzed.
Disease-free survival (DFS) was calculated from the date of randomization (the initial randomization for the Sunbelt Melanoma Trial) to the date of the first recurrence. Overall survival (OS) was calculated from the date of randomization to the date of death. Survival distributions were estimated using the Kaplan-Meier method, and the log-rank test was used to assess the statistical significance of differences in DFS and OS between groups. For continuous variables, P values were calculated with 2-sample, equal-variance t tests. For categorical variables, P values were calculated with chi-square tests for association with unknown categories omitted. P values <.05 were considered significant. We performed univariate and multivariate analyses to evaluate factors that were predictive of tumor-positive SLN, DFS, and OS. All analyses were conducted using JMP software (SAS Institute, Cary, NC).
In total, 1110 patients were identified with cutaneous melanoma that measured between 1 mm and 2 mm in Breslow thickness. The median follow-up was 69 months. There were 575 men and 535 women. Overall, SLNs were positive for metastatic melanoma in 12% of patients. These patients were separated into comparable groups using tumor thickness as a stratifying variable. Partitioning tree software was used to split tumor thickness into 2 groups to maximize the difference in the proportion of patients with SLN metastasis between the 2 groups: Patients in Group A had melanoma that measured from 1 mm to 1.59 mm in thickness (n = 762), and patients in Group B had melanoma that measured from 1.60 mm to 2.0 mm in thickness (n = 348).
Primary Tumor Characteristics and Thickness
Table 1 outlines the clinicopathologic factors relating to melanoma thickness. The Group A cohort consisted of 762 patients (68%) and included 385 men and 377 women. The Group B cohort consisted of 348 patients (31.4%) and included 190 men and 158 women. On univariate analysis, patients in Group B were more likely than patients in Group A to have ulceration (P < .0001), a Clark level of IV or V (P < .0001), and tumor-positive SLNs (P < .0001). SLN biopsy was positive in 8.7% of Group A melanomas (n = 66) and in 19.3% of Group B melanomas (n = 67) (Table 1). Patients in Group A also were more likely to have a superficial spreading histologic subtype (P < .0001) compared with patients in Group B. There was no difference between the 2 groups in age (P = .88), primary tumor site (P = .91), lymphovascular invasion (P = .27), vertical growth (P = .79), or regression (P = .24) on univariate analysis.
Table 1. Clinicopathologic Factors Related to Melanoma Thickness
Sentinel lymph node status
Next, the data were analyzed to determine which factors were associated with SLN metastasis (Table 2). On univariate analysis, Breslow thickness (P < .0001), age (P = .0002), and lymphovascular invasion (P = .006) all were predictive of SLN metastasis. Factors that did not predict SLN metastasis included sex (P = .59), ulceration (P = .44), tumor site (P = .18), histologic subtype (P = .45), vertical growth (P = .48), and regression (P = .31). On multivariate analysis, Breslow thickness (P < .0001), age (P = .0025), and lymphovascular invasion (P = .0081) all remained predictive of positive SLNs (Table 3).
Using data that were identified as predictive of SLN metastasis, we developed a partitioning tree to help identify a subset of patients with thin melanoma who had a low incidence of SLN-positive disease (Fig. 1). The overall rate of SLN-positive disease, taking into account all patients, was 12.3%. We characterized risk further based on tumor thickness: Patients who had melanoma from 1.0 mm to 1.59 mm in thickness had an 8.7% risk of positive SLNs, whereas patients who had melanoma from 1.6 mm to 2.0 mm in thickness had a 19.3% risk of positive SLNs. The data could be stratified further based on lymphovascular invasion: Patients who had melanoma <1.6 mm in thickness and no lymphovascular invasion had an 8.3% rate of positive SLNs. However, the presence of lymphovascular invasion increased that rate to 23.5%. Among patients who had thicker melanoma (>1.6 mm), the presence of lymphovascular invasion predicted SLN metastasis in 28.6% compared with 18.8% of patients without lymphovascular invasion. Further subdividing the data by age highlighted several groups that had a higher incidence of SLN metastasis: Patients aged <59 years who had thicker melanoma (>1.6 mm) and lymphovascular invasion had a 37% risk of positive SLNs. Even younger patients (aged <59 years) who had thicker melanoma (>1.6 mm) without lymphovascular invasion had a 21.5% risk of positive SLNs compared with 11% in a similar group of older patients (aged >59 years). We observed the same difference in younger patients (aged <59 years) who had thinner melanoma (<1.6 mm). There was a lower incidence of positive SLNs (4.9%) in patients aged >59 years who had thinner melanoma (<1.6 mm) and no lymphovascular invasion.
A univariate analysis was conducted to determine which factors were associated with DFS (Table 4). Breslow thickness (P < .0001), Clark level (P = .03), ulceration (P < .0001), histologic subtype (P = .0085), age (P = .02), and SLN status (P = .0005) all were associated with DFS. It is noteworthy that sex (P = .07) and lymphovascular invasion (P = .998) were not associated with DFS. On multivariate analysis, Breslow thickness (P < .0001), SLN metastasis (P = .0005), tumor ulceration (P < .0001), and age (P = .03) remained predictive of DFS. Clark level (P = .31), histologic subtype (P = .29), and regression (P = .91) were not significantly related (Table 5).
Table 4. Univariate Analysis of Clinicopathologic Factors Related to Disease-Free Survival
Breslow thickness, mm
Sentinel lymph node status
Table 5. Multivariate Analysis of Factors Associated With Disease-Free Survival
RR (95% CI)
RR indicates risk ratio; CI, confidence interval.
Sentinel lymph node status
Several different clinicopathologic factors were identified that influenced OS in the same group of patients. Breslow thickness (P < .0001), sex (P = .0017), ulceration (P < .0001), age (P < .0001), tumor site (P = .0016), and SLN status (P < .0001) all were significant in predicting OS on univariate analysis (Table 6). Each of these factors, including Breslow thickness (P = .04), sex (P = .0497), age (P < .0001), tumor site (P = .0023), SLN status (P < .0001), and ulceration (P < .0001), remained clinically significant on multivariate analysis (Table 7).
Table 6. Univariate Analysis of Clinicopathologic Factors Related to Overall Survival
Breslow thickness, mm
Sentinel lymph node status
Table 7. Multivariate Analysis of the Factors Associated With Overall Survival
RR (95% CI)
RR indicates risk ratio; CI, confidence interval.
Sentinel lymph node status
Both DFS (P < .0001) (Fig. 2, top) and OS (P = .0001) (Fig. 2, bottom) were significantly greater for patients in Group A versus Group B. Further analysis of the patients in Group A demonstrated that positive SLN status did not have an impact on DFS (P = .27) (Fig. 3, top). However, the patients in Group A who had metastasis within an SLN did have a shorter OS (P = .0074) (Fig. 3, bottom). For patients who had thicker tumors (Group B), a tumor-positive SLN had an impact on both DFS (P = .0036) (Fig. 4, top) and OS (P = .0064) (Fig. 4, bottom).
The data presented in the current report demonstrate significant heterogeneity in the biologic behavior of melanoma, even in a group of patients with relatively thin primary tumors. On the basis of univariate and multivariate analyses, we established an association between several clinicopathologic factors and lymph node metastasis. Lymphovascular invasion, age, and tumor thickness all were predictive of lymph node metastasis for patients who had melanoma tumor thickness between 1 mm and 2 mm. However, when we stratified patients based on these characteristics, we could not identify a subset of patients that had a negligible rate of lymph node metastasis. On the basis of the data presented, we cannot reasonably predict which patients with melanoma from 1 mm to 2 mm in thickness may be at minimal risk of lymph node metastasis and, thus, may forego SLN biopsy. Consequently, we continue to recommend SLN biopsy for our patients who have melanoma that measure ≥1 mm in thickness. It is this group of patients that would stand to gain the greatest from accurate lymph node staging in terms of adjuvant therapy (should a meaningful agent be developed) and/or enrollment in prospective trials.
Multiple studies have evaluated tumor thickness as it relates to SLN status.2, 11, 21, 23-25 Most of those studies subdivided patients based on current staging schemes. Several studies specifically evaluated intermediate thickness melanoma. Bliecher et al reported a 7.1% SLN-positive rate for melanoma between 1 mm and 1.5 mm in thickness.26 Morton et al produced similar results for melanoma between 0.75 mm and 1.5 mm in thickness with an incidence of SLN metastasis of 10.3%.27 That rate increased to 23% for tumors that measured >1.5 mm thick.27 We chose to evaluate patients who had melanomas between 1.0 mm and 2.0 mm in thickness because the majority of these patients will not have metastatic disease, yet they frequently are offered SLN biopsy as a staging procedure. Our data suggest that 8.7% of melanomas between 1.0 mm and 1.59 mm in thickness have evidence of metastatic disease, whereas 19.7% of melanomas between 1.6 mm and 2.0 mm have lymph node metastasis. This is comparable to our previously published data, which indicated a 22% rate of SLN-positive disease in patients who had tumors >1.0 mm thick.11
We further explored the biologic variability of intermediate thickness melanoma by using Classification and Regression Tree (CART) analysis and developed a partitioning tree to risk stratify patients who had intermediate thickness melanoma. Kesmodel et al previously described decision trees for predicting SLN metastasis based on mitotic index and tumor thickness.28 Those studies suggested that a mitotic rate >0 is a significant risk factor for SLN metastasis. The mitotic rate was not widely known as important in melanoma during design of the Sunbelt Melanoma Trial; therefore, we do not have any data regarding the mitotic rate. Our data demonstrate that tumor thickness, age, and lymphovascular invasion all are independently predictive of SLN-positive melanoma. However, when we stratified our patients using CART analysis, we could not identify a group of patients that was at negligible risk of lymph node metastasis. The lowest risk (<5%) was observed in the subset of patients with Breslow thickness <1.6 mm, aged >59 years, and with no lymphovascular invasion. The decision tree further highlights our previously reported data, indicating that age is related inversely to the presence of SLN metastasis. Several theories have been proposed to explain this finding, including a greater rate of hematogenous metastasis in older patients, varied biologic behavior of melanoma, and decreased sensitivity of SLN biopsy.11, 18, 29, 30
Regional lymph node status has been recognized as 1 of the most powerful indicators of patient outcome.2, 3 The current study of intermediate thickness melanoma corroborates those findings. The results from our survival analyses were similar to those published previously by Balch et al, who reported an OS of approximately 70% at 15 years for melanoma that measured from 1.0 mm to 2.0 mm in thickness.2 On multivariate analysis, we observed that lymph node status, age, tumor thickness, and ulceration were predictive of OS. Both DFS and OS were shorter for patients who had metastatic lymph nodes. It is noteworthy that, for patients who had very thin tumors (1-1.59 mm), the presence of a metastatic SLN did not have an impact on DFS, but it did predict OS. One hypothesis to explain this finding is that the patients died from other nonmelanoma-related causes and that they had not developed recurrent disease, because OS was measured to the date of death, not melanoma-related death. The other possibility is that not all recurrences were captured in this group before death. This effect was lost, however, when the thicker tumors (Group B) were analyzed. The presence of metastatic melanoma in an SLN predicted shorter DFS and OS in this group, most likely reflecting true progression of disease. Yet another possibility is that the variability of different prognostic factors in predicting recurrence and survival may have been the result of subset analysis and the resultant statistical phenomena observed in multivariate analyses, and not a true indicator of biology.
The prognostic importance of SLN biopsy in melanoma has been well established. Like others, we have sought to refine the technology and indications in an attempt to be more selective when evaluating patients. Despite our large cohort of prospectively studied patients, we were not able to define a group of patients that had relatively thin melanoma who had a negligible risk for lymph node metastasis. Patients with melanoma <1.6 mm thick, aged >59 years, and no lymphovascular invasion had a relatively low risk (4.9%) of SLN metastasis, and SLN biopsy may be considered selectively in this group. However, SLN biopsy is recommended for the majority of patients who have melanoma that measures between 1 and 2 mm in thickness.
We thank the coordinators of the Sunbelt Melanoma Trial as well as Deborah Hulsewede, Sherri Matthews, Pam Harlan, Ivan Deyahs, and Alex Scoggins for their commitment and hard work in managing the study. We also thank Advertek, Inc. for statistical analysis.
CONFLICT OF INTEREST DISCLOSURES
This work was supported by a grant from Schering Oncology-Biotech as well as the Center for Advanced Surgical Technologies of Norton Hospital, Louisville, Kentucky.