|Home | About | Journals | Submit | Contact Us | Français|
To determine the survival rates and independent predictors of survival using a contemporary international cohort of patients with stage III melanoma.
Complete clinicopathologic and follow-up data were available for 2,313 patients with stage III disease in an updated and expanded American Joint Committee on Cancer (AJCC) melanoma staging database. Kaplan-Meier and Cox multivariate survival analyses were performed.
Among all 2,313 patients with stage III disease, 81% had micrometastases, and 19% had clinically detectable macrometastases. The 5-year overall survival was 63%; it was 67% for patients with nodal micrometastases, and it was 43% for those with nodal macrometastases (P < .001). Tremendous heterogeneity in survival was observed, particularly in the microscopically detected nodal metastasis subset (from 23% to 87% for 5-year survival). Multivariate analysis demonstrated that in patients with nodal micrometastases, number of tumor-containing lymph nodes, primary tumor thickness, patient age, ulceration, and anatomic site of the primary independently predicted survival (all P < .01). When added to the model, primary tumor mitotic rate was the second-most powerful predictor of survival after the number of tumor-containing nodes. In contrast, for patients with nodal macrometastases, the number of tumor-containing nodes, primary ulceration, and patient age independently predicted survival (P < .01).
In this multi-institutional analysis, we demonstrated remarkable heterogeneity of prognosis among patients with stage III melanoma, especially among those with nodal micrometastases. These results should be incorporated into the design and interpretation of future clinical trials involving patients with stage III melanoma.
The technique of lymphatic mapping and sentinel node biopsy is now widely used as the standard method of staging patients deemed to have significant risk of clinically occult nodal metastases.1–8 This approach greatly facilitates early identification of nodal metastases relative to clinical examination or imaging studies, and it has led to a dramatic shift in the prognosis of patients with melanoma metastatic to regional lymph nodes.2,9–14 By definition, in the American Joint Committee on Cancer (AJCC) staging manual, clinically occult nodal metastases of melanoma found at sentinel node biopsy or elective lymphadenectomy are classified as micrometastases, whereas nodal metastases found by clinical or radiologic evaluation and confirmed pathologically are classified as macrometastases.15,16 Importantly, these designations do not refer directly to the size or visibility of the nodal metastases, but rather they have been defined by the mode of detection. Previously, most patients with stage III melanoma had nodal macrometastases,17–19 whereas most contemporary patients with stage III melanoma have nodal micrometastases.20 To determine whether previously developed staging criteria also apply to contemporary patients whose initial staging demonstrates nodal micrometastases, we performed a comprehensive multivariate analysis to identify and compare independent predictors of survival for patients with melanoma who have nodal micrometastases compared with those with macrometastases. This analysis included mitotic rate as a potential prognostic factor. Mitotic rate was previously demonstrated to be a significant adverse predictor of survival in patients with localized (ie, stage I and II) melanoma.10,21–26
The AJCC melanoma staging database was created in 1999 by an international collaboration that combined prospective melanoma databases from major cancer centers and clinical trial cooperative groups.11,20 A recently updated version of this database used to define the recently published seventh edition of the AJCC staging manual was queried to identify a subset of 2,313 patients with stage III disease who had complete clinical and pathologic data on their primary melanoma and regional nodal metastases. Patients with stage III disease included in this AJCC database were contributed by nine major medical centers, freestanding cancer centers, or cooperative groups (Data Supplement, online only). Definitions for micrometastases and macrometastases were operational and were similar to those used in the AJCC sixth15 and seventh editions of the melanoma staging system.16,20 More than 90% of the micrometastases were detected by the sentinel node biopsy technique.
Data definitions were described previously.11,15,20 As detailed in the seventh edition of the AJCC melanoma staging system, mitotic rate was recorded as number of mitoses per square millimeter of primary tumor.16,20 Of note, the present analysis contended with a mitotic rate that was not reported in all databases, and in some the rate was only available as a categoric rather than as a continuous variable.
Survival curves were calculated from time of diagnosis of primary melanoma and considered censored for patients alive at last follow-up or dead without evidence of melanoma. There were 86 patients for whom the cause of death could not be determined (12.4% of the 694 patients who died). The 31 patients who had metastases at the time of death were included in the survival calculation, whereas the remaining 55 patients were censored in the survival calculations. Standard statistical approaches were used: melanoma-specific survival curves were generated by the Kaplan-Meier product-limit method and were compared using the log-rank test, and multivariate analyses were based on the Cox proportional hazards model.11,15,20 For the Cox multivariate analyses, the relative importance of the prognostic factors was determined according to χ2 values (with associated df and P values).27
Characteristics of the 2,313 patients with stage III disease are listed in Table 1. Most (81%) presented with clinically occult nodal micrometastases and were men (72% of the macrometastatic group and 61% of the micrometastatic group; Table 1). Compared with patients who had nodal macrometastases, patients who had nodal micrometastases were more likely have a truncal primary melanoma (43% v 35%; P = .0012), and they were less likely to have a head and neck melanoma (13% v 22%; P < .001), a primary greater than 4.0 mm thick (23% v 47%; P < .001), a level V primary (11% v 26%; P < .001), an ulcerated primary (43% v 57%; P < .001), or with two or more tumor-bearing nodes (30% v 63%; P < .001). Patients with micrometastases more often had a single tumor-positive node compared with those with nodal macrometastases (70% v 37%), and they were less likely to have four or more metastatic nodes (5% v 27%; Table 1). In fact, most patients (55%) with nodal micrometastases in this database had a nonulcerated primary and one to three tumor-containing nodes (AJCC stage IIIA), whereas 40% had one to three nodal micrometastases arising from an ulcerated primary (AJCC stage IIIB), and only 5% of patients had four or more nodes containing micrometastases (stage IIIC; Table 2).
Survival was notably better for patients with nodal micrometastases than in patients with macrometastases. Among all patients with micrometastases, the 5-year survival was 67%. Five-year survival rates varied by number of tumor-bearing nodes, ranging from 71% for one involved node to 36% for four or more tumor-bearing nodes (Fig 1A). Survival was significantly worse if the primary melanoma was ulcerated (Fig 1B) or thicker than 2.0 mm (Fig 1C). The 5-year survival rate for patients with nodal macrometastases was 43%, ranging from 50% to 36% as number of tumor-bearing nodes increased (Fig 1D). Survival worsened if primary melanomas were ulcerated (Fig 1E) or were thicker than 6.0 mm (Fig 1F).
When all 1,872 patients with micrometastasis were stratified by tumor thickness, ulceration, and number of involved nodes, there was a remarkable variation of 5-year survival rates, ranging from 87% for a single nodal micrometastasis arising from a nonulcerated primary melanoma ≤ 2.0 mm in thickness, to 23% for patients with four or more nodal macrometastases from an ulcerated primary melanoma greater than 6.0 mm in thickness (Table 3).
Multivariate Cox regression analysis of the 2,313 patients with stage III disease was initially performed with eight pathologic and clinical variables; mitotic rate was at first not included, because this factor had not been included in prior editions of the AJCC staging system. In the present analysis, six factors were independent predictors of survival (all with P < .001), rank ordered by χ2 for number of involved nodes (χ2, 50.3; P < .001), patient age (χ2, 49.4; P < .001), primary ulceration (χ2, 33.7; P < .001), nodal tumor burden (micrometastases v macrometastases (χ2, 23.6; P < .001), primary thickness (χ2, 23.2; P < .001), and anatomic location of the primary (χ2, 14.8; P < .001). Sex and Clark level were not significant factors in this analysis.
Multivariate Cox regression analyses were performed within cohorts of patients with nodal micrometastases or nodal macrometastases (Table 4). For the patients with nodal micrometastases (n = 1,872), independent predictors of survival included patient age, sex, tumor thickness, ulceration, primary anatomic site, and number of tumor-bearing lymph nodes (Table 4). For the patients with nodal macrometastases (n = 441), age, ulceration, anatomic site, and number of tumor-bearing lymph nodes independently predicted survival; sex and tumor thickness were not significant. The number of tumor-containing nodes was the most significant predictor of survival in patients with micrometastases, and the second-most significant predictor of survival in patients with macrometastases (Table 4). Clark level of invasion was not significant in either group.
Additional multivariate Cox regression analyses were performed within cohorts of patients with stage III disease with primary tumor mitotic rate incorporated into the model (n = 1,338). In patients with micrometastases, age, tumor thickness, mitotic rate, ulceration, anatomic site, and the number of involved lymph nodes were independent predictors of survival. Notably, among patients with nodal micrometastases, mitotic rate was the second-most powerful predictor of survival after number of tumor-bearing nodes (Table 5) when mitotic rate was included in the model for patients with nodal macrometastases, and only age and number of tumor-containing nodes were significant independent predictors of survival (Table 5).
Number of involved nodes was the most significant prognostic factor in these analyses. Five-year survival correlated inversely with number of involved nodes regardless of micrometastatic or macrometastatic status (Figs. 1A and and1D;1D; Table 2). Specifically, when stratified by number of tumor-positive nodes and tumor burden, 5-year survival rates for patients with one, two, or three tumor-positive nodes (microscopic v macroscopic) were 71% versus 50% (P < .001), 65% and 43% (P < .001), and 61% and 40% (P = .004), respectively. In contrast, 5-year survival was identical in both groups (36%) when four or more nodes contained tumor. When primary ulceration was included in the stratification, there was tremendous variation in 5-year survival, ranging from 29 to 82% (Table 2).
For patients with mitotic rate data available, the multivariate analysis showed that major characteristics of the primary tumor (ie, thickness, ulceration, mitotic rate, and lesion site, but not Clark level) correlated with survival outcome in the 1,072 patients with micrometastases (Table 5). In striking contrast, none of these histopathologic features of the primary independently predicted survival in the 268 patients with macrometastases (Table 5).
Most patients with stage III disease in this study were younger than 70 years. In patients with micrometastases, there was a significant survival difference when patients were stratified by age in a single-factor analysis. The 5-year survival rate for patients younger than 50 years was 74%, whereas it was 65% for those between 50 and 69 years and was 47% for those 70 years and older (P < .001). For patients with nodal macrometastases, age-related survival differences were not statistically significant when examined by decade, but they were significant when comparing patients younger and older than 70 years (47% v 27%, respectively; P = .001).
These results demonstrate a remarkable variation in survival for patients with stage III melanoma; in particular, the results reveal distinctive patterns of independent predictors of survival when patients were stratified according to tumor burden (ie, nodal micrometastases or macrometastases). Within these strata, patients with microscopic nodal metastases had a more heterogeneous range in 5-year survival that varied between 23% and 87% compared with the somewhat narrower range of 5-year survival among patients with nodal macrometastases between 29% and 51.6%.
Our data also confirmed that existing AJCC stage III melanoma criteria remain valid despite the dramatic shift in clinical presentation from the predominantly macroscopic disease to microscopic disease typically encountered in most contemporary patients. We demonstrated, by multivariate analysis, that mitotic rate, primary ulceration, and primary thickness are all independent prognostic factors for patients with nodal micrometastases.
The number of nodal metastases was the most significant independent predictor of survival among all patients with stage III disease, regardless of whether they had micrometastases or macrometastases. These findings are consistent with multiple published studies.10,11,17–20,24,28–36 However, important differences emerged when these data were stratified by nodal tumor burden. In patients with nodal micrometastases, multiple covariates independently predicted survival, including number of involved nodes, patient age, and several primary melanoma features (ie, thickness, mitotic rate, ulceration, and anatomic site of the primary tumor). Of note, these clinical and pathologic features of melanoma correlated with survival even after micrometastatic tumor was identified in regional nodes. In contrast, for patients with nodal macrometastases, primary melanoma characteristics did not predict survival.
In multivariate analyses, mitotic rate has been identified an important prognostic factor in patients with stages I and II melanoma.10,21–26 There are no prior reports of a relationship of primary melanoma mitotic rate in a multivariate survival analysis of stage III micrometastases. Although tumor thickness and ulceration have been reported as univariate predictors of survival in patients with nodal micrometastases,34,35,37,38 their positive interaction among subgroups in a multivariate survival analysis of stage III micrometastases has been reported only in a single institutional study.35
This study was limited by the relatively small sample size in patients with nodal macrometastases with mitotic rate information available. Inclusion of mitotic rate into the revised AJCC melanoma staging system (seventh edition) occurred in 2010.16 Additional refinement of the role of mitotic rate, particularly among patients with nodal micrometastases, will likely emerge as additional relevant data becomes available.
Older age was an independent adverse prognostic factor among patients in this study, confirming past observations.11,17,20,36–40 Although the mechanisms that underlie this observation remain unclear, a decrease in immune surveillance and/or host response in older patients that could contribute to a higher rate of distant metastases has been postulated.36,37 Gershenwald et al36 recently demonstrated in a cohort of patients with stage III disease identified by sentinel node biopsy, advanced age was an independent predictor of additional non–sentinel node involvement.36 Primary melanoma in older patients may also be associated with different biologic features than those in younger patients. Finally, because causes of death were not independently reviewed for the AJCC database, it is possible that misclassification of some non–melanoma-related deaths as melanoma related could result in an ostensibly lower 5-year survival rate for older patients.
The number of tumor-containing nodes and primary tumor ulceration were incorporated into the melanoma N criteria in the sixth edition (ie, 2001) melanoma staging system.11,15 These criteria, as well as associated stage groupings, remain unchanged in the seventh edition.15,16,20 Although it is not possible to incorporate the six independent variables identified in our multifactorial analysis into the current TNM-based AJCC melanoma staging system, we have recently complemented our analyses and have attempted to overcome some of the limitations inherent in using of multiple and continuous covariates in prognostic factor analyses by developing a clinical scoring system model and multivariate predictive tool under the auspices of the AJCC that include several of the variables presented in this study.41,42 This first-generation Web-based predictive tool (http://www.melanomaprognosis.org) can be used to calculate estimates of 1-, 2-, 5- and 10-year survival by using multiple independent predictors of survival for patients with stage III melanoma.
The results of this comparative analysis of patients with melanoma nodal metastases show a remarkable heterogeneity in the survival rates and a distinctive pattern of independent prognostic factors when comparing patients with nodal micrometastases with those with macrometastases. These results are important not only for melanoma staging but also as critical components in the design and interpretation of melanoma clinical trials for stage III disease, especially in the setting of nodal micrometastases.
The following institutions and cancer cooperative groups generously contributed individual patient data to the 2008 American Joint Committee on Cancer (AJCC) Melanoma Database, and the stage III melanoma data included in this study were contributed by the nine institutions and groups with * indicated. Sydney Melanoma Unit,* Sydney, Australia, (John F. Thompson, MD); Istituto Nazionale Tumori, Milan Italy, (Natale Cascinelli, MD); San Pio X Hospital, Milan Italy (Natale Cascinelli, MD); Memorial Sloan-Kettering Cancer Center,* New York, NY, Daniel G. Coit, MD); The University of Texas M. D. Anderson Cancer Center,* Houston, TX (Jeffrey E. Gershenwald, MD, Merrick I. Ross, MD, Marcella Johnson); John Wayne Cancer Institute, Santa Monica, CA (Donald L. Morton, MD); Netherlands Cancer Institute, Amsterdam, the Netherlands (Omgo Niewig, MD); University of Pennsylvania Hospital, Philadelphia, PA (Keith Flaherty, MD and Phyllis A. Gimotty, PhD); University of Michigan,* Ann Arbor, MI (Timothy Johnson, MD); H. Lee Moffitt Cancer Center,* Tampa, FL (Vernon K. Sondak, MD); University of Alabama at Birmingham (UAB),* Birmingham, AL (Charles M. Balch, MD, Seng-jaw Soong, PhD, and Marshall Urist, MD); Eastern Cooperative Oncology Group (John M. Kirkwood, MD and Michael B. Atkins, MD); Southwest Oncology Group (Vernon K. Sondak, MD); European Organization for Research and Treatment of Cancer (Alexander M.M. Eggermont, MD); Sunbelt Melanoma Trial* (Kelly M. McMasters, MD); Sentinel Lymph Node Working Group* (Stanley P. Leong, MD); Intergroup Melanoma Surgical Trial* (Charles M. Balch, MD).
The following were members of the AJCC melanoma staging committee: Charles M. Balch (Chair), Johns Hopkins Medical Institutions, Baltimore, MD; Jeffrey E. Gershenwald (Vice-chair), The University of Texas M. D. Anderson Cancer Center, Houston, TX; Seng-jaw Soong (Vice-chair), University of Alabama at Birmingham, Birmingham, AL; Michael B. Atkins, Beth Israel Deaconess Medical Center, Boston, MA, and Eastern Cooperative Oncology Group; David R. Byrd, University of Washington, Seattle, WA; Antonio C. Buzaid, Hospital Sirio Libanes, Sao Paulo, Brazil; Natale Cascinelli, Istituto Nazionale Tumori; WHO Melanoma Program, San Pio X Hospital, Milan, Italy; Alistair J. Cochran, David Geffen School of Medicine at University of California, Los Angeles, Los Angeles, CA; Daniel G. Coit, Memorial Sloan-Kettering Cancer Center, New York, NY; Alexander M.M. Eggermont, Erasmus University MC—Daniel den Hoed Cancer Center, Rotterdam, the Netherlands (International Union Against Cancer representative); David Frishberg, Cedars Sinai Medical Center, Los Angeles, CA (College of American Pathologists representative); Keith T. Flaherty, University of Pennsylvania, Philadelphia, PA; Phyllis A. Gimotty, University of Pennsylvania, Philadelphia, PA; Allan C. Halpern, Memorial Sloan-Kettering Cancer Center, New York, NY; Alan N. Houghton Jr, Memorial Sloan-Kettering Cancer Center, New York, NY; Marcella M. Johnson, University of Texas M. D. Anderson Cancer Center, Houston, TX; John M. Kirkwood, University of Pittsburgh Cancer Institute, University of Pittsburgh Medical Center, Pittsburgh, PA, Eastern Cooperative Oncology Group; Kelly M. McMasters, University of Louisville Medical Center, Louisville, KY, Sunbelt Melanoma Trial Group; Martin F. Mihm Jr, Massachusetts General Hospital, Harvard Medical School, Boston, MA; Donald L. Morton, John Wayne Cancer Institute at Saint John's Health Center, Santa Monica, CA; Merrick I. Ross, The University of Texas M. D. Anderson Cancer Center, Houston, TX; Arthur J. Sober, Massachusetts General Hospital, Boston MA; Vernon K. Sondak, H. Lee Moffitt Cancer Center, Tampa, FL, Southwest Oncology Group; Kristen Stephens, Roswell Park Cancer Institute, Buffalo, NY; John F. Thompson, Sydney Melanoma Unit, University of Sydney, Sydney, New South Wales, Australia.
The following were associated with Data Management and Analysis for the AJCC Melanoma Database: Seng-jaw Soong, PhD, Professor, and Senior Biostatistician, UAB; Shouluan Ding, PhD, Biostatistician, UAB; Matthew Dickerson, BS, Programmer Analyst II, UAB; Rush Elliott, BS, Data Manager, UAB; Connie Pitts, Program Coordinator, UAB; Marcella Johnson, The University of Texas M. D. Anderson Cancer Center; Carla Warneke, The University of Texas M. D. Anderson Cancer Center.
Supported by a grant from the American Joint Committee on Cancer; by research funds from the University of Alabama at Birmingham; by SPORE Grant No. P50 CA93459 in melanoma at The University of Texas M. D. Anderson Cancer Center in Houston, TX; and by an unrestricted educational grant from Schering Plough (Kenilworth, NJ; for AJCC Melanoma Staging Committee meetings).
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
Although all authors completed the disclosure declaration, the following author(s) indicated a financial or other interest that is relevant to the subject matter under consideration in this article. Certain relationships marked with a “U” are those for which no compensation was received; those relationships marked with a “C” were compensated. For a detailed description of the disclosure categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors.
Employment or Leadership Position: None Consultant or Advisory Role: Charles M. Balch, Schering-Plough (C); Alexander M. Eggermont, Schering-Plough (C); John M. Kirkwood, Schering-Plough (C); Stanley P. Leong, Schering-Plough (C); Martin C. Mihm Jr, Electro-Optical Systems (C); Merrick I. Ross, Schering-Plough (C), Genentech (C); Vernon K. Sondak, Schering-Plough (C) Stock Ownership: None Honoraria: Jeffrey E. Gershenwald, Schering-Plough; Alexander M. Eggermont, Schering-Plough; John M. Kirkwood, Schering-Plough; Merrick I. Ross, Schering-Plough, Genentech; Vernon K. Sondak, Schering-Plough Research Funding: Alexander M. Eggermont, Bristol-Meyers Squibb, Schering-Plough Expert Testimony: Charles M. Balch, Schering-Plough (C); Alexander M. Eggermont, Schering-Plough (C); Vernon K. Sondak, Schering-Plough (C) Other Remuneration: None
Conception and design: Charles M. Balch, Jeffrey E. Gershenwald, Seng-jaw Soong, John F. Thompson, Daniel G. Coit, Alexander M. Eggermont, John M. Kirkwood, Kelly M. McMasters, Martin C. Mihm Jr, Merrick I. Ross
Provision of study materials or patients: Charles M. Balch, Jeffrey E. Gershenwald, John F. Thompson, Natale Cascinelli, Daniel G. Coit, Timothy Johnson, Stanley P. Leong, Kelly M. McMasters, Donald L. Morton, Vernon K. Sondak
Collection and assembly of data: Charles M. Balch, Jeffrey E. Gershenwald, Seng-jaw Soong, John F. Thompson
Data analysis and interpretation: Charles M. Balch, Jeffrey E. Gershenwald, Seng-jaw Soong, John F. Thompson, Shouluan Ding, David R. Byrd, Natale Cascinelli, Alistair J. Cochran, Daniel G. Coit, Alexander M. Eggermont, John M. Kirkwood, Kelly M. McMasters, Martin C. Mihm Jr, Donald L. Morton, Merrick I. Ross, Vernon K. Sondak
Manuscript writing: Charles M. Balch, Jeffrey E. Gershenwald, Seng-jaw Soong, John F. Thompson, Alistair J. Cochran, John M. Kirkwood, Martin C. Mihm Jr, Vernon K. Sondak
Final approval of manuscript: Charles M. Balch, Jeffrey E. Gershenwald, Seng-jaw Soong, John F. Thompson, Shouluan Ding, David R. Byrd, Natale Cascinelli, Alistair J. Cochran, Daniel G. Coit, Alexander M. Eggermont, Timothy Johnson, John M. Kirkwood, Stanley P. Leong, Kelly M. McMasters, Martin C. Mihm Jr, Donald L. Morton, Merrick I. Ross, Vernon K. Sondak