Prognostic Factors Analysis of Melanoma Patients

Abstract & Commentary

Synopsis: Factors predicting melanoma-specific survival rates were analyzed in 17,600 melanoma patients with complete clinical, pathologic, and follow-up information using the Cox proportional hazards regression model. Results for patients with localized melanomas (stages I and II) demonstrated that tumor thickness and ulceration were the most important predictors of survival, and the Clark’s level of invasion had a significant effect only for melanomas less than 1.0 mm in thickness. The most important predictors of survival for patients with nodal metastases (stage III) were the number of metastatic nodes, tumor burden (microscopic vs macroscopic), and ulceration of the primary melanoma. Patients with stage IV disease and nonvisceral metastases had a better survival than did stage IV patients with visceral metastases. The results of this analysis provide an evidence-based foundation for a revised staging system for cutaneous melanoma.

Source: Balch C, et al. J Clin Oncol. 2001;19: 3622-3634.

Accurate and universally accepted cancer staging systems are essential to accurately identify prognosis for the individual patient and to allow for comparison of treatment results among different centers. Staging systems for cutaneous melanoma stratify melanomas based upon local, regional, and distant disease and have correlated with survival.1 Since most melanoma patients present with localized disease, the recent American Joint Committee on Cancer (AJCC) staging system for cutaneous melanoma used stage I and II to designate low- and intermediate-risk patients with localized disease.1 Stage III identified patients with regional disease and stage IV identified patients with distant metastatic disease.1 Several published studies have improved our understanding of prognostic factors of melanoma, and the tumor-node-metastasis (TNM) criteria and stage groupings used in the recent AJCC staging system have been shown to have many inaccuracies.2 The prior AJCC staging system used thickness as a factor for T-stage with thresholds of 0.75, 1.5, and 4.0 mm, and the Clark’s level of invasion was considered an important determinant of T-stage.3 The variable of ulceration was not considered to be an important prognostic factor for T-stage. A primary prognostic factor for stage III disease was the dimension of the nodal involvement, while the number of nodal metastases and the distinction of microscopic vs. macroscopic nodal involvement were not recognized as clinically important.3 Thus, a comprehensive analysis of prognostic factors of cutaneous melanoma was needed to provide a basis for a new and more accurate staging system.

This study by Balch and colleagues reports on prospectively accumulated melanoma patient data from 13 institutions and cooperative groups. The entire melanoma patient database consisted of 30,450 patients, and 17,600 of these patients (58%) had information available for the factors required for the proposed AJCC classification and stage grouping.3 Follow-up on these patients was available for at least 5 years for 12,837 patients (73%), and 8633 patients (49%) had at least 10 years of follow-up. Six experienced clinical statisticians participated in the analysis. The prognostic variables of tumor thickness and ulceration were identified as the 2 most important independent variables in a multivariate analysis of the 13,581 patients with localized disease. The presence or absence of primary tumor ulceration was the second most powerful prognostic indicator in these patients. When the primary melanomas were divided into thickness categories (less than 1.0 mm; 1.01-2.0 mm; 2.01-4.0 mm; and more than 4.0 mm), the survival rate for ulcerated melanomas was virtually the same as for nonulcerated melanomas of the next greater category. Other significant prognostic factors included patient age (worse with increasing decades of life), site of the primary melanoma (trunk and head and neck sites worse than extremities), level of invasion (most important for the subgroup with thin melanomas less than 1.0 mm), and sex (men worse than women). Important variables for patients with lymph node metastases included the number of metastatic nodes, the tumor burden (defined as microscopic vs macroscopic) and the presence or absence of ulceration of the primary melanoma. Somewhat lower correlation with survival was also seen with the site of the primary melanoma as well as with patient’s age. The number of involved nodes (1 vs 2 to 3 vs 4 or more nodes) was the most significant prognostic factor in these patients. For patients with distant metastases, the survival rates were measured in months rather than years. The most significant differences were noted with visceral vs nonvisceral (skin, subcutaneous tissue, and distant lymph nodes) sites. In addition, patients with lung metastases had a better 1-year survival than did patients with other visceral sites of metastases, but no differences were noted when 2-year survival data were compared. Balch et al conclude that the current analysis provides an evidence-based foundation for a revised melanoma staging system.

Comment by Mark R. Albertini, MD

The melanoma prognostic factor analysis in this report is the largest ever conducted and provides a solid evidence-based foundation for a revised melanoma staging system. This revised system will be valuable for improved prognostic information for the individual patient and for better comparison of results from melanoma clinical trials. Balch et al successfully combined prospective databases to provide a large number of patients for this analysis. In addition, the length of follow-up for these patients (73% for at least 5 years; 49 % for at least 10 years; and 14% for at least 20 years) allowed for meaning comparison of survival impacts of the prognostic factors in this study. Thus, the results from this analysis represent an important contribution to our staging of melanoma patients.

The current study supports the powerful prognostic importance of ulceration of the primary melanoma. This variable of ulceration was important both for primary melanomas as well as for melanomas that had metastasized to the regional nodal basin. The biologic activity of an ulcerated melanoma has been compared to that of poorly differentiated cancers of other histologies and is considered to represent a characteristic of tumors with a greater capacity to metastasize.3 Ulceration is clearly identified as the second most important prognostic variable behind tumor thickness in predicting survival. Balch et al also clearly identify the prognostic importance of regional nodal characteristics of number of nodes and tumor burden. Patients with stage III disease represent a heterogeneous group of patients, and accurate comparison of stage III treatment trials will require an understanding of the number of involved lymph nodes, tumor burden, and presence of ulceration of the primary tumor. In addition, ongoing investigation may identify other factors potentially relevant to prognosis and staging of melanoma patients.4 Ongoing evidence-based analysis will be required of these new variables, and the current study provides a model for this type of analysis.


1. Stadelmann WK, et al. Prognostic Clinical and Pathologic Features. In: Balch, Houghton, Sober, Soong, eds. Cutaneous Melanomas. 3rd ed. St. Louis, Mo: Quality Medical Publishing;1998:11-35.

2. Buzaid AC, et al. J Clin Oncol. 1997;15:1039-1051.

3. Balch CM, et al. J Clin Oncol. 2001;19:3635-3648.

4. Bostick PJ, et al. J Clin Oncol. 1999;17:3238-3244.