This study was presented in part at the 2010 Annual Meeting of the American Society of Clinical Oncology; June 4-8, 2010; Chicago, IL.
A history of multiple primary melanomas (PMs) has been associated with improved survival in patients with early stage melanoma, but whether it also is correlated with survival in patients with metastatic melanoma is unknown. The authors sought to address the latter question in the current study.
Patients with metastatic melanoma diagnosed at the Melanoma Institute Australia between 1983 and 2008 were identified. Overall survival (OS) was calculated from date of first distant metastasis. Survival analysis was performed using the Kaplan-Meier method, log-rank tests, and multivariate Cox proportional hazards models.
Of 2942 patients with metastatic melanoma, 2634 (89.5%) had 1 PM and 308 (10.5%) had >1 PM. Factors that were associated independently with shorter OS were site of metastasis, including the brain (hazard ratio [HR], 2.41; 95% confidence interval [CI], 2.07-2.81; P < .001) and nonlung viscera (HR, 1.92; 95% CI, 1.67-2.22; P < .001, vs lymph node/subcutaneous/soft tissue), age >60 years (HR, 1.23; 95% CI, 1.12-1.36; P < .001), shorter disease-free interval from PM to first distant metastasis (≤12 months vs >36 months: HR, 1.62; 95% CI, 1.39-1.89; P < .001), and fewer PMs (1 vs >1; HR, 1.26; 95% CI, 1.08-1.47; P = .004).
Large prognostic and staging studies, notably those used to develop the American Joint Committee on Cancer (AJCC) melanoma staging system,1-3 have clearly established that patients with metastatic melanoma have a dismal prognosis. However, these studies usually exclude patients with multiple primary melanomas, because it is difficult to establish with certainty which of each patient's primary melanomas is responsible for the metastatic disease and, thus, to unambiguously associate clinical and pathologic characteristics of the primary tumor with clinical outcomes. Consequently, the biology of melanoma in these patients is poorly understood, and the prognostic significance of having multiple primary melanomas is not clear.
Prognostic factors associated with survival from the time of diagnosis in patients with early stage primary melanoma are well established and include several clinical and primary tumor characteristics, such as Breslow tumor thickness, Clark level, ulceration, and tumor mitotic rate.2-7 In patients who have developed metastatic melanoma, established prognostic factors include metastasis (M) classification, serum lactate dehydrogenase (LDH) level, and Eastern Cooperative Oncology Group (ECOG) performance status.8, 9 In patients with early stage melanoma, a history of multiple primary melanomas is associated with improved survival,10-12 but it is unknown whether this association also holds true in patients with metastatic melanoma.
The objective of this study in a large cohort of patients with metastatic melanoma who received treatment at a single institution was to determine the clinical and pathologic factors predictive of survival from the date of diagnosis of first metastasis. In particular, we sought to determine the nature of the association of a history of multiple primary melanomas with postmetastasis survival.
MATERIALS AND METHODS
Patient Selection and Data Collection
Clinical and pathologic data from all patients with melanoma who receive treatment at Melanoma Institute Australia (MIA) are recorded in the MIA research database. From this database, patients who were diagnosed with metastatic melanoma between January 1983 and December 2008 and who had a history of at least 1 cutaneous primary melanoma were identified and included in this study. Patients with ocular, mucosal, or uveal primary melanomas were excluded (Fig. 1, top).
Assignment of Culprit Primary Melanoma in Patients With Multiple Primary Melanomas
In patients with multiple cutaneous melanomas, it is important to ascertain whether tumors diagnosed subsequent to the first primary melanoma are independent primary tumors or cutaneous metastases, and to determine which of the primary melanomas is likely to be responsible for causing the metastatic disease (designated the “culprit” primary melanoma). In the current study, patients were designated as having multiple independent primary tumors on the basis of the clinical and pathologic features of the tumors, as described previously.13 In these patients, the culprit primary melanoma was assigned using an algorithm (which has been used in other studies14) based on clinicopathologic criteria, such as regional lymph node or in transit melanoma involvement, time interval between diagnosis of primary and metastatic disease, and AJCC tumor (T) classification (Fig. 1, bottom).
Associations of clinical and pathologic parameters with survival were evaluated. The parameters included clinical details (patient age at diagnosis of first metastasis, sex, ECOG performance status,8 serum LDH level), the number of primary melanomas, details of the culprit primary melanoma (year of diagnosis, anatomic location, and histopathologic subtype, Breslow thickness, Clark level of invasion, dermal mitotic rate, and presence of ulceration), distant disease-free interval (DFI) (defined as the interval between diagnosis of the culprit primary melanoma and first metastasis), follow-up intensity (number of visits per 6-month period during the DFI), and details of the first metastasis (year of diagnosis and anatomic location).
Statistical analysis was carried out using IBM SPSS Statistics 17.0 software (IBM Corporation, Armonk, NY). In those with multiple primary melanomas, characteristics of the culprit primary melanoma were used in the analyses. Overall survival was defined as the interval between diagnosis of the first distant metastasis and death. Univariate associations of each of the parameters with overall survival were assessed using the Kaplan-Meier method, and differences in survival were evaluated with the Mantel-Cox log-rank test. Multivariable Cox regression models were used to assess associations of each of the parameters with survival after adjusting for the other parameters. The criterion for entry of each parameter into the multivariable model was a P value < .1 in the univariate analysis.
In total, 2942 patients (1003 women [34.1%] and 1939 men [65.9%]) were included in this study, of whom 2634 (89.5%) had a single primary melanoma. The culprit primary tumor was assigned in 308 patients (10.5%) who had >1 primary melanoma (Fig. 1), including 229 patients (7.8%) who had 2 melanomas, 51 patients (1.7%) who had 3 melanomas, and 28 patients (1%) who had >3 primary melanomas.
Patients who had a history of ≥2 primary melanomas had significantly improved survival from the time of diagnosis of first distant metastasis compared with those who had a single primary melanoma (median survival, 11.2 months and 7.7 months, respectively; P < .001) (Fig. 2). The number of primary melanomas remained a significant prognostic factor after accounting for clinicopathologic parameters pertaining to the primary and metastatic tumors in a multivariate analysis (Table 1). No other feature of the primary melanoma was associated significantly with survival after the diagnosis of metastatic melanoma, including Breslow thickness, mitotic rate, or ulceration. Other factors that were associated significantly with survival after diagnosis of metastatic disease in multivariate analysis were the DFI from primary diagnosis to first distant metastasis, age at first distant metastasis, and anatomic site of first metastasis (Table 1). Age, sex, higher mitotic rate, and ulceration in the primary melanoma were associated significantly with survival only in univariate analysis. Follow-up intensity was significantly higher in patients with >1 primary melanoma than in patients with 1 primary melanoma (Table 2), but it was not associated significantly with survival when adjusted for other clinicopathologic factors (Table 1). There was no statistically significant difference in the mean DFI between patients with 1 primary melanoma and those with >1 primary melanomas (Table 2).
Table 1. Clinical and Pathologic Factors Associated With Overall Survival
Compared with patients with 1 primary melanoma, patients who had >1 primary melanoma were older and more commonly were men; they had culprit primary melanomas that were diagnosed in more recent decades, more often located on the head and neck, and less often ulcerated; they were reviewed more frequently during the distant DFI; and they had their metastases diagnosed in more recent decades. There was no statistically significant association between the number of primary melanomas and each of the following parameters evaluated at the time of first diagnosis of metastatic disease: anatomic sites of metastasis; number of metastases; and serum LDH level (Table 2).
Because patients with >1 primary melanoma were diagnosed in more recent decades than those with 1 primary melanoma, we separately examined the patients who were alive at the time of last follow-up for any differences, particularly in the DFI, follow-up intensity, follow-up duration, and the age and year of diagnosis of first metastasis, that potentially may have biased the results (Table 3). The only significant differences were in the total duration of follow-up from primary melanoma, the duration of follow-up from first metastasis, and age at diagnosis of first metastasis. All of these factors were significantly higher in patients who had >1 primary melanoma (Table 3).
Table 3. Associations of the Number of Primary Melanomas With Temporal Factors in the Subgroup of Patients Who Were Alive at Last Follow-Up
Follow-up intensity: No. of visits per 6-mo period during DFI
The multivariable models included 2017 patients (68.6%) who had complete information available on all evaluated parameters. However, data on serum LDH level and ECOG performance status was recorded in only 391 patients (13.3%) and 641 patients (21.8%), respectively. Therefore, these parameters were not included in the main multivariate Cox regression model (Table 1), although both of these factors were associated significantly with overall survival in univariate analyses (Table 4, Fig. 3). When multivariable models that included serum LDH alone, ECOG status alone, or both variables were performed, both were associated independently with survival (Table 5).
Table 4. Associations of Serum Lactate Dehydrogenase Level and Eastern Cooperative Oncology Group Status With Overall Survival
Chi-square statistic (3 df) = 108.1(3 df); P < .001
0: Fully active, no restriction
1: Restricted, light work only
2: Up >50% of d, no work
3: Bed/chair >50% of d
Table 5. Clinical and Pathologic Factors Associated With Overall Survival: Models That Included 1) Serum Lactate Dehydrogenase (LDH) Level Alone, 2) Eastern Cooperative Oncology Group Performance Status (ECOG PS) Alone, and 3) Both LDH Level and ECOG PSa
In this study, we evaluated the prognostic significance of a variety of clinical and pathologic factors, some of which relate to the primary tumor. In patients with multiple cutaneous melanomas, it is important to determine whether tumors diagnosed subsequent to the first primary melanoma are independent primary tumors or cutaneous metastases. Correlation of several clinical and pathologic features,15 such as anatomic site, presence of an associated precursor/in situ lesion, lymphatic invasion, and dense lymphocytic inflammation, may assist in accurate classification, although both primary and metastatic melanomas may share some of these characteristics. A recent study demonstrated that, in the majority of patients who are designated as having multiple independent primary tumors on the basis of their clinical and pathologic features, there was no evidence of clonal origin of the multiple melanomas, thereby supporting their designation as independent primary tumors.13 These results validated the criteria used by experienced clinicians and pathologists in accurately distinguishing primary melanomas from cutaneous melanoma metastases. In the current study, these criteria allowed the confident distinction of patients who had a single primary melanoma from those who had multiple primary tumors.
In patients with multiple melanomas, we applied an algorithm (Fig. 1) to determine which of the primary melanomas was most likely to have given rise to metastatic disease. This algorithm was based on the logical application of clinicopathologic criteria, such as lymphovascular drainage, time interval between diagnosis of primary melanoma and metastatic melanoma, and thickness of the primary melanoma. In the absence of clonality studies to definitively establish the biologic relations between the primary melanomas and metastases, we judiciously selected and applied the criteria in the algorithm to identify the culprit primary melanoma (namely, the primary melanoma deemed to have the greatest impact on clinical outcome), as described previously.14
ECOG performance status and serum LDH levels were strongly associated with survival in univariate analysis (Fig. 3). When the analysis was adjusted for other predictive factors, these factors retained their strong prognostic utility (Table 5), as demonstrated in several previous studies.9, 16-18 Although the numbers of patients (471 and 314, respectively) with these data included in the model compare favorably with those in prior studies,3, 8, 9 they represent only a small proportion (16% and 10.7%, respectively) of the total number of patients in the study. Therefore, their inclusion in the main multivariate model (Table 1) would have resulted in the analysis of only a small subgroup of the entire cohort, which, in turn, would have been subject to the biases attendant in such restrictive analyses. This argument also applies to the question regarding whether the number of primary melanomas is a prognostic factor independent of serum LDH level and ECOG performance status. In the multivariate models that included these parameters, although the presence of >1 primary melanomas was a favorable prognostic indicator independent of ECOG performance status alone (n = 471; 16%), it was not independent of serum LDH level alone (n = 314; 10.7%) or ECOG status and serum LDH level together (n = 183; 6.2%) (Table 5). However, these results are derived from an analysis of a small subgroup of the study cohort and, thus, should be interpreted with great caution. Study of large populations in which all 3 parameters are recorded will be required to determine whether the prognostic significance of the number of primary melanomas is independent of serum LDH level and ECOG performance status.
In contrast to the strong prognostic value of primary melanoma features (such as Breslow thickness, ulceration, and mitotic rate) for predicting survival (and the development of metastasis) in patients with localized melanoma,2-4, 7 these characteristics were not associated independently with survival after the development of metastasis. After the development of metastasis, progression to death was slower (equating to better survival) in patients who had >1 primary melanoma than in patients who had 1 primary melanoma. Previous studies have demonstrated an association of multiple primary melanomas with improved survival from the time of primary tumor diagnosis in patients with clinically localized melanoma10-12; however, to our knowledge, ours is the first study to demonstrate an association between the number of primary melanomas and survival in patients with metastatic melanoma. To further explore possible factors that account for the better survival observed among patients with >1 primary melanoma, we analyzed differences between the group with 1 primary melanoma and the group with >1 primary melanoma and explored 3 areas of potential bias: “later in time” diagnosis, lead-time bias, and “missing data” because of shorter follow-up.
Advances in diagnostic and therapeutic options over time may contribute to better survival in patients diagnosed in more recent years (“later in time” diagnosis) than in those diagnosed in earlier years. Although the decade of diagnosis of first metastasis was significantly more recent in patients with >1 primary melanoma than in patients with 1 primary melanoma, the decade in which the first metastasis was diagnosed did not prove to be an independent predictor of survival after adjustment for other parameters in the multivariate model.
Second, we evaluated potential lead-time bias. Although the follow-up intensity differed significantly between those with 1 primary melanoma and those with >1 primary melanoma, it was not correlated independently with survival. To determine whether metastasis was detected earlier in patients with >1 primary melanoma than in those with 1 melanoma (which may artifactually produce the impression of longer postmetastasis survival), we compared the DFI in these groups, and there was no statistically significant difference in the mean DFI between these patient groups. Together, these findings suggest that lead-time bias does not account for the observed survival benefit in patients with multiple primary melanomas.
Finally, we considered the possibility that patients who had >1 primary melanoma may not have been followed for as long after development of their metastases as patients who had 1 primary melanoma, resulting in censoring of a greater proportion of the former patients in the analysis and a false impression of improved survival. However, when we compared the patients who had 1 primary melanoma with those who had >1 primary melanoma who were still alive at the last follow-up, we discovered that follow-up duration after first metastasis was significantly longer in the latter group of patients. DFI and follow-up intensity did not differ not significantly between these groups. These findings support the thesis that these temporal factors are not confounding factors that can explain the improved survival of patients with >1 primary melanoma.
Although there was no significant difference between the patient groups with 1 primary melanoma and those with >1 primary melanoma with regard to the time to progression from primary melanoma to first metastasis (ie, the DFI), there was a significant difference in the rate of progression to death after metastasis (ie, survival). This difference appears to be real, because other potential confounding factors to explain this survival difference have been excluded. It is possible that the biology of metastases in patients who have >1 primary melanoma is such that the metastases behave less aggressively than those in patients who have 1 primary melanoma. For example, patients with v-raf murine sarcoma viral oncogene homolog B1 (BRAF)-mutant, metastatic melanoma are more likely to have a history of single or occult primary melanomas.19 Patients with BRAF-mutant melanoma also have a worse survival from first metastasis despite a disease-free survival similar to that in patients with BRAF wild-type melanoma.19 An alternate, but not mutually exclusive, possibility is that host factors (such as immune response) differ in patients with >1 primary melanoma and those with 1 primary melanoma. This may be related to the priming of the immune system by exposure to multiple clones of tumor cells from more than 1 primary tumor site in patients who have >1 primary melanoma. These clones may have greater antigenic diversity (and, thus, elicit a more broad-ranging host immune response) than those derived from a single primary melanoma. Therefore, after the development of metastasis, the host immune system in patients with >1 primary melanoma may be more capable of slowing tumor progression. Further studies will be required to elucidate the biologic mechanisms that underpin the differences in clinical behavior in these patient groups.
In summary, we report for the first time that the number of primary melanomas is an independent prognostic factor in patients with metastatic melanoma. This factor should be incorporated into future studies of prognosis and staging in patients with metastatic melanoma and when evaluating patient eligibility for clinical trials and interpreting their results. Further mechanistic studies will be required to explain the biologic and immunologic processes underlying the prognostic significance of the number of primary melanomas.
Note Added in Proof
This work was supported by the Harry Lloyd Charitable Trust (Translational Research Award to R.M.); the Center for Microbes, Inflammation, and Cancer at Memorial Sloan-Kettering Cancer Center (Lucille Castori Fellowship to R.M.); Cancer Institute New South Wales (fellowships to R.A.S. and G.V.L.); the University of Sydney (International Visiting Research Fellowship to M.B.A.); an New South Wales Health Department and Sydney West Area Health Service Infrastructure Grant to Westmead Millennium Institute; a National Health and Medical Research Council of Australia Program Grant (402761); and a Cancer Institute NSW Translational Research Program Grant (05/TPG/1-01).