A, Distribution by age at diagnosis of single primary melanoma (SPM) or by age at diagnosis of the selected lesion in multiple primary melanoma (MPM). B, Distribution of melanoma-related deaths by age at death in 2372 SPM patients (152 melanoma-related deaths) and 1206 MPM patients (103 melanoma-related deaths).
The curves show melanoma-specific survival for patients with melanomas thicker than 2.00 mm who received a diagnosis before age 70 years or when age 70 years or older. A, Men and women combined. B, Men. C, Women.
eTable 1. HRs for melanoma-related death associated with personal and pathological characteristics and adjusted for age, sex, center, and melanoma status as SPM or MPM
eTable 2. HRs for melanoma-related death associated with age, sex, and pathological characteristics in separate analyses for SPM and MPM patients
Kricker A, Armstrong BK, Goumas C, Thomas NE, From L, Busam K, Kanetsky PA, Gallagher RP, Marrett LD, Groben PA, Gruber SB, Anton-Culver H, Rosso S, Dwyer T, Berwick M, for the GEM Study Group. Survival for Patients With Single and Multiple Primary MelanomasThe Genes, Environment, and Melanoma Study. JAMA Dermatol. 2013;149(8):921–927. doi:10.1001/jamadermatol.2013.4581
Little is known about survival after a diagnosis of a second or higher-order (multiple) primary melanoma, and no study has explored survival in a population-based sample that included patients with single primary melanomas (SPMs) and multiple primary melanomas (MPMs) of any stage. Because people with a first primary melanoma are known to have an increased risk of being diagnosed with another, evidence for prognosis is needed.
To determine whether survival after diagnosis was better in patients with MPMs than with SPMs, as suggested in a recent study.
Survival analysis with median follow-up of 7.6 (range, 0.4-10.6) years.
The Genes, Environment, and Melanoma Study enrolled incident cases of melanoma from population-based cancer registries in Australia, Canada, Italy, and the United States. Multiple primary melanomas were ascertained during a longer period than SPM.
Two thousand three hundred seventy-two patients with SPM and 1206 with MPM.
Diagnosis with melanoma.
Main Outcomes and Measures
Melanoma-specific fatality hazard ratios (HR) and 95% confidence intervals associated with clinical and pathological characteristics of SPM, MPM, and both in Cox proportional hazards regression models.
Melanoma thickness was the main determinant of fatality (HR for >4 mm, 7.68 [95% CI, 4.46-13.23]); other independent predictors were ulceration, mitoses, and scalp location. After adjustment for these other predictors, we found little difference in fatality between MPM and SPM (HR for MPM relative to SPM, 1.24 [95% CI, 0.91-1.69; P = .18]). Thicker SPM, however, had higher fatality (HR for >4 mm, 13.56 [95% CI, 6.47-28.40]) than thicker MPM (2.93 [1.17-7.30]).
Conclusions and Relevance
Although overall fatalities due to SPM and MPM were similar, relative fatality for thicker SPM was greater than that for thicker MPM. This finding may offer support for a difference in outcome between patients with SPM and MPM related to factors other than closer surveillance and earlier diagnosis. The better outcomes are worth further exploration.
Subsequent melanomas in patients with multiple primary melanomas (MPMs) are well known to be thinner, on average, than the prior melanomas and might therefore have a better outcome.1- 3 Whether survival is better or worse for patients diagnosed with MPMs rather than a single primary melanoma (SPM) is important for prognosis. Results from earlier studies4- 7 examining survival of patients with SPMs and MPMs offer little consistent information; the studies included few patients, limited analysis to selected stage categories, or presented crude survival estimates only. A recent assessment based on 4952 patients with stage I or II SPMs and 298 with stage I or II MPMs from a melanoma clinic suggested that patients with MPMs survived longer than those with an SPM after taking account of other prognostic factors.8 Among possible mechanisms for an apparent survival benefit, the authors discuss whether multiple melanomas may provoke a stronger host immune response,8 which might also explain better survival for patients with MPMs than for those with SPMs after diagnosis of metastatic disease.9 In addition, the possibility that initial and subsequent melanomas are thinner in MPMs than in SPMs has given rise to speculation that MPM may differ biologically from SPM.2,10
No study has yet explored survival in a population-based sample that included patients with an SPM and patients with MPMs of any stage. Our analysis was undertaken to determine whether survival after diagnosis with MPM was better than after diagnosis with SPM in the large, population-based Genes, Environment, and Melanoma (GEM) Study. Specifically, we analyze detailed pathological features of the melanomas to identify their contribution to overall survival and to survival in MPM and SPM separately. We also explored whether differences in these contributions suggest biological differences between MPM and SPM that affect outcome.
The GEM Study enrolled incident cases of primary melanoma registered in population-based cancer registries in Australia, Canada, Italy, and the United States.1,11 Briefly, SPMs were diagnosed as first invasive primary melanomas from January 1 through December 31, 2000; MPM, as second or higher-order invasive or in situ melanomas from January 1, 1998, through December 31, 2003 (hereinafter referred to as the recent MPM). All patients provided written informed consent. Institutional review boards at the Memorial Sloan-Kettering Cancer Center and each contributing center approved the study protocol.
A total of 2372 SPMs and 1206 MPMs were eligible for the survival analyses. Breslow thickness and anatomic site were abstracted from reports of the community pathologists and expert reviewers,1 whereas ulceration and mitoses were available only from the reviewers; more SPMs (86.8%) than recent (79.4%) or prior (64.8%) MPMs had slides available for review. Information on nodal metastases was unavailable. Among other items, patients were asked at interview for educational level, used as a surrogate for socioeconomic status in these analyses, and history of melanoma in first-degree relatives.
We analyzed melanoma-specific survival in SPM and MPM patients by tumor pathological characteristics. For MPM patients, we included in the analysis the pathological characteristics of the melanoma considered most likely to cause death, which were selected using an algorithm based on stage at and the time between diagnosis of the melanomas.9 Characteristics of the recent MPM were used in statistical models when melanomas were diagnosed more than 5 years apart, except when the most recent MPM was in situ, because the prior melanoma is less likely to be responsible for clinical outcome in patients who survive more than 5 years after diagnosis. For melanomas less than 5 years apart, we used the characteristics of the MPM with the highest known tumor category according to the American Joint Committee on Cancer (AJCC). When both melanomas were at least 1 mm thick, the presence of mitoses was considered to indicate the higher T category. Patient follow-up to ascertain deaths finished at the end of 2007 in most centers and 2008 in British Columbia and Torino, Italy.
Patient age at diagnosis of SPM or the selected MPM was designated for analysis. Cumulative percentages of incidence and mortality were plotted by 5-year age groups to investigate the apparent rarity of deaths at younger ages in MPM compared with SPM patients. For survival analyses, melanoma-specific survival time was accumulated from the diagnosis date of the SPM or the most recent MPM to melanoma-related death; all other cases were censored at the end of follow-up or at death from other causes. We used Kaplan-Meier methods to estimate survival proportions in SPM and MPM separately and to test for differences using the log-rank test. We also used Kaplan-Meier methods to estimate survival proportions in SPM and MPM separately and to construct survival curves for all SPM and MPM by thickness (≤2.00 and >2.00 mm) and age (<70 and ≥70 years) and tested the associations between survival and these factors using the log-rank test; all P values less than .05 denoted a statistically significant difference between the two.
Hazard ratios (HRs) and 95% confidence intervals associated with SPM or MPM status, Breslow thickness, Clark level, presence of vertical growth, ulceration, mitoses, and site were estimated in Cox proportional hazards regression models for all SPMs and MPMs, including as covariates age as a continuous variable, sex, and study center. To display the influence of age, we present estimates for age grouped in decades. Cases with missing values were included in these models by using missing values categories for relevant variables. We constructed a variable for AJCC tumor subcategories by using thickness and ulceration or mitoses to define T1 and thickness and ulceration for T2, T3, and T4. Information on metastases was unavailable.
We also estimated HRs associated with SPM or MPM status, thickness, ulceration, mitoses, and site in multivariable models for all SPMs and MPMs, adjusted for age, sex, and center and including educational level and family history as potential confounders; the models excluded patients with missing information except where family history was missing, for which a missing value category was used to retain 6% of patients lacking this information. To assist in evaluating whether a difference existed in effect estimates for thickness, ulceration, mitoses, and site by SPM or MPM status, we estimated P values for interaction by using the Wald test to compare the multivariable model with main effects only and a model with main effects and the cross-product term.
The associations of pathology variables with survival were evaluated separately for men and women and for SPM and MPM, and we examined the influence of time between MPMs by using 3 time intervals (<3 months, 3 months to 5 years, and >5 years). All P values were 2-sided, and P < .05 was considered statistically significant. Tests based on Schoenfeld residuals and graphical methods using Kaplan-Meier curves in commercially available software (STATA/SE, version 8.2; StataCorp) showed little evidence that the proportional hazards assumption was violated. We used a different software package (SAS, version 9.1; SAS Institute, Inc) for all other statistical analyses.
A total of 563 deaths occurred, including 255 deaths from melanoma (152 SPM and 103 MPM deaths). The median follow-up times were 8 years from diagnosis of SPM and 6.6 years from diagnosis of the most recent MPM (7.6 years for all patients; range, 0.4-10.6 years). Kaplan-Meier survival estimates were more favorable at 7 years for SPM than for MPM (93.8% vs 90.3%; P = .001), although they were identical at 3 years (97.2%) and very similar at 5 years (95.1% for SPM and 94.1% for MPM). Compared with SPM, dates of diagnosis of the most recent MPM and death from melanoma were shifted toward older ages; no deaths occurred before 50 years of age among MPM patients (Figure 1). The MPM patients were older (mean age, 63 [range, 7-95] years) than SPM patients (mean age, 55 [range, 9-97] years), and the mean age at death was 65 (range, 23-93) years in SPM and 73 (55-90) years in MPM. The MPM patients were more likely to be men (65.9% vs 34.1% women) than were those with SPM (51.7% men vs 48.3% women). More of the MPMs thicker than 2 mm occurred in men than women (73.2% vs 26.8% for women).
The most recent MPM was generally thinner than the previous one,1 although lesions thicker than 4 mm were as likely in MPM (4.9%) as SPM (5.1%; P = .70). The mean tumor thickness in men was similar for SPMs and MPMs (1.45 and 1.43 mm; P = .11), but women had thicker SPMs than MPMs (mean thickness of 1.20 mm for SPMs vs 1.11 mm for MPMs; P < .001).
As a preliminary to comparing survival in SPM and MPM, we analyzed the effects of other clinical and pathological factors on survival in models, including age, sex, and study center. Deeper invasion, whether assessed by Breslow thickness, Clark level, or the presence of vertical growth, increased the risk of melanoma-related death strongly (eTable 1 in the Supplement). Unadjusted 5-year survival in patients with melanoma thicker than 2.00 mm was 72.4% at 70 years or older and 84.8% before 70 years of age and similar for men and women, although deaths were rare in women younger than 70 years (Figure 2).
We constructed a multivariable model to include all variables with P < .1 in single-variable analyses, adjusted for age, sex, and center and including family history and educational level; the multivariable model included 2770 patients with complete information. Increasing thickness (HR for >4.00 mm, 7.68 [95% CI, 4.46-13.23]), ulceration, presence of mitoses, and scalp location (HRs, ≤2.25) were each independently associated with increased melanoma fatality (P < .05 in each case), and being younger than 40 years was associated with reduced risk (HR, 0.44 [95% CI, 0.23-0.84]) (Table 1). Statistical evidence suggested that the relationships between thickness and fatality were different for SPM and MPM (P = .004 for interaction) but not between ulceration or mitoses and fatality. Although fatality was much lower in women when adjusted for age and center only (HR, 0.56 [95% CI, 0.43-0.75] relative to men), we found no evidence of a difference in fatality in the fully adjusted model (HR, 0.86 [95% CI, 0.62-1.20]). In separate analyses for men and women, the patterns of increasing risk with increasing thickness and an increased HR with ulceration were also present, whereas mitoses increased fatality in men (by 80%) but not in women (data not shown).
The HRs for AJCC Tb subcategories were approximately 1.5- to 2.0-fold higher than for Ta subcategories for each thickness category relative to melanomas no thicker than 1.00 mm with no ulceration or mitoses (Table 2). The patterns of risk by AJCC tumor category were very similar in separate analyses in men and women as in both sexes combined (data not shown). Fewer women than men had ulcerated MPMs (5.4% of all melanomas in women and 9.3% in men); 5-year survival proportions by ulceration in both sexes were 77% with ulceration and 96% in its absence.
When we examined SPM and MPM in separate multivariable models, fatality due to SPM increased with increasing thickness to an HR of 10.45 (95% CI, 5.21-20.97) for 2.01 to 4.00 mm and 13.56 (6.47-28.40) for greater than 4.00 mm. For MPM, however, HRs for all categories of thickness ranged from 2.13 to 2.93, with no evidence of a trend (HR for >4 mm, 13.56 [95% CI, 6.47-28.40]) (eTable 2 in the Supplement). Ulceration and mitoses had HRs ranging from 1.50 to 2.02 for SPM and MPM, but the estimates were statistically significant only for ulceration in SPM and mitoses in MPM. Melanoma on the scalp increased risk 2.5-fold, and being younger than 60 years at diagnosis reduced risk substantially for SPM and MPM. We found no evidence of variation in fatality due to MPM by time interval between the melanomas (results not shown).
Contrary to the finding by Doubrovsky and Menzies8 of more favorable survival for MPM than SPM after controlling for prognostic factors, we found the fatality after a diagnosis of MPM to be overall a little higher than that after a diagnosis of SPM but not significantly so. Our study, however, was population based and included melanomas of any stage, whereas the previous study included stages I and II melanomas only. Thickness was the most important determinant of survival in the GEM Study for patients with SPM and MPM, as it is for melanoma generally.12,13 Ulceration and the presence of mitoses among pathological characteristics and location on the scalp were also independently associated with increased fatality due to SPM and MPM, as was increasing age.
Thickness is a well-established prognostic factor for survival in primary melanomas, with relative survival proportions of 50% for thickness of 4.00 mm or greater but nearly 100% for melanoma thinner than 0.75 mm at 5 and 10 years after diagnosis.12 Similarly in the GEM Study, survival was much poorer for patients with melanoma thicker than 4.00 mm compared with 1.00 mm or less; in addition, we observed a relative risk of death after a melanoma thicker than 4.00 mm that was 4.6 times higher for SPM than for MPM despite similar proportions of lesions thicker than 4.00 mm in both groups of patients. Doubrovsky and Menzies8 reported a significant effect of thickness on melanoma-related death in their study but did not report on SPM and MPM separately. Another study investigated whether a less aggressive biology in MPM, represented by the mitotic rate, was a possible explanation for the generally thinner lesions in MPM than in SPM, but they concluded that the mitotic rate was not an explanation and that other factors, such as increased surveillance or genetic constitution, probably contribute.10
We evaluated the prognostic significance for SPM and MPM of a number of other pathological factors known to be associated with a poor outcome in primary melanomas. Risks of death were higher for a given thickness in the presence of ulceration than in its absence, which is consistent with the increased severity ulceration confers at each stage in AJCC melanoma tumor category definitions.14 In the GEM Study, ulceration was as likely in SPM and MPM (7.1% and 7.7%, respectively) as elsewhere,15 but we found little evidence of a difference between SPM and MPM in its effect on survival. Similarly, mitotic rate is an accurate and independent predictor of survival,14,16- 20 but its effect showed little difference between SPM and MPM. That similar mitotic rates have been observed for SPM and MPM suggests little difference in underlying biological aggressiveness between SPM and MPM lesions.10
We also examined other clinical factors. A poorer outcome for scalp than for facial sites, usually in older men,12,21,22 has not been investigated fully in previous relevant MPM studies4,8 but is considered to be due to tumor factors.12,20- 22 Scalp melanomas in the GEM Study were similarly fatal in SPM and MPM and apparently had a higher fatality in women than in men, consistent with possible late-stage diagnosis in women (35.7% of scalp SPMs in women were >2.00 mm).
Survival was more favorable in younger people in our study and worsened with increasing age. Being 70 years or older has been suggested to worsen melanoma prognosis,12,16,17,23 an outcome that is independent of thicker lesions in older patients,24 probably because age is a surrogate for a declining host defense system.14,17
The poorer survival in men than in women reported for primary melanomas12 is suggested to be the result of late-stage diagnosis in men.23,25 We found limited evidence, however, of poorer outcomes in men in our data, consisting mainly of somewhat poorer survival for men with thicker melanomas (Figure 2), and men had more melanomas with ulceration. The relatively small number of women 70 years or older with MPM (23.5% of MPM occurred in women) is not unexpected given that melanoma incidence at older ages is lower in women than in men (http://seer.cancer.gov/csr/1975_2007/).
Major strengths of our study include the population-based ascertainment, the extensive clinicopathological information on prognostic variables, and the expert pathological review for many slides.1 Information on ulceration and mitoses was missing more often for MPM than for SPM (32.2% vs 18.4%) because tumors diagnosed before 2000, many of which were prior MPM, were more often unavailable for expert review (19.2% after 1999 vs 45.5% before 2000; P < .001). In addition, the fact that the GEM Study did not identify SPM patients who had another melanoma diagnosed between the close of the initial study enrollment and the end of follow-up may be a weakness. The resulting misclassification of SPM as MPM would have tended to reduce any observed difference between them in overall fatality and in fatality associated with lesions thicker than 4.00 mm. The study also lacked information about nodal or metastatic disease at diagnosis and time to relapse or recurrence. The different health care systems in place in GEM Study centers in the United States, Canada, Italy, and Australia could have been a source of differences in access to diagnosis or care for patients. We sought to address this issue by including center as a covariate in all analyses and educational level as a surrogate for socioeconomic status in all multivariable models.
In summary, we found no strong evidence of a difference in survival between SPM and MPM patients despite the evidence of other researchers8 and suggestions that MPM may have a less aggressive biology than SPM.2,8,10 However, to our knowledge, we report for the first time a greater increase in the risk of death with increasing tumor thickness for SPM than for MPM. This main feature indicated a possible difference between outcomes in patients with single vs multiple melanomas.
Doubrovsky and Menzies8 reported that an increasing number of melanomas influenced survival favorably. They suggested that the effects of immunological surveillance against melanomas may be substantial in MPM patients.8 The results of our study suggest that preference be given to further investigation of this hypothesis rather than the effects simply of closer surveillance and earlier diagnosis as a potential explanation for generally thinner lesions—and an apparently better outcome for the thickest lesions—in MPM patients than in SPM patients.
Accepted for Publication: March 13, 2013.
Corresponding Author: Anne Kricker, PhD, Sydney School of Public Health, University of Sydney, Sydney, New South Wales 2006, Australia (email@example.com).
Published Online: June 19, 2013. doi:10.1001/jamadermatol.2013.4581.
Author Contributions: All authors had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Kricker, Armstrong, Thomas, Gallagher, Marrett, Gruber, Anton-Culver.
Acquisition of data: Kricker, Armstrong, Thomas, From, Busam, Gallagher, Marrett, Groben, Gruber, Anton-Culver, Rosso, Dwyer, Berwick.
Analysis and interpretation of data: Kricker, Armstrong, Goumas, Thomas, Kanetsky, Gallagher, Gruber, Anton-Culver, Rosso, Dwyer.
Drafting of the manuscript: Kricker, Armstrong.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Kricker, Armstrong, Goumas, Gruber, Rosso.
Obtained funding: Kricker, Armstrong, Thomas, Gallagher, Marrett, Anton-Culver, Berwick.
Administrative, technical, and material support: Kricker, Armstrong, Thomas, From, Gruber.
Study supervision: Kricker, Armstrong, Busam, Gallagher, Gruber, Anton-Culver, Rosso.
Conflict of Interest Disclosures: None reported.
Funding/Support: This study was supported by grants R01 CA112243, R01 112524, R01 CA112243-05S1, R01 CA112524-05S2, U01 CA83180, K05 CA CA 131675-02, and ES014635 from the National Cancer Institute at the National Institutes of Health; and by a Health Research Infrastructure Award from the Michael Smith Foundation (Mr Gallagher).
Role of the Sponsors: The sponsors had no role in the design and conduct of the study; in the collection, analysis, and interpretation of data; in the preparation of the manuscript; or in the review or approval of the manuscript.
GEM Study Group: The following are members of the GEM Study Group: Marianne Berwick, PhD (principal investigator [PI]), University of New Mexico); Colin Begg, PhD (co-PI), Irene Orlow, PhD (coinvestigator), Urvi Mujumdar, MPH (project coordinator), Klaus Busam, MD (dermatopathologist), and Pampa Roy, PhD (laboratory technician), Memorial Sloan-Kettering Cancer Center; Bruce K. Armstrong, MD, DPhil (PI), Anne Kricker, PhD (co-PI), and Melisa Litchfield, MPH (study coordinator), University of Sydney and The Cancer Council New South Wales, Sydney, Australia; Terence Dwyer, MD, PhD (PI), Paul Tucker, MD (dermatopathologist), Alison Venn, PhD (coinvestigator), and Nicola Stephens, M Clin Epid (study coordinator), Menzies Research Institute, University of Tasmania, Hobart, Australia; Richard P. Gallagher, MA (PI), and Teresa Switzer (coordinator), British Columbia Cancer Agency, Vancouver; Loraine D. Marrett, PhD (PI), Elizabeth Theis, MSc (coinvestigator), Lynn From, MD, PhD (dermatopathologist), and Noori Chowdhury, MPH (coordinator) Cancer Care Ontario, Toronto, Ontario, Canada; Stefano Rosso, MD (PI), Roberto Zanetti, MD, PhD (co-PI), and Carlotta Sacerdote, MD, PhD (coordinator), Piedmont Cancer Registry, Cancer Prevention Center, Turin, Italy; Hoda Anton-Culver, PhD (PI), and Nancy Leighton, MS (coordinator), University of California, Irvine; Stephen B. Gruber, MD, PhD (PI), and Joanne Jeter, MD (coordinator), University of Michigan, Ann Arbor; Judith Klotz, PhD (PI), Homer Wilcox, PhD (co-PI), and Helen Weiss, RN (coordinator), New Jersey Department of Health and Senior Services, Trenton; Robert Millikan, PhD (PI [deceased]), and Nancy E. Thomas, MD, PhD (co-PI), The University of North Carolina at Chapel Hill; and Timothy Rebbeck, PhD (PI), Peter A. Kanetsky, PhD (coinvestigator), Amy Walker, BSc (laboratory manager), and Saarene Panossian, MS (laboratory technician), University of Pennsylvania, Philadelphia.