A population-based analysis of risk factors for a second primary cutaneous melanoma among melanoma survivors
Article first published online: 17 JAN 2003
Copyright © 2003 American Cancer Society
Volume 97, Issue 3, pages 639–643, 1 February 2003
How to Cite
Goggins, W. B. and Tsao, H. (2003), A population-based analysis of risk factors for a second primary cutaneous melanoma among melanoma survivors. Cancer, 97: 639–643. doi: 10.1002/cncr.11116
- Issue published online: 17 JAN 2003
- Article first published online: 17 JAN 2003
- Manuscript Accepted: 16 SEP 2002
- Manuscript Revised: 19 AUG 2002
- Manuscript Received: 19 JUN 2002
- American Cancer Society. Grant Number: CRTG 99-249-01-CCE
- Surveillance, Epidemilogy, and End Results (SEER);
- second tumor;
- cutaneous melanoma;
- Cox proportional hazards model;
The results of several studies have provided evidence that patients diagnosed with cutaneous melanoma (CM) are at a higher risk of developing a second primary CM than the general population. In this study, the authors examined how the risk of a second primary tumor varied with time from diagnosis of CM and examined the patient-specific factors that modify a CM patient's risk of developing a second primary tumor.
Survival curves for time from first CM to second CM were calculated using the Kaplan–Meier method. The Cox proportional hazards model was used to determine which demographic- and disease-related factors influence the risk of a second CM.
Approximately 0.5% of Surveillance, Epidemiology, and End Results (SEER) CM patients were found to have synchronous second primaries. The estimated cumulative probability of having a second primary CM was 0.99% at 1 year after initial CM diagnosis, 2.06% at 5 years, 3.17% at 10 years, and 5.34% at 20 years. Risk was significantly greater for males; older patients; patients with first CM on the face, neck, or trunk; those from the Atlanta, Hawaii, or Connecticut registries; and more recently diagnosed patients. Risk was lower for patients from the Utah registry and those with Stage IV disease.
The elevated risk for CM among CM survivors appears to be greatest in the first few months, and then subsequently declines. However, the risk for a second CM among CM survivors was found to remain substantially higher than the risk for a first CM in the general population throughout the observation period (> 20 years). Demographic- and disease-related factors substantially modify the risk of a second primary CM. Cancer 2003;97:639–43. © 2003 American Cancer Society.
Although the incidence of cutaneous melanoma (CM) has increased steadily over the past several decades, earlier detection has led to a greater proportion of curable lesions and thus larger numbers of long-term survivors. This year alone, the American Cancer Society estimates 51,400 cases of CM with 7800 deaths attributable to the disease.1
Several studies2–8 have reported an increased risk (standardized incidence ratios [SIRs] ranging from 4.5–25.6) of developing a second primary CM. A recent case series calculated the 5- and 10-year risks of developing a second primary melanoma among CM patients to be 2.8% and 3.6%, respectively.8 Most studies in the U.S. have relied on the experience of a single institution and may thus suffer from potential selection, referral, or detection biases. To minimize these effects, we set out to calculate the risk of developing a second primary melanoma for CM survivors using the National Cancer Institute Surveillance, Epidemiology, and End Result (SEER) database. We also used survival analysis methodology to determine specific patient factors that can predict the development of a second primary CM. The advantage of using survival analysis techniques for this type of study is that it allows us to ascertain the effect of patient characteristics on the risk for a second CM while adjusting for differential follow-up times for the patients.
MATERIALS AND METHODS
We utilized the SEER public use diskette that includes data from nine SEER cancer registries, including the states of Connecticut, Hawaii, Iowa, New Mexico, and Utah and the metropolitan areas of Atlanta, Detroit, San Francisco–Oakland, and Seattle. The SEER public-use dataset contains many demographic- and disease-related variables with one record for each known cancer diagnosis that occurred in these geographic areas from 1973 on. Ascertainment and data quality are considered to be excellent. The cases included in this study were restricted to cancers that were: 1) malignant, 2) microscopically confirmed, 3) newly diagnosed from January 1, 1973 to December 31, 1997. The survival time for each patient was defined as the time from diagnosis of the first primary melanoma to the time of diagnosis of the second primary CM. Patients were considered censored if they died or were lost to followup without having a diagnosis of a second primary. Only the times from the first primary CM to the second were considered in the analysis; all subsequent primaries were ignored.
Survival probabilities were calculated using the Kaplan–Meier method. To determine which patient characteristics affected the probability of developing a second primary, the Cox proportional hazards model was used. This model assumes that the hazard (risk) of failure (in this case failure is defined as the development of a second primary tumor) is affected multiplicatively by the covariates.
For the entire 1973–1997 SEER data, the following variables were entered into univariate analyses: race (white and nonwhite), age, gender, year of diagnosis, site of first primary CM, stage at diagnosis (I, II, IV; SEER historic Stage A: I [localized], II [regional], and IV [distant]), and registry. Age was entered into the models as a continuous covariate. For the categorical covariates (except for registry and tumor location), indicator variables were created for each category with one category being chosen as the baseline. For the registry variable, indicator variables were created for five registries: Utah, Connecticut, Atlanta, Hawaii, and Iowa. Patients from the San Francisco, New Mexico, Detroit, and Seattle registries then served as the baseline group because preliminary analysis indicated that the patients in these areas had rates of second CM similar to each other and to the average for all the patients. For tumor locations, indicator variables were created for patients with tumors of the face (except ear and eyelid), neck, trunk, and arms, with patients with all other tumor locations serving as the baseline group.
Because of potential confounding, all variables were included in the initial multivariate models, even if they were not statistically significant in the univariate models. A backwards-stepwise procedure was then used to determine the final multivariate Cox models. Covariates were dropped one at a time, starting with the least significant, until all remaining variables had P values < 0.15.
A total of 2.32% of the SEER CM patients eventually developed at least 1 additional primary CM during their time of follow-up. Approximately 22% of these patients had 2 synchronous primary CMs. The Kaplan–Meier cumulative probabilities of a second CM, separated by gender and age groups, are shown in Figure 1. Males were at a considerably higher risk of developing a second CM, although the effect of older age at diagnosis of the first CM varied between males and females. For males, risk of a second CM was found to increase sharply with a higher age at the time of diagnosis of the first CM, but for females the increase in risk was found to be slight.
The estimated monthly hazard function, defined as the probability of developing a second CM in a particular month after diagnosis given that the patient has not had one prior to that month, is plotted in Figure 2. These hazard rates correspond to mean rates of 271.3 cases per 100,000 per year from 1–4 years after diagnosis, and mean rates of 227.8, 229.3, and 224.4 for 5–10 years, 10–15 years, and 15–20 years postdiagnosis, respectively.
The results of the univariate and multivariate Cox model analyses are shown in Table 1. A total of 61,245 patients with complete data for all covariates were used for the multivariate analyses. Males were at a considerably higher risk for a second CM than females both in the univariate analysis and after adjustment. The finding that the male age interaction and age at diagnosis were both significant suggests that older age at first CM diagnosis was a significant risk factor and that the age effect was stronger for males than for females. In both the univariate and multivariate models, the presence of a first CM on the face (except eyelid and ear), neck, and trunk was found to carry a significantly higher risk for a second CM compared with patients whose first CM was located elsewhere. Patients with SEER Stage IV disease had a lower risk of a second primary tumor, whereas no significant difference was found with Stage I and Stage II patients. Patients who were diagnosed with a first primary CM between 1983–1987, 1988–1992, or 1993–1997 had significantly higher rates of second CM than those diagnosed before 1983, with a clear trend toward higher rates for those diagnosed more recently. Geographically, patients with CM in Connecticut, Atlanta, and Hawaii were at a significantly higher risk for developing a second CM than those in other registries, whereas those in Utah were at a significantly lower risk.
|Covariate||Univariate RR (95% CI)||P value||Adjusted RR (95% CI)||P value|
|Male gender||1.66 (1.49, 1.85)||< 0.0001||1.50 (1.34, 1.68)||< 0.0001|
|Age||1.014 (1.01, 1.02)||< 0.0001||1.01 (1.00, 1.01)||0.0178|
|Age-gender interaction||1.01 (1.00, 1.01)||0.0327|
|Location: referent group: All other locations|
|Face||1.80 (1.50, 2.16)||< 0.0001||1.37 (1.15, 1.63)||0.0004|
|Neck||1.70 (1.36, 2.12)||< 0.0001||1.34 (1.09, 1.65)||0.0075|
|Trunk||1.37 (1.19, 1.58)||< 0.0001||1.10 (0.97, 1.24)||0.124|
|Arms||1.24 (1.06, 1.45)||0.006||a|
|Nonwhite: referent group: Whites||0.695||a|
|1.10 (0.69, 1.75)|
|Year of diagnosis: Referent group: 1973–1977|
|1978–1982||1.12 (0.89, 1.42)||0.339||a|
|1983–1987||1.69 (1.34, 2.12)||< 0.0001||1.50 (1.27, 1.77)||< 0.0001|
|1988–1992||2.24 (1.78, 2.82)||< 0.0001||1.96 (1.66, 2.31)||< 0.0001|
|1993–1997||2.87 (2.27, 3.63)||< 0.0001||2.40 (2.02, 2.85)||< 0.0001|
|Registry: referent group: San Francisco, Seattle, New Mexico, Detroit|
|Connecticut||1.39 (1.22, 1.60)||< 0.0001||1.35 (1.18, 1.54)||< 0.0001|
|Atlanta||1.42 (1.20, 1.67)||< 0.0001||1.41 (1.20, 1.67)||< 0.0001|
|Iowa||0.99 (0.98, 1.00)||0.0586||0.99 (0.98, 1.00)||0.046|
|Hawaii||1.52 (1.20, 1.94)||< 0.0001||1.52 (1.19, 1.93)||0.0001|
|Utah||0.68 (0.52, 0.88)||< 0.0001||0.68 (0.53, 0.88)||0.0023|
|Stage: referent group: Stage I|
|II||1.09 (0.90, 1.32)||0.392||a|
|IV||0.33 (0.09, 0.63)||0.0008||0.33 (0.17, 0.63)||0.0009|
Using the large population-based SEER database, we found that the rates of second primary CM among CM survivors were much higher than the rates of a first primary CM among the general population and that the risk remains elevated up to 20 years after diagnosis of the first CM. The rate of second CM for the SEER CM survivors was > 10 times the rate for first primary CM among the general SEER population. Although a comparison of the rates for the SEER CM survivors and the SEER general population is complicated by the fact that the 2 groups have different age distributions, we note that the rates we observed for the SEER CM survivors, 271.3 per 100,000 1–4 years postdiagnosis and approximately 225 per 100,000 ≥ 5 years postdiagnosis, were elevated relative to even the highest risk segments of the general population. For instance, for white males age > 75 years in the SEER population, the rate was approximately 80 per 100,000 per year from 1993–1997.
Based on the 61,245 CM patients in the SEER registries, we calculated the 1-year, 5-year, and 10-year probabilities of developing a second primary to be 1.0%, 2.1%, and 3.2%, respectively. DiFronzo et al. reported similar probabilities (2.8% at 5 years and 3.6% at 10 years) for the 8928 melanoma patients seen at the John Wayne Cancer Institute,7 whereas Slingluff et al. found slightly higher risks (approximately 2% at 1 year, 3.4% at 5 years, and 5.3% at 10 years) for the 7816 patients followed at Duke University Medical Center.8 Variations in rates between our population-based analysis and the single institutional findings may reflect more intensive post-CM surveillance, better reporting to institutional data registries, and less geographic drift at the medical centers compared with the SEER registries. The actuarial cumulative risk was greater for men than for women and for older CM patients compared with younger CM patients. These results are consistent with those reported for first CM diagnoses and may reflect the trend in the general population.
Our hazard function found the greatest risk of a second CM to be within the first 2 years after diagnosis followed by a marked drop in risk from Years 2–5 and even a slightly lower risk from Years 5–20. There are several possibilities for this observed hazard profile. Multiple primary CMs may be either produced or detected in clusters. Concurrent development of multiple tumors can result from a single, profoundly deleterious exposure to ultraviolet light that directly affects a cohort of melanocytes on the skin. In mice, neonatal pups exposed to a single dose of ultraviolet irradiation are reported to develop melanomas approximately 1 year later.9
Simultaneous detection, conversely, most likely reflects metachronous lesions that are observed together at one time. Several studies have found that up to 40% of multiple primary CMs occur in a synchronous (simultaneous or within 2 months) manner.5, 7, 8 Finally, epidermotropic metastases within the first 2 years of diagnosis may simulate secondary primaries. The finding that 99.75% of the SEER CM cases used in this analysis were histologically confirmed reduces, but cannot eliminate, this possibility. The demographic and clinical characteristics that confer an increased risk for a second primary CM are consonant with those found for the general CM population (i.e., males, older individuals, and persons living in sunny locations such as Hawaii and Atlanta). In addition, the age-specific CM rates for males rise quickly with age, whereas those for females rise quite slowly; this pattern is consistent with the gender-age interaction observed for second primary CMs. Because the incidence and cumulative risk of CM have risen continuously over the past few decades, the risk for a second primary CM also was higher for more recently diagnosed patients.
However, not all demographic factors predicting second CM were consistent with those predicting CM in the general population. The white population in Iowa, Detroit, and New Mexico had slightly lower rates of CM than in Utah, but unlike Utah these areas did not have a statistically significant lower risk of a second CM. Whites in Seattle and San Francisco had slightly higher rates of CM than those in Connecticut but, unlike Connecticut, did not have statistically significantly higher adjusted rates of second primary CMs. Finally, nonwhite CM survivors did not have a significantly lower risk of a second primary CM despite the fact that they are at a considerably lower risk of CM in general. There are other potentially strong risk factors for second primary melanomas, such as presence of dysplastic nevi and cumulative sun exposure. Unfortunately, to our knowledge, data concerning these factors are not available for individual patients in the SEER database. Therefore, they could not be considered in this analysis.
We analyzed the SEER data from 1988–1997 separately for extent of disease including Clark level and Breslow thickness and found no consistent trends associated with either parameter (data not shown). Patients with advanced disease appear to be at lower risk for developing a second primary, even though the Cox model adjusts for a shortened survival. It is possible that individuals with a documented metastasis have a greater need for other follow-up regimens and a reduced need for cutaneous examinations with a concomitant drop in the rate of secondary detection. Alternatively, clinicians may be reluctant to engage the patient in additional surgery for suspicious lesions if metastatic disease is progressive. Moreover, clinicians also may interpret new suspicious melanocytic growths as cutaneous metastases in the context of visceral disease.
The etiologies underlying the increased risk of a second CM are manifold. Although CM patients most likely are examined more frequently than the general population, the magnitude of risk enhancement (>10× risk) argues against a detection bias being solely responsible. Furthermore, the increased risk with age suggests that cumulative sun exposure may contribute to the likelihood of another CM. Moreover, genetics also may modulate this risk. The development of multiple primary tumors is a hallmark of many genetic cancer syndromes. As such, several recent studies have found an increased prevalence of germline CDKN2A mutations among patients with multiple primary melanomas.10–14 Thus, the presence of multiple CMs may one day identify carriers of mutations in melanoma predisposition genes.
The population served by the SEER registries is, in general, more affluent and more urban than the U.S. population as a whole.15 This may affect the generalizability of these results to the U.S. population.
In summary, to our knowledge the current study is one of the largest population-based studies on second primary CMs performed to date using the U.S. SEER database. From our analysis, we found a highly significant increased risk for a second CM among CM survivors, especially in the first 2 years after the initial CM diagnosis. These results suggest that melanoma patients should undergo thorough cutaneous examinations, especially during the first 2–3 years after diagnosis. Because the risk remains elevated for at least 20 years, a long-term risk-reductive strategy, including sun-protective measures and ongoing surveillance, also must be implemented.