Figure. Kaplan-Meier (A) and Cox-adjusted (B) survival curves for patients diagnosed as having cutaneous melanoma (CM) as first cancer and with CM diagnosed subsequent to a previous cancer. Surveillance, Epidemiology, and End Results (SEER) 9 Registries, 1988-2002.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Yang GB, Barnholtz-Sloan JS, Chen Y, Bordeaux JS. Risk and Survival of Cutaneous Melanoma Diagnosed Subsequent to a Previous Cancer. Arch Dermatol. 2011;147(12):1395–1402. doi:10.1001/archdermatol.2011.1133
Objective To understand the risk of cutaneous melanoma (CM) following a previous cancer.
Design Using the Surveillance, Epidemiology, and End Results database (1988-2007), we compared a cohort of patients diagnosed as having CM as a first cancer with a cohort of patients diagnosed as having CM following a previous cancer.
Participants We included 70 819 patients with CM as a first primary cancer and 6353 patients with CM following a previous cancer.
Main Outcome Measures We calculated the relative risk (RR) for development of primary CM following a previous cancer and used Cox modeling to examine survival characteristics of the 2 cohorts.
Results Patients younger than 45 years at first cancer diagnosis had significantly higher risk of CM following cutaneous melanoma (RR, 11.89; 95% CI, 10.83-13.03), other nonepithelial skin cancer (RR, 2.81; 95% CI, 1.13-5.79), Kaposi sarcoma (RR, 3.26; 95% CI 1.41-6.42), female breast cancer (RR, 1.38; (1.11-1.70), and lymphoma (RR, 1.79; (95% CI, 1.30-2.41). Patients 45 years or older at first cancer diagnosis had significantly higher risk of CM following cutaneous melanoma (RR, 8.36; 95% CI, 7.93-8.81), other nonepithelial skin cancer (RR, 2.00; 95% CI 1.35-2.86), ocular melanoma (RR, 5.34; 95% CI 3.42-7.94), female breast cancer (RR, 1.12; 95% CI, 1.03-1.21), prostate cancer (RR, 1.08; 95% CI, 1.03-1.13), thyroid cancer (RR, 1.40; 95% CI, 1.06-1.82), lymphoma (RR, 1.34; 95% CI, 1.16-1.55), and leukemia (RR, 1.79; 95% CI, 1.49-2.13). Characteristics associated with better survival in both cohorts included female sex, age younger than 45 years at melanoma diagnosis, being married, being white vs black, decreasing Breslow depth, lack of tumor ulceration, no nodal involvement, and absence of metastases.
Conclusion Given that cutaneous melanoma is the most common second primary cancer in patients with a first CM (a risk that remains elevated for over 15 years), our results suggest the need for continued skin surveillance in melanoma survivors.
Cutaneous melanoma (CM), one of the most aggressive form of skin cancer, is the fifth and seventh most commonly diagnosed cancer among US men and women, respectively.1 Not only is its incidence climbing rapidly, mortality rates of CM have not significantly diminished. The 10-year survival is 92% for primary tumors 1.00 mm or less in depth and 50% for primary tumors greater than 4.00 mm in depth.2
The greatest risk factor for CM development is UV radiation exposure, though this risk is affected by patients' race and genetics.3 In the United States, European Americans have 10 times the risk of developing melanoma compared with African Americans. Other known risk factors include increased nevi counts, presence of atypical nevi, and immunosuppressed states.4-7
Approximately 10% of patients with melanoma also have a positive family history of the disease.8 Studies have demonstrated the role of numerous genes, including CDKN2A, CDK4, XP, and BRCA2, in the pathogenesis of melanoma.9-15 Individuals with germline mutations in CDKN2A and BRCA2 are at increased risk of also developing pancreatic cancer and breast cancer, respectively.9-12 In addition, a personal history of nonmelanoma or melanoma skin cancer is associated with an increased risk of acquiring CM.16 Recent studies have further investigated the increased risk of second primary cancers among cutaneous melanoma survivors.17-20 One study from Ireland found that the subsequent risk of any cancer was more than doubled in melanoma survivors.21
Using the Surveillance, Epidemiology and End Results (SEER) program of the National Cancer Institute, we examined the characteristics and survival associated with CM diagnosed as a first cancer compared with CM diagnosed subsequent to a previous cancer. We also calculated the relative risk (RR) for development of primary CM following a previous cancer relative to the general US population.
We used publicly available data from the original SEER 9 registries to identify 2 cohorts of individuals diagnosed as having CM between 1988 and 2007. The SEER 9 registries included Atlanta, Connecticut, Detroit, Hawaii, Iowa, New Mexico, San Francisco –Oakland, Seattle-Puget Sound, and Utah.22 The first cohort consisted of individuals diagnosed as having CM as their first cancer, while the second cohort consisted of individuals diagnosed as having CM subsequent to a previous cancer. Only microscopically confirmed malignant cases of melanoma were included in our study. Melanoma in situ cases, cases with unknown age or race, or cases diagnosed at autopsy and on death certificate were excluded. To account for surveillance bias in the second cohort, we included only those individuals with subsequent CM that developed at least 2 months after their first cancer diagnosis. Only microscopically confirmed malignant first cancers were included.
Data collected from the SEER registries included age, sex, race, marital status, month and year of melanoma diagnosis, histologic subtype of melanoma, primary tumor location, Breslow depth, presence or absence of ulceration, lymph node status, presence or absence of distant metastases, and vital status.22 We classified race as white, black, or other and categorized marital status into the following 3 groups: married, unmarried (single, divorced, separated, and widowed), or unknown. Using the International Classification of Diseases for Oncology, Third Edition (ICD-O-3),23 we divided the histologic subtypes of CM into superficial spreading (ICD-O-3 histologic code 8743), nodular (code 8721), lentigo maligna (code 8742), acral lentiginous (code 8744), malignant —not otherwise specified (NOS) (code 8720), or other (codes 8722, 8723, 8730, 8740, 8741, 8745, 8761, and 8770-8774). The primary site of melanoma presentation was categorized into the following 5 major anatomic areas: (1) head and neck (ICD-O-3 site codes C440-444), trunk (code C445), upper limb/shoulder (code C446), lower limb/hip (code C447), or other, including overlapping areas of skin and NOS (codes C448-C449). We divided Breslow depth of the primary CM into the following 5 groups: 0.01 to 1.00 mm, 1.01 to 2.00 mm, 2.01 to 4.00 mm, 4.01 mm or larger, or unknown. We classified ulceration as absent, present, or unknown. Lymph node status was categorized as no node involvement, positive regional node involvement, positive distant node involvement, or NOS/unknown; it is important to note that SEER revised its nodal staging codes in 2004 and the 2 SEER codes —Extent of Disease coding (1988-2003) and CS Lymph Nodes coding (2004-2007) —were recoded to form our 4 categories. Distant metastasis was defined as absent, present, or unknown.
This study received institutional review board approved exemption because it is based on publically available SEER data.
We compared the demographic and tumor-related characteristics between the 2 cohorts. To calculate the RR (including the 95% confidence interval [CI]) of subsequent CM in the second cohort, we used SEER*Stat 6.6.2 to compare the observed number of subsequent CM with the expected number in the general US population.22 The expected number of melanoma developed relative to each first cancer site was based on incidence rates in the United States stratified by race, sex, age at diagnosis, and calendar year of diagnosis. Person-years at risk for developing subsequent CM included the time beginning 2 months after first cancer diagnosis to the date of death, the date of last known follow-up, or the end of study, whichever occurred first.
For the survival analysis, we analyzed patients diagnosed as having CM from 1988 through 2002, with follow-up through 2007. Survival time was calculated as years following CM diagnosis, with death being defined as death from any cause. The Kaplan-Meier method was used to estimate the overall survival difference between the 2 cohorts. For each cohort, we performed univariate and multivariate Cox proportional hazards models of patient and tumor characteristics to calculate unadjusted and adjusted hazard ratios (and their 95% CIs). All variables of interest were tested for agreement with the proportional hazards assumption of the Cox model. Analyses were performed using SAS version 9.1.3 statistical software (SAS Institute Inc).
A total of 70 819 patients were diagnosed as having CM as a first primary cancer, and a total of 6353 patients were diagnosed as having CM subsequent to a previous cancer. Demographic and tumor characteristics for both cohorts are presented in Table 1. The median age at melanoma diagnosis was 54 years for the first cohort and 70 years for the second cohort. More than 98% of patients in both groups were white. A larger proportion of patients in the second cohort were men (67.1% vs 54.2%) and older than 45 years (92.1% vs 70.1%). Although superficial spreading melanoma was the most common histologic diagnosis in both cohorts, lentigo maligna melanoma was much more common in the second cohort (11.8% vs 6.6%). Among the first cohort, the most common site of first primary CM was on the trunk (34.3%), followed by the upper limb (23.9%), and in the second cohort, the most common site of subsequent CM was on the trunk (30.6%), followed by the head and neck (27.9%). Most melanoma cases in both cohorts were less than 1.00 mm in depth, nonulcerated, and without lymph node or distant organ involvement.
The risk of melanoma development in patients younger than 45 years was more than doubled for survivors of melanoma of the skin (RR, 11.89; 95% CI, 10.83-13.03); other nonepithelial skin cancer, which consists of many histologic subtypes including Merkel cell carcinoma, dermatofibrosarcoma, and sebaceous adenocarcinoma (RR, 2.81; 95% CI, 1.13-5.79); and Kaposi sarcoma (RR, 3.26; 95% CI, 1.41-6.42). Risks were also significantly increased following female breast cancer and lymphoma (specifically, non-Hodgkin lymphoma, data not shown) (Table 2). The risk of melanoma development in patients older than 45 years was more than doubled for survivors of melanoma of the skin (RR, 8.36; 95% CI, 7.93-8.81), other nonepithelial skin cancer (RR, 2.00; 95% CI, 1.35-2.86), and eye and orbit melanoma (RR, 5.34; 95% CI, 3.42-7.94). Risks were also significantly increased following female breast cancer, prostate cancer, thyroid cancer, lymphoma, leukemia (specifically, chronic lymphocytic leukemia, data not shown), and other cancers, and significantly decreased following cancers of the colon/rectum, lung and bronchus, other respiratory tract, and urinary bladder (Table 2).
To account for surveillance bias, we further stratified these results by time since diagnosis of first primary cancer (Table 3). For patients younger than 45 years, the significantly elevated risk of melanoma following Kaposi sarcoma was only seen 2 to 11 months after initial cancer diagnosis; the RR following other nonepithelial skin cancers became nonsignificant after stratifying according to year since first cancer diagnosis. For patients older than 45 years, the significantly elevated risk of melanoma following cancer of the female breast and thyroid was only seen 2 to 11 months after initial cancer diagnosis. For all first melanoma survivors, the risk for subsequent melanoma remained significantly elevated for the entire study period.
The survival analysis was restricted to those patients diagnosed between 1988 and 2002 with follow-up until 2007, which resulted in 48 818 patients in the first cohort and 3378 patients in the second cohort. The overall 10-year survival rate was 74.5% (95% CI, 74.1%-74.9%) for the first cohort and 51.9% (95% CI 50.2%-53.6%) for the second cohort (Figure, A). Univariate analysis showed that sex, age, race, marital status, tumor site, histologic subtype, Breslow depth, tumor ulceration, lymph node status, and metastatic disease all had a significant impact on survival (all P < .001). After adjustment for all these factors, having a diagnosis of melanoma subsequent to a previous cancer was still associated with worsened survival when compared with having a diagnosis of melanoma as a first cancer (Figure, B).
Multivariate analysis revealed that female sex, younger age at melanoma diagnosis ( <45 years), and being married were associated with better survival (Table 4). White individuals also had a better survival rate compared with black individuals and individuals of other race, although this was significant only for the first cohort. Compared with superficial spreading melanoma, all other histologic subtypes had decreased survival in the first cohort, while in the second cohort, only nodular, acral lentiginous, and other histologic subtypes had significantly decreased survival. Additional factors associated with worst survival in both cohorts included increasing Breslow depth, tumor ulceration, nodal involvement, and metastatic disease. When comparing risk of death between the 2 cohorts, Breslow depth of 2.01 to 4.00 mm and 4.01 mm or larger played a more significant role in the first cohort. Metastatic disease in the first cohort was also 1.6 times more likely to result in death than metastatic disease in the second cohort.
Although previous studies have evaluated second cancer risks after a diagnosis of melanoma, we provide the first large scale study, to our knowledge, comparing patients diagnosed as having CM as a first cancer (first cohort) with patients diagnosed as having CM subsequent to a previous cancer (second cohort). The finding that a greater number of patients in the second cohort were older and male compared with the first cohort is likely because many of the observed melanoma cases in the second cohort followed prostate cancer. In addition, men are at increased risk of developing cancers in general compared with women, perhaps because of certain occupational exposures and social risk factors such as drinking and smoking.24 The greater age at melanoma diagnosis in the second cohort may also explain the increased rate of lentigo maligna melanoma, which usually occurs on the head and neck regions of older individuals with chronic sun-damaged skin.25
Our study showed that patients younger than 45 years and 45 years or older with any initial cancer diagnosis were at an overall 192% and 36% increased risk for developing CM, respectively. The greatest number of melanomas developed in individuals already with a first melanoma diagnosis, which is consistent with other published studies17-20,26; the markedly elevated incidence of second melanomas within the first 2- to 11-month latency period is likely due to increased initial surveillance of patients with cancer. Surveillance bias may also explain the significant risk of melanoma in the first year following a diagnosis of Kaposi sarcoma in patients younger than 45 years and cancer of the female breast and thyroid in patients 45 years or older. Nevertheless, the risk for subsequent melanoma remained significantly elevated for first melanoma survivors throughout the study period, demonstrating the importance of other factors in addition to medical surveillance.
Inherited susceptibility is often implicated with the development of multiple cancers and of cancer at earlier stages in life. Although the marked risk of multiple melanomas is predominantly from UV radiation exposure, the higher risk in those patients younger than 45 years is also due to patient characteristics (including light skin and multiple nevi) and genetics (such as familial atypical mole and melanoma syndrome and atypical mole syndrome).4,6,7,27 In our study, cancer of the breast and prostate (the most common cancer in women and men, respectively, other than nonmelanoma skin cancer) were also frequently found preceding CM; other studies have reported breast and prostate cancer to be the most common non –skin cancer following primary melanoma diagnosis.17 Many genes have been proposed to underlie these bidirectional associations, including germline mutations of BRCA2, which predispose to CM and breast cancer,9,10 and increased levels of TSPY, which promotes prostate cancer and melanoma tumorigenesis.28 The lack of genetic information in the SEER registry prohibited us from further investigating the contribution of genetic risk.
Many reports have also documented the increased risk of melanoma following and preceding non-Hodgkin lymphoma and chronic lymphocytic leukemia. Possible mechanisms include immunosuppressive effects of cancer and cancer treatment, genetic factors, and UV radiation exposure.29-31 Similarly, UV radiation has been debated as a risk factor for ocular melanoma.32,33 Given that the risk of ocular melanoma is not increased following CM,33 surveillance bias may better explain the increased incidence of CM following ocular melanoma. Many physicians believe in a common risk factor between ocular and cutaneous melanoma and may examine patients with ocular melanoma more thoroughly for any skin lesions.
The decreased risk for melanoma development following cancer of the colon/rectum, lung, other respiratory tract, and urinary bladder for patients 45 years or older is possibly due to unshared risk factors. For instance, light skin and sun tanning, which are strong risk factors for melanoma, are more prevalent in high socioeconomic classes,34,35 while smoking and occupational exposures, which are strong risk factors for cancers of the respiratory tree and bladder, are more prevalent in low socioeconomic classes.36 Recent epidemiological studies have also found that decreased vitamin D levels correlate with increased risk of cancers, particularly of the gastrointestinal system.37,38 Because over 90% of our vitamin D requirements is through UV-induced cutaneous synthesis,39 this leaves much room for discussion on the appropriate amount of sun exposure to balance the risk of melanoma and other cancers.
Findings from our survival analysis demonstrated better survival for patients in the first cohort, even after adjusting for all demographic and tumor-related variables. In both cohorts, significantly better survival was seen in female patients compared with male patients, possibly because of differences in lifestyle choices, early detection, and treatment patterns.40 A recent study by Joosse et al41 also demonstrated that women with melanoma had a lower risk of lymph node and visceral metastases. In addition, we found that married persons had better survival than unmarried persons, likely owing to better socioeconomic status and greater support systems,42 as did younger patients relative to older patients, probably owing to fewer comorbidities. Compared with white patients, black patients had worse survival in both cohorts, though only statistically significant in the first; this is possibly owing to differences in socioeconomic status and access to care, which could not be assessed using SEER data. In a study by Zell et al,43 black patients had worse survival even after controlling for differences in socioeconomic status and treatment, suggesting a different cancer biological profile between racial groups.
Important tumor-related prognostic factors included tumor location, histologic subtype, and extent of disease. As in other studies, patients with melanoma located on the head and neck had the worse prognosis40,44; a recent study by Bradford et al17 also found that these patients had increased risk of developing subsequent melanoma. Compared with superficial spreading melanoma, nodular melanoma and acral lentiginous melanoma had the highest risk of death in both cohorts —findings which are consistent with other reports.40,45 However, even with all cases in our study being histologically confirmed, a large proportion still consisted of malignant melanoma —NOS, which may have limited the strength of the analysis of survival and histologic subtype. Additional factors associated with worst survival included increasing Breslow depth, tumor ulceration, nodal involvement, and metastatic disease. It is important to remember though that many patients in the second cohort died from complications of their first cancer, thereby diluting the effect of increasing Breslow depth and metastatic disease on survival compared with the first cohort.
One of the greatest strengths of the present study is the large sample size and high quality of the SEER database. The SEER-9 registries represent approximately 10% of the US population, and individual registries ensure accurate collection of patient and tumor information. We also limited the cases in our study to only those that were histologically confirmed. However, SEER does not include a central pathology review and has frequent missing data on tumor histologic subtype and location, Breslow depth, ulceration status, and lymph node and distant organ involvement. SEER also does not record nonsurgical treatment for cancers, including chemotherapy and biologic therapy, which prevented us from studying the contribution of initial cancer treatment on the risk of melanoma development. In addition, studies have shown that many melanoma cases are diagnosed by private physicians and not reported to cancer registries until many years later46; hence, the number of melanomas for the 2 cohorts may have been underestimated. We were also unable to account for those patients who may have moved out of a SEER area after their first cancer diagnosis, which would further falsely lower the incidence and RR of melanoma in the second cohort. Lastly, we were unable to assess the impact of comorbidities, nutritional status, socioeconomic status, genetic factors, and family history on risk of melanoma and on survival because the publicly available SEER data do not include this information.
In conclusion, our findings revealed an increased risk of CM in cancer survivors compared with the general population. Genetic susceptibility, UV exposure, immunologic compromise, and surveillance bias are likely to contribute to this increased risk. Given that cutaneous melanoma is the most common second primary cancer in patients with a first CM (a risk that remains elevated for over 15 years), our results suggest the need for continued skin surveillance in melanoma survivors.
Correspondence: Jeremy S. Bordeaux, MD, MPH, Department of Dermatology, Case Western Reserve University School of Medicine, University Hospitals Case Medical Center, 11100 Euclid Ave –Lakeside 3500, Cleveland, OH 44106 (Jeremy.Bordeaux@uhhospitals.org).
Accepted for Publication: August 7, 2011.
Author Contributions: All authors had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Mr Yang and Dr Barnholtz-Sloan contributed equally to this work. Study concept and design: Yang, Barnholtz-Sloan, and Bordeaux. Acquisition of data: Yang and Chen. Analysis and interpretation of data: Yang, Barnholtz-Sloan, Chen, and Bordeaux. Drafting of the manuscript: Yang. Critical revision of the manuscript for important intellectual content: Yang, Barnholtz-Sloan, Chen, and Bordeaux. Statistical analysis: Yang and Barnholtz-Sloan. Obtained funding: Bordeaux. Study supervision: Bordeaux.
Financial Disclosure: None reported.
Funding/Support: Dr Bordeaux is supported by the Dermatology Foundation Clinical Career Development Award in Dermatologic Surgery.
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; or in the preparation, review, or approval of the manuscript.
Additional Contributions: We are indebted to the staff at the Case Comprehensive Cancer Center and Case Department of Dermatology, as well as the cancer registries that participate in the SEER program.
Create a personal account or sign in to: