Association between female breast cancer and cutaneous melanoma
Version of Record online: 20 MAY 2004
Copyright © 2004 Wiley-Liss, Inc.
International Journal of Cancer
Volume 111, Issue 5, pages 792–794, 20 September 2004
How to Cite
Goggins, W., Gao, W. and Tsao, H. (2004), Association between female breast cancer and cutaneous melanoma. Int. J. Cancer, 111: 792–794. doi: 10.1002/ijc.20322
- Issue online: 8 JUL 2004
- Version of Record online: 20 MAY 2004
- Manuscript Accepted: 20 FEB 2004
- Manuscript Revised: 22 JAN 2004
- Manuscript Received: 22 MAY 2003
- Dermatology Foundation
- American Skin Association
- Hong Kong Baptist University Faculty Research Grant. Grant Number: 99-00/II-84
- cutaneous melanoma;
- breast cancer
Epidemiologic studies have provided suggestive evidence of a link between cutaneous melanoma (CM) and breast cancer (BC). Moreover, carriers of mutations in the breast cancer predisposition gene, BRCA2, have an increased risk of melanoma while carriers of mutations in the melanoma susceptibility gene, CDKN2A, exhibit a higher than expected risk of breast cancer. These findings raise the possibility that pathways involved in the development of CM and BC overlap and that survivors of one cancer may be prone to develop the other. To this end, we set out to determine if survivors of female BC in the Surveillance, Epidemiology and End Result (SEER) database are at increased risk for CM and vice versa. We followed female BC patients registered in the 1973—1999 SEER database for development of a second CM and female CM patients for the development of a second BC. The expected number of cases was then compared to the observed number of cases using standardized incidence ratios. Overall, we found a modest but statistically significant increased risk of CM among female BC survivors and vice versa. Among young BC patients, we observed a 46% elevated risk of a second CM. Women who underwent radiation therapy exhibited a 42% increased risk for CM. The risks of BC among female CM survivors and CM among BC survivors were also elevated, albeit to a much lesser degree (overall, 11% and 16%, respectively). We found a mutual association between female BC and CM. The elevated risk for CM, especially among younger BC patients, suggests that the genetic observations from high-risk groups may also be operative at a much lower level in the general BC population. © 2004 Wiley-Liss, Inc.
Observed associations between cancers at disparate sites often lend clues to possible etiologic links or reveal recurrent patterns of bias. Several previous studies have examined the risk of second cancers among women with a history of breast cancer. Although earlier registry-based analyses of second neoplasms after breast cancer (BC) did not detect an increased risk of cutaneous melanoma (CM),1, 2 several more recent registry-based3, 4 and hospital-based5 studies have documented a statistically significant increased risk of CM after BC with standardized incidence ratios (SIRs) ranging from 1.4 to 2.7. On the other hand, several studies have found only small nonsignificant elevations in BC risk among female CM survivors.6, 7, 8, 9, 10 Wassberg et al.11 reported a small but significant (SIR = 1.4) risk of BC after melanoma in situ lesions. More recently, a genetic relationship between CM and BC has been suggested by reports of a higher risk of melanoma (relative risk = 2.58)12 among BRCA2 mutation carriers and an elevated frequency of breast cancer in CDKN2A families (SIR = 3.8).13 Although familial BC and CM account for only small fraction of the cancers observed in the general population, these studies speak to the possibility of shared heritable susceptibilities and common pathways of evolution for these 2 tumor types. We thus hypothesized that survivors of BC have an increased risk of CM and vice versa. To test this hypothesis, we examined the rate of female BC among CM survivors and the rate of CM among female BC survivors using data from the U.S. National Cancer Institute (NCI) Surveillance, Epidemiology, and End Results (SEER) database.
MATERIAL AND METHODS
We utilized the SEER public-use diskette that includes data from 9 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 cases included in this study were restricted to cancers that were malignant, microscopically confirmed and newly diagnosed between 1 January 1973 and 31 December 1999. Because of the low incidence of CM among nonwhites, the analysis was restricted to whites. To determine whether patients with a diagnosis of CM were at an increased risk of BC, we followed CM patients who fit our criteria through the SEER system for a subsequent diagnosis of BC. The follow-up time for each patient was defined as the time from CM diagnosis until either the date of BC diagnosis, the date of last follow-up, or the date of death, whichever came first. Person-years of follow-up were stratified by gender, 5-year age groups and calendar time groups (1973–1978, 1979–1984, 1985–1989, 1990–1994 and 1995–1999) and were multiplied by the corresponding age, gender and calendar time-specific incidence rates to obtain the expected number of BC cases for each stratum. These then were summed to obtain the overall expected numbers of BC cases. The expected number then was compared with the observed number of BC cases in CM survivors using the SIR, the ratio of the observed to expected cases. The same procedure was used to estimate the relative risk of CM after BC diagnosis. Cases with a BC diagnosis and CM diagnosis in the same month were excluded from the analysis because the order of diagnosis could not be determined. The incidence rates used were obtained from the SEERStat program provided by SEER with the public-use data set. The 95% confidence intervals for the SIRs are based on the Poisson distribution and were calculated using tabulated values.14 Byar's approximation to the exact Poisson test was used to test the statistical significance of the SIRs. The effect of age was assessed in subgroup analyses by using 50 years as a cutoff.
To assess the possibility of detection bias, we compared the characteristics (SEER tumor thickness and SEER stage at diagnosis) of the post-CM BC patients and the post-BC CM patients with the overall SEER population of BC and CM cases, respectively. Multivariate analyses (linear regression for thickness and ordinal logistic regression for stage) were used to control for the confounding effects of age and gender. The extents of CM and BC are reported to the SEER registry as in situ (not included in this analysis), localized, regional, distant and unstaged.
The observed and expected numbers of BC cases among CM survivors are shown in Table I. Overall, we found a small (11%) but statistically significant increased risk of CM, with somewhat higher increase in risk for patients diagnosed with CM at or before age 50, for those diagnosed more recently and for the first 3 years following the CM diagnosis.
|Age at CM diagnosis (years)|
|Time from CM diagnosis (years)|
A comparison of the characteristics of the post-CM BC patients with the overall population of SEER BC patients is shown in Table II. There were more post-CM BC patients diagnosed with localized breast cancer, and fewer with regional or distant spread, than among the overall group of SEER BC patients. In addition, the post-CM patients had statistically significant thinner tumors. Female BC survivors exhibited a 16% elevation in risk for a subsequent CM (Table III). This increase in risk was considerably greater for younger patients, patients who received radiation therapy and during the first 3 years following BC diagnosis, and slightly higher for patients diagnosed with BC between 1987 and 1999. There were a total of 47 patients who had diagnoses of both CM and BC in the same month and were not included as either post-BC CM patients or post-CM BC patients.
|Stage||Post-CM BC patients number (%)||All SEER BC patients (%)||p-value||p-value (adjusted)|
|Localized (%)||528 (67.0)||(58.6)||< 0.0001||< 0.0001|
|Regional (%)||195 (24.8)||(32.8)|
|Distant (%)||35 (4.4)||(5.7)|
|Mean BC thickness (mm)||18.32||20.93||0.001||0.0022|
|Median BC thickness (mm) interquartile ranges (IQR)||15 (10–24)||16 (10–25)||0.001|
|Age at BC diagnosis (years)|
|≤ 50||178||121.6||1.46||1.26–1.70||< 0.00001|
|Time from BC diagnosis (years)|
A comparison of characteristics of the post-BC CM patients with the SEER population of CM patients is summarized in Table IV. Slightly more of the post-BC CM patients were diagnosed at more advanced stages, but the difference was not statistically significant after adjusting for age. The post-BC patients also had a statistically significant greater mean tumor thickness, but the difference was no longer significant once adjustment was made for the age of the patients. This suggests that there may be an element of detection bias in the ascertainment of BC after CM but not CM after BC.
|Stage||Post-CM BC patients number (%)||All SEER BC patients (%)||p-value||p-value (adjusted)|
|Local (%)||565 (86.5)||(88.7)||0.048||0.17|
|Regional (%)||58 (8.9)||(8.3)|
|Distant (%)||30 (4.6)||(3.0)|
|Mean CM thickness (mm)||1.22||1.04||0.004||0.78|
|Median CM thickness (mm) interquartile ranges (IQR)||0.65||0.60||0.043|
We have found modest but statistically significant increased risks of cutaneous melanoma among female breast cancer patients and of breast cancer among female cutaneous melanoma survivors. However, since BC is quite prevalent, the small risk still translates to a reasonable excess of CM cases in the general population. This mutual interaction suggests a bona fide association, although the clinical and/or biologic nature of this link is unknown.
Young female BC patients (≤ 50 years of age) experienced a 46% increased risk for a subsequent CM. Although there are inherent risks in comparing SIRs between different age structures, the greater risk among the younger group is consistent with the notion that some element of genetic influence may be operative. For instance, the Breast Cancer Linkage Consortium found a 2.58-fold enhanced risk of CM among 3,728 BRCA2 carriers from 173 breast-ovarian cancer families.12 Age appeared to play a role even among BRCA2 carriers as there were no melanoma diagnoses after age 65.12 We also detected a 19% increased chance of a second BC among younger CM patients. While it is true that our cohort would have more follow-up for younger patients, this would not account for the difference in SIRs between the ≤ 50 at diagnosis patients and the > 50 patients since the SIR adjusts for age group. Breast cancer has also been found to be increased 3.8-fold among 52 Swedish melanoma-prone kindreds who harbor a recurrent 113insArg mutation in CDKN2A.13 Although the prevalence of mutations of high-risk genes (such as BRCA2 and CDKN2A) is low in the general population, the frequency of low- and medium-risk modifier genes can be significant and may coordinately modulate the risk of both BC and CM in the general population.
In both the genetic studies and our population-based analysis, the possibility of a detection bias cannot be eliminated. It is more likely that a skin cancer is identified during routine post-BC physical examination than a BC during post-CM surveillance since breast examinations are not part of the routine post-CM follow-up. One assay for detection bias is the recognition of earlier lesions among the test cohort compared to the general SEER population. We did not find that post-BC melanomas were significantly thinner, after adjustment, when compared to all melanomas in the SEER database (p = 0.78). We did find more localized BC cases among our post-CM cohort when compared to other SEER BC. This earlier disease may possibly be attributable in part to the use of routine chest radiographs for post-CM surveillance or greater medical attention after the first CM.
We also found a 42% increased risk of CM among BC patients who underwent radiation therapy. This increased risk of CM is consistent with a large recent institutional case series of 1,884 patients who underwent radiation therapy for early-stage breast cancer.15 In that analysis, 14 melanomas were detected over 8 years of follow-up (p = 0.0002). However, the tumors did not appear to correlate with the anatomic distribution of irradiation. Although external radiation may lead to increased skin examinations, a more systemic effect cannot be eliminated. One caveat, however, is that our analysis is subject to limitations in the reliability of therapeutic data from the SEER.
An interaction between age and radiation therapy is also possible. We found that the SIRs for patients ≤ 50 who had or had not received radiation therapy were 1.78 and 1.36, respectively, and the SIRs for patients > 50 who had or had not received radiation therapy were 1.30 and 1.00, respectively. These SIRs are in line with expected values (Table III) if both age and radiation therapy affected the probability of CM but no synergistic interaction existed between the 2 parameters.
In summary, we found evidence for a bidirectional association between breast cancer and cutaneous melanoma. Although enhanced detection is always a possible confounding factor, there is increasing genetic evidence that, mechanistically, the 2 cancers may be linked. Our data does not provide grounds for enhanced breast cancer screening after a primary melanoma, although a routine skin examination could be incorporated into a postbreast cancer regimen.
Supported in part by the Dermatology Foundation and the American Skin Association (to H.T.) and Hong Kong Baptist University Faculty Research Grant 99-00/II-84 (to W.G.).
- 14Statistical methods in cancer research: the design and analysis of cohort studies. IARC Sci Publ 1987; 82: 1–406., .