Fax: (650) 721-3464
Aspirin is associated with lower melanoma risk among postmenopausal Caucasian women
The Women's Health Initiative
Version of Record online: 11 MAR 2013
Copyright © 2012 American Cancer Society
Volume 119, Issue 8, pages 1562–1569, 15 April 2013
How to Cite
Gamba, C. A., Swetter, S. M., Stefanick, M. L., Kubo, J., Desai, M., Spaunhurst, K. M., Sinha, A. A., Asgari, M. M., Sturgeon, S. and Tang, J. Y. (2013), Aspirin is associated with lower melanoma risk among postmenopausal Caucasian women. Cancer, 119: 1562–1569. doi: 10.1002/cncr.27817
- Issue online: 8 APR 2013
- Version of Record online: 11 MAR 2013
- Manuscript Accepted: 10 AUG 2012
- Manuscript Revised: 3 AUG 2012
- Manuscript Received: 14 JUN 2012
- anti-inflammatory agents;
Nonsteroidal anti-inflammatory drugs (NSAIDs) have been associated with decreased risk of gastric, colorectal, and breast cancer. However, the impact of NSAIDs on the risk of melanoma has been inconsistent. The authors evaluated the association between NSAID use and cutaneous melanoma risk in the Women's Health Initiative (WHI) Observational Study (OS).
At study entry, use of aspirin (acetylsalicylic acid [ASA]) and nonaspirin NSAIDs was assessed among 59,806 postmenopausal Caucasian women ages 50 to 79 years. Cox proportional hazards models were constructed after adjusting for participant skin type, sun exposure history, and medical indications for NSAID use among other confounders.
During a median follow-up of 12 years, 548 incident melanomas were confirmed by medical review. Women who used ASA had a 21% lower risk of melanoma (hazard ratio, 0.79; 95% confidence interval, 0.63-0.98) relative to nonusers. Increased duration of ASA use (<1 year, 1-4 years, and ≥5 years) was associated with an 11% lower risk of melanoma for each categorical increase (Ptrend = .01), and women with ≥5 years of use had a 30% lower melanoma risk (hazard ratio, 0.70; 95% confidence interval, 0.55-0.94). In contrast, use of non-ASA NSAIDs and acetaminophen were not associated with melanoma risk.
Postmenopausal women who used ASA had a significantly lower risk of melanoma, and longer duration of ASA use was associated with greater protection. Although this study was limited by the observational design and self-report of NSAID use, the findings suggest that ASA may have a chemopreventive effect against the development of melanoma and warrant further clinical investigation. Cancer 2013. © 2012 American Cancer Society.
Melanoma incidence has been rising steadily,1 which has prompted the investigation of primary prevention strategies.2 The use of nonsteroidal anti-inflammatory drugs (NSAIDs), including aspirin (acetylsalicylic acid [ASA]) and nonaspirin NSAIDs, has been associated with a decreased risk of a variety of cancers, including gastric,3 colorectal,4 and breast cancer.5 Thus, interest in the chemopreventive benefits of NSAIDs for other malignancies, such as melanoma, has grown.
NSAIDs, which inhibit cyclooxygenase (COX) enzymes, may prevent carcinogenesis through both COX-dependent and COX-independent mechanisms.6 COX-1 is constitutively expressed in human tissue, whereas COX-2 is an indicator of inflammation and has been implicated in the development of cancer.7 Human melanoma cell lines over-express COX-2,8 and high COX-2 levels are associated with melanoma progression.9 Thus, COX-2 inhibition by NSAIDs may reduce melanoma development and progression. In addition, NSAIDs inhibit the activation of nuclear factor-kappa-B (NFκB), a transcription factor that promotes inflammation and reduces apoptosis, through a COX-independent mechanism.10
Evidence from observational studies investigating the association between NSAID use and the risk of melanoma has been inconsistent. Some case-control studies11-13 have reported a significant association between intake of NSAIDs and lower melanoma risk. In contrast, a randomized trial of alternate-day, low-dose ASA14 and 2 large cohort studies15, 16 failed to demonstrate a significant association between NSAID use and melanoma.
By using the Women's Health Initiative (WHI) Observational Study (OS), which was designed to evaluate new risk indicators and biomarkers for disease in postmenopausal women,17 we investigated whether NSAIDs are associated with a lower risk of cutaneous melanoma. Because light skin pigmentation is the major risk factor for melanoma and approximately 95% of cutaneous melanoma cases occur in Caucasians,18 we focused on this population in the WHI.
MATERIALS AND METHODS
The design of the WHI OS (National Clinical Trial identifier NCT00000611) has been described previously along with the eligibility criteria and recruitment methods.17, 19 In brief, the OS enrolled 93,676 postmenopausal women ages 50 to 79 years between 1993 and 1998 at 40 clinical centers throughout the United States. Most participants in the cohort were initially screened for the randomized trials and identified as ineligible or unwilling to participate, but they still were interested in and eligible for participating in the OS. Enrollment in the OS required women to have a minimum predicted survival of 3 years.19
Consistent with the typical predominance of melanoma cases among Caucasians, only 7 non-Caucasian women developed melanoma during follow-up in the WHI OS. Thus, we limited our cohort to OS women of Caucasian race/ethnicity (N = 78,413) and further excluded participants with missing covariates that were included in the fully adjusted scientific model, yielding a final cohort of 59,806 women. Participants were followed for as long as possible (eg, ≤10 years for women enrolled in the OS but not in the Extension Study, and ≤15 years for those women who also enrolled in the Extension Study). All procedures and protocols were approved by the institutional review boards at each participating institution, and all participants provided written informed consent.
Demographics, medical history, diet and supplement use, physical activity, smoking status, and physical measures were obtained by questionnaire, interview, or physical examination at baseline. Regional solar radiation, reported in langleys (1 langley = 1 g-cal/cm2), was based on the amount of sunlight reaching the clinical center sites as measured by the US Weather Bureau.20
Sun-exposure variables were collected in a questionnaire 4 years after study entry. Women reported average daily time spent in the sun (<30 minutes, 30 minutes to 2 hours, ≥2 hours) during summer and the remainder of the year in their childhood, teens, thirties, and at their current age. Skin reaction to the sun (burn/tan) after 45 to 60 minutes in first summer sun was recorded. Sunscreen use in the past year, if outdoors >10 minutes, was reported along with strength (eg, sun protection factor [SPF]). The possibility of recall bias of sun-exposure history in participants with melanoma was explored by comparing the distribution of self-reported exposure before and after diagnosis.
Measurement of Exposure
Baseline medication intake was collected using an interview-administered questionnaire, which may be accessed at cleo.whi.org (Form 44, Current Medications; [accessed July 28, 2012]).21 Participants were asked, “Do you take aspirin pills or powders, for example, Anacin, Bufferin? Do you take ibuprofen tablets or capsules, for example, Advil, Motrin, or Nuprin?” If participants reported a minimum of twice weekly use for the prior 2 weeks, then the drug type, strength, and duration of use were recorded. Medication data were validated by checking prescription records and bottle labels.5 Medication use was recorded again at year 3 when participants returned for a second clinical visit.
To minimize multiple comparisons, 3 mutually exclusive exposure groups were defined to allow for a single, 3-way comparison. Participants were divided into those who reported no NSAID use at baseline (N = 35,529) versus those who used NSAIDs. Among the women who used NSAIDs, those who reported ASA use were defined as ASA users (N = 15,089), including women who reported concurrent use of non-ASA NSAIDs (n = 2857). Finally, non-ASA NSAID users were defined as participants who reported only non-ASA NSAID use (N = 9188).
Outcome of Interest
Our primary outcome of interest was the time from study entry to development of melanoma, with participants censored at death or at their last date of follow-up. Participants reported medical outcomes, including melanoma, in an annual questionnaire. Cutaneous melanoma cases were confirmed by adjudication of pathology and medical records and were coded following the International Classification of Diseases for Oncology, second edition coding scheme.
Cox proportional hazards techniques were used to address: 1) whether the risk of melanoma differs among the 3 exposure groups (ASA users, non-ASA NSAID users, and NSAID nonusers) and 2) whether the duration of NSAID use affects melanoma risk. The first model compared the hazard ratio (HR) of melanoma among the user groups. The second model investigated the duration of NSAID use (<1 year, 1-4 years, ≥5 years) as an ordered, categorical variable. For both models, we report corresponding HRs with 95% confidence intervals (CIs) that adjust for the following potential confounders: age, education, body mass index, smoking status, vitamin D intake, physical activity, history of nonmelanoma skin cancer, history of melanoma, skin reaction to the sun, regional solar radiation, childhood and current summer sun exposure, sunscreen use, time since last medical visit, NSAID and acetaminophen categories, and medical indication for NSAID use (including history of cardiovascular disease, arthritis, and migraine). To identify potential residual confounding by analgesic use, we also performed the above analyses for acetaminophen use, which is a non-NSAID analgesic.
To assess whether the effect of NSAID use on the hazard of melanoma varied by prespecified baseline risk factors for skin cancer, interaction effects between the potential modifier and the user groups were assessed. Previous studies have indicated that melanoma risk is associated with history of skin cancer, smoking status,22 intermittent and potentially ambient ultraviolet exposure,23, 24 and perhaps vitamin D intake.25 Thus, 8 potential prespecified modifiers included: 1) age, 2) history of nonmelanoma skin cancer, 3) history of melanoma, 4) smoking status, 5) childhood summer sun exposure, 6) current summer sun exposure, 7) regional solar radiation, and 8) vitamin D intake.
All statistical analyses were completed using SAS 9.2 (SAS Institute Inc., Cary, NC). All statistical tests were 2-sided and were conducted at the .05 level of significance.
In this cohort, 59,806 Caucasian women were deemed eligible for our study. Table 1 presents select descriptive statistics on our sample by exposure group. At baseline, 25% of the women reported use of ASA at least twice in the prior 2 weeks, 15% reported exclusive use of non-ASA NSAIDs at least twice in the prior 2 weeks, and 59% of the women reported no regular use of NSAIDs in the prior 2 weeks. Of the ASA users, 25% reported low-dose (≤81 mg) ASA use. Among the non-ASA NSAID users, ibuprofen and naproxen were the most commonly used medications, and none reported use of selective COX-2 inhibitors. At the year 3 clinical visit, 59% of ASA users and 38% of non-ASA NSAID users reported continued use of the same medications.
|Percentage of Patients|
|Covariate||Aspirin Users||Nonaspirin NSAID Users||NSAID Nonusers||Total No. of Patients||P|
|No. of patients (%)||15,089 (25.2)||9188 (15.4)||35,529 (59.4)||59,806|
|Median follow-up, y||11.3||11.9||11.9||N/A|
|Age, y||< .0001|
|BMI, kg/m2||< .0001|
|Total vitamin D intake, IUb||< .0001|
|200 to <400||35.1||36.2||35.7||21,298|
|400 to <600||23.5||23.4||22.7||13,764|
|History of NMSC||.07|
|History of melanoma||.23|
|Last medical visit within 1 y||< .0001|
|Cardiovascular disease ever||< .0001|
|Ever had arthritis||< .0001|
|Ever had migraine||< .0001|
|Regional solar radiation, langleysc||< .0001|
|Skin reaction to the sun||< .0001|
|Tans, does not burn||37.2||35.1||35.1||21,313|
|Burns, then tans||24.3||25.1||26.2||15,288|
|Burns, then tans minimally||27.1||28.2||27.1||16,295|
|Burns, does not tan||11.5||11.7||11.6||6910|
|Average daily time outdoors in summer as a child||.04|
|30 min to 2 h||25.9||25.1||26.1||15,471|
|Average daily time outdoors in summer currently||< .0001|
|30 min to 2 h||50.2||49.9||50.3||30,049|
Although the 3 exposure groups had statistically significant differences in covariates (Table 1), these were small percentage differences and are more reflective of the large sample size of the WHI. Nonetheless, these factors were included in our multivariate models. The most important predictors of melanoma risk, prior nonmelanoma skin cancer and melanoma history, did not differ among the groups. Overall, women who used ASA were older than women in the other exposure groups (Table 1). Non-ASA NSAID users had the highest proportion of obesity, and NSAID nonusers had the lowest. Not surprisingly, more ASA users reported a history of cardiovascular disease, and more non-ASA NSAID users reported a history of arthritis. NSAID users were more likely to have had a medical visit within the past year than NSAID nonusers. It is noteworthy that ASA and non-ASA NSAID users tended to reside in areas with lower regional solar radiation (langleys) than NSAID nonusers, and this was adjusted for in the multivariate models.
During a median follow-up of 12 years, 548 incident melanomas developed and were physician-adjudicated. There were 289 melanomas in situ and 255 invasive melanomas, and 4 cases were classified as unknown. In total, 115 cutaneous melanoma cases (0.76%) were reported among women who used ASA compared with 89 cases (0.97%) in non-ASA NSAID users and 344 (0.97%) in NSAID nonusers (global P = 0.06) (Table 2). The majority of melanoma cases (94%) occurred in women without a history of melanoma.
|NSAID Typea||No. of Melanoma Cases/Total Users||Incidence Per 100,000 Person-Years||Age-Adjusted HR (95% CI)||Fully Adjusted HR (95% CI)b|
|NSAID nonusers||344/35,529||87.1||1.00 (Ref)||1.00 (Ref)|
|ASA users||115/15,089||69.8||0.80 (0.65-0.99)||0.79 (0.63-0.98)|
|Non-ASA NSAID users||89/9188||87.9||1.02 (0.81-1.29)||1.05 (0.83-1.34)|
Type of Nonsteroidal Anti-Inflammatory Drug Use, Duration of Use, and Risk of Melanoma
ASA use was associated significantly with a 21% lower risk of melanoma relative to NSAID nonusers (HR, 0.79; 95% CI, 0.63-0.98) (Table 2). Although the aforementioned global P value for overall category of use indicated only borderline significance, the HR for aspirin users versus NSAID nonusers indicated that there was a significant difference in melanoma risk between these 2 groups. There was an effect of duration of ASA use on the hazard of melanoma as well. Specifically, each incremental increase in duration of ASA use (<1 year, 1-4 years, and ≥5 years) was associated with an 11% lower risk of melanoma (Ptrend = .01) (Table 3). The risk of melanoma was 30% lower with ASA use ≥5 years (HR, 0.70; 95% CI, 0.55-0.94). The association between ASA use and risk of melanoma did not differ by known and potential skin cancer risk factors, such as age, history of nonmelanoma skin cancer, history of melanoma, smoking status, childhood summer sun exposure, current summer sun exposure, regional solar radiation, or vitamin D intake at baseline in prespecified subgroup analyses (data not shown; all Pinteraction > .1).
|NSAID Type||No. of Patients With Melanoma||Age-Adjusted HR (95%CI)||Fully Adjusted HR (95%CI)a|
|ASA use, yb|
|None||433||1.00 (Ref)||1.00 (Ref)|
|<1||22||0.90 (0.82, 0.98)||0.89 (0.82, 0.98)|
|1-4||51||0.81 (0.67, 0.96)||0.79 (0.67, 0.96)|
|≥5||42||0.73 (0.55, 0.94)||0.70 (0.55, 0.94)|
|Non-ASA NSAID use, y|
|None||440||1.00 (Ref)||1.00 (Ref)|
|<1||26||0.87 (0.59-1.30)||0.92 (0.62-1.37)|
|1-4||50||1.10 (0.82-1.47)||1.15 (0.85-1.55)|
|≥5||32||0.97 (0.67-1.38)||0.94 (0.66-1.36)|
In contrast, non-ASA NSAID use was not associated with risk of melanoma relative to NSAID nonusers (HR, 1.05; 95% CI, 0.83-1.34) (Table 2). There was no association between duration of non-ASA NSAID use and melanoma risk (Pcategorical = .8) (Table 3). In addition, there was no interaction between non-ASA NSAID use and the 8 prespecified subgroups (data not shown; all Pinteraction > .1).
We also determined the relation between acetaminophen use (a non-NSAID analgesic) and melanoma risk. Acetaminophen use was not associated with risk of melanoma relative to acetaminophen nonusers (HR, 0.89; 95% CI, 0.68-1.17), nor was there an association with duration of use (Pcategorical = .2).
For a sensitivity analysis, we also performed a 2-way comparison of NSAID use (eg, any ASA use vs no ASA use and any non-ASA NSAID use vs no non-ASA NSAID use). These groups were not mutually exclusive and included concurrent use of other NSAIDs. Similar to findings from the primary analysis, ASA use was associated with lower melanoma risk (HR, 0.78; 95% CI, 0.63-0.96; P = .02), and non-ASA NSAID use was not associated with melanoma risk (HR, 1.02; 95% CI, 0.82-1.27; P = .9).
To evaluate potential recall bias of sun-exposure history, we compared reports of sun exposure before and after diagnosis of melanoma. Chi-square tests of association did not suggest significant differences in reporting of past sun exposure (all P > .1).
In the WHI OS, aspirin use was associated significantly with a 21% lower risk of melanoma relative to NSAID nonusers among postmenopausal Caucasian women who were followed for a median of 12 years. Increased duration of ASA use was associated with greater protection, and ASA use for ≥5 years was associated with a 30% lower risk of melanoma (Ptrend = .01). In contrast, neither non-ASA NSAID use nor acetaminophen use was associated with melanoma risk.
To our knowledge, the WHI OS is the first cohort to demonstrate an association between aspirin use and the risk of melanoma in postmenopausal Caucasian women, and the results are consistent with prior case-control studies demonstrating a relation between NSAID use and lower melanoma risk.11-13 Similar to our findings, 2 large case-control studies12, 13 reported different correlations among aspirin use, nonaspirin NSAID use, and the risk of melanoma. In a Dutch population-based case-control study by Joosse et al12 that used national pharmacy and pathology databases, women with continuous low-dose ASA use were half as likely to develop melanoma. No association was evident between non-ASA NSAID use and melanoma risk. However, those authors noted that non-ASA NSAIDs were commonly prescribed “as needed,” and there were few continuous users, which may be a limitation of their study.12 A US case-control study by Curiel-Lewandrowski et al13 similarly demonstrated that continuous ASA use for >5 years was associated with a 56% lower melanoma risk among women. In contrast, non-ASA NSAID use was not associated with risk of melanoma among women in that study, although continuous use >5 years was associated with lower melanoma risk among men and women combined.13 Like in the Dutch case-control study, the authors commented that few non-ASA NSAID users reported frequent use compared with aspirin users. Similar to those studies, infrequent non-ASA NSAID use in the WHI may provide an explanation for the lack of an association between non-ASA NSAID use and melanoma risk in our current analysis. Another potential explanation is that aspirin may have properties that other non-ASA NSAIDs lack.26 It has been demonstrated that aspirin down-regulates antiapoptotic, genes such as B-cell leukemia/lymphoma 2 (BCL2), and also up-regulates several tumor suppressor genes (eg, arachidonate 15-lipoxygenase [ALOX15] and prostate apoptosis response protein 4 [PAWR]).6 Furthermore, aspirin induces reactive oxygen species and mitochondrial toxicity in melanoma cell lines, promoting cell apoptosis.27 Aspirin may have selective protection against melanoma, specifically mediated by tyrosinase, which is highly expressed in melanoma cells.27
Our findings differ from those of a randomized clinical trial,14 the Women's Health Study, and 2 cohort studies.15, 16 The Women's Health Study is a randomized 2 × 2 factorial trial of alternate-day, low-dose aspirin (100 mg) and vitamin E in the primary prevention of cardiovascular disease and total cancer among women aged ≥45 years.14 During an average intervention period of 10 years, there were 138 melanoma cases among nearly 40,000 women, but there was no effect of alternate-day, low-dose aspirin on melanoma risk in post hoc analyses.14 In contrast to other randomized data,4, 28, 29 that trial also indicated that there was no effect of aspirin on total cancer and colorectal cancer incidence or total cancer mortality.14 The authors14 and others6, 28 noted that the alternate-day, low-dose aspirin may have been inadequate for cancer chemoprevention. It is noteworthy that 75% of women in our study reported regular or extra-strength aspirin use and had a longer follow-up period. In the Vitamin and Lifestyle (VITAL) cohort study, aspirin use (baby or regular strength) was not associated with melanoma risk in 63,809 men and women, although aspirin exposure was recorded from any period within the previous 10 years, and use at study baseline was not required.15 Similarly, in the Cancer Prevention Study II Nutrition Cohort, aspirin use was not associated with the risk of melanoma, but those investigators combined less than daily use, low-dose use, and past use, because there was insufficient statistical power to examine each individually, and their analyses were not adjusted for known skin cancer risk factors (eg, family history, skin type, sun exposure).16
Our finding that increased duration of aspirin use was associated with lower risk of melanoma after extended follow-up is consistent with studies of other malignancies, particularly colorectal cancer, that reported benefit after several years of aspirin use and extended follow-up (eg, 10-20 years).4, 30 In 2009, an expert group at the International Conference on Cancer Prevention considered an antitumor effect of aspirin and sulindac “very probable” and considered an effect of other NSAIDs “possible.”31 Although aspirin use is not recommended for cancer prevention in the general population, it has been recommended as part of the standard of care for chemoprevention in patients with Lynch syndrome, a form of hereditary colorectal cancer, after a trial indicated a significant protective effect of aspirin intervention at 10 years of follow-up, but not before.30 The combined cardiovascular benefits and potential antitumor effects of aspirin are particularly intriguing in the context of preventive medicine, but they must be weighed against well known risks, such as gastrointestinal bleeding.
Limitations of our current study include the self-report nature of medication use. However, medication data were obtained by an interview-administered questionnaire that included specific questions regarding NSAID use, and the data were validated with pill-bottle labels and/or prescription records. Although we adjusted for many potential confounders, including skin cancer risk factors and medical indications for NSAID use in this observational study, we lacked data on family history of melanoma and eye and hair color, and we cannot completely exclude residual confounding as an explanation for these findings. Strengths of this study include detailed information on participant skin type, sun-exposure and sun-protection habits, and prior history of skin cancer, which are important melanoma risk factors that other studies may not have captured.12, 14, 16 It is noteworthy that self-reports of past sun exposure were similar among those who completed the questionnaire before and after melanoma diagnosis in year 4, supporting prior arguments that recall bias in patients with melanoma may be minimal.32, 33 Additional strengths of the study include the large cohort with wide geographic diversity (high vs low ultraviolet exposure sites), a relatively large number of melanoma cases that were physician-adjudicated, and the lengthy follow-up.
In conclusion, the current results indicate that postmenopausal Caucasian women who used aspirin had a significantly lower risk of melanoma, and increased duration of use was associated with greater protection against melanoma. These findings suggest that aspirin may have a chemopreventive effect against the development of melanoma, and further clinical investigation is warranted.
This work was supported by the Women's Health Initiative program, which is funded by the National Heart, Lung, and Blood Institute, National Institutes of Health, US Department of Health and Human Services (grants N01WH22110, 24152, 32100-2, 32105-6, 32108-9, 32111-13, 32115, 32118-32119, 32122, 42107-26, 42129-32, and 44221). This study was supported in part by the National Institute of Arthritis and Musculoskeletal and Skin Diseases (grant 1K23AR056736-01) and by the Damon Runyon Clinical Investigator Award to Dr. Tang. A Selected Professions Fellowship from the American Association of University Women was held by Ms. Gamba.
CONFLICT OF INTEREST DISCLOSURES
Jean Tang is a consultant for Genentech.
- 21National Institutes of Health. Women's Health Initiative. Form 44—Current Medications. Available at: https://cleo.whi.org/studydoc/Central%20Forms/F044.pdf. Accessed July 28, 2012.