A Large Cohort Study of Nonsteroidal Anti-inflammatory Drug Use and Melanoma Incidence

Epidemiological studies have demonstrated chemopreventative benefits of nonsteroidal anti-inflammatory drugs (NSAIDs) for total cancer incidence and mortality (1,2) and for site-specific cancers such as those of the colon (3,4), breast (5), or prostate (6). Few epidemiological studies have examined the association of NSAIDs with melanoma risk and have found conflicting results (1,7–9). NSAIDs block cyclooxygenase enzymes, which inhibit the synthesis of proinflammatory molecules that play a role in tumor growth and development (10,11). Cyclooxygenase expression has been associated with melanoma invasion (12) and metastasis (13). Thus, NSAIDs could play a role in the chemoprevention of melanoma.

We conducted a prospective study of the Vitamins and Lifestyle (VITAL) cohort to investigate whether commonly used over-the-counter and prescription NSAIDs are associated with melanoma risk. The VITAL study recruited 77 719 men and women aged 50–76 years in western Washington state from October 1, 2000, through December 31, 2002, and ascertained detailed information about lifestyle, dietary, and cancer risk factors by use of a baseline questionnaire. Further study details have been published previously (14). This study was approved by the Fred Hutchinson Cancer Research Center Institutional Review Board.

Participants were excluded if they reported a melanoma diagnosis at baseline (n = 1557), had missing data on NSAID use (n = 6652), or were nonwhite or did not report their race (n = 5701), leaving a total of 63 809 participants for this analysis. Participants self-reported NSAID exposure in the previous 10 years, noting years and frequency of their use of low-dose (81 mg) and regular- (325 mg) or extra-strength aspirin, ibuprofen, naproxen, and celecoxib or rofecoxib. Women also reported their intake of piroxicam and indomethacin. Ever use was defined as taking the medication at least once a week for a year in the 10-year period before baseline. Dose was computed as the average number of days per week of use for the 10 years before enrollment (ie, [days per week x years of use]/10).

Through linkage of the VITAL database to the NCI Surveillance, Epidemiology, and End Results cancer registry, 349 patients with incident cases of malignant melanoma were identified between baseline and December 31, 2005. Cox proportional hazards models were used to estimate age- and multivariable-adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for melanoma risk. Age was used as the time scale. We observed no violation of the proportional hazards by age with the Wald test. Nonusers of any NSAID were the common reference for all four drug exposures (ie, any NSAID; any NSAID excluding low-dose aspirin; regular- or extra-strength aspirin; and nonaspirin NSAIDs). The censor date was defined as the earliest date of withdrawal from the study (0.03%, n = 21), death (3.6%, n = 2720), out-of-area move (4.6%, n = 3486), or the end of follow-up on December 31, 2005. To control for confounding by indication, we adjusted for self-reported medical conditions that indicated (cardiovascular disease, chronic pain, arthritis) or contraindicated (kidney disease, gastric ulcer) NSAID use (as described in Table 1). Established or possible risk factors for melanoma were also included in the model; other lifestyle factors such as body mass index, total physical activity, and smoking history were not found to be confounding factors. To examine the association between participant characteristics and NSAID use (Table 1), odds ratios were calculated by logistic regression. All statistical tests were two-sided.

Among the 63 809 cohort participants, 40 200 (63%) regularly used any type of NSAIDs at least once a week for 1 year in the 10-year period before enrollment. Factors associated with regular NSAID use are given in Table 1. Multivariable associations between NSAID use and the risk of melanoma are summarized in Table 2. None of our four NSAID exposures examined (whether expressed as overall use, duration, or dose) was associated with melanoma risk. Compared with nonuse, no melanoma risk reduction was detected with use (≥4 d/wk on average over the past 10 years) for any NSAID dose (HR = 1.12, 95% CI = 0.84 to 1.48), any NSAID excluding low-dose aspirin (HR = 1.03, 95% CI = 0.74 to 1.43), regular- or extra-strength aspirin (HR = 1.10, 95% CI = 0.76 to 1.58), or nonaspirin NSAIDs (HR = 1.22, 95% CI = 0.75 to 1.99). We also observed no difference in melanoma risk among males (HR = 1.21, 95% CI = 0.85 to 1.71) or among females (HR = 1.01, 95% CI = 0.63 to 1.60; P_interaction = .63) for dose of any NSAID.

There were 157 patients with in situ melanoma and 191 with invasive (including local, regional, and distant) melanoma. The association between dose of any NSAID use and the risk of in situ melanoma (HR = 1.17, 95% CI = 0.77 to 1.78, >4 days a week vs none) was not appreciably different than the risk of invasive melanoma (HR = 1.09, 95% CI = 0.75 to 1.57; P_interaction = .38). Furthermore, we did not observe an association between dose of any NSAID use and the risk of metastatic melanoma, defined as regional or distant disease (HR = 1.09, 95% CI = 0.33 to 3.62; n = 23). Lastly, we did not detect an association between dose of any NSAID use and tumor thickness (n = 172 and P_trend = .98).

In conclusion, we found no evidence to support an association between NSAID use and decreased melanoma risk. The lack of effect was persistent even though our NSAIDs variables accounted for frequency and duration of NSAID use and we considered the major classes of NSAIDs separately and combined. Our findings are consistent with a recently published large prospective study (6) that also found no association between duration of daily regular-strength aspirin use and the risk of incident melanoma. Although clinical reports of patients with surgically incurable or metastatic melanoma who demonstrated regression or complete remission with NSAID treatment have been published (15,16) and the rate of melanoma metastasis has been shown to be lower for patients using NSAIDs than nonusers (8), we did not observe an association between dose of any NSAID and tumor invasion, tumor thickness, or metastatic potential.

This study had several limitations, including the absence of detailed information on some known melanoma risk factors such as sunlight exposure and number of nevi. However, adjusting for risk factors of melanoma did not alter the risks in the multivariable model, and so it is unlikely that including residual risk factors would appreciably change the results. Also, the actual NSAID pills per day was not ascertained. Finally, the follow-up time between the 10-year period for which we had NSAID use data and cancer incidence was only 5 years on average, which may not be sufficient if the lag time between the initiation and diagnosis of melanoma is on the order of decades.

Strengths of this investigation include its prospective design, large cohort size, and the availability of baseline information on major potential confounding factors (including constitutional, personal, and family skin cancer history and patient numbers). In addition, we obtained data on long-term use of NSAIDs including commonly used over-the-counter NSAIDs.

We found no association between self-reported NSAID use and melanoma risk. Given the potential side effects of NSAIDs, including gastric ulcers, the search must continue for a good chemopreventive agent for melanoma.