Background  Epidemiologic evidence indicates that aspirin use is associated with reduced risks of colon cancer and possibly several other cancers, including prostate and breast cancers. Recent results from the Women's Health Study randomized trial indicate that long-term use of low-dose aspirin (100 mg every other day) does not substantially reduce cancer risk. However, the potential effect of long-term daily use of higher doses of aspirin on cancer incidence remains uncertain.
Methods  We examined associations between long-term daily use of adult-strength aspirin (≥325 mg/day) and both overall cancer incidence and incidence of 10 types of cancer among 69810 men and 76303 women participating in the Cancer Prevention Study II Nutrition Cohort, a relatively elderly population. Aspirin use was reported at enrollment in 1992–1993 and updated in 1997, 1999, and 2001. Multivariable Cox proportional hazards regression was used to calculate rate ratios (RRs).
Results During follow-up through June 2003, 10931 men and 7196 women were diagnosed with cancer. Long-term (≥5 years) daily use of adult-strength aspirin, compared with no use, was associated with lower overall cancer incidence in men (multivariable-adjusted RR = 0.84, 95% confidence interval [CI] = 0.76 to 0.93) and non–statistically significantly lower overall cancer incidence in women (multivariable-adjusted RR = 0.86, 95% CI = 0.73 to 1.03). Overall cancer incidence per 100000 person-years (standardized to the age distributions of men and women in the study) with long-term daily aspirin use and no aspirin use was 1858 and 2163, respectively, among men and 1083 and 1169, respectively, among women. Long-term daily aspirin use was associated with lower incidence of colorectal cancer (RR = 0.68, 95% CI = 0.52 to 0.90 among men and women combined) and prostate cancer (RR = 0.81, 95% CI = 0.70 to 0.94) and a non–statistically significant lower risk of female breast cancer (RR = 0.83, 95% CI = 0.63 to 1.10).
Conclusions  Long-term daily use of adult-strength aspirin may be associated with modestly reduced overall cancer incidence in populations among whom colorectal, prostate, and breast cancers are common.
CONTEXT AND CAVEATS
Aspirin use is associated with reduced risk of colon cancer.
Risks of 10 types of cancer were estimated among mainly elderly men and women from the Cancer Prevention Study II Nutrition Cohort from surveys of aspirin use and cancer diagnosis.
Daily use of adult-strength aspirin for 5 or more years was associated with reduced risks of overall cancer and prostate cancer among men and of colorectal cancer among men and women combined.
Daily use of adult-strength aspirin for 5 or more years may be associated with reductions in cancer incidence among populations in which incidence of colorectal, prostate, and breast cancers is high.
 Being an observational study, it is subjected to confounding by factors that are associated with aspirin use and cancer risk that were not measured and/or may not be measurable. No data on potentially harmful side effects from daily long-term use of adult-strength aspirin were gathered.
Recently, the Women's Health Study, a large 10-year randomized trial of alternate-day low-dose aspirin (100 mg every other day) showed no reduced risk of any cancer in aspirin users (1). However, the potential effect of daily adult-strength aspirin (i.e., ≥325 mg/day) on overall cancer incidence or on incidence of specific individual cancers remains uncertain. Considerable evidence suggests that aspirin use could reduce risk of several cancers. Aspirin and other nonsteroidal anti-inflammatory drugs (NSAIDs) inhibit the development of many different types of cancers in rodent models, including cancers of the colorectum, breast, prostate, lung, skin, and bladder (2). It has been hypothesized that aspirin may inhibit carcinogenesis by reducing the synthesis of prostaglandins by cyclooxgenase (COX)-1 and COX-2 enzymes (3). Aspirin reduced risk of colorectal polyp recurrence in two randomized trials (4,5) and has been consistently associated with lower risk of colorectal cancer in observational studies (2), although the optimal dose associated with reduced colorectal cancer risk remains unclear. Meta-analyses of observational studies have reported considerably reduced risk of gastric and esophageal cancer with aspirin use (6), although these cancers are not common in the United States. Meta-analyses of observational studies have also reported smaller, but statistically significant, reductions in risk of breast (6), prostate (7), and lung cancers (8) with aspirin use, and some studies have also suggested reduced risk of other cancers (9).
Although many epidemiologic studies have examined the association between aspirin use and individual cancers, it is difficult to use results of these studies to assess the overall potential cancer prevention benefit of any particular aspirin regimen. Studies of individual cancers have examined widely varying levels of aspirin use. In addition, publication bias is a concern in that researchers may prioritize the publication of results for individual cancers that show associations with aspirin use in their study population. Studies of overall cancer incidence may be more useful with respect to assessing aspirin's overall potential cancer prevention benefits, but few studies have examined overall cancer incidence. In a randomized trial of aspirin for cardiovascular disease prevention among male British doctors (10), cancer incidence rates (examined as a secondary outcome) were not statistically significantly reduced among men randomly assigned to 500 mg aspirin daily. However, this trial included only 5139 men followed for a maximum of 6 years and, therefore, may have been too small and short in duration to provide important information about the potential effects of long-term aspirin use on cancer. Use of prescription low-dose aspirin (≤150 mg/day) was associated with slightly increased overall cancer incidence in a Danish pharmacy database analysis (11), although there was no evidence of increasing risk with duration of use, and results could not be adjusted for smoking. In the National Health and Nutrition Examination Survey (NHANES)-I cohort (12), use of any aspirin during the month before enrollment was associated with a modest but statistically significant reduction in overall cancer incidence, but duration of use was not examined. To our knowledge, no previous studies have examined the association between long-term daily aspirin use and overall cancer incidence.
 Any meaningful effect of aspirin use on overall cancer incidence could be of considerable importance. Unlike other traditional NSAIDs or COX-2 inhibitors, aspirin has been proven in numerous randomized trials to reduce risk of cardiovascular events, although this cardiovascular benefit is accompanied by increased risk of serious gastrointestinal bleeding, both at low doses (e.g., approximately 80 mg, a common dose in tablets intended for children or specifically for cardiovascular disease prevention) and at “adult-strength” doses (defined here as at least 325 mg, the lowest standard dose in tablets intended for adults in the United States) (13–15). The US Preventive Services Task Force recommends consideration of aspirin therapy in individuals for whom the coronary heart disease benefits are likely to exceed the risks of serious bleeding events but does not specify an optimum dose (16). If daily use of adult-strength aspirin were to reduce overall cancer risk, there could be important implications with respect to who should be taking aspirin and at what dose. We therefore examined the association between long-term daily use of adult-strength aspirin and overall cancer incidence in a large cohort of predominantly elderly US men and women, using detailed information on aspirin use reported at several different time points.
Subjects and Methods
Men and women in this analysis were drawn from the 86404 men and 97786 women in the Cancer Prevention Study II Nutrition Cohort (hereafter called the Nutrition Cohort), a prospective study of cancer incidence and mortality among US men and women that was established in 1992–1993 and has been described in detail elsewhere (17). The Nutrition Cohort is a subgroup of the approximately 1.2 million participants in the Cancer Prevention Study II (CPS-II), a prospective study of cancer mortality that was established by the American Cancer Society in 1982. The Emory University Institutional Review Board approves all aspects of the Nutrition Cohort. At enrollment in 1992–1993, Nutrition Cohort participants completed a mailed self-administered questionnaire that included information on demographic, medical, and lifestyle factors. Follow-up questionnaires to update exposure information and to ascertain newly diagnosed cancers were mailed in 1997, 1999, 2001, and 2003. The response rate among living participants for each of these follow-up questionnaires was at least 89%. Written informed consent to participate in the study was implied by return of the completed questionnaires.
For this analysis, we excluded 3232 men and 3212 women who were lost to follow-up (those who were alive at the time of the first follow-up questionnaire in 1997 but did not return the 1997 follow-up questionnaire or any later questionnaire). We also excluded participants with a history of cancer other than nonmelanoma skin cancer at enrollment (9761 men and 13084 women). In addition, we excluded participants with incomplete information on smoking status or use of aspirin or other NSAIDs at enrollment (3566 men and 5171 women) or with missing information on year of cancer diagnosis (35 men and 16 women). After these exclusions, a total of 69810 men and 76303 women remained for analysis.
We documented 18127 participants (10931 men and 7196 women) who were diagnosed with cancer between enrollment in 1992–1993 and June 30, 2003. Of these, 14703 diagnoses were initially identified by self-report on the 1997, 1999, 2001, or 2003 follow-up questionnaires and subsequently verified by obtaining medical records or, when complete medical records could not be obtained, through linkage with state registries (17). A previous comparison (18) of self-reports with information from state cancer registries has demonstrated that participants in the Nutrition Cohort can accurately self-report a cancer diagnosis (sensitivity = 0.93). An additional 3424 participants were identified as having died from cancer through linkage of the cohort with the National Death Index (19). For these 3424 case subjects, the death certificate listed cancer as the primary cause of death between the date of enrollment and December 31, 2002. A total of 2894 self-reports of cancer diagnoses could not be verified and therefore were not counted as cases.
Ascertainment of aspirin use.
Aspirin use was reported on questionnaires in 1982 (at enrollment into the larger CPS-II mortality cohort), 1992–1993 (at enrollment into the Nutrition Cohort), 1997, 1999, and 2001. The 1982 questionnaire asked for “times per month” that aspirin was used in the last month. The questionnaire completed in 1992–1993 (hereafter referred to as the 1992 questionnaire) asked for average days per month of aspirin use during the past year, the average number of pills taken on days used, and the number of years of use. Follow-up questionnaires in 1997, 1999, and 2001 asked similar questions about days per month and pills per day but asked separately about use of low-dose (or “baby”) aspirin and adult-strength aspirin.
We used Cox proportional hazards modeling (20) to calculate rate ratios (RRs) for cancer incidence associated with aspirin use adjusted for age and potential cancer risk factors. The time axis used was follow-up time since enrollment in 1992–1993. For analyses of overall cancer incidence, diagnosis date was defined as the date of first cancer diagnosis. In addition to examining overall cancer incidence, we also examined the association between aspirin use and the incidence of the 10 individual cancers for which statistical power was greatest (based on the number of expected cases among long-term daily aspirin users). For analyses of individual cancer sites, date of diagnosis was defined as the diagnosis date of the individual cancer of interest, regardless of whether a different cancer had been diagnosed earlier in the follow-up period.
Our analyses were designed specifically to examine long-term daily use of adult-strength aspirin. We had insufficient statistical power to examine long-term use of low-dose aspirin. Although aspirin dose was not reported on the 1982 or 1992 questionnaires, we considered all aspirin use reported in 1982 or 1992 to be adult-strength aspirin. It is unlikely that substantial numbers of our participants were taking daily low-dose aspirin in or before 1992 because the efficacy of low-dose aspirin for cardiovascular disease prevention was not well established in 1992 (21). Daily use of adult-strength aspirin was defined as use at least 30 “times” per month in 1982 and as 30 or 31 days per month in 1992, 1997, 1999, or 2001. Although years of aspirin use was reported at enrollment, frequency of past use was not assessed. We therefore defined duration of daily use based on having reported current daily aspirin use on two or more consecutive questionnaires, as detailed below.
We created a time-dependent variable for aspirin use with four categories: 1) never reported use; 2) low dose, less than daily, or past use only; 3) current daily use of adult-strength aspirin for less than 5 years; and 4) current daily use of adult-strength aspirin for 5 or more years. This comparison is similar to that used in previous analyses of aspirin and other NSAIDs in this cohort (22,23). Participants were categorized as daily users of adult-strength aspirin for 5 or more years if they met the following criteria: 1) for the 1992–1997 follow-up interval, if they had reported at least 5 years of aspirin use on the 1992 questionnaire and daily aspirin use on the 1982 and 1992 questionnaires; 2) for the 1997–1999 follow-up interval, if they had reported daily aspirin use on the 1992 and 1997 questionnaires (not including low-dose aspirin); 3) for the 1999–2001 follow-up interval, if they had reported daily aspirin use on the 1992, 1997, and 1999 questionnaires (not including low-dose aspirin); 4) for the 2001–2003 follow-up interval, if they had reported daily aspirin use on the 1997, 1999, and 2001 questionnaires (not including low-dose aspirin).
During each follow-up interval, participants who had not reported any aspirin use on either the questionnaire at the start of that follow-up interval or on any previous questionnaire were categorized as never users; those who were neither never users nor current daily users of adult-strength aspirin were grouped into a mixed-use category that included less than daily, low-dose, and past users. At the start of each follow-up interval, we censored participants whose aspirin use status could not be updated due either to missing or invalid aspirin data on the follow-up questionnaire or failure to return the follow-up questionnaire.
Age was adjusted for using the stratified Cox procedure with 1-year age strata (24). Additional potential confounders that were included in all multivariable models were race (white, black, other or missing); education (less than high school, high school graduate, some college, college graduate, graduate school, or unclassifiable); smoking status (never, former: <20 years, ≥20 years, or years unknown, current: <40 or ≥40 years); physical activity level (<3.5, 3.5 to <4.5, 4.5 to <14.0, 14.0 to <24.5, ≥24.5 metabolic equivalents/wk, or unclassifiable); body mass index (BMI, <22.5, 22.5 to <25, 25.0 to <27.5, 27.5 to <30, ≥30 kg/m2, or unclassifiable); history of heart attack, diabetes, or hypertension (yes or no); colorectal endoscopy (yes or no); and use of NSAIDs other than aspirin (none, 1–29, or ≥ 30 pills per month). Use of nonaspirin NSAIDs (in pills per month) was modeled as frequency of use during the 1992–1997 follow-up interval (when duration of NSAID use could not be calculated) and duration of regular NSAID use during the remainder of follow-up, as described for a previous analysis in this cohort (23). Models that included men were also adjusted for history of prostate-specific antigen (PSA) testing, and models that included women were also adjusted for use of hormone replacement therapy and for history of mammography. Models that included both men and women were also adjusted for sex. Mammography was modeled using a time-dependent variable for having had a mammography in the previous 2 years. Follow-up after 1997 (the first time information on PSA testing and colorectal endoscopy was collected) was adjusted for PSA testing using time-dependent variables for having had PSA test during the previous follow-up interval and for ever having had a colorectal endoscopy. Except for the time-dependent variables described above, potential confounders were modeled based on status at enrollment. Categories were chosen to be similar to those used in previous analyses in this cohort. Further adjustment for nutritional factors, including use of alcohol, multivitamins, and calcium supplements, and intake of saturated fat and vegetables had negligible effects on results.
We examined whether the association between long-term daily aspirin use (defined as daily use of adult-strength aspirin for ≥5 years) and overall cancer incidence varied by cigarette smoking status, BMI, attained age, and follow-up time by modeling multiplicative interaction terms between long-term daily aspirin use and variables for BMI, attained age, follow-up time (all modeled as continuous variables), and cigarette smoking status (never, former, current). Two-sided P values for interaction were calculated using the likelihood ratio statistic (25); P value of less than .05 was considered to be statistically significant.
At enrollment in 1992–1993, only 2.3% of male participants and 1.3% of female participants met our definition for being long-term daily aspirin users. However, the prevalence of long-term daily aspirin use increased during the follow-up period. Long-term daily aspirin users contributed 4.6% of the person-time included in this analysis among men and 2.1% of that among women. Most long-term daily aspirin users reported using either one aspirin pill per day (79% of person-time among men, 56% of person-time among women) or two aspirin pills per day (11% of person-time among men, 20% of person-time among women).
At enrollment in 1992–1993, participants who were long-term daily aspirin users were, on average, older and slightly more likely to be white than participants who reported no aspirin use (Table 1), although nearly all participants in this cohort were white and older than age 50 years, regardless of aspirin use. Long-term daily aspirin users were slightly more likely than nonusers to be highly educated, to be former smokers rather than never smokers, to have a high BMI, and to use nonaspirin NSAIDs. In addition, long-term daily aspirin users were considerably more likely than nonusers to have had a history of heart attack, diabetes, or hypertension, presumably reflecting the use of aspirin to prevent cardiovascular disease. Among women, long-term daily aspirin users were more likely than nonusers to use hormone replacement therapy (41% versus 30%) but were not more likely to have reported a mammogram in the last year (64% in both groups). Information on PSA testing and colorectal endoscopy was first collected in 1997. At that time, long-term daily aspirin users were slightly more likely than nonusers to have ever had a PSA test (77% versus 71% among men) or a colorectal endoscopy (61% versus 55% among men, 52% versus 50% among women).
Selected cancer risk factors by aspirin use at enrollment of the Cancer Prevention Study II Nutrition Cohort in 1992–1993*
Long-term daily aspirin use was associated with lower overall cancer incidence than no use in men (multivariable-adjusted RR = 0.84, 95% confidence interval [CI] = 0.76 to 0.93) and women (multivariable-adjusted RR = 0.86, 95% CI = 0.73 to 1.03), although the association was not statistically significant in women (P = .10) (Table 2). These multivariable-adjusted rate ratios were slightly lower than those adjusted only for age (in men, age-adjusted RR = 0.85, 95% CI = 0.77 to 0.94; in women, age-adjusted RR = 0.91, 95% CI = 0.77 to 1.08). Current use of adult-strength aspirin for less than 5 years was not associated with overall cancer incidence in either men or women. The absolute cancer incidence rate per 100000 person-years (standardized to the age distributions of men and women in the study using 5-year age categories) was 1858 among men with long-term daily aspirin use, 2163 among men who never reported using aspirin, 1083 among women with long-term daily aspirin use, and 1169 among women who never reported using aspirin.
Overall cancer incidence by duration of daily adult-strength aspirin use, Cancer Prevention Study II Nutrition Cohort, 1992–2003*
The association between long-term daily aspirin use and overall cancer incidence did not change substantially when we excluded participants who had ever reported use of NSAIDs other than aspirin (RR = 0.88, 95% CI = 0.79 to 0.98 among men; RR = 0.85, 95% CI = 0.69 to 1.04 among women). We found no statistically significant differences in the association between long-term daily aspirin use and overall cancer incidence in either men or women, by attained age, BMI, smoking status, or follow-up year. In analyses restricted to participants who were current smokers at enrollment, long-term daily aspirin use was not associated with reduced risk of cancer among either men (RR = 1.08, 95% CI = 0.80 to 1.45) or women (RR = 1.13, 95% CI = 0.72 to 1.76), although statistical power was limited by the low prevalence of smoking, and these results should therefore be interpreted cautiously. In analyses by follow-up interval, the rate ratio for cancer associated with long-term daily aspirin use among men was 0.89 (95% CI = 0.75 to 1.06) during the 1992–1997 interval and 0.81 (95% CI = 0.71 to 0.92) during the 1997–2003 interval. The rate ratio for cancer associated with long-term daily aspirin use among women was 0.94 (95% CI = 0.70 to 1.26) during the 1992–1997 interval and 0.79 (95% CI = 0.64 to 0.99) during the 1997–2003 interval.
In analyses of individual cancers (Table 3), long-term daily aspirin use was associated with a statistically significant reduction in risk of colorectal cancer among both sexes combined (RR = 0.68, 95% CI = 0.52 to 0.90). This association was not statistically significantly different by sex (RR = 0.76, 95% CI = 0.55 to 1.04 in men, RR = 0.45, 95% CI = 0.24 to 0.86 in women). Long-term daily aspirin use was also associated with a statistically significant reduction in risk of prostate cancer (RR = 0.81, 95% CI = 0.70 to 0.94). The risk of female breast cancer was also reduced but not statistically significantly so (RR = 0.83, 95% CI = 0.63 to 1.10). These results for prostate and breast cancers are similar to those reported in previous analyses of this cohort (22,23), although this analysis includes 2 additional years of follow-up. Shorter term daily aspirin use (use for <5 years) was not associated with incidence of any cancer examined. Results for each individual cancer were not substantially changed when we excluded follow-up time occurring after a diagnosis of a different type of cancer (data not shown).
Incidence of individual common cancers by duration of daily adult-strength aspirin use, Cancer Prevention Study II Nutrition Cohort, 1992–2003*
In this study, daily use of adult-strength aspirin for 5 or more years was associated with approximately 15% less overall cancer incidence, as compared with no use, in both men and women, although among women, this association was not statistically significant. The lower overall cancer incidence associated with long-term daily aspirin use was primarily the result of approximately 30% less colorectal cancer incidence, approximately 20% lower prostate cancer incidence, and approximately 15% lower incidence of female breast cancer, although the reduction in female breast cancer was not statistically significant.
Comparing our results with those of previous studies is difficult because we specifically examined aspirin use that was both daily and long term, whereas in nearly all previous studies, aspirin use was defined as including less frequent and/or short-term usage. In the NHANES-I cohort, any aspirin use during the 30 days before enrollment was associated with a statistically significant 17% reduction in overall cancer incidence, but duration of aspirin use was not examined (12). In a meta-analysis of observational studies of colon cancer, daily aspirin use was associated with an approximately 50% reduction in risk (9). Meta-analyses that included a large number of observational studies have reported aspirin use to be associated with a 10% reduction in risk of prostate cancer (7) and a 23% reduction in risk of breast cancer (6). Results from the meta-analyses of breast and prostate cancers did not consider duration of use and were based on combining risk estimates from individual studies that used different measures of aspirin use (generally ranging from at least once a week to daily use). Given that we specifically examined daily aspirin use for at least 5 years, we might have expected to observe larger reductions in cancer risk than those reported by the meta-analyses. Instead, our results suggest that even long-term daily aspirin use is likely to be associated with relatively small, although still potentially clinically relevant, reductions in cancer risk.
We found no association between long-term daily aspirin use and risk of lung cancer or other individual cancers examined (bladder cancer, melanoma, leukemia, non-Hodgkin lymphoma, pancreatic cancer, and kidney cancer). Our null results with respect to lung cancer differ from those of a recent meta-analysis that reported “regular” aspirin use to be associated with a 27% reduction in risk (8). However, results of this meta-analysis were influenced by substantial reductions in risk observed in two hospital-based case–control studies (26,27), whereas results from prospective studies have been inconsistent (8,28). Most previous studies of the other individual cancers we examined produced null results, although statistically significant associations with aspirin use have been observed in at least one study of each of these cancers (9). There were limited numbers of these less common cancers in our study; therefore, moderate-sized associations with long-term aspirin use remain plausible.
A limitation of our study, and of all observational studies of aspirin use, is that we cannot rule out confounding by unmeasured factors associated with both aspirin use and cancer risk. However, adjustment for measured risk factors strengthened, rather than attenuated, our results. We did not have information on the reason for daily aspirin use. However, it is likely that many daily users were taking aspirin for cardiovascular disease prevention, given that the prevalence of cardiovascular risk factors was high among daily aspirin users and that most daily aspirin users were taking only one aspirin per day. An additional limitation is that we did not have sufficient statistical power to examine long-term daily use of low-dose aspirin.
Strengths of this study include its prospective design, large size, and the availability of detailed information on aspirin use updated at several different time points. Because of these strengths, we could comprehensively examine the association between cancer risk and a relatively well-defined, and potentially clinically usable, regimen of aspirin use (long-term daily use of adult-strength aspirin) that has not previously been studied with respect to cancer.
The results in our study with respect to overall cancer incidence were strongly influenced by the three most common cancers in this cohort, prostate cancer, breast cancer, and colorectal cancer. Therefore, our results with respect to overall cancer incidence cannot be generalized to populations in which these cancers do not account for a substantial proportion of overall cancer incidence.
Our results do not have immediate clinical implications. Confirmation from randomized trials is necessary before a reduction in cancer risk could be considered a benefit of using adult-strength aspirin. Our results indicate that a randomized trial examining the effect of aspirin on cancer incidence would need to be both large and long term, probably lasting a minimum of 10 years. More evidence is needed before any such trial can be justified. This evidence could come from additional large observational studies with detailed and prospectively collected information on dose, frequency, and duration of aspirin use. In addition, if appropriate biomarkers of breast and prostate carcinogenesis could be identified, the effect of aspirin on these biomarkers could be tested in relatively short-term randomized trials, potentially providing further evidence to justify a long-term randomized trial examining cancer incidence. Further research to clarify the potential influence of aspirin on cancer risk will be challenging. However, if daily adult-strength aspirin use is ultimately found to meaningfully reduce overall cancer risk, there could be important clinical implications with respect to who should be taking aspirin and at what dose.