Aspirin Use and the Risk for Colorectal Cancer and Adenoma in Male Health Professionals

Methods

Study Population

We analyzed data from an ongoing cohort of men, the Health Professionals Follow-up Study. This cohort was initiated in 1986 to study various potential causes of cardiovascular disease and cancer, particularly diet [26]. At baseline, 51 529 U.S. male dentists (58%), optometrists (7%), osteopaths (4%), podiatrists (3%), pharmacists (8%), and veterinarians (20%) who were 40 to 75 years of age responded to mailed questionnaires on aspirin and other NSAID use and on history of cancer and other clinically diagnosed medical conditions. In 1988, 1990, and 1992, we mailed a follow-up questionnaire to update exposure information and ascertain newly diagnosed medical conditions [27].

Current use of aspirin and other NSAIDs was assessed at baseline and at follow-up using a list that included the following categories: 1) aspirin, 2+ times per week (for example, Anacin, Bufferin, Alka-Seltzer); 2) acetaminophen, 2+ times per week (for example, Tylenol); and 3) other anti-inflammatory medications (for example, Motrin, Indocin, Naprosyn, Dolobid). Advil was used as an example rather than Dolobid on the 1988 and 1990 questionnaires. Reasons for use were not assessed at baseline. In 1993, however, a randomly selected sample of 211 participants who reported aspirin use from 1986 to 1990 were mailed a brief supplementary questionnaire. Of the 211 men, 185 (88%) responded and reported one or more of the following reasons: cardiovascular disease, 25.4%; to decrease risk for cardiovascular disease, 58.4%; headaches, 25.4%; joint or musculoskeletal pain, 33.0%; and other reasons, 7.0%. From the supplementary questionnaire, we inferred that the median duration of use during the 1990 to 1992 follow-up period was 9 years.

At baseline, we also inquired about age, current and past smoking, weight and height, family history of colorectal cancer, leisure-time physical activity, and dietary and alcohol intake using a semiquantitative food frequency questionnaire [28, 29]. On the follow-up questionnaires, we included questions about the participant's history of colonoscopy or sigmoidoscopy before and during the study. We also assessed the indications for endoscopy (bleeding in stool, positive test result for occult fecal blood, abdominal pain, diarrhea or constipation, family history of colon cancer, routine screening [no symptoms], or follow-up).

Identification of Patients with Colorectal Cancers

The follow-up procedures for this cohort have been described in detail [26]. Briefly, on each of the follow-up questionnaires we asked whether a diagnosis of cancer had been made. The response rate was 95% of total possible person-years through 31 January 1992. Most deaths in the cohort were reported by family members or the postal system in response to the follow-up questionnaires. In addition, we used the National Death Index, a highly sensitive method of identifying deaths among nonrespondents [30]. When a patient (or next of kin for decedents) reported a diagnosis of cancer of the colon or rectum on our follow-up questionnaire, we asked for permission to obtain hospital records and pathology reports. A study physician, blinded to exposure information, reviewed all medical records and extracted data on histologic type, anatomic location, and stage of the cancer. We confirmed 251 new patients with colorectal adenocarcinoma (excluding carcinoma in situ), 225 (90%) by medical records and 26 with corroborating information on diagnosis and treatment from the cohort member.

Identification of Patients and Controls with Colorectal Adenomas

Because more than 90% of the adenomas were diagnosed in men who had had an endoscopic procedure for routine screening or for unrelated gastrointestinal conditions, we restricted the adenoma analysis to 12 854 men who reported having had a colonoscopy or sigmoidoscopy between 1986 and 1992. Because aspirin-related bleeding could influence the detection of polyps, we limited our major analyses to the 10 521 men who did not have bleeding as an indication for endoscopy. Most procedures were sigmoidoscopies; thus, we studied only patients who had adenomas of the distal colon and rectum.

Of the 10 521 men, 1070 reported a diagnosis of polyp of the colon or rectum. Ninety-six percent (1028 of 1070) of the men responded to a follow-up letter; 94 of the 1028 men denied having had a polyp, 3 had died, and all but 35 of the remaining respondents granted us permission to review medical records, 96% of which were available. A diagnosis of adenomatous or hyperplastic polyp was confirmed in 756 men, and, after excluding those with hyperplastic and adenomatous polyps proximal to the descending colon, 472 patients who had distal colorectal adenomas (including carcinoma in situ) remained for analysis. Twenty-eight percent of these adenomas were 1 cm in diameter or larger.

In 1993, we surveyed a random sample of 200 men who reported a negative endoscopic result. After one mailing, 140 (70%) men granted permission to review medical records of their endoscopic procedure. No patients with histologically confirmed adenomas were discovered on review, but minute polyps were often destroyed in situ without histologic diagnosis. This practice, along with the dearth of current smokers, who are at considerable risk for small adenomas [31], accounted largely for the relatively high percentage of patients with large adenomas.

Data Analysis

Before doing the cancer and adenoma analyses, we excluded 3629 men in total, including those whose reported caloric intake was outside the range of 800 to 4200 kcal/d because most of these men probably completed the questionnaire incorrectly, those with 70 or more items left blank on the questionnaire, and those who reported previous cancer (other than nonmelanoma skin cancer), ulcerative colitis, or a familial polyposis syndrome. The remaining 47 900 men contributed follow-up time beginning at the month of return of the initial questionnaire in 1986 and ending at the month of diagnosis of colorectal cancer, month of death from other causes, or at the end of the study period, 31 January 1992. We computed incidence rates for aspirin use by dividing the number of incident cases by the number of person-years in that category and computed the relative risk by dividing this incidence rate by that of nonusers. We used the Mantel-Haenszel summary estimator and proportional hazards modeling to adjust for age (using 5-year categories) and potentially confounding variables [32]. Men diagnosed with adenomas and cancers were included only in the cancer analyses.

For the adenoma analyses, we also excluded men who had had a prestudy diagnosis of colorectal polyp. We analyzed patients with separately prevalent adenomas (polyps diagnosed at the time of the first endoscopic procedure), patients with incident adenomas (those discovered in men who had had a negative endoscopic test result before 1986), and patients with a combination of these end points. Multiple logistic regression analysis was used to control for potentially confounding variables [32].

Results

Table 1 shows characteristics of cohort members according to aspirin use in 1986. Men who reported regular aspirin use were slightly older, heavier, and less physically active; smoked slightly more; and were more likely to use multivitamin tablets. Dietary patterns were quite similar, but consumption of alcoholic beverages was slightly higher among aspirin users. Intakes of other major nutrients were similar among users and nonusers of aspirin (data not shown), except for vitamins and minerals contained in multivitamin tablets.

About one third of the men had had an endoscopic procedure before the study period. During the study period, more men who took aspirin had an endoscopic procedure for overt or occult fecal blood (5.9% compared with 4.8%) or for other indications including screening (23.2% compared with 21.8%); both were significant (P < 0.01). Among men having an endoscopy with fecal blood as an indication, an adenoma was found in 5.6% of aspirin users and in 8.1% of nonusers (percentages were age-standardized), suggesting a 31% decrease in risk among aspirin users.

We observed a lower risk for colorectal cancer among users of aspirin compared with nonusers (relative risk [RR], 0.70; 95% CI, 0.53 to 0.92; P = 0.008). The decreased risk was noted for the colon (RR, 0.72; CI, 0.53 to 0.97) and rectum (RR, 0.61; CI, 0.31 to 1.20), and this decreased risk was similar in the proximal and distal colon. As shown in Table 2, the inverse association between aspirin use and colorectal cancer became progressively stronger with evidence of more consistent use of aspirin, that is, regular use of aspirin indicated on more than one questionnaire (consecutive RRs of 0.70, 0.54, and 0.38). The lower risk for diagnosis of advanced (metastatic and fatal) colorectal cancer among aspirin users was particularly strong. The relative risks did not alter appreciably when we controlled for history of colorectal polyp, endoscopy, parental history of colorectal cancer, pack-years of smoking (0, 1 to 9, 10 to 19, 20 to 29, 30 to 39, and 40+ pack-years), body mass index (in quintiles), leisure-time physical activity (in quintiles), and intakes of red meat and vitamin E (in quintiles), and past and current alcohol use (0, 0.1 to 0.25, 0.26 to 0.50, 0.51 to 1.0, 1.1 to 2.0, and 2+ drinks per day). In addition, calcium, vitamin D, fiber, folate, total fat, and animal fat were not confounders (data not shown). The age-adjusted RR (during the 1988 to 1992 follow-up period) for men who reported aspirin use in 1988 but not in 1986 compared with nonusers was 0.66 (CI, 0.41 to 1.07). Among men who reported aspirin use in 1986 but reported no use in 1988 (or use in 1988 was unknown), the age-adjusted RR was 0.84 (CI, 0.56 to 1.25). We do not know why these men stopped taking aspirin.

Among patients initially diagnosed after the start of follow-up in 1986, 53 patients who had fatal colorectal cancer had been diagnosed by 1992. Among men who used aspirin in 1986, the RR for fatal colorectal cancer was 0.55 (CI, 0.29 to 1.05; P = 0.07) when compared with nonusers. Among 22 patients who died in those initially diagnosed after 1988, men who used aspirin in 1986 and 1988 had a RR of 0.10 (CI, 0.01 to 0.78; P = 0.03). In the 1990 to 1992 follow-up, 11 new patients died among 16 162 nonusers of aspirin (1986, 1988, and 1990) compared with no deaths among 6024 men who reported using aspirin regularly on all three questionnaires (P < 0.01). The RR for total NSAID use including aspirin (1986) was 0.70 (CI, 0.54 to 0.91; P = 0.01) for all patients and was 0.48 (CI, 0.30 to 0.75; P = 0.001) for patients with advanced disease. Too few men were taking nonaspirin NSAIDs consistently to provide stable individual estimates.

We examined the possibility that men with undiagnosed cancers or polyps might have had bleeding, which would have led them to avoid aspirin use and thus artifactually create an inverse association between aspirin use and cancer. Men using aspirin regularly in 1986, regardless of their subsequent use, had a decreasing risk for colorectal cancer over follow-up time to 1992 (Table 3). Thus, any discontinuation of aspirin use because of preclinical disease for 6 or more years before the diagnosis is unlikely to account for the association between aspirin use and colorectal cancer.

We used age-specific incidence rates of colorectal cancer based on data from the Surveillance, Epidemiology, and End-Result Program (SEER) [33] to calculate the expected number of patients (given the SEER or national rates) in our cohort. The observed number of patients in our cohort was then divided by the expected number of patients (based on SEER rates) to calculate a standardized rate ratio [34]. The total cohort was at lower risk for colorectal cancer compared with the SEER rates (251 observed patients compared with 358 expected patients; standardized incidence ratio, 0.70). The standardized rate ratio was 0.78 (184 observed of 236 expected) for nonaspirin users in 1986, and the standardized rate ratio was 0.55 for aspirin users (67 observed of 123 expected). Users of aspirin in 1986, 1988, and 1990 had an even lower standardized rate ratio for colorectal cancer in the 1990 to 1992 follow-up (standardized rate ratio, 0.33; 8 observed of 24 expected), whereas the rate among nonusers was close to the SEER rate (standardized rate ratio, 0.90; 42 observed of 47 expected).

Excluding men who had endoscopy for overt or occult fecal blood, participants who took aspirin were at a decreased risk for being diagnosed with an adenoma (Table 4). The decrease in risk among aspirin users was stronger for incident cases, which would tend to reflect polyps of more recent onset, than for prevalent cases. Among men who reported aspirin use in 1986 and 1988, the multivariate RR (1989 to 1992) was 0.72 (CI, 0.54 to 0.97) when compared with nonusers. The percentage of total adenomas removed that were 1 cm or larger in diameter was 27.6% among aspirin users and 28.1% among nonusers.

Discussion

Our findings support the hypothesis that long-term use of aspirin may substantially decrease the incidence of colorectal cancer. The finding of a particularly strong inverse association between aspirin use and metastatic cancer suggests that aspirin-related bleeding could further decrease mortality by leading to earlier diagnosis and treatment of cancers. However, the results are also consistent with an influence of aspirin during one or more stages of colorectal carcinogenesis; that is, aspirin use may prevent adenoma occurrence or progression and may decrease the probability that an existing malignancy would progress and metastasize. Based on previous reports [8-19], the effect of aspirin on colorectal cancer is likely to be similar for women. Too few men were long-term users of NSAIDs other than aspirin to provide useful information.

Several features distinguish our study from previous reports: 1) Data were available on endoscopies done and colorectal polyps diagnosed before and during the study period; 2) many potentially confounding factors, including dietary factors, alcohol intake, body mass, physical activity, smoking, family history of cancer, and screening behavior could be controlled; and 3) use of aspirin and other NSAIDs was assessed at various times throughout the study. The relatively homogeneous nature of our sample of patients may decrease the likelihood of any residual confounding by a factor we may have failed to consider, but this possibility may limit the generalizability of our results to other populations.

The adenoma analyses were prone to detection bias because most of these tumors are asymptomatic and only a portion of the population was examined. By restricting the adenoma analysis to men who had had endoscopy for reasons other than bleeding, essentially the only way a polyp exhibits itself, we decreased the likelihood of detection bias. Surgical ablation of very small polyps without retrieval would lead to an underestimation of the number of small adenomas; however, this is unlikely to bias the odds ratio. We observed an inverse association between aspirin use and risk for adenoma even after controlling for indications for endoscopy. Most importantly, because fewer adenomas (including those >1 cm) were removed from aspirin users, the inverse association between aspirin use and risk for colorectal cancer was probably not caused solely by aspirin-induced gastrointestinal bleeding leading to the earlier diagnosis and removal of adenomas.

Having three separate assessments of aspirin use allowed more refined comparisons of long-term consistent users and nonusers. Of respondents to the 1988 questionnaire, 23.8% of the men reported regular use on the 1986 and 1988 questionnaires, 5.7% reported use in 1986 only, and 16.9% reported use in 1988 only. The decrease in risk for colorectal cancer was markedly stronger among men who reported regular use of aspirin in two or three consecutive time periods. Using a similar approach, Greenberg and colleagues [15] found a decreased risk for recurrent adenomas among individual patients who reported aspirin use on two consecutive questionnaires (RR, 0.52) but not among those who reported use on one questionnaire only (RR, 0.95).

The study was limited by the absence of data on duration of NSAIDs use. However, supplementary data indicated that a substantial proportion of consistent aspirin users (throughout 1986 to 1990) had been taking aspirin for at least 10 years by 1990. These men were at a markedly lower risk for total colorectal cancer during the 1990 to 1992 follow-up period (RR, 0.35; CI, 0.16 to 0.75). The only two other studies with details on duration showed similar relative risks associated with 10+ years of use for colon cancer incidence (RR, 0.3; CI, 0.1 to 0.7) [12] and mortality (RR, 0.35 for colon cancer and 0.39 for rectal cancer for aspirin use 16+ times per month) [18].

Beyond regular use of at least 2 days per week, we had no additional data on dosage of aspirin at baseline. However, based on the supplementary questionnaire, more than 92% of the participants reported taking aspirin on at least 3 days per week and 51% reported daily intake. On days when aspirin was used, the daily mean number of pills was 2.4. In the study by Rosenberg and colleagues [12], regular NSAID use was defined as at least 3 days per week, and virtually all were daily users. In the study by Thun and colleagues [18], there was evidence of a dose-response relation by frequency of aspirin use; rather surprisingly, even participants using aspirin less than once per month were at lower risk. Among patients with arthritis, who appear to be at lower risk for colorectal cancer, NSAID use tends to be relatively high.

Although most of the epidemiologic evidence supports a protective effect of aspirin on incidence of colorectal cancer [8-19], one notable exception is from a prospective cohort study of elderly residents from California [20]. One possible contributing factor was that the median age of 73 years in the California study [20] was considerably higher than that in the others that had available data (mean age, our study, 57 years; Kune and colleagues [11], 65 years, Thun and colleagues [18], 57 years, and Schreinemachers and Everson [19], 49 years).

The most important evidence from a randomized study comes from a trial of U.S. physicians randomly assigned to take an aspirin pill (325 mg) every other day for 4.7 years. Those assigned to receive aspirin were not at a substantially lower risk for developing colorectal cancer or hyperplastic and adenomatous polyps combined [21]. However, the authors acknowledged potential limitations of the study, including the relatively low dosage, short duration of treatment, and small number of cases. In fact, a marginally significant trend (P = 0.09) of decreasing risk for colorectal cancer over time existed among aspirin users, and the confidence intervals during years 5 to 6 of follow-up overlap with those of studies showing an inverse association.

Our findings support the need for further research to determine the requisite aspirin dosage and duration of use. Although randomized studies of recurrent nonfamilial adenomas would provide important data, these would not provide direct information about dosage and duration on cancer risk. Future clinical trials of cancer may be hindered by ethical concerns, given the likelihood that a beneficial effect of aspirin on cardiovascular disease [34] would occur long before an effect on colorectal cancer. Before general recommendations can be made, we need to consider adverse effects of aspirin use, the possibility that an optimum dose for preventing colorectal cancer may not coincide with that for preventing cardiovascular disease, the role of alternative approaches to decrease colorectal cancer mortality, including sigmoidoscopic screening, dietary and other lifestyle changes, and treatment. Nonetheless, the potential role of aspirin use cannot be ignored, given the enormity of the problem of colorectal cancer.