Dr. Sullivan suggests that the mechanism by which long-term aspirin use may protect against colorectal cancer involves the progressive reduction of blood stores caused by aspirin-induced gastrointestinal blood loss. As mentioned in our report [1], however, anti-inflammatory sulindac causes a temporary regression of the polyps in patients with familial polyposis [2], an effect that is reversible immediately after discontinuation of sulindac therapy. Such a pattern is not consistent with a progressive loss of blood stores. Although the iron hypothesis is intriguing, some evidence suggests that fecal iron could influence the risk for colon cancer through the generation of hydroxyl radicals [3]. If this were true, any gastrointestinal bleeding induced by aspirin would be expected to increase the risk for colorectal cancer by increasing the fecal iron content. Aspirin is a potent, irreversible inhibitor of cyclo-oxygenase, and prostaglandins appear to influence tumor growth; this indicates that the effect of aspirin involves inhibition of prostaglandin synthesis [4]. Alternatively, aspirin also has been shown to inhibit phospholipase activity [5], which is important in various aspects of intracellular signaling. Numerous potential mechanisms for a causal relation between aspirin and colorectal cancer have been discussed in detail by Marnett [4], although no mechanism has been established.

Dr. Burkhart raises concerns about differential baseline risks for colorectal cancer between aspirin users and nonusers and suggests some analyses to address this issue. As presented in our report [1], the age-standardized percentage of men who had had endoscopy before follow-up was similar among aspirin users (34.3%) and nonusers (32.6%). The percentage of persons in the cohort who had reported a diagnosis of colorectal polyp before follow-up was 3.60% for nonusers and 3.85% for aspirin users. Viewed as a proportion of those who had had endoscopy, this represents 11.0% of users and 11.2% of nonusers. The relative risks of aspirin users compared with nonusers were nearly identical among those who had not been screened before follow-up (relative risk, 0.69 [95% CI, 0.49 to 0.98]) and those who had been screened (relative risk, 0.68 [CI, 0.36 to 1.27]) before follow-up. Given that more than 96% of persons in the cohort did not have a polyp diagnosed at baseline, the age-adjusted relative risk of 0.71 (CI, 0.53 to 0.95) for those without a polyp diagnosed before baseline was almost identical to the relative risk for the entire cohort (relative risk, 0.70 [CI, 0.53 to 0.92]). Among the few patients with previous diagnoses of polyps, a similar nonsignificant reduction in risk was seen (relative risk, 0.57 [CI, 0.22 to 1.48]). These additional analyses suggest that aspirin users and nonusers were at similar risk for colorectal cancer at baseline.