Nonsteroidal anti-inflammatory drugs (NSAIDs) inhibit the synthesis of prostaglandins, which play a role in cell proliferation and neoplasia. Data from experiments indicate that the administration of NSAIDs after cancer induction by chemicals inhibits carcinogenesis in the rodent large bowel, which is thought to be a good model for human large bowel carcinogenesis. We tested whether NSAIDs may have the same effect on the human large bowel in an analysis of data from a hospital-based case-control drug surveillance study: 1326 incident cases of large bowel cancer were compared with 1011 cancer controls and 3880 noncancer controls. There was a statistically significant inverse association between regular NSAID use that had lasted into the year before interview and risk of large bowel cancer (multivariate RR = 0.5, 95% CI 0.4-0.8). Regular NSAID use that had been discontinued more than one year previously was not associated with a reduction, nor was short-term or sporadic use. Virtually all NSAID use was of aspirin-containing drugs. The data are compatible with an inhibitory effect of aspirin on human large bowel carcinogenesis. Preliminary data from a large follow-up study conducted by the American Cancer Society support our findings. To test the hypothesis in a study that overcomes methodologic limitations of our first study, we propose to conduct a population-based case-control study in Massachusetts, with [1200] incident cases identified through the Massachusetts Cancer Registry and [1200] age, sex, and precinct matched population controls identified from town lists. The study is designed to collect more complete and detailed information on NSAID use, and to minimize potential sources of bias that may have been present in our first study so that noncausal explanations can be ruled out: selection bias, uncontrolled confounding, and bias from symptoms in the cases affecting their recent NSAID use. The study will be able to test, separately, the effects of aspirin-containing and nonaspirin-containing NSAIDs. The availability of the Cancer Registry and town lists in Massachusetts make it feasible to carry out this population-based study quickly and at moderate expense. Large bowel cancer is one of the most commonly occurring cancers in American men and women. If a preventive were identified, this could have a considerable impact on the public health.