The analysis of the ongoing Health Professionals Follow-up Study (HPFS), focusing on the risks for colorectal cancer and adenoma during aspirin exposure, was timely and well conceived [1]. The findings provide compelling evidence that colorectal cancer developing during aspirin use may be diagnosed at a less advanced stage. Although this effect was probably due to earlier diagnosis, a direct effect on cancer cell growth has not been excluded. Further confirmation and clarification of this effect could result in a substantial public health benefit.

Because aspirin users in the HPFS may have had a lower risk for colorectal cancer at baseline, evidence of a decreased incidence of colorectal cancer or adenoma resulting from regular aspirin use is less compelling. If regular aspirin use resulted in a shortened lead time, the earlier detection of existing colorectal cancers could cause the incidence of colorectal cancer to appear increased early in exposure but decreased during latter exposure. If the period of early exposure was not available for observation, then overall incidence could appear to be reduced, even without direct effect of aspirin on colorectal cancer incidence. Earlier detection of colorectal adenomas could also reduce the baseline risk for colorectal cancer. Such biased estimates of risk in aspirin users could occur when a population defined by prevalent aspirin use is being followed, thus assuming that recipients of long-term therapy have the same risk as new users. On the basis of the random sample of men in the HPFS who reported aspirin use in 1986, 1988, and 1990. a significant number of men reporting current use actually started this use several years before the study began. This period of early use was not available for observation in the HPFS, which raises the question of comparable baseline risks.

The observation in the HPFS that aspirin exposure was associated with earlier diagnosis of colorectal cancer suggests a shortened lead time with aspirin exposure. In addition, previous endoscopy was reported slightly more often at baseline in aspirin users, suggesting more opportunity for diagnosis of cancer and adenomas before study enrollment. Although the investigators do not state the number of patients with a history of colorectal polyps or adenomas—or the number of colonoscopies (sigmoidoscopies)—they state that a history of polyps did not substantially affect the relative risk estimates.

In the HPFS, the ideal way to evaluate the comparability of baseline risks may be to stratify patients by duration of baseline aspirin use and to estimate relative risk in new users compared with long-term users. If the relative risk estimates were significantly reduced in long-term users with identical periods of observation (effect modification by duration), the comparability of baseline risks should be questioned. Because data on duration or type of aspirin use may not have been collected at baseline, stratification by baseline history of adenoma or polyp may be informative. If the relative risk estimates are much lower in patients with a history of adenoma and aspirin use at baseline, long-term aspirin users may not have had the same opportunity to develop colorectal cancer as new aspirin users and nonusers. Controlling for a history of polyp in the analysis does not imply that such modification was absent.

Although this study provides important findings on aspirin use and colorectal cancer, the findings may be even more relevant if the authors can assure readers that aspirin users and nonusers had the same baseline risk for either colorectal cancer or adenoma.