In vitro mutagenic activity and carcinogenic potential of metronidazole in certain animals raised concerns about its possible carcinogenicity in humans. We studied the late incidence of cancer after metronidazole use among persons enrolled in the Group Health Cooperative of Puget Sound, Seattle, a health maintenance organization. Randomly selected nonusers were matched on a one-to-one basis for age, gender, and year of enrollment to persons who used metronidazole on an outpatient basis during the period January 1975 to December 1983; 5,222 metronidazole user/nonuser pairs, for whom the median follow-up was 12.6 years, were analyzed. Forty-nine percent, 39.2%, 9.8%, and 2% of users had 1, 2-4, 5-9, and > or = 10 prescriptions or refills of metronidazole filled, respectively. The late (after the first 7 years of follow-up) incidence of cancer was nearly identical among users and nonusers (652 and 662 per 100,000 person-years, respectively; relative risk, 0.98; 95% confidence interval, 0.80-1.20). Age-gender stratified analysis did not reveal any association between metronidazole use and cancer. These data support no association between short-term exposure to metronidazole and cancer in humans. Although the results are reassuring, they may not extend to subjects who have used metronidazole for prolonged periods; further epidemiological studies should focus on these individuals.