A case-control study in the Côte d'Or area of France used the multistep concept of colorectal carcinogenesis to compare lifetime tobacco consumption and present alcohol consumption in patients with small adenomas (less than 1 cm, n = 154) or large adenomas (n = 208) and in polyp-free controls (n = 427). Cancer patients (n = 171) were compared with population controls (n = 309). In men, smoking was associated with the risk of adenomas (odds ratio = 3.6 over 20 pack-years vs. nonsmokers, p < 0.001). Alcohol was a risk factor for large adenomas only, with relative risks of 4.2 (p < 0.01), 3.0 (p < 0.05), and 4.4 (p < 0.01) for consumptions of 20-39, 40-59, and 60 g/day compared with less than 10 g/day. When patients with large adenomas were compared with polyp-free controls, both alcohol and tobacco were independently related to the risk of tumor. There was no association between tobacco or alcohol intakes and cancer risk. In women, consumption was much lower in all groups, and no significant association with either risk factor was observed. These data suggest for the first time that there is an independent effect of alcohol and tobacco in men at different early steps of the adenoma-carcinoma sequence. They demonstrate the usefulness of such a model for etiologic studies on cancer.