We reported previously that subjects homozygous for the cytochrome P450 2A6 (CYP2A6) (*)4 have a lower risk of lung cancer. The purpose of this study was to clarify whether or not the alterations of smoking behavior and risk for lung cancer could be found in subjects possessing novel CYP2A6 variants discovered recently. An epidemiological study was performed with 1094 cases and 611 controls in male Japanese smokers. It was found that the amounts of daily cigarette consumption in subjects who harbored CYP2A6(*)4/(*)7, (*)4/(*)10, (*)7/(*)7, (*)7/(*)9 and (*)4/(*)4 genotypes were significantly less than those in subjects carrying the (*)1/(*)1 genotype (P < 0.01). Even after adjustment with cigarette consumption, the adjusted odds ratios (ORs) for lung cancer were significantly lower in subjects who harbored CYP2A6(*)1/(*)4, (*)1/(*)7, (*)1/(*)9, (*)1/(*)10, (*)4/(*)4, (*)4/(*)7, (*)4/(*)9, (*)7/(*)7 and (*)7/(*)9 genotypes than those who possessed the (*)1/(*)1 genotype (P < 0.05). When participants were classified into four groups according to the CYP2A6 genotypes, group 1 ((*)1/(*)1), group 2 (heterozygotes for the (*)1 and a variant allele), group 3 (heterozygotes and homozygotes for variant alleles except for (*)4/(*)4) and group 4 ((*)4/(*)4), lung cancer risk was found to be less in subjects with the variant of CYP2A6 alleles [group 2, OR of 0.59 [95% confidence interval (CI), 0.44-0.79]; group 3, OR of 0.52 (95% CI, 0.37-0.72); group 4, OR of 0.30 (95% CI, 0.16-0.57)]. The reduced risk for lung cancer was seen more clearly in heavy smokers than in light smokers. Additional stratification analysis showed that the ORs for squamous cell carcinoma (OR of 0.07) and small cell carcinoma (OR of 0.10) were lower than that of adenocarcinoma (OR of 0.39) in group 4. These results suggest that the CYP2A6 is one of the principal determinants affecting not only smoking behavior but also susceptibility to tobacco-related lung cancer.