We investigated whether the presence of (+)-anti-benzo(a)pyrene diolepoxide adducts to serum albumin (BPDE-SA) among workers exposed to benzo(a)pyrene (BaP) and unexposed reference controls was influenced by genetic polymorphisms of cytochrome P4501A1 (CYP1A1), microsomal epoxide hydrolase (EHPX), glutathione S-transferases M1 (GSTM1) and P1 (GSTP1), all involved in BaP metabolism. Exposed workers had significantly higher levels of adducts (0.124 ± 0.02 fmol BPTmg(-1) SA, mean ± SE) and a higher proportion of detectable adducts (40.3%) than controls (0.051 ± 0.01 fmol BPT mg(-1) SA; 16.1%) (p = 0:014 and p = 0:012). Smoking increased adduct levels only in occupationally exposed workers with the GSTM1 deletion (GSTM1 null) (p = 0:034). Smokers from the exposed group had higher adduct levels when they were CYP1A1 *1/*1 wild-type rather than heterozygous and homozygous for the variant alleles (CYP1A1 *1/*2 plus *2/*2) (p = 0:01). The dependence of BPDE-SA adduct levels and frequency on the CYP1A1 *1/*1 genotype was most pronounced in GSTM1-deficient smokers. Exposed workers with GSTM1 null/GSTP1 variant alleles had fewer detectable adducts than those with the GSTM1 null/GSTP1*A wild-type allele, supporting for the first time the recent in vitro finding that GSTP1 variants may be more effective in the detoxification of BPDE than the wild-type allele. Logistic regression analysis indicated that occupational exposure, wild-type CYP1A1*1/*1 allele and the combination of GSTM1 null genotype+EHPX genotypes associated with predicted low enzyme activity were significant predictors of BPDE-SA adducts. Though our findings should be viewed with caution because of the relatively limited size of the population analysed, the interaction between these polymorphic enzymes and BPDE-SA adducts seems to be specific for high exposure and might have an impact on the quantitative risk estimates for exposure to polycyclic aromatic hydrocarbons.