In this large confirmatory study of 803 patients with squamous cell carcinoma of head and neck (SCCHN) and 839 controls frequency matched by age, sex, and ethnicity, we further examined potential predictors of benzo[a]pyrene diol epoxide (BPDE)-induced adduct levels and their associations with SCCHN risk. BPDE-DNA adduct levels were determined by the (32)P-postlabeling method in peripheral lymphocytes after in vitro challenged by BPDE. We also genotyped for GSTM1 null, GSTT1 null, GSTP1 Ile(105)Val, and GSTP1 Ala(114)Val. Potential predictors of BPDE-DNA adducts were evaluated by stratification and multivariate linear regression analyses and the association between adduct levels and SCCHN risk by multivariate logistic regression analyses. We found that mean BPDE-DNA adduct levels (the relative adduct labeling x 10(7) +/- SD) were significantly higher in cases (77.6 +/- 111.8) than in controls (57.3 +/- 98.3; P < 0.001). Using the median control value (29.22) as a cutoff, 63% of the cases were distributed above this level (adjusted odds ratio, 1.71; 95% confidence interval, 1.39-2.10). A significant dose-response relationship was observed between adduct quartiles and SCCHN risk (P(trend) < 0.001). Multivariate linear regression analysis revealed that ethnicity and smoking were significant predictors of BPDE-DNA adduct levels in controls. In conclusion, we confirmed the previously reported association between in vitro BPDE-induced DNA adduct levels and SCCHN risk, and the assay may help identify individuals at high risk of developing smoking-related cancers.