An integrated approach based on ambient and biological monitoring, the latter including both biomarkers of exposure and susceptibility, was applied to characterize benzene exposure in a group of 37 taxi drivers of the city of Parma (Italy). Airborne benzene concentrations were assessed by 24 h personal sampling and work-shift sampling inside the taxicab using passive samplers (Radiello). Benzene metabolites, trans,trans-muconic acid (t,t-MA) and S-phenylmercapturic acid (S-PMA), and urinary cotinine as biomarker of smoking habits were measured by isotopic dilution liquid chromatography tandem mass spectrometry in both pre-shift (PS) and end-of-shift (EOS) samples. Urinary benzene (U-B) levels were determined by solid-phase microextraction gas chromatography-mass spectrometry in EOS samples. Relevant polymorphisms of microsomal epoxide hydrolase, NAD(P)H:quinone oxidoreductase, glutathione S-transferases M1-1 (GSTM1), T1-1, and A1 were characterized by PCR-based methods. Mean airborne benzene concentration was 5.85 +/- 1.65 microg/m3, as assessed by 24 h personal sampling integrating for work-shift, indoor or general environment activities. Significantly, higher benzene concentrations were detected in the taxicab during the work-shift (7.71 +/- 1.95 microg/m3, p < 0.005). Smokers eliminated significantly higher concentrations of U-B and S-PMA than non-smokers in EOS samples [geometric mean (geometric S.D.): 2.58 (4.23) versus 0.44 (1.79) microg/l for U-B; 3.79 (1.50) versus 2.14 (1.87) microg/gcreat. for S-PMA, p < 0.002]. Within smokers, S-PMA concentrations significantly increased at the end of the work-shift compared to pre-shift values (p < 0.05). t,t-MA showed a similar behaviour, although differences were not significant. In the narrow range examined, no correlation was observed between air benzene concentration and urinary biomarkers. All benzene biomarkers but EOS t,t-MA were correlated with U-cotinine (p < 0.05). GSTM1 polymorphism significantly modulated S-PMA excretion, as subjects bearing the GSTM1pos genotype [3.61 (1.15) microg/gcreat.] excreted significantly higher S-PMA concentrations than GSTM1null subjects [2.19 (1.18) microg/gcreat., p < 0.05].