Exposure to gasoline vapors is classified by the International Agency for Research on Cancer as possibly carcinogenic to humans, mainly on the basis of the established carcinogenicity of some component chemicals such as benzene. The mechanism of benzene toxicity, particularly its leukemogenic effects, is far from being fully understood. Different studies, aimed at evaluating the risk associated with exposure to benzene through fuels and coordinated by the Istituto Superiore di Sanità, are in progress in Italy. In an environmental monitoring survey on a sample of 111 service stations, conducted in Rome (Italy) in 1992, average yearly personal exposure to benzene, toluene and xylenes were estimated. Chemical determination of benzene and methylbenzene was carried out by GL-gas chromatography. From a sample of 27 service stations 34 fuel samples were collected, and their benzene content was measured by hr-gas chromatography. Subgroups of the filling station attendants undergoing the exposure assessment study, were included in biological monitoring surveys of early indicators of genotoxicity. In particular, 65 subjects were enrolled in a study aimed at evaluating the urinary concentrations of 8-hydroxydeoxyguanosine (8-OHdG), a biological marker of oxidative DNA damage, and 23 filling station attendants were selected for a survey of the frequencies of sister chromatid exchanges (SCE) and micronuclei (MN) in peripheral T lymphocytes. In the exposure assessment survey levels of 0.53, 0.71 e 0.32 mg/m3 in the average yearly personal exposure to benzene, toluene and xylenes, respectively, were estimated (individual means based on 6.5 repeated samples per employee). The daily quantities of super premium gasoline sold proved to be associated with the average yearly personal exposure to benzene, and current smokers showed a significantly lower exposure intensity compared with non-smokers. Among the latter, an increase of 0.11 ln mg/m3 in benzene exposure per unit increase (100 l) in gasoline sold (p < 0.001) was estimated by a multiple regression analysis with some personal characteristics of the subjects included in the model as potentially predictive variables (R2 = 0.17, p (F) < 0.05). Among smokers, however, only the age and the length of employment were able to predict the intensity of benzene exposure. On a sample of 27 filling station attendants, furthermore, the relationship between personal exposure to benzene and benzene fuel content was evaluated and an increase of 0.01 mg/m3 in the personal benzene exposure per unit increase (100 g) in the absolute quantity of benzene in the fuel sold was estimated (p < 0.0001, R2 = 0.50).(ABSTRACT TRUNCATED AT 400 WORDS)