Urinary benzene (UB) was investigated as a biomarker of exposure among benzene-exposed workers and unexposed subjects in Shanghai, China. Measurements were performed via headspace solid phase microextraction of 0.5 ml of urine specimens followed by gas chromatography-mass spectrometry. This assay is simple and more sensitive than other methods (detection limit 0.016 microg benzene/l urine). The median daily benzene exposure was 31 p.p.m. (range 1.65-329 p.p.m.). When subjects were divided into controls (n = 41), those exposed to < or =31 p.p.m. benzene (n = 22) and >31 p.p.m. benzene (n = 20), the median UB levels were 0.069, 4.95 and 46.1 microg/l, respectively (Spearman r = 0.879, P < 0.0001). A linear relationship was observed between the logarithm of UB and the logarithm of benzene exposure in exposed subjects according to the following equation: ln(UB, microg/l) = 0.196 + 0.709 ln (exposure, p.p.m.) (r = 0.717, P < 0.0001). Considering all subjects, linear relationships were also observed between the logarithm of UB and the corresponding logarithms of four urinary metabolites of benzene, namely t,t-muconic acid (r = 0.938, P < 0.0001), phenol (r = 0.826, P < 0.0001), catechol (r = 0.812, P < 0.0001) and hydroquinone (r = 0.898, P: < 0.0001). Ratios of individual metabolite levels to total metabolites versus UB provide evidence of competitive inhibition of CYP450 enzymes leading to increased production of phenol and catechol at the expense of hydroquinone and muconic acid. Among control subjects UB was readily detected with a mean level of 0.145 microg/l (range 0.027-2.06 microg/l), compared with 5.63 microg/l (range 0.837-26.38 microg/l) in workers exposed to benzene below 10 p.p.m. (P < 0.0001). This suggests that UB is a good biomarker for exposure to low levels of benzene.