Higher blood levels of beta-carotene have been found to be associated with reduced risk of lung cancer, but large intervention trials have failed to demonstrate reduced lung cancer incidence after prolonged high-dose beta-carotene supplementation. Data on blood levels of specific carotenoids other than beta-carotene in relation to lung cancer are scarce. Little is known about the relationship between prediagnostic serum levels of carotenoids, retinol, and tocopherols, and risk of lung cancer especially in non-Western populations. Between January 1986 and September 1989, 18,244 men ages 45-64 years participated in a prospective study of diet and cancer in Shanghai, China. Information on tobacco smoking and other lifestyle factors was obtained through in-person interviews. A serum sample was collected from each study participant at baseline. During the first 12 years of follow-up, 209 lung cancer cases, excluding those diagnosed within 2 years of enrollment, were identified. For each cancer case, three cancer-free control subjects were randomly selected from the cohort and matched to the index case by age (within 2 years), month and year of blood sample collection, and neighborhood of residence. Serum concentrations of retinol, alpha- and gamma-tocopherols, and specific carotenoids including alpha-carotene, beta-carotene, beta-cryptoxanthin, lycopene, and lutein/zeaxanthin were determined on the 209 cases and 622 matched controls by high-performance liquid chromatography methods. A high prediagnostic serum level of beta-cryptoxanthin was significantly associated with reduced risk of lung cancer; relative to the lowest quartile, the smoking-adjusted relative risks (95% confidence intervals) for the 2nd, 3rd, and 4th quartile categories were 0.72 (0.41-1.26), 0.42 (0.21-0.84), and 0.45 (0.22-0.92), respectively (P for trend = 0.02). Increased serum levels of other specific carotenoids including alpha-carotene, beta-carotene, lycopene, and lutein/zeaxanthin were related to reduced risk of lung cancer although the inverse associations were no longer statistically significant after adjustment for smoking. A statistically significant 37% reduction in risk of lung cancer was noted in smokers with above versus below median level of total carotenoids. Serum retinol levels showed a threshold effect on lung cancer risk. Compared with the lowest quartile (<40 microg/dl), the smoking-adjusted relative risk (95% confidence interval) was 0.60 (0.39-0.92) for men in the 2nd-4th quartiles of retinol values combined; no additional decrease in risk was observed between individuals from the 2nd to 4th quartiles. There were no associations between prediagnostic serum levels of alpha- and gamma-tocopherols and lung cancer (all Ps for trend > or =0.4). The present data indicate that higher prediagnostic serum levels of total carotenoids and beta-cryptoxanthin were associated with lower smoking-related lung cancer risk in middle-aged and older men in Shanghai, China. Low level of serum retinol (with a threshold effect) is associated with increased lung cancer risk in this oriental population.