The purpose of our study was to investigate the relation in between physical activity in different periods of life at work and in leisure-time and prostate cancer risk. We conducted a case-control study on prostate cancer in Italy between 1991 and 2002, which included 1,294 incident cases of histologically confirmed prostate cancer below 75 years of age and 1,451 controls, who were admitted to hospital for acute nonneoplastic conditions. Odds ratios (OR) of prostate cancer according to physical activity in different periods of life were obtained by unconditional multiple logistic regression models, including terms for age, study centre, education, social class, body mass index, energy intake, family history and other selected covariates. Compared to the lowest level of occupational physical activity, the multivariate ORs for prostate cancer for the highest level were 0.94 (95% confidence interval, CI, 0.75-1.17) at age 15-19, 0.78 (95% CI, 0.63-0.97) at age 30-39 and 0.75 (95% CI, 0.61-0.93) at age 50-59. A significant inverse trend in risk was found for activity at work at ages 30-39 and 50-59. The inverse associations were consistent in strata of age at diagnosis, body mass index, education and social class. No significant association was found for leisure-time physical activity. The inverse association between occupational physical activity and prostate cancer risk may reflect favorable hormonal correlates of physical activity, but residual confounding by socioeconomic covariates cannot be excluded.