To determine whether serum dehydroepiandrosterone sulfate (DHEAS) levels could predict longevity in residents. Prospective community-based cohort study. Community. Nine hundred forty subjects (396 men, 544 women; aged 21 to 88) underwent a health examination in 1978. Serum DHEAS levels were measured according to radioimmunoassay at baseline in all subjects, and subjects were followed periodically until 2005. Baseline DHEAS levels were higher in men than in women and decreased with age in both sexes. In a Cox proportional hazards model, age, DHEAS (inversely), blood pressure, and fasting plasma glucose were significantly associated with shorter longevity in men but not in women. Of these variables, high DHEAS levels in men were the strongest predictor of longevity (beta=-2.032, hazard ratio=0.131, 95% confidence interval=0.029-0.584 in the Cox proportional hazards model after adjustment for age). The Kaplan-Meier survival curve, stratified according to tertiles of DHEAS levels, in men after adjustments for age, systolic blood pressure, and fasting plasma glucose showed significantly (log-rank stat =10.6; P<.001) greater longevity in the highest group (200 microg/dL) than in the moderate (130-199 microg/dL) or lowest groups (129 microg/dL). This 27-year study in a community-based cohort indicated that DHEAS level may be a predictor of longevity in men, independent of age, blood pressure, and plasma glucose.