The decrease with age of the adrenal-secreted dehydroepiandrosterone sulfate (DHEAS) in serum has suggested that it may be causally related to longevity. For the PAQUID [People (Personnes) Aged (Agées) About What (Quid, in Latin)] cohort of elderly subjects, we have previously reported higher DHEAS in men than in women, a decrease with age and, among men, a negative correlation between the DHEAS level and mortality at 2 and 4 years. Here, with an 8-year followup in 290 subjects, we show a global decrease of 2.3% per year for men and 3.9% per year for women. However, in approximately 30% of cases, there was an increase of DHEAS. We observed no relationship between the evolution of DHEAS level and functional, psychological, and mental status, possibly because of selection by death. In women, no association was found between mortality and DHEAS level. In men, the relative risk (RR) of death was higher for the lowest levels of DHEAS (RR = 1.9, P = 0.007), with RR = 6.5, P = 0.003 for those under 70 years old, a result indicating heterogeneity of the population. There was an effect of subjective health on mortality that disappeared after adjustment of DHEAS levels, suggesting its relation with these DHEAS levels. Death RR was much higher in smokers with a low DHEAS level than in nonsmokers with high DHEAS (RR = 6.7, P = 0.001). We submit that the involvement of DHEAS is possibly different according to gender, that association between low DHEAS level and mortality only for men under 70 years old possibly reflects heterogeneity of the population, and that DHEAS level is a reliable predictor of death in male smokers.