OBJECTIVES: Previous studies have reported controversial findings regarding the association of testosterone with mortality in older men. This heterogeneity might be partially explained by comorbidities and the presence of metabolic syndrome, as well as differential associations according to causes of death. METHODS: We used data from a random subsample of the Three-City study, in which hormone levels were measured in 338 men ≥65 years without metabolic syndrome who were followed-up for 12 years. Vital status was determined for all participants from different sources. We used inverse-probability-weighted Cox regression to estimate the hazard ratios (HRs) of cause-specific mortality and 95% confidence intervals (CIs). RESULTS: Over the follow-up period, 130 men died (30 from cardiovascular disease, 45 from cancer, 55 from other causes). The association of testosterone with mortality showed significant heterogeneity across causes of death (p=0.027 and p=0.022 for total and bioavailable testosterone, respectively). Higher testosterone levels were associated with increased cardiovascular mortality (HR for 1-standard deviation increase, 1.86; 95% CI, 1.28 to 2.71 and 1.50; 95% CI, 1.04 to 2.17 for total and bioavailable testosterone, respectively). By contrast, there were no significant associations of testosterone with mortality from cancer and other causes. CONCLUSIONS: Our data suggest that the association of testosterone with mortality in men without metabolic syndrome might be differential according to the cause of death. These findings may partially explain the heterogeneity across studies on the relationship between testosterone levels and mortality.< Réduire