P>Objective To examine the association of serum hormone levels with all-cause mortality in older community-dwelling men. Design Single centre cohort study. Subjects Men aged 50 and older, insured by Societe de Secours Miniere de Bourgogne (Montceau les Mines, France). Among 3400 men invited to participate, 782 volunteers had serum hormone measurements and were followed up for 10 years. No exclusion criteria were used. Results Nonsurvivors (n = 182) were older, had more comorbidities and lower physical performance. The lowest quartile of 25-hydroxycholecalciferol (25OHD) level predicted mortality [HR = 1 center dot 44, 95% confidence interval (CI): 1 center dot 03-2 center dot 03, P < 0 center dot 05] regardless of age, BMI, smoking, physical activity, vitamin D supplementation, and health status; mainly for the first 3 years. The 17 beta-E-2 level predicted mortality independent of confounders after the third year (HR = 1 center dot 21 per 1 SD increase, 95% CI: 1 center dot 09-1 center dot 35, P < 0 center dot 001). In the fully adjusted models, risk of death increased per quartiles of 17 beta-E-2 (trend -P < 0 center dot 001) and was higher in the third and the fourth quartiles compared with the lowest quartile (HR = 1 center dot 80, 95% CI: 1 center dot 09-2 center dot 98, P < 0 center dot 05 and HR = 2 center dot 83, 95% CI: 1 center dot 71-4 center dot 67, P < 0 center dot 001). Concentrations of testosterone and PTH did not predict mortality independent of the model. Conclusions In older men, increased 17 beta-E-2 level predicted mortality after 3 years of follow-up. Thus, high 17 beta-E-2 level may reflect presence of risk factors precipitating development of diseases. Low 25OHD level predicted mortality more weakly, mainly for the first 3 years of the follow-up, and was strongly influenced by the confounding variables. Thus, low 25OHD level may reflect poor current health status and unhealthy lifestyle.