The aim of this study was to describe the association between glucose tolerance status and mortality in an elderly (> 70 years) Finnish community-living population (n = 379, of whom 141 were men). The baseline examination in 1991-1992 consisted of a postal questionnaire, a physical examination, and a 2-h oral glucose tolerance test. The follow-up was continued until death or 31 May 1996. The deaths were ascertained from official death certificates; 31% (n = 44) of the men and 18% (n = 40) of the women died within 4 years. The cumulative 4-year mortalities were 42% in previously diagnosed diabetic men, 25% in previously undiagnosed diabetic men, 27% in men with impaired glucose tolerance (IGT), and 31% in men with normal glucose tolerance (NGT). The corresponding figures for women were: 26%, 29%, 19%, and 8%, respectively. After adjusting for age, gender, smoking, body mass index and exercise by the Cox proportional hazards model in the total population, the estimated relative mortality rate was 2.2 (95% confidence interval (CI) 1.2 to 4.1) in previously diagnosed diabetic subjects, 1.7 (95% CI 0.8 to 3.7) in subjects with previously undiagnosed diabetes, and 1.2 (95% CI 0.7 to 2.1) in the IGT group, respectively, all in comparison to the NGT subjects. In separate modelling by gender these relative mortalities were higher in women than in men, but the gender difference could well be explained by chance variation. In conclusion, the results suggest that previously diagnosed diabetes is associated with increased mortality among the elderly and, particularly among elderly women, undiagnosed diabetes and even IGT may be associated with increased mortality.