In 1979, all the known diabetic subjects (849) were identified from a community (population 81851), of whom 717 (85%) were reviewed by a single observer. Using the NHS Central Register, follow-up was completed for 98% of subjects. After 11 years, 306 (42.7%) diabetic subjects had died, of whom 65 were insulin treated and 241 were non-insulin treated. Circulatory disease accounted for 168 (54.9%) deaths, of which 124 (73.8%) were due to ischaemic heart disease. The standardized mortality ratio (SMR) for all causes of death, based on data from England and Wales, was significantly raised for both insulin-treated and non-insulin-treated patients (1.75, 95% CI 1.35 to 2.24 and 1.32, 95% CI 1.15 to 1.50, respectively). SMRs for all cause mortality were significantly greater for diabetic subjects in the 45-64 (SMR, 1.97, 95% CI 1.34 to 2.80), 65-74 (SMR 1.59, 95% CI 1.27 to 1.97 and 75 years and over (SMR 1.26, 95% CI 1.08 to 1.45) age ranges. Using a proportional hazards model, after adjusting for age and gender, systolic blood pressure and vibration threshold were significant predictors of all cause mortality in insulin-treated subjects. For non-insulin-treated subjects, blood glucose, systolic blood pressure, glycated haemoglobin, retinopathy, proteinuria, coronary artery disease, and stroke were significant baseline predictors of mortality. No association was found for serum cholesterol, body mass index, diastolic pressure or cigarette smoking in either treatment group.