In 1981-1982 urinary albumin excretion rates were determined in 211 diabetic and 216 non-diabetic subjects aged 60-74 years. By April 1992 122 diabetic and 58 non-diabetic probands had died. Dividing the two study populations at an albumin excretion rate of 15 micrograms/min showed that 69.3% of diabetic subjects with values at or above the limit, and 49.9% of those with values below (log rank test p = 0.0082) had died. The corresponding values for non-diabetic subjects were 44.4% and 21.0%, respectively (log rank test p = 0.0002). In single factor log rank tests ischaemic heart disease and a low value of HDL were also predictive of death in the diabetic population during a 10-11-year observation period. In the non-diabetic population ischaemic heart disease, hypertension, and a serum creatinine level above the median value were predictive. In further log rank analyses probands dying during the first years, (e.g. the first 2 years) were removed from the calculations. The prognostic value of the above-mentioned factors diminished with time. In a Cox Regression analysis we found that the predictive value of urinary albumin excretion rate to mortality had disappeared when subjects who had died during the first 5 years were removed from the analysis, whereas HDL in the diabetic patients and blood pressure and serum creatinine in non-diabetic subjects were still of significant predictive value. We therefore conclude that urinary albumin excretion rate is a more short-term predictor of mortality than previously thought, in contrast to HDL, hypertension and serum creatinine.