The impact of microalbuminuria on mortality as well as other risk factors was investigated in a 10-year follow-up study of 503 predominantly non-insulin-dependent diabetic patients of whom 265 had died. Using Cox's regression analysis the prognostic influence of age, sex, age at diagnosis, known diabetes duration, blood pressure, fasting plasma glucose, relative weight, serum creatinine, retinopathy, and treatment was evaluated as well as morning urine albumin concentration (UAC) in four categories, i.e. UAC less than or equal to 15 micrograms/ml (normal), 15 micrograms/ml less than UAC less than or equal to 40 micrograms/ml, 40 micrograms/ml less than UAC less than or equal to 200 micrograms/ml and UAC greater than 200 micrograms/ml. Age, UAC, known duration, and serum creatinine were the only significant risk factors. After correction for the other three independent risk factors, the hazard ratios in the elevated UAC categories relative to the group with UAC less than or equal to 15 micrograms/ml were 1.53 (p = 0.007), 2.28 (p = 0.000002), and 1.82 (p = 0.02). The statistically significant correlations with UAC were: age (r = 0.09, p less than 0.05), duration (r = 0.14, p less than 0.01), systolic blood pressure (r = 0.12, p less than 0.01), serum creatinine (r = 0.33, p less than 0.001), and fasting plasma glucose (r = 0.12, p less than 0.01). Increased UAC was associated also with retinopathy (p = 0.01). Fifty-eight per cent of the deaths were caused by cardiovascular disease or stroke; only 3% died from uraemia. A reinvestigation including blood pressure, fasting plasma glucose, and UAC was made on 208 survivors.