OBJECTIVE To describe risk factors associated with microalbuminuria (MA) in subjects with diabetes, investigate the predictive value of MA as a marker of risk for diabetic nephropathy (DN), and define risk factors associated with the development and progression of MA. METHODS We conducted a prospective longitudinal study of 23 diabetic subjects with persistent MA and 209 diabetic subjects without MA who attended diabetes clinics at the University of Michigan Medical Center in 1989 and 1990. Both groups were examined at baseline and after 7 years. At baseline, urinary albumin-to-creatinine ratios were studied in random, first morning, and 24-h urine samples. At follow-up, a 12-h overnight urine sample was collected and analyzed for albumin and creatinine. At baseline, MA was defined by at least two separate urine specimens with albumin-to-creatinine ratios between 30 and 299 microg albumin per milligram of creatinine. RESULTS MA regressed in 56% of subjects with baseline MA without systematic application of corrective measures and developed in 16% of subjects without baseline MA. The predictive value positive of MA as a marker of risk for DN was 43%, and the predictive value negative was 77%. In the combined cohort, the incidence and progression of MA were significantly associated with poor glycemic control and duration of diabetes between 10 and 14 years. CONCLUSIONS MA may not be as sensitive and specific a predictor of DN as previously suggested. Other markers of risk for DN are needed for optimal clinical management.