In diabetic patients, measurement of glycohemoglobin (HbA1C) is widely accepted as the standard method for determining long-term glycemic control. Another test, the serum fructosamine test, has been suggested as a less costly alternative. To compare these two tests, we evaluated how well each correlated with fasting blood glucose and how well each could predict the degree of glycemic control perceived by treating physicians. Among 222 diabetic subjects, fructosamine (r=0.74) and fructosamine corrected for serum albumin (c-fructosamine) (r=0.79) correlated better with fasting blood glucose than did HbA1C (r=0.68) (P<0.05). Among 450 diabetic subjects, fructosamine, c-fructosamine, and HbA1C showed similar error rates (23%-26%) when discriminating between subjects who had either poor vs not-poor control or poor-to-fair versus good-to-excellent control. However, receiver operating characteristic curves for these tests indicated that HbA1C was the best discriminator because it showed a 9% to 10% greater area under the curve (P<0.05).