The usefulness of urine tests for glucose was compared with that of blood sugar determinations in detecting diabetes mellitus in 2000 medical outpatients. Eighty-five patients proved to be diabetic, but 33 of these had no glycosuria on their first visit and would not have been detected by laboratory tests had the blood glucose levels not been measured. Exactly one-half of the new diabetics discovered would have been missed (15 of 30) had only urine tests been performed. The new diabetics who were aglycosuric at their initial examination had a mild form of the disease, were predominantly elderly, and were controlled, in nearly every case, by diet alone. Most of these patients had evidence of an elevated renal threshold for glucose. These results indicate that there is a large group of patients with mild, asymptomatic, diabetes mellitus who remain undetected unless blood tests are employed routinely. Because they have little or no glycosuria these patients do not have polydipsia, polyphagia, polyuria, or loss of weight. Automation in the medical laboratory has made routine blood tests for glucose feasible and rewarding in detecting diabetes mellitus.