Sensitivity and specificity of three methods (random urine glucose (RUG) > 25 mg. per deciliter; positive Clinitest; and random plasma glucose (RPG) > 140, > 120, and > 100 mg. per deciliter) for detecting the unknown diabetic were compared. In 1,952 screenees, RUG of 96.7 per cent was 3 to 25 mg. per deciliter (normoglucosuria), RUG of 2.0 per cent was 26 to 100 mg. per deciliter (moderate hyperglucosuria), and of 1.3 per cent was > 100 mg. per deciliter (marked hyperglucosuria). Twenty-five randomly selected from each of the first two groups and 18 from the third group had a glucose tolerance test (GTT); seven in the third group had fasting plasma glucose tests > 150 mg. per deciliter. GTTs were evaluated by United States Public Health Service (USPHS), Fajans-Conn (F-C), and summation (S) criteria, and tested screenees divided into nondiabetic (ND) and diabetic (DM). NDs with RUG > 25 mg. per deciliter were false positive screens: renal hyperglucosurics (RHG) if GTT urine glucose ≥ 80 mg. per deciliter, nonrenal hyperglucosurics (NRHG) if GTT urine glucose ≤ 25 mg. per deciliter. The prevalence of DM was 1.24 per cent (USPHS), 1.37 per cent (F-C), and 1.45 per cent (S). The prevalence of false positives (USPHS) was 2.04 per cent (1.19 per cent RHG and 0.85 per cent NRHG). In the normoglucosuria group, 100 per cent who had a GTT were true negatives. In the moderate hyperglucosuria group, 16 per cent (USPHS) were true positives, 84 per cent false positives (44 per cent RHG, 40 per cent NRHG). In the marked hyperglucosuria group, 72 per cent (USPHS) were true positives, 28 per cent false positives (24 per cent RHG and 4 per cent NRHG). By Clinitest, 23 per cent of the DMs (USPHS) were false negatives, and 1.8 per cent in the normoglucosuria group were false positives. Nine per cent of the DMs were false negatives by RPG > 140, and 4.5 per cent were false negatives by RPG > 120.
RUG > 25 mg. per deciliter was as sensitive as the GTT in detecting unknown DMs but was less specific. Random normoglucosuric screenees are unlikely to have diabetes; about 40 per cent of random hyperglucosuric screenees have diabetes.