The predictive value of the 1-h 50-g glucose screen for diagnosing gestational diabetes mellitus in a high-risk population

Abstract

OBJECTIVE: To determine a value, for a gestational diabetes mellitus (GDM) screening test, above which the glucose tolerance test is obviated. METHODS: A database search of patients delivered at the Medical College of Virginia Hospital (MCV) between April 1991 and April 2002 was undertaken. Subjects were screened using standard methodology: blood glucose level 1 h after a 50-g oral glucose load (1OGT). Subjects with values meeting/exceeding 140 mg/dl underwent 3-h 100-g oral glucose tolerance tests (3OGTT). GDM was diagnosed using criteria of the National Diabetes Data Group (NDDG), with Carpenter-Coustan (CC) criteria for comparison. Receiver-operator characteristic (ROC) curves were generated; areas under the curve (AUC) were calculated. RESULTS: 1OGT results were available for 16898 subjects; 2770 (16.4%) had values meeting/exceeding 140 mg/dl. The NDDG and CC criteria were applied to 1972 subjects with both 1OGT and 3OGTT results available: 419 (21%) and 614 (31%) subjects had GDM, respectively. Positive predictive values for results > or =180 mg/dl and values at 20 mg/dl increments up to 260 mg/dl were: 36, 47, 55, 57 and 63% (NDDG) and 45, 54, 62, 61 and 66% (CC). AUC for NDDG=0.68; AUC for CC=0.64. CONCLUSIONS: GDM cannot be diagnosed with the 1OGT; predictive values are low. A cut-off of 200 mg/dl predicts only 47-54% of GDM cases correctly, and may lead to over-diagnosis. It is inappropriate for GDM to be diagnosed based on the 1OGT.