Undetected gestational diabetes mellitus (GDM) is associated with a two- to fivefold increase in perinatal morbidity and mortality. Widespread screening of the obstetric population (resulting in identification and treatment) should reduce these rates. Seven hundred ninety-eight women were examined during a 13-mo period of universal glucose challenge testing (GCT). A total of 2.8% of the population had an abnormal oral glucose tolerance test (OGTT). Thirty percent of those with an abnormal OGTT were less than 25 yr old. The specificity of a 1-h GCT (50-g carbohydrate load) using a threshold of either 140 or 150 mg/dl was compared with that of a 2-h specimen using a threshold of 118 mg/dl to determine whether the cost of screening could be reduced. One- and 2-h specimens were obtained in 347 of these women. A 34% reduction in the number of follow-up OGTTs required would have been achieved if a 2-h specimen had been used as the index instead of a 1-h specimen (P less than .05). As a result, the (direct and indirect) cost per patient identified with GDM would have declined 23.5%--from $866 to $662. No comment concerning the actual false-negative rate of either the 1- or 2-h GCT can be made because only select women underwent an OGTT. To assess the validity of the 2-h threshold, an OGTT was performed in an additional 190 women if either the 1- or 2-h screen was abnormal. The results were confirmatory: the 2-h screen would have reduced the cost per case identified by 32% in this small group. Screening on the basis of past medical history clearly lacked sensitivity and cost efficacy in comparison with the GCT and should be abandoned as a practice.