Random plasma glucose was determined in 276 apparently healthy pregnant women attending our antenatal clinic at 28 to 32 weeks gestation. Mean and standard deviation values within 2 h and greater than 2 h after a meal were calculated. A standard 75-g oral glucose tolerance test was then given to 250 of the 276 pregnant women. Three patients were found to be diabetic and 46 had post-load concentrations indicative of impaired glucose tolerance according to the criteria of the World Health Organization (1980). Using a cut-off point whereby 15% of the population would be tested, we would have identified only 2 of the 3 diabetics and 12 of the 46 with impaired glucose tolerance. This poor predictive power cannot be resolved by altering cut-off points for screening, or by altering the criteria for abnormal glucose tolerance. The basic problem is lack of a close relation between 2-h glucose tolerance value and random glucose when this has been taken greater than 120 min after a meal. In this population with a high prevalence of abnormal glucose tolerance, random plasma glucose is not an efficient screening test.