We tested 32 routine clinical parameters for their ability to discriminate between a high risk and a low risk of non-insulin-dependent diabetes mellitus (NIDDM) within 5-7 years after pregnancies complicated by gestational diabetes mellitus (GDM). Latino women (n = 671) with GDM who did not have diabetes 4-16 weeks after delivery returned for at least one 75-g oral glucose tolerance test (OGTT) within 7.5 years. Multivariate analysis was used to identify parameters ascertained during or immediately after the index pregnancy that were independently associated with the development of diabetes during follow-up. Life table analysis revealed a 47% cumulative incidence rate of NIDDM 5 years after delivery for this cohort of patients who did not have diabetes at the initial postpartum examination. Four variables were identified as independent predictors of NIDDM: the area under the OGTT glucose curve at 4-16 weeks postpartum, the gestational age at the time of diagnosis of GDM, the area under the OGTT glucose curve during pregnancy, and the highest fasting serum glucose concentration during pregnancy. Examination of relative risks (RRs) of NIDDM between the highest and lowest quartiles of the cohort for each variable, adjusted for the other three variables, revealed that the postpartum OGTT provided the best discrimination between high-risk and low-risk individuals (adjusted RR = 11.5 [95% confidence interval 4.5-29.1] compared with adjusted RRs of only 0.5-2.5 for the other three variables).(ABSTRACT TRUNCATED AT 250 WORDS)