OBJECTIVE: Use of Carpenter-Coustan compared with National Diabetes Data Group criteria increases the number of women diagnosed with gestational diabetes mellitus (GDM) by 30-50%, but whether treatment of this milder GDM reduces adverse outcomes is unknown. We explored the effects of the diagnostic criteria used on the benefits of GDM treatment. METHODS: This was a secondary analysis of a randomized trial for treatment of mild GDM diagnosed using Carpenter-Coustan criteria. We evaluated the effect of treatment within two mutually exclusive diagnostic groups: 1) women who met the stricter National Diabetes Data Group as well as Carpenter-Coustan criteria (National Diabetes Data Group), and 2) those diagnosed by Carpenter-Coustan but not meeting National Diabetes Data Group criteria (Carpenter-Coustan only). Maternal outcomes examined were pregnancy-induced hypertension, shoulder dystocia, maternal weight gain, and cesarean delivery. Neonatal outcomes were large for gestational age, macrosomia (greater than 4,000 g), fat mass, small for gestational age, and a composite outcome of perinatal death, birth injury, hypoglycemia, hyperbilirubinemia, and hyperinsulinemia. Analysis of variance or the Breslow-Day test, as appropriate, was used to test for the interaction between diagnostic criteria and GDM treatment on the outcomes of interest. RESULTS: Of 958 patients, 560 (58.5%) met National Diabetes Data Group criteria and 398 (41.5%) met Carpenter-Coustan only. Compared with untreated women, the direction of treatment effect did not differ by diagnostic criteria used and was consistent with the original trial. The P value for interaction between diagnostic criteria and treatment status was not significant for any outcome. CONCLUSION: The overall beneficial treatment effect on pregnancy-induced hypertension, shoulder dystocia, cesarean delivery, and macrosomia was seen in patients diagnosed by the higher National Diabetes Data Group and by the lower thresholds of the Carpenter-Coustan criteria.