OBJECTIVE - To evaluate the role of metabolic control in the progression of diabetic retinopathy during pregnancy. RESEARCH DESIGN AND METHODS - We conducted a prospective cohort study of 155 diabetic women in the Diabetes in Early Pregnancy Study followed from the periconceptional period to 1 month postpartum. Fundus photographs were obtained shortly after conception (95% within 5 weeks of conception) and within 1 month postpartum. Glycosylated hemoglobin was measured weekly during the 1st trimester and monthly thereafter. RESULTS - In the 140 patients who did not have proliferative retinopathy at baseline, progression of retinopathy was seen in 10.3, 21.1, 18.8, and 54.8% of patients with no retinopathy, microaneuryms only, mild nonproliferative retinopathy, and moderate-to-severe nonproliferative retinopathy at baseline, respectively. Proliferative retinopathy developed in 63% with mild and 29% with moderate-to-severe baseline retinopathy. Elevated glycosylated hemoglobin at baseline and the magnitude of improvement of glucose control through week 14 were associated with a higher risk of progression of retinopathy (adjusted odds ratio for progression in those with glycohemoglobin ≥ 6 SD above the control mean versus those within 2 SD was 2.7; 95% confidence interval was 1.1.-7.2; P = 0.039). CONCLUSIONS - The risk for progression of diabetic retinopathy was increased by initial glycosylated hemoglobin elevations as low as 6 SD above the control mean. This increased risk may be due to suboptimal control itself or to the rapid improvement in metabolic control that occurred in early pregnancy. Excellent metabolic control before conception may be required to avoid this increase in risk. Those with moderate-to-severe retinopathy at conception need more careful ophthalmic monitoring, particularly if their diabetes was suboptimally controlled at conception.