To determine how much insulin-dependent diabetes increases a woman's risk of giving birth to a malformed infant and how that risk is influenced by metabolic control, we followed 347 diabetic and 389 control women who enrolled in the study within 21 days of conception (the early-entry group) and 279 diabetic women who entered later (the late-entry group). We detected major malformations in the infants of 4.9 percent of the early-entry diabetic women, 2.1 percent of the controls, and 9.0 percent of the late-entry diabetic women. Malformation rates were significantly higher among offspring of early-entry diabetic women than among those of controls (odds ratio, 2.45; lower one-sided 95 percent confidence limit, 1.12; P = 0.027), and higher among late-entry than among early-entry diabetic women (odds ratio, 1.91; lower one-sided 95 percent confidence limit, 1.07; P = 0.032). Mean blood glucose and glycosylated hemoglobin levels during organogenesis were not significantly higher in women whose infants were malformed. Hypoglycemia (glucose, less than or equal to 50 mg per deciliter [2.8 mmol per liter]) was not significantly more common in the same group. Hyperglycemia and glycosylated hemoglobin were not correlated with malformation. The data suggest that more sensitive measures are needed to identify the teratogenic mechanisms, or that not all malformation can be prevented by good glycemic control. Despite the increased malformation rate among infants of the early-entry diabetic women, as compared with the controls, the more favorable outcome seen in the former group as compared with the late-entry group justifies the attempt to achieve good metabolic control around the time of conception.