Whether pregnant women with insulin-dependent diabetes mellitus have an increased risk of spontaneous abortion is controversial. To address this question, we enrolled 386 women with insulin-dependent diabetes and 432 women without diabetes before or within 21 days after conception and followed both groups prospectively. Sixty-two diabetic women (16.1 percent) and 70 control women (16.2 percent) had pregnancy losses (odds ratio, 0.99; 95 percent confidence interval, 0.67 to 1.46). After adjustment for known risk factors for spontaneous abortion, the rate was still not significantly higher in the diabetic group (odds ratio, 0.91; 95 percent confidence interval, 0.59 to 1.40). Nonetheless, among the diabetic women, most of whom had good metabolic control, those who had spontaneous abortions had higher fasting and postprandial glucose levels in the first trimester than those whose pregnancies continued to delivery (P = 0.01 for fasting glucose levels and P = 0.005 for postprandial levels). In the small subgroup of diabetic women with poor control, who had elevated values for glycosylated hemoglobin in the first trimester, each increase of 1 SD above the normal range was associated with an increase of 3.1 percent in the rate of pregnancy loss (95 percent confidence interval, 0.6 to 5.6). We conclude that diabetic women with good metabolic control are no more likely than nondiabetic women to lose a pregnancy, but that diabetic women with elevated blood glucose and glycosylated hemoglobin levels in the first trimester have a significantly increased risk of having a spontaneous abortion.