The incidence of spontaneously occurring premature labor in insulin-dependent diabetic pregnancies is unclear, because previous studies have been confounded by a high rate of iatrogenic prematurity. The purpose of this study was to determine, in a large population of insulin-dependent diabetic pregnant women, the rate of spontaneous occurrence of premature labor and the various factors that may affect it. We hypothesized a priori that spontaneously occurring premature labor occurs at a high rate in insulin-dependent diabetic pregnant women, mainly because of poor control of diabetes during pregnancy, and is related to the presence of polyhydramnios and hypomagnesemia. One hundred forty-five insulin-dependent diabetic women undergoing 181 pregnancies were recruited since 1978 in an interdisciplinary prospective study. The goals of glucose control were a fasting blood glucose less than 100 mg/dL and a 90-minute postprandial glucose less than 140 mg/dL. The rate of spontaneous premature labor, 31.1%, was significantly higher (P less than .01) than that in a control population managed by the same obstetricians in similar clinical settings (20.2%). The following variables were not significantly associated with the onset of premature labor: maternal age, parity, gravidity, diabetic class according to White, presence of renal disease or retinopathy, previous elective abortion, chronic hypertension or pregnancy-induced hypertension, cigarette smoking, first-trimester or post-20 weeks' gestation vaginal bleeding, maternal serum magnesium concentration, or polyhydramnios.(ABSTRACT TRUNCATED AT 250 WORDS)