Sixty-two consecutive diabetic women hospitalized from one to eight weeks prior to delivery were included in this study. Some 1,100 simultaneous unconjugated plasma estriol, total plasma estriol, 24 hour urinary estriol, and creatinine assays were performed on an almost daily schedule. Clinical management was based upon a weekly oxytocin challenge test and daily 24 hour urinary estriol and creatinine determinations. Twenty patients had spontaneous onset of labor and 32 were delivered electively at 38 weeks while three were delivered for maternal and seven for fetal indications. Gestational age at delivery averaged 38 weeks and ranged from 35 to 42 weeks. A total of 840 day-to-day variations of unconjugated plasma estriol, total plasma estriol, and the urinary estriol/creatinine ratio were computed as the per cent rise or fall from the highest mean of three consecutive preceding values. Observed were 369, 419, and 428 decreases in unconjugated plasma estriol, total plasma estriol, and urinary estriol/creatinine, respectively, averaging 12.8 ± 9.6 (S.D.), 13.4 ± 10.1, and 14.4 ± 10.6 per cent. One stillbirth occurred, which was preceded by a 42 per cent decrease in unconjugated plasma estriol but unheralded by either a drop in total plasma estriol or the urinary estriol/creatinine ratio. There were fewer falls of more than 40 per cent unassociated with perinatal morbidity and death with unconjugated plasma estriol (No. = 3) than with total plasma estriol (No. = 8) and urinary estriol/creatinine (No. = 8). These data suggest that unconjugated plasma estriol is the most predictive test among presently available estriol assays for managing the pregnant diabetic patient.