The value of prenatal care has been obscured by multiple factors, including the limitations of birth certificate data, large socioeconomic disparities between women who seek prenatal care and those who do not, and the "preterm delivery bias", ie, the reduced pregnancy duration and opportunity for prenatal care among women who give birth prematurely. Perinatal mortality and morbidity (neonatal intensive care unit admission; ventilator therapy) were carefully assessed in an indigent population (28,838 deliveries at Parkland Memorial Hospital). To avoid the preterm delivery bias, a cohort of all women whose pregnancy reached a specific week of gestation was identified and their prenatal care status (zero vs one or more visits) by that week was related to pregnancy outcome. Separate cohorts were defined at 26, 30, 34, 38, and 42 weeks. Prenatal care was associated with improved pregnancy outcomes in only the 34-, 38-, and 42-week cohorts (P less than .01). Findings suggest substantial benefit from prenatal care after 30 weeks' gestation but not from early prenatal care. Unfortunately, it may not be possible to assess prenatal care accurately in observational studies even if cohort analyses are used. Clinical trials are needed to assess the effects of strategies for increasing or improving prenatal care, especially in early pregnancy.