In a randomized blinded trial of antenatal thyrotropin-releasing hormone(TRH) done at 13 North American Centers between October 1992 and December 1996 we failed to demonstrate any effect of TRH in incidence of RDS, chronic lung disease or death at 28 days (CLD28) or at 36 weeks postmenstrual age (CLD36) or other morbidities in preterm infants delivered before 32 weeks gestation. There was, however, a wide variation between centers in CLD28 (18-63%) and CLD36 (25-41%) in the at-risk infants (≤ 32 weeks gestation). As expected, birth weight (BW) and gestational age (GA) were predictive of CLD in the total population and in individual institutions, but gender and race were not. The 6 largest centers had Clinical Risk Index for Babies (CRIB) Score data available on 403 infants < 1500 grams. CRIB Score predicted CLD in the whole group and in each institution both with and without adjustment for GA, race and gender, indicating that the initial clinical status is an important factor in CLD. TRH administration was not associated with CRIB Score, BW, GA or CLD. We found no significant difference in BW, GA, race, or CRIB Score between centers to explain differences in CLD. Adjusting for CRIB Score and GA, institution continued to be a strong predictor of CLD28 and CLD36 (p < 0.01). Post-natal steroid administration was a significant predictor of CLD above and beyond gestational age and CRIB and varied among the centers from 17% in the center with the lowest incidence of CLD to 75% in the center with the highest CLD. When adjustment is made for post-natal steroid use in a logistic regression model, the institutional predictive effect becomes non-significant, suggesting that management approaches associated with the development of CLD include increased use of steroids. We conclude that center differences in incidence of CLD are not due to population characteristics or the condition of infants at birth, and are likely related to differences in clinical management of these infants.