Objective. To assess the relationship between clinical, demographic, and site-of-care factors and the use of tocolysis and corticosteroid therapy in the treatment of premature labor. Data Source. Secondary clinical and demographic data collected for the five-center March of Dimes Prematurity Prevention clinical trial, 1983-1986. Study Design. We used logistic regression analysis in assessing the clinical, patient, and care site factors associated with the use of tocolysis and corticosteroid therapy during episodes of premature labor occurring to women enrolled in the trial. The two interventions were not subject to control in the trial, but were provided according to customary practice at the care site. Data Extraction. A total of 4,625 episodes of labor occurring before 37 weeks gestation were identified from either preterm labor or preterm delivery records recorded for the 33,792 women enrolled in the trial. Principal Findings. The use of tocolysis, an intervention that attempts to control premature labor contractions and that was widely used in high-risk obstetrics, varied almost exclusively by clinical factors. The use of corticosteroid therapy, a little used but effective intervention that reduces respiratory complications in premature infants, varied significantly by site of care and was used less frequently across sites and clinical conditions for minority group patients. Conclusion. This study confirms the premise that practice variation on the basis of nonclinical factors occurs more commonly for interventions were there is more uncertainty about clinical indications and effectiveness. The study also identifies another area of clinical care in which the use of aggressive and relatively uncertain interventions is provided less frequently to minority group patients.