OBJECTIVE To identify modifiable obstetric factors associated with the failure of zidovudine chemoprophylaxis to prevent perinatal human immunodeficiency virus type 1 (HIV-1) transmission. METHODS We analyzed data from Pediatric AIDS Clinical Trials Group protocol 076, a randomized, double-masked, placebo-controlled trial that demonstrated that a zidovudine regimen could prevent perinatal HIV-1 transmission. We estimated the zidovudine treatment effect using the relative reduction in transmission risk among women randomized to treatment with zidovudine compared with women randomized to receive placebo. Univariate and multivariate statistical analyses were used to assess whether the treatment effect differed in magnitude according to potential antepartum or intrapartum risk factors. RESULTS In the univariate analysis, the zidovudine treatment effect was found to differ significantly in magnitude according to quartile of maternal weight at the time of study entry (interaction test, P = .03); among women in the heaviest-weight quartile (weight more than 82 kg), there was a 26% relative reduction in transmission risk, compared with a 79% relative reduction among the other three quartiles (interaction test, P = .05). In the zidovudine treatment group, women who transmitted HIV-1 were significantly more likely than nontransmitters to have had antepartum procedures or conditions associated with increased risk of fetal exposure to maternal blood or cervicovaginal secretions (43% compared with 19%, P = .04). In the multivariate analysis, adjustment for the plasma HIV-1 RNA level and CD4+ cell percentage did not eliminate the differential treatment effect according to these factors. CONCLUSIONS High maternal weight and conditions associated with fetal exposure to maternal blood or cervicovaginal secretions may diminish the efficacy of zidovudine chemoprophylaxis.