OBJECTIVE: To determine the predictors for early versus later (breastfeeding) transmission of HIV-1. METHODS: Secondary data analysis was performed on HIV Network for Prevention Trials 012, a completed randomized clinical trial assessing the relative efficacy of nevirapine (NVP) versus zidovudine in reducing mother-to-child transmission (MTCT) of HIV-1. We used Cox regression analysis to assess risk factors for MTCT. The ViroSeq HIV genotyping and a sensitive point mutation assay were used to detect NVP resistance mutations. RESULTS: In this subset analyses, 122 of 610 infants were HIV infected, of whom 99 (81.1%) were infected early (first positive polymerase chain reaction < or =56 days). Incidence of MTCT after 56 days was low [0.7% per month (95% confidence interval, CI: 0.4 to 1.0)], but continued through 18 months. In multivariate analyses, early MTCT "factors" included NVP versus zidovudine (hazard ratio (HR) = 0.57, 95% CI: 0.38 to 0.86), pre-entry maternal viral load (VL, HR = 1.76, 95% CI: 1.28 to 2.41), and CD4 cell count (HR = 1.16, 95% CI: 1.05 to 1.28). Maternal VL (6-8 weeks) was associated with late MTCT (HR = 3.66, 95% CI: 1.78 to 7.50), whereas maternal NVP resistance (6-8 weeks) was not. CONCLUSIONS: Maternal VL was the best predictor of both early and late transmission. Maternal NVP resistance at 6-8 weeks did not predict late transmission