Background. Adverse pregnancy outcomes for women who conceive on antiretroviral therapy (ART) may be increased, but data are conflicting. Methods. Human immunodeficiency virus-infected, nonbreastfeeding women with pre-ART CD4 counts ?400 cells/?L who started ART during pregnancy were randomized after delivery to continue ART (CTART) or discontinue ART (DCART). Women randomized to DCART were recommended to restart if a subsequent pregnancy occurred or for clinical indications. Using both intentto- treat and as-treated approaches, we performed Fisher exact tests to compare subsequent pregnancy outcomes by randomized arm. Results. Subsequent pregnancies occurred in 277 of 1652 (17%) women (CTART: 144/827; DCART: 133/825). A pregnancy outcome was recorded for 266 (96%) women with a median age of 27 years (interquartile range [IQR], 24-31 years) and median CD4+ T-cell count 638 cells/?L (IQR, 492-833 cells/?L). When spontaneous abortions and stillbirths were combined, there was a significant difference in events, with 33 of 140 (23.6%) in the CTART arm and 15 of 126 (11.9%) in the DCART arm (relative risk [RR], 2.0 [95% confidence interval {CI}, 1.1-3.5]; P = .02). In the as-treated analysis, the RR was reduced and no longer statistically significant (RR, 1.4 [95% CI, .8-2.4]). Conclusions. Women randomized to continue ART who subsequently conceived were more likely to have spontaneous abortion or stillbirth, compared with women randomized to stop ART; however, the findings did not remain significant in the as-treated analysis. More data are needed on pregnancy outcomes among women conceiving on ART, particularly with newer regimens.