Background: Two recent analyses of HIV-1-infected mothers' mortality according to their children's feeding modality have produced conflicting results. Methods: An individual patient data meta-analysis was conducted using data regarding HIV-1-infected women from eligible clinical trials. Analyses included Cox proportional hazards regression modeling, with children's feeding modality treated as a time-dependent covariate. Results: Of 4237 HIV-1-infected women, 162 (3.8%) died within 18 months after delivery. The risk of mortality during the 18-month period after delivery did not differ significantly by children's feeding modality (ever vs. never breast-fed), with or without adjustment for maternal CD4(+) count. Treating children's feeding modality as a time-dependent covariate, the risk of mortality was lower among women still breast-feeding (hazard ratio = 0.05, 95% confidence interval. 0.03, 0.09; P < 0.0001) than among those who had ceased, with similar results observed with adjustment for maternal CD4(+) count. Conclusions: HIV-1-infected women with lower CD4(+) counts were less likely to initiate breast-feeding. Mothers' mortality during the 18-month period after delivery did not differ significantly according to children's feeding modality (ever vs. never breast-fed). Of those women who initiated breast-feeding, the lower mortality risk among those still breast-feeding compared with those not breast-feeding likely represents better overall maternal health (with healthier women being able to breast-feed longer).