OBJECTIVE To examine growth of children by maternal and infant HIV status allowing for infant feeding mode. METHODS Women enrolled into a nonrandomized intervention cohort. METHODS Children of HIV-infected and uninfected women weighed and assessed for HIV status, monthly: from birth to 9 months; quarterly: 10-24 months. Daily infant feeding practices recorded at weekly intervals. Weight-for-age z-scores of children born to HIV-infected mothers compared with the reference population of children of HIV-uninfected mothers. Changes in z-scores over age were examined by HIV infection status and infant feeding practice using linear mixed effects models. RESULTS The 1261 children of HIV-infected mothers grew as well as the reference group of 1061 children of HIV-uninfected mothers, irrespective of feeding mode. z-scores for HIV-infected children were consistently lower than those of HIV-exposed but uninfected children: a difference of 420 g for male children and 405 g for female children at 52 weeks of age. Breastfed HIV-infected infants had consistently higher z-scores for weight, especially during first 6 weeks (difference of 130 g for male children; 110 g for female children). In an adjusted regression analysis, maternal mid-upper arm circumference, CD4 cell count, infant birth weight and HIV status had the biggest impact on infant growth (z-score coefficient: 0.38 for mid-upper arm circumference > or =28.35 vs. <25.7 cm; P < 0.001; -0.32 for CD4 cell count <200 vs. > or =500; P = 0.001; -2.01 for birth weight <2500 vs. > or =2500 g; P < 0.001; -0.20 for infected vs. uninfected children; P < 0.001). CONCLUSIONS Optimal early feeding practices ameliorate the effect of being born to an HIV-infected mother and strengthen the recommendation of exclusive breastfeeding for HIV-infected women in terms of long-term child health.