Background:  A 3-fold higher mortality risk was observed among HIV-infected women randomized to breast feed vs formula feed in a Kenyan trial, but subsequent observational studies in Africa did not find this.

Methods:  We randomized 1200 HIV-infected pregnant women in the Mashi prevention of mother-to-child transmission (PMTCT) study in Botswana to breast feeding for 6 months with prolonged infant zidovudine (ZDV), or to formula feeding. HAART became available to participants meeting WHO treatment criteria. Mothers were followed for 7 years. Serum 6 months post partum from a random sample of 131 women (65 breast feeding, 66 formula feeding) was tested for micronutrients (vitamins B12, A, and E and selenium), albumin, and C-reactive protein (CRP) levels. We report the proportion of women reaching a composite endpoint of CD4 decline to <200, AIDS illness, or death, by randomized feeding arm (intent-to-treat, Kaplan and Meier method). Variables with p ≤0.15 for association with endpoint were included in a step-wise Cox proportional hazards regression model.

Results:  We followed the 1200 women for a median of 36 months’ follow-up; baseline characteristics did not differ by randomized arm. The median entry CD4 was 364 and median HIV-1 RNA was 4.3 log₁₀. 372 (31.0%) of women experienced an endpoint—204 (34.1%) of 598 women in the breast feeding arm and 168 (27.9%) of 602 in the formula feeding arm (p = 0.08). Of the 372 endpoints, 280 were CD4 decline alone (151 breast feeding, 129 formula feeding), and 25 AIDS diagnoses alone (18 breast feeding, 7 formula feeding); 35 (2.9%) women died (16 breast feeding, 19 formula feeding). Factors associated with reaching an endpoint included age <median (p = 0.02), baseline CD4 <350 (p <0.01) or HIV-1 RNA ≥median (p <0.01), any income (p = 0.02), no education (p = 0.04), or being married or co-habiting (p = 0.01). In MV analysis, the adjusted HR (95%CI) for reaching an endpoint was 0.82 (0.67 to 1.01) for formula-feeding; low CD4, high HIV-1 RNA, and lack of education remained significant predictors. The trend toward worse outcome in the breast feeding arm emerged after end of breast feeding. Median HIV-1 RNA did not differ by arm at 6, 12, and 60 months. Median CRP levels were higher among breast feeding than formula feeding women (2.28 mg/L vs 1.05 mg/L; p <0.01) and among women experiencing an endpoint (1.77 mg/L vs 0.77 mg/L; p <0.01). Micronutrient and albumin levels did not differ between these groups

.

Conclusions:  Breastfeeding was not associated with maternal mortality among women with access to HAART. We found higher CRP and a trend (appearing after cessation of breast feeding) toward faster progression to AIDS among women randomized to breast feeding for 6 months (vs to formula feeding).