Background: There are few longitudinal data on the natural history of pediatric HIV infection in Africa.

Methods:  Children with HIV Antibiotic Prophylaxis (CHAP) is a randomized controlled trial comparing daily cotrimoxazole with placebo in HIV-infected children aged 1 to 14 years in Lusaka. Baseline information included age, sex, CD4%, height, weight, details of prior and current illnesses (malaria, oral candida, bacterial infections, tuberculosis, malnutrition, diarrhea, developmental delay), and details of primary carer. The prognostic importance of baseline characteristics for mortality was investigated using Cox models based on time from enrolment to the earliest of death, last seen alive, or September 15, 2003. Randomized allocation was not considered.

Results:  Between March 2001 and January 2003, 541 children were enrolled, contributing 541 follow-up years (maximum 2.3 years). Half were boys, 68% were cared for by their mother, 15% were >10 years old, and median CD4% was 11% (n = 482).  Height and weight were less than the third centile compared to WHO growth standards in 56% and 45%, respectively. During the study, 166 children died (31.0 per 100 years at risk [95% CI 26.4 - 35.7]). The variation in mortality with age was non-linear, with infants and older children having the highest death rates (p <0.001). In contrast, death rates decreased as baseline CD4% and age-adjusted growth scores increased (all p <0.001). However, baseline age-adjusted height was not an independent predictor after adjusting for baseline CD4% (p <0.001) and age-adjusted weight (p <0.001). After adjusting for age, sex, and primary carer, only clinically assessed malnutrition  (RR=2.43 [1.70 - 3.48], p <0.001) and current Candida (RR=1.60 [1.12 - 2.28], p = 0.01) independently predicted death. However, most of the effect of Candida was accounted for by baseline CD4%, with only malnutrition remaining an independent predictor after adjusting for CD4%.

Conclusions:   More than 90% of children still alive are continuing in follow-up. Candida and malnutrition are the best clinical predictors of mortality; malnutrition appears to predict HIV- and non-HIV-related death, whereas Candida appears to predict HIV-related death directly through low CD4 percentage.

Keywords: Natural history; Mortality; African Children