Predictors of TB disease in HIV-exposed children from Southern Africa.

BACKGROUND: P1041 was a randomised, placebo-controlled isoniazid prophylaxis trial in South Africa. We studied predictors for TB in HIV-exposed children participating in the P1041 trial.METHODS: We included data from entry until Week 108. Predictors considered were type of housing, overcrowding, age, sex, ethnicity, tobacco exposure, weight-for-age percentile Z-score (WAZ), CD4%, viral load (VL), antiretroviral therapy (ART) and number of household smokers.RESULTS: Of 543 HIV-positive (HIV+) and 808 HIV-exposed uninfected (HEU) infants at entry, median age was 96 days (interquartile range: 92-105). Of 1,351 caregivers, 125 (9%) had a smoking history, and 62/1,351 reported current smoking. In 594/1,351 (44%) households, there was at least one smoker. Smoking caregivers consumed 1-5 cigarettes daily. In the HIV+ cohort, significant baseline TB predictors after adjusting covariates were as follows: WAZ (adjusted hazard ratio [aHR] 0.76, P = 0.002) and log10 HIV RNA copies/ml (aHR 1.50, P = 0.009). Higher CD4% (aHR 0.88, P = 0.002) and ART (aHR 0.50, P = 0.006) were protective. In the HEU cohort, smoking exposure was associated with reduced TB-free survival on univariate analysis, but not after adjustment in the multivariate model.CONCLUSION: Low WAZ and high VL were strong predictors of TB disease or death. Rising CD4 percentage and being on ART were protective in the HIV+ cohort.