The Original Study
Havlir et al. (2011) randomly assigned 809 HIV-1 infected and ART naive patients with CD4+ T-cell counts of less than 250 per cubic millimeter and suspected tuberculosis to an earlier ART arm (within 2 weeks after the initiation of treatment for tuberculosis; n=405) and later ART arm (between 8 and 12 weeks after the initiation of treatment for tuberculosis; n=401). The enrollment of participants in the study at 26 clinical-research sites in four continents went from September 2006 to August 2009. The authors find that earlier ART did not reduce the rate of new AIDS-defining illness and death, as compared with later ART (12.9% vs.16.1% of HIV positive patients in each group; 95% confidence interval [CI] -1.8 to 8.1; P=0.45). However, in the pre-specified subgroup analyses, the authors find that for HIV-1 positive patients with CD4+T cell count strata <50 cells/mm3, the rate of new AIDS-defining illness or death was significantly lower in the earlier-ART than in the later-ART group (15.5% vs. 26.6%; 95%CI, 1.5 to 20.5; P=0.02).
The Replication
Three recent studies find that early ART initiation within 2 weeks after the initiation of treatment for tuberculosis reduces the rate of new AIDS-defining illness and death only for HIV positive TB patients with a CD4 count of less than 50cells/mm3. These results led World Health organization in 2011 to recommend that the provision of ART begins within 8 weeks of initiation of antituberculosis treatment in TB patients with a CD4 count of more than 50cells/mm3 and within 2 weeks after the onset of antituberculosis treatment for TB patients with a CD4 count of less than 50cells/mm3. We propose to replicate one of the studies conducted at 26 clinical research sites in four continents to ensure validity and potentially provide additional insights related to optimal timing for ART initiation in HIV-TB co-infected patients, in a context, where the initiation of ART will be at higher CD4 count. We will use the raw data and methods used to produce the results presented in the original paper. In addition, we will assess the robustness of authors’ results through the use of different analytical methods borrowed from epidemiology and econometrics.