This study investigates the prevention of early mortality in patients initiating antiretroviral therapy (ART) in sub-Saharan Africa where 79% of the co-infected cases of TB reside. Many published studies have shown a surprisingly high proportion of all patients initiated on ART dying within 6 months (8-26%) with increasing risk with decreasing CD4 T cell count. The majority (median 70%) occur in the first 3 months with the greatest proportion of deaths due to previously undiagnosed tuberculosis (TB). The investigators will enroll patients from 4 geographically diverse countries (Gabon, Mozambique, South Africa, and Uganda) in a randomized open label clinical trial targeting a population of people with high mortality risk; patients with CD4 T cell count \< 50 cells/μl and body mass index (BMI) \< 18 kg/m2. Severely immunocompromised patients with low BMI in the intervention arm will receive presumptive anti-TB 4-drug chemotherapy and subsequently initiate ART within 2 weeks compared to ART alone. The main objective is to measure and compare early mortality in the group presumptively treated for TB in addition to ART. Other sub-objectives are to determine the predictors of early mortality and the causes of death by autopsy (traditional and verbal), to determine if presumptive anti-TB treatment affects viral suppression with ART, and to assess incidence rates and characterize drug toxicity in patients dually treated. Because of the high rates of TB co-infection in sub-Saharan Africa in the HIV-infected, the investigators expect that patients presumptively treated for TB in addition to HIV will have a lower mortality rate than patients receiving ART only. This trial is expected to be of great public health benefit and generalisability.