Background:  To evaluate survival and to investigate causes of death among HIV-1-infected adults receiving HAART in Senegal.

Methods:  Mortality was assessed in the patients enrolled between August 1998 and April 2002 in the Senegalese antiretroviral drug-access initiative. Patients were included according to the consensus reports on HAART in Africa issued in 1997 and 2000. First-line regimen combined 2 nucleoside reverse transcriptase inhibitor (NRTI) and either a non-NRTI (NNRTI) or a protease inhibitor (PI). Follow-up visits were done every 2 months with a complete biological assessment every 6 months. Survival analysis and Cox regression with fixed and time-dependent covariates were performed. The most likely causes of death were ascertained through medical records or post-mortem interviews (verbal autopsy).

Results:  We enrolled 404 patients (54.7% women) and followed them for a median of 46 months (interquartile range 32 to 57 months) after HAART initiation as of September 30, 2005. At baseline, 5% were ART-non-naive, 39% and 55% were respectively at CDC stage B and C, median age, CD4 count, and viral load were respectively 37 years, 128 cells/μL, and 5.2 log copies/mL. During follow-up, 93 patients died and the overall incidence rate of death was 6.2/100 person-years (95%CI 5.0 to 7.6). During the first year after HAART initiation, 47 patients died and 7 were lost to follow-up, yielding to a probability of dying of 11.7% (8.5% to 15.3%). The death rate, highest during the first year after HAART initiation, decreased with time yielding to a cumulative probability of dying of 17.4% (13.9 to 21.4%), 21.0% (17.3 to 25.4%), 22.9% (19.1 to 27.5%), 24.6% (20.4 to 29.4%) and 25.7% (21.1 to 31.0%) at the end of the subsequent years. A Cox’s model retained a body mass index ≥19 kg·m^­2, an hemoglobin level ≥10 g/dL, and a CD4 cells count ≥200 cells/μL as baseline-independent predictors of a better survival. A model allowing for time-dependent co-variates retained body mass index, CD4, and a PI-containing regimen as predictors of survival. Causes of death were ascertained in 76 instances. Mycobacterial infections, neurotropic infections, and septicemia were the most frequent likely causes of death.

Conclusions:  This study underlines the early mortality pattern after HAART initiation, reports long-term trend in mortality and its predictors among patients under HARRT in an African setting, and highlights the leading role of mycobacterial infections in the causes of death.