Background:  The optimal time to initiate ART in HIV-infected patients with tuberculosis (TB) is unknown. Although early treatment may complicate clinical management, delayed initiation may increase mortality risk. We determined mortality rates among patients with TB in an ART program and examined the relationship between the time to initiation of ART and outcome.

Methods:  Mortality occurring both before initiation of ART and during the first 16 weeks of ART was prospectively studied among patients with (n = 213) and without (n = 675) TB accessing an ART program in South Africa.

Results:  The mortality rate among those with TB was 1.82-fold (95%CI 1.62 to 2.80) greater than that of patients who remained TB-free (39.7 vs 21.7 deaths/100 person-years; p = 0.003). However, multivariate analysis revealed that mortality in the overall cohort was not independently associated with TB but was only associated with CD4 count <100 cells/μL (AHR = 2.85, 95%CI 1.52 to 5.34) and baseline WHO stage 4 disease (AIDS) (AHR = 2.94, 95%CI 1.80 to 4.82). Of patients with TB diagnosed within the program but before ART initiation (n = 73), 48 had received ART by data censorship after a median of 42 days (IQR 28 to 67) from TB diagnosis. A total of 14 deaths occurred of which 10 (71%) were among patients waiting to commence ART and most (n = 7) were within the first 4 weeks of anti-TB treatment. Just 4 deaths occurred after initiation of ART of which 2 were due to immune reconstitution disease. Risk of death was not associated with baseline patient or disease characteristics but only ART status.

Conclusions:  Advanced immunodeficiency accounts for the extremely high early mortality risk among patients with TB accessing ART in this setting. Pending the results of randomised controlled trials, these data strongly support a policy of initiation of ART within the first month from TB diagnosis among patients with CD4 cell counts <100 cells/μL or AIDS.