BACKGROUND:

Studies of the effect of highly active antiretroviral therapy (HAART) on the risk of HIV-1-associated tuberculosis have had variable results. We set out to determine the effect of HAART on the risk of tuberculosis in South Africa.

METHODS:

We compared the risk of tuberculosis in 264 patients who received HAART in phase III clinical trials and a prospective cohort of 770 non-HAART patients who were attending Somerset Hospital adult HIV clinic, University of Cape Town, between 1992 and 2001. Poisson regression models were fitted to determine risk of tuberculosis; patients were stratified by CD4 count, WHO clinical stage, and socioeconomic status.

FINDINGS:

HAART was associated with a lower incidence of tuberculosis (2.4 vs 9.7 cases per 100 patient-years, adjusted rate ratio 0.19 [95% CI 0.9-0 38]; p<0.0001). This finding was apparent across all strata of socioeconomic status, baseline WHO stage, and CD4 count, except in patients with CD4 counts of more than 350 cells/microL. The number of tuberculosis cases averted by HAART was greatest in patients with WHO stage 3 or 4 (18.8 averted cases per 100 patient-years, adjusted rate ratio 0. 22 [0.09-0.41]; p=0.03) and in those with CD4 counts of less than 200 cells/microL (14.2 averted cases per 100 patient-years, adjusted rate ratio 0.18 [0.07-0.47]; p,0.0001).

INTERPRETATION:

HAART reduced the incidence of HIV-1-associated tuberculosis by more than 80% (95% CI 62-91) in an area endemic with tuberculosis and HIV-1. The protective effect of HAART was greatest in symptomatic patients and those with advanced immune suppression.