Background:  Evidence of the effect of HAART from randomized controlled trials is limited because most trials use surrogate endpoints such as CD4 count or virologic response, and follow-up is often for short periods. There has been no placebo-controlled randomized trial of HAART. The effectiveness of HAART over several years of treatment is therefore unknown at present. Most HIV-infected persons now considering starting HAART have not been treated with other antiretroviral drugs.

Methods:  We used data from the Swiss HIV Cohort Study, an observational study of HIV-infected residents in Switzerland aged ³16 years. All participants followed-up after January 1996 were potentially eligible for analysis. Standard statistical methods (e.g., Cox proportional hazards models) are likely to yield biased estimates of the effect of HAART in the presence of time-dependent confounders affected by prior treatment, so we used weighted Cox models (marginal structural models) to estimate hazard ratios for progression to AIDS or death. These control confounding by weighting according to the probability of starting HAART (estimated using logistic regression). In separate models, we estimated the effect of HAART compared with no treatment and compared with dual therapy.

Results:  Lower CD4 count, increasing HIV-1 viral load and previous CDC stage B events were associated with an increased risk of starting HAART. Overall hazard ratios for the effect of HAART on progression to AIDS or death were 0.07 (95% CI 0.03 to 0.16) compared with no treatment, and 0.48 (0.31 to 0.74) compared with dual therapy (consistent with results from randomized controlled trials). Compared with no treatment, HAART became more beneficial with increasing time since initiation. In patients with presumed transmission via intravenous drug use the hazard ratio was 0.12 (0.04 to 0.33); in other patients the hazard ratio was 0.05 (0.02 to 0.12) (interaction p = 0.096). Compared with both no treatment and dual therapy, the beneficial effect of HAART was greater in patients whose CD4 cell count at baseline was < 200 cells/mL.

Conclusions:  To our knowledge, this is the first study to estimate the effect of HAART compared with no treatment in reducing rates of progression to AIDS and death. The substantial beneficial effects of HAART are consistent with observed declines in rates of AIDS and death in HIV-infected persons observed in developed countries since the introduction of HAART, and provide a context for considering adverse effects.

Keywords: Antiretroviral therapy; AIDS; cohort studies