CD4 cell count as a surrogate endpoint in HIV clinical trials: a meta-analysis of studies of the AIDS Clinical Trials Group.

OBJECTIVE: To evaluate initial changes in CD4 cell count as a surrogate endpoint for clinical outcome in HIV-infected patients.
DESIGN: Meta-analysis of all relevant Phase II and III randomized clinical trials undertaken by the Adult AIDS Clinical Trials Group.
METHODS: Individual patient data were obtained from each clinical trial, and the difference between a pair of treatments in their effect on clinical outcome (AIDS or death, or death alone) during 2 years of follow-up was evaluated. The proportion of treatment effect explained (PTE) was the proportion of this difference explained by the change in CD4 cell count 6 months after starting treatment, evaluated using proportional hazards models. A weighted average PTE across treatment comparisons was obtained. The association between the difference between treatments in clinical outcome, expressed as hazard ratio, and the difference in mean change in CD4 cell count was evaluated using regression analysis.
RESULTS: There were 15 clinical trials involving 24 treatment comparisons. The weighted average PTE for both progression to AIDS or death was 0.16 [95% confidence interval (Cl), 0.07-0.26] and for death was 0.10 (95% Cl, 0.00-0.20). There were significant associations between treatment differences in effect on AIDS or death, and on death alone, and the difference in mean change in CD4 cell count. A difference in mean change in CD4 cell count of 30 or 40 x 10(6)/l or more in favor of the test treatment indicated with high probability that there was a corresponding difference in progression to AIDS or death.
CONCLUSIONS: The small PTE suggest that other mechanisms of drug action not captured by initial change in CD4 cells are important. CD4 cell count is a weak surrogate endpoint, but has some value as an aid for screening treatments for drug development or preliminary regulatory approval.