Background:  The question of exactly when HAART should be started is unresolved. Previous analyses that compared disease progression according to immunodeficiency at HAART initiation may have been biased as they generally ignored the lead time (the time it takes for patients starting later to reach the lower CD4 count) and unseen events that occur in the deferred group. In particular, patients who deferred treatment, but progressed to AIDS or died while untreated are not represented in analyses comparing rates of progression to AIDS in those starting HAART at different CD4 counts.

Methods:  We estimated hazard ratios (HR) comparing individuals starting HAART when CD4 count fell between 500 and 351 cells/mL with those starting between 350 and 201 cells/mL; we also compared those starting between 350 and 201 cells/mL with those starting £200 cells/mL. We estimated the time to cross the CD4 thresholds and the time to AIDS using data from the Multicenter AIDS Cohort Study (MACS) during the pre-HAART era (1989-1995). Data from 10,885 individuals in the ART Cohort Collaboration (ART-CC) who were AIDS-free and ART-naïve at HAART initiation were used to estimate rates of progression to AIDS after HAART initiation. For individuals starting HAART below the CD4 threshold, lead time was imputed 25 times by randomly sampling from the pre-HAART-era distribution. Data on “unseen” individuals progressing from the upper CD4 threshold to AIDS in the absence of therapy were also imputed 25 times. Standard methods for combining results from multiple imputed datasets were used to derive overall HR and 95% CI.

Results:  CD4 count at initiation of therapy was £200, 201 to 350, and 351 to 500 cells/mL for 40%, 37%, and 23%, respectively. Median length of follow up was 2.7 years (IQR 1.2 to 4.4 years). There was clear evidence that individuals starting £200 had higher progression rates than those starting between 201 and 350 cells/mL (HR for AIDS 3.30, 95%CI 2.51 to 4.33) after accounting for lead time and unseen events. Comparing individuals starting at 201 to 350 with those starting at 351 to 500 cells/mL, the corresponding HR was 1.46 (95%CI 0.96 to 2.21).

Conclusions:  Our results provide evidence that over several years, AIDS rates are lower in individuals who initiate HAART at higher CD4 counts. However, these results must be discounted for possible disadvantages of early initiation of therapy, including adverse effects of HAART and potential exhaustion of future treatment options.