Background:  Initiating HAART at low CD4 cell counts/mm³ (CD4) increases mortality but the effects of delayed initiation on CD4 responses and causes of death have not been extensively examined.

Methods:  We studied patients with >6 months of follow-up after starting HAART seen at 10 US clinics during 1996 to 2007. We analyzed CD4 trajectories (compared using the Jonckheere-Terpstra test) and mortality rate (MR) trends by CD4 at time of HAART start (baseline, BL). We assessed factors associated with mortality using Cox proportional hazards models.

Results:  Among 1,378 patients with a CD4 recorded at HAART initiation, median follow-up was 4.2 years and 82 died within 6 months of last contact. By baseline CD4 strata (< 50, 50 to 199, 200 to 349, ≥350), we found: i) median peak CD4 achieved after BL was progressively higher at each higher BL CD4 stratum: 392, 443, 644, and 956, P <0.001 for trend; ii) the proportions of surviving patients with CD4 ≥350 at 4 years increased: 46%, 59%, 79%, and 95%, P <0.001 for trend; iii) crude MR per 100 person-years decreased: 2.80, 1.52, 0.60, and 0.53, P <0.001 for trend; and iv) among deaths, higher BL CD4 was associated with higher CD4 at death: 68, 186, 245, and 516, P <0.001 for trend. In multivariable analyses, factors independently associated with mortality were CD4<50 (hazard ratio [HR] = 4.6, 95%CI 2.7 to 7.9) and CD4 50 to 199 (HR = 2.6, 95%CI 1.5 to 4.8) compared with CD4 >350, and public insurance (HR = 1.73, 95%CI 1.1-2.7). For persons with known primary cause(s) of death (n = 69), crude mortality rates for patients with AIDS-related (n = 33) and non-AIDS-related causes (n = 36) decreased with increasing BL CD4 (P <0.001 and P =0.005, respectively, for trend); in the 2 higher BL CD4 strata (200 to 350 and >350) deaths from non-AIDS-related causes predominated (77%). Median CD4 near death was 27 for patients with AIDS-related causes and 193 for non-AIDS-related causes (P <0.001). Median CD4 nadir-to-peak increases were greater among persons who survived than among persons who died in the BL CD <50 (401 vs 135, P <0.001; median follow-up = 57 months) and BL CD4 50 to 199 (351 vs 158, P <0.001, median follow-up = 52 months) strata, but not in the 2 higher BL CD4 strata.

Conclusions:  Lower BL CD4 at HAART start was associated with lower peak CD4 achieved while on HAART, lower CD4 at time of death, and increased risk of death from AIDS-related and non-AIDS-related causes. Among persons starting HAART with BL CD4>200 and CD4>350, non-AIDS-related causes of death predominated.