Background: Greater initial HAART success has been attributed to the third drug used, to baseline viral load and CD4 count, and pill burden. Unevaluated parameters include: study design, phase, sponsor, duration, eligibility criteria, and type of intention-to-treat analysis; NRTI backbone; pill frequency and dosing relative to food; and participant characteristics. Why participants cease ART has not been studied.

Methods: We reviewed all prospective HAART studies (64 randomized clinical trials; 14 cohorts) in naïve adults, except 3-class, 4-drug, and non-recommended triple-NRTI ART. We hypothesized that successful HAART would be affected by additional study, pill and participant factors. Analysis of parameters associated with efficacy was by forward, stepwise, logistic regression.

Results: In 140 groups, 138 (99%) received 2 NRTI, most with an NNRTI (39%), an unboosted protease inhibitor (PI) (30%), a boosted PI (ritonavir, r)/PI; 20%), or a third NRTI (8%). Common NRTI backbones were zidovudine-lamivudine (ZDV-3TC; 34%), stavudine (d4T) -3TC (16%), d4T-didanosine (ddI) (15%), abacavir (ABC) -3TC (11%), tenofovir (TDF) -3TC/emtricitabine (FTC) (11%), and ddI-3TC/FTC (6%). Participants took a mean 8.7 pills/day in 2.2 doses; 47 (34%) groups took ART without regard to food and 39 (28%) groups received placebo. 22,635 participants (24% women, 13% with AIDS, 36% non-white, mean 36 years, CD4 287 cells/mm³, HIV RNA 4.8 logs) were recruited in the Americas (47%), Europe/Australia (43%), Asia (4%), and Africa (5%). Mean intent-to-treat success (RNA<50 copies/mL in 132 groups) was 60% over a mean 14.3 months. The factors independently associated with higher success were:  third-drug class (favoring r/PI; p <0.001); dosing without regard to food (p <0.001); NRTI backbone (favoring ddI-3TC/FTC; p = 0.008); and race (favoring non-whites; p <0.001). Overall model r² was 0.76. ART cessation (26%) was for adverse events (10.1%), lost to follow-up (6.4%), participant choice (3.9%), virologic failure (3%) or other reasons (3%). Greater incidence of any grade 2+ adverse events was inversely correlated with ART success (r = –0.41, p = 0.007); the individual adverse events most related was any grade diarrhea (Pearson's r = –0.26, p = 0.06). Notably, cessation for "other reasons" correlated with a greater incidence of any grade fatigue (r = 0.42, p = 0.02), headache (r = 0.38, p = 0.009), and nausea (r = 0.33, p = 0.01).

Conclusions: Initial ART success was influenced by the drug classes used, but also by pill dosing relative to food and by race; comparisons of trials should consider these factors. Adverse events are the most common cause of ART failure. Some ART cessation for non-adverse event reasons may reflect unrecognized ART toxicity.