Background:  Thus far, efforts to develop HIV vaccines capable of eliciting broadly cross-reactive neutralizing antibodies have been unsuccessful; therefore, much recent attention has focused on vaccine-induced HIV-specific cell-mediated immune  (CMI) responses. Such vaccines have shown partial success in some animal models, but this concept has not been previously tested in humans.

Methods:  The STEP study is a phase II, randomized, multi-center, double-blind, placebo-controlled test-of-concept study in 3000 HIV-seronegative high-risk volunteers. Volunteers were randomized (1:1) to receive 3 injections of either the MRKAd5 HIV-1 gag/pol/nef vaccine (a replication-defective Ad5 vector) or placebo, stratified by 4 baseline levels of Ad5 neutralizing antibody. Volunteers self-reported demographics, risk in the previous 6 months, and circumcision status. HIV testing was conducted ~every 6 months; plasma viral load was measured at multiple post-infection timepoints. 

Results:  In the modified intent-to-treat primary analysis (population with baseline Ad5 ≤200), there were 24 infections among 741 vaccinees and 21 infections among 762 placebo recipients. Geometric mean plasma vRNA level was similar in infected vaccine vs placebo recipients (~40,000 vs ~26,000 copies/mL, respectively). In ongoing, post-hoc, multivariate analyses the following were associated with HIV acquisition:  treatment (vaccine vs placebo), hazard ratio (HR) 1.9, 95%CI (1.1 to 3.2); region (North America vs other), HR 3.2 (1.5 to 6.7); and baseline unprotected receptive anal sex (yes vs no), HR 4.4 (2.4 to 8.7). With interaction terms in the model, the treatment HR (vaccine vs placebo) was 3.1 (1.5 to 6.5) in Ad5 >18 vs HR 1.0 (0.5 to 2.0) in Ad5 ≤18; the treatment HR was 4.5 (1.8 to 11.4) in uncircumcised men vs 1.0 (0.6 to 1.8) in circumcised men.

Conclusions:  There was no evidence that the Merck Ad5 trivalent vaccine prevented infection or lowered viral set-point. In multivariate analysis of baseline risk factors, vaccination appeared associated with an increased risk of HIV acquisition in men with pre-existing Ad5 immunity and in uncircumcised men. We are awaiting additional data on herpes simplex virus-2 (HSV-2) status, HLA typing, and sexual network clustering to explore possible confounding factors for HIV acquisition in STEP study volunteers. Because of the implications for the development of other CMI-based vaccines, it will be important to understand the potential mechanisms underlying these results.