A case for preART-adjusted endpoints in HIV therapeutic vaccine trials.

Background: In a randomized, double-blind, placebo-controlled phase 2 clinical trial of Vacc-4x, a peptide-based therapeutic HIV-1 p24(Gag) vaccine candidate, 135 HIV-infected participants (vaccine:placebo=92:43) received a series of six immunizations while on combination antiretroviral therapy (cART). At week 28, all participants underwent an analytical treatment interruption (ATI) for up to 24 weeks. preART VL appeared to be higher among Vacc-4x recipients. Based on a previous analysis, during ATI viral load (VL) appeared to be lower in Vacc-4x recipients, but no difference in CD4 level was observed between Vacc-4x and placebo groups. We propose fold-change-based endpoints and report comparative analyses accounting for imbalanced preART VL and missing data.

Methods: All analyses included per-protocol (PP) participants who received the full immunization and underwent ATI. Linear regression models were used to identify predictors of study endpoints and to estimate the vaccine effect based on fold changes in CD4 counts or VL over preART values at week 40 or at set-point (geometric mean over weeks 48 and 52 values). We adjusted for potential baseline factors and used a multiple imputation approach to account for missing endpoints due to cART resumption or dropout. P-values were adjusted for multiple comparisons using q-values.

Results: preART VL and CD4 count were significant predictors of study endpoints. The vaccine recipients had a higher fold change in week 40 CD4 counts (vaccine vs. placebo mean fold-change difference=0.08; p=0.02; q=0.03), a higher fold change in CD4 count set-point (0.06; p=0.06; q=0.07), a lower fold change in week 40 VL (-0.47; p=0.03; q=0.05), and a lower fold change in VL set-point (-0.50; p=0.02; q=0.03).

Conclusions: These exploratory analyses consistently suggested that Vacc-4x provided positive effects on both CD4 counts and VL. Future HIV therapeutic vaccine studies may adopt similar preART-adjusted endpoints and missing data imputation methods in vaccine effect evaluations.