Accounting for Adherence in Comparisons of Pediatric Antiretroviral Therapy [thesis]

Dwight Edward Yin
Near-perfect adherence to antiretroviral therapy (ART) is necessary to prevent treatment failure and resistance. In children, protease-inhibitor (PI)-based regimens appear more potent than non-nucleoside reverse transcriptase inhibitor (NNRTI)-based regimens, but poor PI taste and tolerability may lead to worse adherence. We aimed to disentangle relationships among ART regimen, adherence, and viral control. In Aim 1, we assessed treatment-naïve children for differences in time to treatment
more » ... e to treatment disruption across randomly assigned PI versus NNRTI-based initial ART. At 4 years, the hazard ratio (HR) for treatment disruptions on PI versus NNRTI-based regimens was 1.19 (95% confidence interval [CI] 0.88-1.61). By study end, treatment disruption probabilities converged with HR 1.11 (95% CI 0.84-1.48). Reported reasons for treatment disruptions suggested that PI participants experienced greater tolerability problems. In Aim 2, we estimated the per-protocol effect of initial PI versus NNRTI-based ART on time to treatment failure in children with HIV in a setting of ideal adherence. After an intention-to-treat (ITT) analysis, we generated per-protocol estimates by administratively right-censoring participants experiencing protocol deviations (non-medically indicated treatment disruption or dropout) and using inverse-probability of censoring weights to correct for imparted informative censoring. In ITT analysis, PI participants experienced 4-year treatment failure probabilities of 41.3% versus 39.5% (NNRTIs), risk difference (RD) 1.8%, HR 1.09 (95% CI 0.74-1.60). In per-protocol analysis, PI participants experienced treatment failure probabilities of 35.5% versus 29.5% (NNRTI), RD 6.4%, HR 1.30 (95% CI 0.80-2.12). Protocol deviations were non-differential across arms. Shifts in failure probabilities from the ITT to per-protocol analysis were 5.7% (PI) versus 10.3% (NNRTI). In conclusion, children experienced similar time to treatment disruption for initial PI- and NNRTI-based ART, despite greater PI tolerabilit [...]
doi:10.17615/r499-mq22 fatcat:42oaqoqa7neuflkdn3opkehxbq