Model based design and analysis of phase II HIV-1 trials

Abstract

This work explores the advantages of a model based drug development (MBDD) approach for the design and analysis of antiretroviral phase II trials. Two different study settings were investigated: (1) a 5-arm placebo-controlled parallel group dose-finding/proof of concept (POC) study and (2) a comparison of investigational drug and competitor. Studies were simulated using a HIV-1 dynamics model in NONMEM. The Monte-Carlo Mapped Power method determined the sample size required for detecting a dose-response relationship and a significant difference in effect compared to the competitor using a MBDD approach. Stochastic simulation and re-estimation were used for evaluation of model parameter precision and bias given different sample sizes. Results were compared to those from an unpaired, two-sided t test and ANOVA (p≤0.05). In all scenarios, the MBDD approach resulted in smaller study sizes and more precisely estimated treatment effect than conventional statistical analysis. Using a MBDD approach, a sample size of 15 patients could be used to show POC and estimate ED50 with a good precision (relative standard error, 25.7%). A sample size of 10 patients per arm was needed using the MBDD approach for detecting a difference in treatment effect of ≥20% at 80% power, a 3.4-fold reduction in sample size compared to a t test. The MBDD approach can be used to achieve more precise dose-response characterization facilitating decision making and dose selection. If necessitated, the sample size needed to reach a desired power can potentially be reduced compared to traditional statistical analyses. This may allow for comparison against competitors already in early clinical studies.