Power to identify exposure‐response relationships in phase IIa pulmonary tuberculosis trials with multi‐dimensional bacterial load modeling

Abstract

Adequate power to identify an exposure‐response relationship in a phase IIa clinical trial for pulmonary tuberculosis (TB) is important for dose selection and design of follow‐up studies. Currently, it is not known what response marker provides the pharmacokinetic‐pharmacodynamic (PK‐PD) model more power to identify an exposure‐response relationship. We simulated colony‐forming units (CFU) and time‐to‐positivity (TTP) measurements for four hypothetical drugs with different activity profiles for 14 days. The power to identify exposure‐response relationships when analyzing CFU, TTP, or combined CFU + TTP data was determined at 60 total participants, or with 25 out of 60 participants in the lowest and highest dosing groups (unbalanced design). For drugs with moderate bactericidal activity, power was low (<59%), irrespective of the data analyzed. Power was 1.9% to 29.4% higher when analyzing TTP data compared to CFU data. Combined analysis of CFU and TTP further improved the power, on average by 4.2%. For a drug with a medium‐high activity, the total sample size needed to achieve 80% power was 136 for CFU, 72 for TTP, and 68 for combined CFU + TTP data. The unbalanced design improved the power by 16% over the balanced design. In conclusion, the power to identify an exposure‐response relationship is low for TB drugs with moderate bactericidal activity or with a slow onset of activity. TTP provides the PK‐PD model with more power to identify exposure‐response relationships compared to CFU, and combined analysis or an unbalanced dosing group study design offers modest further improvement.