182-31: Empirical Power Estimation for Phase I Dose-Proportionality Studies Based on Power-Law Model Using Confidence Interval Criteria

A Phase I dose-proportionality study is an essential tool to understand drug pharmacokinetic dose-response relationship in early clinical development. There are a number of different approaches to the assessment of dose proportionality. The confidence interval (CI) criteria approach, a staitistically sound and clinically relevant approach, has been proposed to detect dose-proportionality (Smith, et al. 2000), by which the proportionality is declared if the 90% CI for slope is completely contained within the pre-determined critical interval. This method, enhancing the information from a clinical dose-proportionality study, has gradually drawn attention. However, exact power calculation of dose proportinality studies based on CI criteria poses difficulity for practioners since the methodology was essentailly from two one-sided tests (TOST) procedure for the slope, which should be unit under proportionality. It requires sophisticated numerical integration, and it is not available in statistical software packages. This paper presents a SAS Macro to compute the empirical power for the CI-based dose proportinality studies. The resulting sample sizes and corresponding empirical powers suggest that this approach is powerful in detecting dose-proportionality under commonly used sample sizes for phase I studies.