Abstract

Concentration-effect (CE) models applied to early clinical QT data from healthy subjects are described in the latest E14 Q&A document as promising analysis to characterise QTc prolongation. The challenges faced if one attempts to replace a TQT study by thorough ECG assessments in Phase I based on CE models are the assurance to obtain sufficient power and the establishment of a substitute for the positive control to show assay sensitivity providing protection against false negatives. To demonstrate that CE models in small studies can reliably predict the absence of an effect on QTc, we investigated the role of some key design features in the power of the analysis. Specifically, the form of the CE model, inclusion of subjects on placebo, and sparse sampling on the performance and power of this analysis were investigated. In this study, the simulations conducted by subsampling subjects from 3 different TQT studies showed that CE model with a treatment effect can be used to exclude small QTc effects. The number of placebo subjects was also shown to increase the power to detect an inactive drug preventing false positives while an effect can be underestimated if time points around t max are missed.