Keywords: Concentration-QTc modelling, Model building, Pre-specification
Following the Q&A document R3 on ICH E14, concentration-QTc analysis is increasingly being used as primary analysis for the assessment of QTc-prolonging potential of a new drug. In such a case pre-specification and formalization of this analysis becomes important to exclude selection bias. This applies to the form of the model, the way in which spontaneous diurnal variability is taken into account, and the specification of any covariates or external factors to be accounted for. Likewise, the principles for model checking need to be pre-specified together with principles for how to proceed if the pre-specified model does not fit. The assumptions underlying the selection of a model, i.e. the absence of hysteresis and the appropriateness of a linear relationship are of particular interest here. The working group on the White Paper "Best practices in Concentration-QTc modelling" has proposed some standard proceeding and agreed on a number of principles for this process. In my presentation I will illustrate these and, in addition, elaborate on the situation where potentially active metabolites also need to be taken into account. In particular I will present a way of model selection based on the size of the treatment effect and the AIC.