The Reference-Corrected Visual Predictive Check: A More Intuitive Diagnostic for Non-Linear Mixed Effects Models

The prediction-corrected visual predictive check (pcVPC) is an informative model diagnostic that can offer advantages over the standard visual predictive check (VPC) when heterogenous study designs and adaptive dosing are used. However, a drawback with these plots is that prediction correction often results in y-axis values and trends that are unintuitive, difficult to explain, and challenging to communicate even among experts. The reference-corrected visual predictive check (rcVPC) offers a solution to these problems by leveraging a user-defined set of independent variables, for a more intuitive model diagnostic and an efficient communication of results to a wider audience. The rcVPC methodology is based on the definition of a reference dataset. Simulations are conducted with this reference dataset and the observed dataset, and then the simulated and the observed dependent variables are normalized by the population prediction for the user-defined independent variables in the reference dataset. The opportunity to manipulate time in the reference dataset is a unique feature that gives rcVPC the ability to visually characterize exposure-response relationships with delayed effect onset. The rcVPC approach was compared to pcVPCs and traditional VPCs for a range of examples inspired by real data. The rcVPC methodology was demonstrated to offer a more intuitive interpretation and more effective guidance to model development in a way that is not possible for VPC or pcVPC plots.