Impact of Preclinical Verification in Physiologically Based Pharmacokinetic Modelling for First-in-Human Predictions:...

Physiologically based pharmacokinetic modelling is routinely used in the pharmaceutical industry and has an impact on drug labels. New applications have emerged, one of which is first-in-human pharmacokinetic predictions. Scientists in this field often believe that verification of models in preclinical species is essential for accurate predictions, but consensus has not been reached and animal use across the industry deserves continual examination with the aim to reduce use.

A published model-building strategy was used to assess the accuracy added by preclinical verification for human PK prediction. Three sequential approaches were explored using five compounds covering a range of physicochemical properties and chemical classes. Approach 1 (QSPR FIH prediction), uses parameter inputs predicted using in silico Quantitative Structure Property Relationship models; Approach 2 (In vitro FIH prediction), supplements predicted parameters with in vitro measurements where available data exist; Approach 3 (Verified FIH prediction), uses preclinical in vivo pharmacokinetic data to verify the in vitro measurements.

Preclinical verification models were able to provide predictions of first-in-human pharmacokinetics within two-fold of the observed data for all five compounds studied, which was a significant improvement compared with predictions using in silico or in vitro inputs without verification in preclinical species. In most cases, the predictions generated purely from structure were superior to those supplemented with in vitro data without preclinical validation. Prediction of human clearance via in vitro in vivo extrapolation proved challenging and was identified as the most common cause of poor predictions of human PK from in vitro data.

The work here supports the continued considered use of preclinical verification in PBPK modelling.