In vitro reversible enzyme inhibition assays to predict drug-drug interactions: Current state and industry...

Inhibition of drug-metabolizing enzymes such as P450s and uridine 5'-diphosphoglucuronosyltransferase is routinely evaluated in drug discovery and development. Substantial efforts have been made over the years to standardize in vitro assay conditions and data interpretation. More recently, increasing importance has been given to the approaches for improving translation of in vitro data to predict in vivo outcomes. The International Consortium for Innovation and Quality in Pharmaceutical Development (IQ) Enzyme Inhibition Working Group conducted a survey across IQ member companies to interrogate on the current state of inhibition assessment in the industry, including the strategies used as compounds progress through the pipeline, the kinetic endpoints utilized, and the accuracy of clinical drug-drug interaction predictions. Focus was also placed on how companies applied correction for unbound fraction of potential inhibitors as required by the various recent regulatory guidance documents. Results showed that most companies follow similar set-ups to identify strong inhibitors in discovery and to fully characterize drug candidates in development. Although there are some minor differences in evaluating and correcting for incubational binding, there is almost universal alignment regarding the use of kinetic endpoints with K_i = IC₅₀/2 widely accepted to determine the inhibition constant. A majority of predictions align with clinical drug-drug interaction results, although instances of over- or underprediction were reported, and examples of these are discussed here as case studies. SIGNIFICANCE STATEMENT: Practices utilized to assess reversible inhibition of drug-metabolizing enzymes were evaluated based on 25 survey responses from IQ member companies. Results show how assay conditions, endpoints, and modeling approaches evolve as compounds advance through discovery to development. The accuracy of drug-drug interaction prediction is discussed, and recommendations for best practices are provided.