P129 ASSESSING MODELING AND SIMULATION TOOLS AND METHODS FOR PREDICTING INDUCTION-BASED DRUG-DRUG INTERACTIONS: A COLLABORATIVE EFFORT BETWEEN ACADEMIC, GOVERNMENT REGULATORY, AND PHARMACEUTICAL SCIENTISTS FROM THE IQ CONSORTIUM (IQC)

Heidi Einolf , Drug Metabolism and Pharmacokinetics, Novartis Institutes for BioMedical Research Inc., East Hanover, NJ
Liangfu Chen , Drug Metabolism and Pharmacokinetics, GlaxoSmithKline, King of Prussia, PA
Odette A. Fahmi , Pharmacokinetics and Drug Metabolism, Pfizer, Groton, CT
Christopher R. Gibson , Drug Metabolism, Merck Research Laboratories, West Point, PA
R. Scott Obach , Pdm, Pfizer Global Research & Development, Groton, CT
Mohamad Shebley , Abbott Laboratories, Abbott Park, IL
Michael Sinz , Metabolism and Pharmacokinetics, Bristol-Myers Squibb Co, Wallingford, CT
Jose Silva , Drug Metabolism Pharmacokinetics, Johnson & Johnson Pharmaceutical Research and Development, LLC., Raritan, NJ
Jashvant D. Unadkat , Pharmaceutics, University of Washington, Seattle, WA
Lei Zhang , Office of Clinical Pharmacology, U.S. Food and Drug Administration, Silver Spring, MD
Ping Zhao , Office of Clinical Pharmacology, CDER, FDA, Silver Spring, MD
Drug-drug interactions (DDI) that arise by one drug causing an increase in activity of drug-metabolizing enzymes and/or transporters can result in decreased efficacy and even toxicity of a second drug.  The most common underlying biochemical mechanism for this phenomenon is the activation of the pregnane X receptor (PXR) resulting in increased transcription of several drug metabolizing enzymes and transporters, in particular cytochrome P450 3A4 (CYP3A4).  An ability to prospectively predict clinical DDI caused by induction of CYP3A4 is highly sought to aid in the research and development of new drugs.  In this collaborative study, the objective was to explore the performance of various DDI prediction models across a set of drugs for which clinical DDI data were available in the scientific literature of their interaction with selected CYP3A4 cleared drugs (e.g. midazolam).  Input used in the predictions were derived from in vitro induction studies in cultured human hepatocytes (Emax and EC50), pharmacokinetic parameters of the inducer and probe substrate, and specific aspects of the clinical study design.   The methods were judged based on two criteria: (a) overall accuracy in predicting the average magnitude of the interaction, and (b) the ability to categorize drugs into those which cause a 20% or greater decline in exposure to a CYP3A4 cleared drug vs. those that do not.  The following methods and approaches were tested: the correlation between DDI magnitude and (1)  Cmax/EC50 or (2) the relative induction score (RIS), (3) the net-effect model (“static” model), and (4) physiologically-based pharmacokinetic modeling (“dynamic” model) using the SimCYPTM program.  The findings suggest that several of these methods offer reliable predictions of DDI.  Average fold-error for the methods were in the range of 2-fold and was influenced by the type of inducer plasma concentration values used in the methods (e.g. free vs. total; systemic vs. estimated portal).  The fidelity of the methods was high with only a couple of mispredictions based on the cutoff of 20% reduction in exposure.  The practical pros and cons of the various approaches will be discussed.