INTRODUCTION

1. DEVELOPMENT OF AN INTEGRATED POPULATION PHARMACOKINETIC MODEL FOR IMMEDIATE RELEASE AND CONTROLLED RELEASE FORMULATIONS OF AN INVESTIGATIONAL COMPOUND IN HEALTHY SUBJECTS AND SCHIZOPHRENIA PATIENTS Vikas Kumar*1, Kevin Sweeney1, Diane Mould2, Jing Liu11Pfizer Global Research and Development, Groton/New London, CT 06320, 2Projections Research Inc., Phoenixville, PA RESULTS RESULTS • INTRODUCTION • Schizophrenia is a severe, chronic mental illness and ranked as fifth for the leading cause of years lived with a disability by WHO • Available treatments for schizophrenia are often only partially effective; residual symptoms are present in the majority of stabilized patients • PF‑XYZ is being developed to treat negative and cognitive symptoms in adjunctive treatment of patients with schizophrenia • PF-XYZ is currently in Phase-II development for schizophrenia • Currently, both immediate release (IR) and controlled release (CR) formulations are used for the development of PF-XYZ • Data from six studies (FIH to Phase-IIa stage) is used for the development of PopPK model Individual Model Predicted Clearance vs Patient Status (Left) and CYP2D6 Metabolizer Status (Right) Population PK Parameter Estimates from Final Model OBJECTIVES • To develop an integrated population PK model for immediate release and controlled release formulations of PF-XYZ • To describe PF-XYZ pharmacokinetics in healthy subjects and stable chronic schizophrenia patients on background therapy of second generation of antipsychotics • To evaluate intrinsic and extrinsic covariates that may be important determinants of PF-XYZ exposure • o provide post-hoc predictions from the PopPK model that will be used in sequential PK/PD analyses • Food was estimated to increase the bioavailability of CR formulation by 18% • High correlation (91%) was observed between random effects of CL/F and Vc/F (Figure below); therefore shared eta was used in final model for parsimonious  structure • No difference in CL/F was observed between HV and patient population STUDIES INCLUDED FOR ANALYSIS PHARMACOKINETIC MODEL METHODS & MODELING STRATEGY DIAGNOSTIC PLOTS VISUAL PREDICTIVE CHECK Goodness of Fit Plots: Observed vs Predicted in Normal Scale (Left) and Log Scale (Right): Final Model Visual Predictive Check using Observed Concentrations from One Dose (symbols) and 90% Prediction Interval (shaded area) The population PK model was developed by pooling the data from four phase I studies and two studies in schizophrenia patients. Pharmacokinetic models were fit to the data using First Order Conditional Estimation (FOCE) method in NONMEM version 7. Both one- and two-compartment models were evaluated. Absorption profiles obtained from IR and CR formulations were modeled simultaneously. First order absorption with or without lag time was explored for IR formulation and a transit compartment model was used for the CR formulation. As PF-XYZ is primarily metabolized by CYP2D6, the effect of CYP2D6 phenotype was added in the base model. For the final model development, various covariates-parameter were explored graphically and then added to the final model with care to avoid co-linearity in predictors. In the final model, one dichotomous covariate (gender) was included to quantify the impact on CL/F. Also, an allometric model with weight normalized to 70 kg was used for the estimation of CL/F, Vc/F, Vp/F and Q/F. Proportional or additive + proportional error models for both normal and log transformed data were explored for residual variability. Log normal distributions were used to characterize the inter-individual variability of pharmacokinetic parameters. Selection of the appropriate structural, variance, and error models was based on the change in the objective function value, visual inspection of goodness of fit plots, precision of the parameter estimates, as well as decreases in both between‑subject variability and residual variability. The final population PK model was evaluated using visual predictive check. Goodness of Fit Plots: CWRES vs Time (Left: Final Model) & Correlation of ETAs (CL/F vs Vc/F) (Right: Base Model) CONCLUSIONS • Overall, the final population PK model provided a good description of the IR and CR data • No differences were observed in the PK of healthy subjects and patient population • CYP2D6 phenotype and gender were found to be significant covariates for oral clearance • Food increases the bioavailability of PF-XYZ CR formulation by 18% • Predicted concentrations from the final PopPK model were used to build sequential PK-PD model