Prediction of the outcome of a phase 3 clinical trial of an antischizophrenic agent (quetiapine fumarate) by simulation with a population pharmacokinetic and pharmacodynamic model.

Abstract

A completed phase 3 trial result was simulated 100 times on the basis of a simulation model of quetiapine fumarate (Seroquel), an antischizophrenic agent. The simulation was executed by analysts who were completely blinded from results of the actual trial until after the simulations were submitted to the holder of the trial results. Data from two clinical investigations of quetiapine in patients with schizophrenia were analyzed by use of nonlinear mixed effects modeling to derive a population pharmacokinetic- and pharmacodynamic-based simulation model. The time course of quetiapine concentrations was described by use of a one-compartment open linear pharmacokinetic model with first-order absorption and elimination. The combination of an inhibitory maximum effect pharmacodynamic model for the active treatment effect and a linear function of time for the placebo effect characterized the observed time course of change in the Brief Psychiatric Rating Scale. Simulation results were compared with those in the actual trial to evaluate how well the simulations predicted the outcome. The actual trial results for all doses except the placebo group fell within the predicted Brief Psychiatric Rating Scale scores +/- 1 SE. Unlike the phase 2 trial, from which the pharmacokinetic/pharmacodynamic model was developed, the placebo group in the actual phase 3 trial showed deterioration of Brief Psychiatric Rating Scale scores with time. We conclude that variable placebo responses observed in short-term studies of schizophrenia provide an inadequate basis for the modeling and simulation of placebo subjects in clinical trials. Knowledge of the range of placebo response observed in other studies may have provided an improved basis for the placebo effect model. The model for active drug produced adequate predictions of the actual trial outcomes.