Modeling and Simulation of Placebo Response and Dropout Patterns in Treatment of Schizophrenia
Venkatesh Pilla Reddy*(1), Magdalena Kozielska(1), Martin Johnson(1), An Vermeulen(2), Rik de Greef(3), Jing Liu(4), Geny M.M. Groothuis(5), Meindert Danhof(6) and Johannes H. Proost(1)
(1) Dept. of Pharmacokinetics, Toxicology and Targeting, University of Groningen, The Netherlands, (2) Advanced PK/PD Modeling and Simulation, Johnson & Johnson Pharmaceutical Research and Development, Beerse, Belgium, (3) Pharmacokinetics, Pharmacodynamics & Pharmacometrics (P3), Merck Research Labs, Oss, The Netherlands, (4) Pfizer Global Research and Development, New London, CT 06320, USA, (5) Leiden/Amsterdam Center For Drug Research, Dept. of Pharmacology, Leiden, The Netherlands
Objectives: Unpredictable variation in placebo response within and among clinical trials can substantially affect conclusions about the efficacy of new antipsychotic medications. Developing a robust placebo model accounting for factors like dropouts, disease progression and trial design is crucial in order to facilitate better quantification of drug effect. The objectives of this study were i) to develop a model for placebo response in schizophrenia as measured with the Positive and Negative Syndrome Scale (PANSS) under varying clinical trial conditions, accounting for dropout and other relevant predictors of the placebo response, ii) to compare different Time to Event (TTE) modeling approaches used to describe the dropout patterns following placebo treatment in schizophrenia.
Methods: Pooled PANSS data from 15 clinical trials (n=1338), which included both acute and chronic schizophrenic patients with different study periods (6, 8, 12 and 54 weeks), were used to describe the time course of PANSS using NONMEM VI. Several placebo models with determinants of placebo response were tested . Influence of dropouts was investigated by exploring three TTE hazard models in conjunction with the best performing placebo models. Different patterns of dropping out were examined by †exploring Missing Completely At Random (MCAR), where dropout does not depend on PANSS; Missing At Random (MAR), where dropout depends on last observed PANSS; Missing Not At Random (MNAR), where dropout depends on predicted PANSS .
Results: The Weibull model†and an indirect response model (IRM) with exponential infusion type of kinetic-PD function described the PANSS data well compared to other placebo models. Proportional, Weibull and Gompertz hazard models (GHM) performed equally well for short-term trials, while for long-term trials and for the entire pooled dataset, GHM was shown to be superior. Preliminary covariate analysis indicated that females, subjects in long-term studies and chronic patients had lower probability of dropping out from trials compared to males, subjects of short-term studies and acute schizophrenic patients.
Conclusions: Placebo-associated change in PANSS was well described by Weibull and the IRM model. Results of simultaneous modeling of dropout model with placebo model indicated that the probability of patients dropping out from a clinical trial is associated with both the last observed PANSS measurement and unobserved PANSS score as predicted by the placebo model.
 Friberg LE, de Greef R, Kerbusch T, Karlsson MO. Modeling and simulation of the time course of asenapine exposure response and dropout patterns in acute schizophrenia. Clin. Pharmacol. Ther. 2009;86:84-91
 Hu C, Sale M. A joint model for nonlinear longitudinal data with informative dropout. J. Pharmacokinet. Pharmacodyn. 2003;30:82-103