PK/PD Model of skin toxicity in animal reported as binary outcome
Christophe Meille, Antje-Christine Walz, Koji Yamaguchi and Thierry Lavé
1)Preclinical pharmacokinetic/pharmacodynamic modeling and simulation, F. Hoffmann-La Roche Ltd, 4070 Basel, Switzerland 2)Preclinical Research Department., Research Division, Chugai Pharmaceutical Co., Ltd., 1-135 Komakado, Gotemba, Shizuoka 412-8513, Japan
Objectives: Develop a PK/PD model for skin toxicity in animal and provide a quantitative basis for preclinical safety assessment.
Material and Methods: The PK and tolerability of drug X targeting the ERK pathway was investigated in rats (n = 150) in a repeated oral dose toxicology study with 4 week recovery period. Drug X was administered once daily for 4 weeks orally by gavage at dosage levels 0.25, 1, 4, 16 mg/kg/day. For each individual, the skin toxicity was reported daily as a binary outcome. The number of incidences of skin toxicity was dependant of the administration protocol. Decrease of skin toxicity frequency was observed during the recovery period. PK samples were collected in a satellite group on the first and last administered dose. A PK/PD model was used for describing damage kinetics function of time and binary outcome. The structural model was built of 1) the PK model, 2) the skin damage model and 3) the probabilistic model. For PK a 1 compartment model with first order absorption and elimination was used. The skin damage model is represented by an indirect response model. In this model, concentration of the drug blocks damage compartment elimination through an Imax model (Dc50 parameter). The damage value is the input function to a Logit model which describes the probability of the toxic event function of the damage value. Model parameters were identified to the observed data using population approach with Monolix 3.2 software [1].
Results: All PK and PD parameters were well identified, with a Dc50 of 0.1 ug/ml and kout of 0.0012 h-1. Gender was identified as a relevant covariate on clearance. An overlay of the predicted probability of skin toxicity and observed frequency for each dosing group showed the model flexibility to describe the observations. Simulations are done to show risk profile for different protocols.
Conclusion: This example shows the fit of a PKPD model on binary outcome data. Logit model can easily be extended to more categories, describing different grades with ordered categorical data. The model can be applied across various species.
References:
[1] Kuhn E., Lavielle M. "Maximum likelihood estimation in nonlinear mixed effects models" Computational Statistics and Data Analysis, vol. 49, No. 4, pp 1020-1038, 2005.
