Integrating distribution to tumor tissue into a dynamic PK/PD model to evaluate the anti-cancer effect of erlotinib in patient-derived LXFA 677 tumor xenograft mice
Nicolas Frances, Christophe Meille, Gerhard Hoffmann, Thierry Lavé, Antje Walz
Preclinical pharmacokinetic/pharmacodynamic modeling and simulation, F. Hoffmann-La Roche Ltd, 4070 Basel, Switzerland
Objectives: Development of a dynamic PK/PD model to describe the anticancer effect of erlotinib in patient-derived LXFA 677 tumor xenograft mice as a function of drug concentration in tumor tissue.
Methods: Two independent experiments were designed in female NMRI nu/nu mice implanted with human LXFA677 primary patient tumors. For assessing tumor growth inhibition, a repeated oral dose study with 100, 25, 6.25 mg/kg/d of erlotinib was conducted. Tumor volume was monitored twice weekly during and after drug treatment and a sparse plasma PK sampling scheme (2 observations per mouse) was applied. In a second study aiming to assess the drug distribution to the tumor tissue, a staggered sampling approach was applied. For each mouse, a single sampling time point for drug concentration in plasma and tumor was obtained after oral administration of 100 mg/kg/d for single and repeated dose.
As a first step, the area under the tumor growth curve was described as a function of the cumulative dose. Next we developed a dynamic PK/PD model relating the time course of the tumor volume to the exposure in the tumor. Concentration in the tumor was directly linked to the effect. Population analyses were performed using MONOLIX v3.2  and simulation analyses using Matlab v9b.
Results: The distribution of erlotinib was described by a ‘hybrid' PK model consisting of a one compartment model describing the plasma erlotinib concentration profile and a tumor compartment describing the tumor concentration profile. The parameters estimated from this model included clearance (CL), volume of distribution of the central compartment (Vc), absorption rate constant (ka), inter-compartmental Clearance (Q) and volume of distribution in the tumor compartment (Vt). Tumor growth was described by a Gompertz model. The drug effect was described by an interface model following signal transduction model.
Conclusions: Linking the effect to the exposure at the tumor compartment improved the PKPD modeling first by accounting separately for the delay due to distribution to the biophase and the delay triggered by biological cascade. Second, the anticancer effect can be related to any kind of PK profile in the biophase.
 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.