Abstract B214: Use of a pharmacokinetic model derived from early clinical studies to assess the relevance of covariates on cilengitide pharmacokinetics

Abstract

Abstract Background: Cilengitide is a selective competitive inhibitor of αvβ3 and αvβ5 integrins. It demonstrates anti-angiogenic, direct anti-tumor and anti-migratory properties and is under investigation for use in treating glioblastoma, cancer of the head and neck, NSCLC and other indications. The primary objective of this analysis was to characterize the population pharmacokinetic behavior of cilengitide from data obtained from three studies in patients with advanced solid cancer, NSCLC or advanced unresectable pancreatic cancer using a nonlinear mixed effects modeling technique. Methods: A stepwise approach was used on the data (992 concentration time points from 136 patients) beginning with exploratory analysis to verify the database for the three studies and target covariate relationships. A two compartment pharmacokinetic structural model was developed in NONMEM® to describe the concentration-time profile of cilengitide using initial parameter estimates determined from preliminary analysis using KINETICA®. Main individual covariates were tested to estimate their impact on PK parameters using two procedures: i) a stepwise selection procedure in NONMEM and ii) generalized additive modeling with XPOSE® software. The model was validated by testing the stability of the base and final model via a bootstrap procedure (1000 replicate data sets generated from the original data set). Results: A pharmacokinetic model was developed and validated for cilengitide from data obtained from three different studies. Study was not a significant covariate for the model. Estimated pharmacokinetic parameters for the structural model were 2.79 L/h/m2 for CL (clearance from central compartment), 6.75 L/m2, for V1 (volume of distribution for central compartment), 1.3 L/h/m2 for Q (intercompartmental clearance) and 3.85 L/m2 for V2 (volume of distribution for peripheral compartment). The half-life was estimated as 0.9 h and 3.8 h for and -phase. The model PK parameters are consistent with individual study derived PK parameters (2 to 4 L/h/m2 for mean systemic clearance (CL), 0.84 to 3.36 for renal clearance and 3 to 5 hours for terminal half-life). From the covariates analyzed (age, weight, BSA, albumin, bilirubin, serum creatinine, aspartate amino transferase, alanine amino transferase, alkaline phosphatase, hepatic status, gender and study) only weight and body surface area were highly correlated (r=0.95, p<0.01). Although the albumin and bilirubin covariates were statistically significant for the final model, their inclusion is not clinically relevant based on the information obtained from the studies. Information generated by using the model to predict cilengitide profiles in different patient populations can speed identification of optimized dosing regimens. Citation Information: Mol Cancer Ther 2009;8(12 Suppl):B214.