Mélanie Wilbaux (1) and David Demanse (1), Astrid Jullion (1), Yi (Gary) Gu (2), Andrea Myers (2), Christophe Meille (1)
(1) Novartis, Basel, Switzerland; (2) Novartis, Shanghai, China
Objectives:
(i) Develop a PKPD model of longitudinal tumor size in hepatocellular carcinoma (HCC) patients receiving FGF401, an oral FGFR4 kinase inhibitor under evaluation in a phase I/II study (NCT02325739); (ii) identify clinical baseline characteristics and derive a continuous risk score predictive of time to progression (TTP) using multivariate analysis; (iii) evaluate this score as a covariate on PKPD model parameters, and (iv) simulate tumor kinetics at different patients characteristics and support dosing regimen optimization for future studies.
Methods:
Plasma concentrations (n=2188) time course, longitudinal tumor size (n=273) and individual baseline characteristics data were collected from 71 HCC patients, with FGF401 once-daily doses from 50 to 150 mg. Pharmacokinetics and tumor growth models were developed using non-linear mixed-effects modeling implemented in Monolix 2016R1. A sequential PKPD model approach was used with PK parameters fixed at an individual level in the PKPD modeling analysis. A total of 82 patients’ baseline characteristics were included to develop a parsimonious predictive model for TTP using robust statistical approach with cross-validation methodology [1]; All statistical computations were carried out in R-3.2.3. A continuous risk score was derived and evaluated as a covariate on key parameters of the PKPD tumor growth inhibition model. Final PKPD model evaluation and selection were based on statistical criteria, goodness-of-fit plots and simulations based diagnostics. The final PKPD model was implemented in a Shiny application [2, 3] to simulate tumor growth inhibition at different doses, regimens and given patients’ baseline characteristics.
Results:
A two compartment model with a delayed 0-order absorption and linear elimination was adequately describing PK data. Unperturbed tumor growth was best characterized by a fist-order process. Plasmatic PK was linked to the tumor-killing rate through an effect compartment to reproduce a delay before drug effect. A resistance component with parameter λ was added to describe the tumor regrowth under treatment [4]. Multivariate analysis resulted in three baseline predictive factors of TTP: (1) adjacent organ invasion (e.g. gallbladder and peritoneum), (2) number of target lesions, and (3) number of metastatic sites. The dose was also found as a predictive factor of TTP, reinforcing the value of a PKPD model development. A continuous risk score was derived excluding dose to avoid any bias with PKPD model integration. The risk score was found to be a significant covariate on the resistance parameter λ (p=5e-05) and baseline tumor size (p=6e-07). Inter-individual variability on λ parameter was reduced from ω = 0.7 to ω = 0.5. All goodness of fit plots were improved after inclusion of the risk score. Simulations from the PKPD tumor response model supported the selection of 120mg daily as the recommended dosing regimen for future studies.
Conclusions:
Multivariate analysis using a robust statistical approach suggested a possible set of covariates for a PKPD tumor response model. The multivariate composite risk score was highly significant as a covariate on the resistance parameter, resulting in reduction of variability and PKPD model improvement. The final PKPD model implemented in a simulation tool was used to support dosing regimen selection for given patient populations. The proposed methodology, combining multivariate analysis and PKPD modeling on related endpoints (e.g. TTP and tumor size), can be applied for other efficacy and safety endpoints.
References:
[1] Bøvelstad HM, Nygård S, Størvold HL, Aldrin M, Borgan Ø, Frigessi A, Lingjærde OC. Predicting survival from microarray data—a comparative study. Bioinformatics. (2007) 15;23(16):2080-7
[2] https://CRAN.R-project.org/package=shiny
[3] https://CRAN.R-project.org/package=mlxR
[4] Claret L, Girard P, Hoff PM, Van Cutsem E, Zuideveld KP, Jorga K, Fagerberg J, Bruno R. Model-based prediction of phase III overall survival in colorectal cancer on the basis of phase II tumor dynamics. J Clin Oncol. (2009) 27(25):4103-8
Reference: PAGE 27 (2018) Abstr 8641 [www.page-meeting.org/?abstract=8641]
Poster: Methodology - Covariate/Variability Models