Population K-PD modeling of CA125 and tumor size kinetics in relapsed ovarian cancer patients
Mélanie Wilbaux (1), Benoit You (1), Olivier Colomban (1), Amit Oza (2), Gilles Freyer (1), Emilie Henin (1), Michel Tod (1)
(1) EMR 3738 CTO, UCBL - HCL Faculté de Médecine Lyon-Sud, Université Lyon 1, Oullins, France, (2) NCIG – CTG, Princess Margaret Hospital, Toronto, Canada
Objectives: Ovarian cancer (OC) is the leading cause of death among cases of gynecologic cancer. Indeed, the majority of OC patients presents with advanced stage of disease at diagnosis's time, and most of them relapse. CA 125 (Cancer Antigen) is used as a biomarker for epithelial ovarian cancer, representing 90% of all ovarian cancers' types.
The aim of this work is to characterise CA 125 and tumor size kinetics using a K-PD model with a population nonlinear mixed-effects modeling approach.
Methods: Data: Data stems from the database of CALYPSO trial: a randomized phase III study, comparing two chemotherapies, the standard's treatment Carboplatin-Paclitaxel (CP) versus Carboplatin-Pegylated Liposomal Doxorubicin (C-PLD), in relapsed ovarian cancer. Medians of 10 CA 125 concentrations values and 4 tumor size observations per subject were assessed for 533 OC patients receiving chemotherapy. Moreover, some covariates were available.
Model: Due to absence of PK data, CA 125 concentrations and tumor sizes were described by a Kinetic-Pharmacodynamic model (KPD model). The population analysis was performed using the non-linear mixed effects modeling approach implemented in Monolix 3.2. Criterion used for selection of best model were the likelihood ratio test for nested models, and AIC for non-nested models. Model evaluation was performed using classical Goodness Of Fits plots and simulation-based diagnostics.
Results: The PK was described by two virtual compartments: one central compartment, receiving the dose of treatment, and one transit compartment. The PD was an indirect effect model with inhibition of tumor growth, which was linked with to a CA125 compartment. Eight random parameters were estimated, with a low standard error for each random effect. In addition, 2 covariates were found: lesion size and number of cycles. The CA125 half life was 1.16 days. The individual estimation's median of the random parameter Q50, describing treatment potency, was equal to 0.388 for C-P, and 0.456 for C-PLD.
Conclusion: A joint model for describing CA 125 and tumor size profiles was built, and some covariates were found, in relapsed ovarian cancer patients receiving chemotherapy. This K-PD model is the first characterizing CA 125 and tumor size variations. It leads to better understanding of CA125 as biomarker and its use for treatment monitoring and evaluation.