Population PK and exposure-response modeling to support the design of trials with Debio 0123, a WEE1 inhibitor, in cancer patients
Vito Dozio (1), Philippe B. Pierrillas (2), Kyriakos P. Papadopoulos (3), Hans Gelderblom (4), Sandrine Micallef (1), Camille Riff (1), Hashim Michla (1), Tri Tat (1), Marie-Claude Roubaudi-Fraschini (1), Noemie Luong (1), Mokhtar Omar (1), Rikke Frederiksen Franzen (1), Melanie Wirth (1), Caroline Mathon (1), Victor Rodriguez Freixinos (1), Esteban Rodrigo Imedio (1), Anne Bellon (1)
(1) Debiopharm International SA, Lausanne, Switzerland, (2) Certara, Basel, Switzerland, (3) START San Antonio, TX, USA, (4) Department of Medical Oncology, Leiden University Medical Center, Leiden, The Netherlands
Introduction: Debio 0123 (D0123) is an oral, brain-penetrant, highly selective WEE1 inhibitor that has demonstrated significant cell growth inhibition in vitro and anti-tumor activity in vivo. Preliminary anti-tumor activity was also observed with the combination of D0123 and carboplatin (CBDCA) in patients (pts) with solid tumors who progressed with prior platinum-based chemotherapy.
D0123 is currently in Phase 1 clinical development in pts with advanced solid tumors, both as monotherapy and in combination with other therapeutic agents. Results from the dose escalation part of two trials were used to develop the population pharmacokinetics (popPK) model: D0123-101 (NCT03968653) evaluating two dosing schedules of D0123 in combination with CBDCA in 21-day cycles, D0123-102 (NCT05109975) evaluating D0123 when administered as monotherapy with repeated dosing.
Objectives: A popPK model was built to characterize the pharmacokinetics (PK) of D0123 and its major active metabolite (M1) when administered as monotherapy or in combination with CBDCA, and to quantify sources of PK variability in pts. The popPK model was combined with exposure-response analyses from the two trials to support the dosing schedule of further D0123 trials.
Methods: PK data from 82 pts with advanced solid tumors were included in the analysis. Nine dose levels were evaluated from 30 mg to 720 mg and more than 4800 observations were included in the dataset. The popPK analysis was performed using a nonlinear mixed-effects modeling approach with NONMEM. First-order conditional estimation method with interaction and user-defined subroutine ADVAN6 were used to estimate the population typical values of the PK parameters, interindividual variability, and residual variability for D0123 and M1. Body weight, age, body mass index, gender, renal function, ethnicity, co-administration of CBDCA, and cancer type were tested as potential covariates. Visual predictive checks and bootstraps were used to evaluate the predictive performance and stability of the final model.
D0123 target engagement was assessed based on the phosphorylation levels of the cell division cycle protein 2 (pCDC2), the downstream target of WEE1, in skin biopsies at baseline and after treatment. Exposure-response analysis was conducted using regression analysis relating exposure (AUC over the first cycle) to the change from baseline of pCDC2.
Results: A two-compartment model with additive residual error on log scale was shown to best characterize D0123 and M1 plasma concentrations. A one-transit compartment model was used to describe D0123 absorption process while a first-pass effect was necessary to model M1 formation. To rule out the occurrence of parameter identifiability issues, the value of the central volume of distribution of M1 was fixed to the value of the parent. The fraction of metabolite was fixed to 19% of the total drug molar concentration. The preliminary population parameters (and inter-individual variability, IIV) were estimated to CL/F= 11.2 L/h (12.7%), Vc /F= 412 L (62%), Ka= 0.371 h-1 (26.7%) for D0123 and as CL/F= 20.5 L/h for M1. Covariate analysis showed that gender and CBDCA co-administration were two statistically significant variables affecting exposure. The exposure-response analysis highlighted a non-linear relationship between D0123 exposure over the first cycle and decrease in pCDC2 levels.
Conclusion and perspectives: The popPK model captured well the typical PK and inter-individual variability of D0123 and its major active metabolite following oral administration in 82 pts with solid tumors enrolled from two trials with three dosing schedules and 12 dose levels. It delineated the influence of significant covariates such as co-administration with CBDCA and was used for population simulations to predict the range of exposures for different doses and schedules. The relationship between individual predicted exposures and the related pharmacodynamic response is being used to support the selection of dosing regimens for ongoing and future trials.