Arwa Alghaith 1, Karen Strathdee 2, Katrina Stevenson 2, Anthony J. Chalmers 2, Joseph F. Standing 3, Jamie A. Dean 1
1 Department of Medical Physics and Biomedical Engineering, University College London (, UK), 2 Institute of Cancer Sciences, University of Glasgow (, UK), 3 Institute of Child Health, University College London (, UK)
Introduction: Clinical studies combining molecularly-targeted drugs with radiotherapy (RT) have largely failed to improve patient outcomes in cancer. A potential reason is suboptimal treatment scheduling. Often, schedules are selected based on previous monotherapy or combination regimens with chemotherapy. Mathematical modelling can be used to explore alternative schedules to enhance synergy between therapies. PARP inhibitor-RT combinations inhibiting repair of RT-induced DNA damage are currently in clinical trials. Olaparib has demonstrated activity with RT in preclinical studies. However, schedules are typically selected without quantitative assessment of optimal drug-RT scheduling. Drug-RT intervals are often unspecified or aligned with peak plasma concentration. We investigated the impact of dosing intervals and OD versus BID on olaparib-RT efficacy to inform future trials.
Objectives: To (i) develop a Bayesian mixed-effects joint PKPD model to predict olaparib exposure in plasma and tumour, and its effect on RT-induced DNA damage, in a glioblastoma patient-derived xenograft model; and (ii) apply this PKPD model to predict the efficacy of different olaparib-RT dosing schedules, and experimentally validate predictions in mice.
Methods: A Bayesian PKPD model was developed in Stan. PK component was a three-compartment oral model (depot, central, tumour, peripheral). PD component modelled the fraction of gamma-H2AX positive cells as a baseline of DNA damage, with transient increases induced by RT. DNA damage repair was modelled mono-exponentially, with the repair rate inhibited by tumour olaparib concentration via a saturable Emax function. Inter-individual variability was modelled for all structural PK parameters and RT-induced damage magnitudes; log-normal and beta residual error models were used for concentration and gamma-H2AX data, respectively. The model was fit to plasma and tumour olaparib concentrations, and gamma-H2AX levels from irradiated mice (olaparib administered 0.5h pre-RT). Concentrations were obtained after single oral doses of olaparib 15 mg/kg (n=12) or 30 mg/kg (n=12), sampled at 1, 2, 4h post-administration. Gamma-H2AX levels were obtained from RT-alone (4 Gy; n=15) and olaparib-RT-treated mice with single oral doses of 15 mg/kg (n=12) or 30 mg/kg (n=12), sampled at 0, 1, 2, 4h post-RT. To predict the efficacy of different olaparib-RT schedules, PKPD simulations were performed for 15 virtual mice with parameters sampled from estimated population distributions. RT schedule was 6 x 2 Gy (Mon/Wed/Fri; 2-weeks). Simulations assessed the impact of olaparib-RT intervals and OD vs BID dosing. All schedules were compared using plasma and tumour Cmax at RT, gamma-H2AX AUC and peak values. The predicted efficacies were validated in a 2-week survival study. Mice (n=15/arm) received RT 3x/week as RT alone (4 Gy) or RT (2 Gy) with olaparib: 50 mg/kg OD 0.5h pre-RT, 50 mg/kg OD 6h pre-RT, or 25 mg/kg BID with the first dose ~4 h pre-RT.
Results: The model accurately captured olaparib plasma and tumour PK, and gamma-H2AX dynamics for RT-alone and olaparib-RT. Shorter drug-RT intervals produced higher tumour concentrations at RT: highest mean tumour Cmax was 1.08 µM at 0.5h pre-RT, declining to 1.03 (1h), 0.63 (2h), 0.24 (4h), 0.12 (6h), 0.06 µM (8h); an 18-fold range. Plasma Cmax ranged from 13.06 (0.5h) to 0.27 (8h) µM. Optimal interval (0.5h) was selected to compare OD and BID schedules; OD (50 mg/kg 0.5h pre-RT) yielded tumour Cmax 1.06 µM; BID (25 mg/kg 0.5h pre-RT and 3h post-RT) yielded lower tumour Cmax at RT (0.53µM). The experimental validation study showed no significant difference in survival between the OD and BID schedules, but a modest trend favouring shorter pre-RT intervals. At 100 mg/kg 1h pre-RT, simulated tumour Cmax was 2.04 µM, below the 3 µM concentration used in key in vitro assays which showed radiosensitisation effects (Guo et al. 2025). Similar gamma-H2AX peaks and AUC seen across schedules (~0.986, 1.00-1.01).
Conclusions: A Bayesian joint PKPD model linking olaparib tumour PK to gamma-H2AX dynamics was developed and applied to evaluate olaparib-RT scheduling. Simulations predicted that minimising the drug-RT interval maximises tumour exposure at RT, but tumour concentrations achieved are insufficient to significantly alter gamma-H2AX AUC/peaks compared with RT alone. Simulated tumour exposures versus published in vitro radiosensitisation experiments suggest tested schedules may be below effective concentrations, providing a plausible explanation for minimal PD effects despite exposure differences. Plasma PK alone may not reliably reflect tumour exposure at RT, DNA damage effects, or in vitro-active ranges, supporting tumour-PK-informed schedule evaluation for PARP inhibitor-RT combinations.
References:
Guo Y, Li Z, Parsels LA, Wang Z, Parsels JD, Dalvi A, The S, Hu N, Valvo VM, Doherty R, Peterson E, Wang X, Venkataraman S, Agnihotri S, Venneti S, Wahl DR, Green MD, Lawrence TS, Koschmann C, Morgan MA, Zhang Q. H3K27M diffuse midline glioma is homologous recombination defective and sensitized to radiotherapy and NK cell-mediated antitumor immunity by PARP inhibition. Neuro-Oncology. 2025;27(8):2129–2146. doi: 10.1093/neuonc/noaf097.
Reference: PAGE 34 (2026) Abstr 11948 [www.page-meeting.org/?abstract=11948]
Poster: Drug/Disease Modelling - Oncology