(T-063) Advancing Modeling of Hematologic Safety using a Semi-mechanistic Multivariate PK/PD Approach: Application to the ATR Inhibitor Tuvusertib in Early Phase Oncology Development

Paul Diderichsen, PhD – Senior Director, Certara, Certara USA, Inc., Radnor, PA, USA; Farina Hellmann, PhD – Scientist, Certara USA, Inc., Radnor, PA, USA; Ioannis Gounaris, MD, PhD – Vice President, Merck Serono Ltd., Feltham, UK, an affiliate of Merck KGaA, Darmstadt, Germany; Wei Gao, PhD – Senior Director, EMD Serono, Billerica, MA, USA; Rainer Strotmann, MD, PhD – Senior Director, The healthcare business of Merck KGaA, Darmstadt, Germany; Karthik Venkatakrishnan, PhD – Vice President, EMD Serono, Billerica, MA, USA

Background: The modeling of drug effects on myelopoiesis is widely used to inform dose and schedule optimization in drug development. While many examples exist for neutropenia, applications for describing anemia are fewer, with delayed attainment of peak drug effect on hemoglobin (HGB) introducing additional complexity when analyzing short-term data in early clinical development. Tuvusertib is a potent, selective, orally administered ATR inhibitor currently in phase1/2 clinical development. Anemia has been reported as a dose limiting toxicity during the dose escalation phase of the DDRiver Solid Tumors 301 study (NCT04170153), indicating dose and exposure dependency. To provide quantitative metrics to guide the dosing regimen selection through model-based drug development, a semi-mechanistic multivariate PK/PD model was developed to link tuvusertib exposure to hematology data.

Methods: Data was analyzed from the DDRiver Solid Tumors 301 Phase 1 dose escalation study (Part A1) [1] in patients with advanced solid tumors (N=55). A non-linear mixed effects model was used to describe the PK of tuvusertib, and subsequently, the PK/PD relationship between individual predicted tuvusertib concentrations and observed reticulocytes (RET), red blood cells (RBC), and HGB levels. Both RET and RBC data were included to dynamically reinforce the model for reliable prediction of long-term effects on HGB during multi-cycle treatment, even if peak effects may not have been observed during dose escalation.

Results: A 2-compartment population PK (popPK) model was developed with an estimated absorption lag and non-linear clearance via a “clearance compartment”. The semi-mechanistic, multivariate PK/PD model consisted of 3 serial cell systems, i.e., for progenitor cells (PC), RET, and RBC [2] (each with 4 transit compartments). The model included an EMAX drug effect on PC production, and two negative feedback mechanisms of RET and RBC counts on PC. HGB was predicted as a scaled sum of RET and RBC.
The popPK model adequately described the observed PK profile of tuvusertib. Estimated apparent central volume of distribution was 30.0 L; apparent central clearance was 55.7 L/h. In the PK/PD model, the drug effect on PC was estimated with EC50 of 736 ng/mL. Models were qualified based on standard diagnostics, evaluated using pcVPC and bootstrap, and adequately described the observed data. No statistically significant covariates were identified. Multi-cycle simulations using PK/PD model suggested that tuvusertib 180 mg QD 2w on/1w off allowed partial recovery of HGB during a one-week break; however, 180 and 130 mg QD showed no recovery and resulted in lower HGB levels than the 180 mg QD 2w on/1w off regimen.

Conclusions: The developed models adequately described the observed PK and RET, RBC and HGB data in a joint PK/PD model, allowing multi-cycle simulations of various continuous and intermittent dosing regimens to inform dose optimization strategy of tuvusertib.

Citations: 1. Yap TA, et al. Clin Cancer Res. 2024. doi: 10.1158/1078-0432.CCR-23-2409.
2. Zhang X, et al. CPT Pharmacometrics Syst Pharmacol. 2017 Dec;6(12): 804-13