Neha Thakre 1, Burak Kuersad Guenhan 1, Andreas Becker 1, Sathej Gopalakrishnan 1, Stuart M Bailey 2, Jatinder Kaur Mukker
1 Merck KGaA (Darmstadt, Germany), 2 EMD Sereno (Billerica, USA)
Objective: Traditional oncology dose-finding trials primarily focus on dose-limiting toxicity (DLT) data to guide dose-escalation decisions. Bayesian Logistic Regression Models (BLRM) have been utilized and recommended to guide dose escalation in combination treatment dose-finding trials [1]. Integration of pharmacokinetic/pharmacodynamic (PKPD) relationships with dose-limiting toxicity (DLT) data and BLRM model can enhance dose selection strategies [2]. Selecting combination doses in oncology is especially challenging when adverse events are expected to overlap; here, we consider a combination dose-finding trial investigating an experimental drug with cytotoxic chemotherapy. The objective is to establish a framework to leverage a population PKPD model for hematological parameter such as absolute neutrophil count (ANC) to inform the prior distribution of the interaction parameter used in the BLRM model.
Methods: A nonlinear mixed effects population PKPD model characterizing the relationship between drug exposure and ANC, accounting for inter-individual variability was developed using data from patients treated with monotherapy and combination regimens. Simulated probabilities of neutropenia across dosing regimens were calculated. A dual-agent BLRM model was developed for the combination dose-finding trial with an interaction parameter. An informative prior is used for the interaction parameter. This enabled the incorporation of both DLT and dynamics of ANC suppression in the BLRM to guide dose escalation decisions and estimate the optimal dose range. Additionally, an informative prior was constructed based on the irinotecan phase I dose-escalation trial [3].
Results: The population PKPD model adequately described ANC time-course across patients and dosing regimens. Model-based simulations predicted dose-dependent neutropenia risk, and probabilities of Grade 3/4 neutropenia were determined. The collaborative approach enabled real-time dose adjustments based on both toxicity and PK/PD endpoints.
Conclusions: Integration of PKPD modeling with BLRM represents a paradigm shift from DLT-centric to exposure-response informed dose selection during phase-1 escalation steps. Similar framework can be extended to other adverse events in combination drug development. This quantitative pharmacology-biostatistics collaboration enhances dose-finding efficiency by incorporating mechanistic understanding of drug-induced adverse events, demonstrating the value of model-informed drug development (MIDD) in early-phase oncology trials.
References:
[1] Neuenschwander, B et al. (2015). A Bayesian industry approach to phase I combination trials in oncology. In Statistical methods in drug combination studies.
[2] Bailey, S et al. (2018) “Beyond MTD: Integrating non-safety endpoints into Oncology dose-finding” FDA-ISoP Public Workshop – MIDD for Oncology
[3] Rothenberg ML, et al (2001). Phase I dose-finding and pharmacokinetic trial of irinotecan (CPT-11) administered every two weeks. Ann Oncol.
Reference: PAGE 34 (2026) Abstr 12102 [www.page-meeting.org/?abstract=12102]
Poster: Methodology - Other topics