(M-071) Dose optimization framework to estimate benefit-risk for oncology compounds: an integrated population model framework incorporating PK, safety, efficacy and impact of dose modifications
Senior Director, Pharmacometrics - Oncology Pfizer Inc San Diego, California, United States
Disclosure(s):
Joanna C. Masters: No relevant disclosure to display
Objectives: To guide dose optimization for a novel oncology compound using a multi-model-informed drug development (MIDD) framework which integrated population models of PK along with longitudinal exposure-response of safety and efficacy for predicting clinical utility across dose levels, while adjusting for expected dose modifications through adaptive dosing trial simulations.
Methods: Utilizing dose escalation and expansion data, a popPK model, longitudinal ER models for tumor dynamics and neutropenia, and logistic regression models for safety and efficacy endpoints of interest were built and refined throughout early-stage development. Tumor size and derived response rate (≥ 30% decrease from baseline) were markers of efficacy, and the model was refined for the indication of interest. Neutropenia was the adverse event of interest linked to dose modifications, with the model being refined and used in an adaptive dosing approach where future dosing events are dictated by the neutropenia event predictions from the longitudinal PK-neutropenia model to more closely mimic clinical experience [1]. Additional endpoints modeled via logistic regression were excluded from the integrated framework if they did not yield ER relationships or were less clinically meaningful. The adaptive dosing clinical trial simulations allowed for various starting doses with associated dose modification schema to be simulated according to the expected safety profile, and the subsequent concentrations were used to predict efficacy through the tumor dynamic ER model at 24 weeks. To compare the predicted benefit-risk of various starting dose regimens (including those untested clinically), the clinical utility index was calculated based on the rates of simulated ≥Grade 3 neutropenia and tumor response (derived from the adaptive-dosing based longitudinal models) as the key safety and efficacy endpoints. The clinical utility index of each dosing regimen was calculated dependent on desired weighting of efficacy vs safety [2].
Results: The framework was successfully implemented utilizing dose escalation and expansion data from Phase 1 to select the recommended Phase 3 dose (RP3D) of the compound based on totality of data for benefit-risk across various doses including potential intermediate starting doses not evaluated in the study. Reliable models were built and the simulated rates of ≥Grade 3 neutropenia and tumor response were generally aligned with clinical observations, where available, providing confidence that the utility index reflected trends observed, and was reliable to compare benefit-risk across doses for RP3D selection.
Conclusions: This integrated MIDD framework with adaptive dosing presents a more realistic and integrated quantitative approach to dose selection while leveraging the totality of relevant available clinical data and evaluating the potential impact of dosing modifications. It is particularly useful for oncology compounds in early phase development but can be applied to other indications or stages of development and may be expanded to incorporate additional models/endpoints or aid in dose selection for novel-novel combinations.
Citations: [1] Jermain B, et al. ACoP13 2022. [2] Zhu R, et al. CPT Pharmacometrics Syst Pharmacol. 2019 Apr;8(4):240-248.
