Objectives: To guide recommended Phase 3 dose (RP3D) selection for a novel solid tumor oncology compound through development of a clinical utility index (CUI) that incorporated longitudinal safety and efficacy model-based clinical trial simulation output.
Methods: Longitudinal safety and tumor models were utilized for clinical trial simulation along with an adaptive dosing platform in mrgsolve1 to understand the interplay of safety events resulting in dose holds/reductions and simulated efficacy response for an oncology compound in Phase 1. A CUI was designed in R2 to incorporate the output of the clinical trial simulation (safety and efficacy endpoints of interest) to quantitatively assess several proposed dose levels – including intermediate doses not evaluated previously in clinic. Clinical trial simulation was conducted using an adaptive dosing platform. The output of the simulation was designed to mimic the expansion cohort of 40 patients, where 4000 virtual patients were simulated and split into 100 different trials. Point estimates and confidence intervals for safety and efficacy output were calculated by taking the median and 90% confidence interval of the median endpoint of interest within each simulated trial of 40 patients. Sensitivity analysis was performed on the weightings of the safety and efficacy output (e.g. 20/80 to 55/45 safety/efficacy) to understand how changes in prioritization of safety and efficacy would impact RP3D selection and inform which dose levels may not warrant further clinical testing.
Results: The adaptive dosing clinical trial simulation platform was utilized for deriving summary statistics for predicted ≥Grade 3 neutropenia rates (safety) and simulated tumor response (efficacy). The CUI initially used higher weighting of efficacy compared to the safety, given the seriousness of disease and leveraged a target product profile based on the compound characteristics and line of therapy consistent with a previously published CUI in oncology. 2 Output from other logistic regression models for additional safety endpoints were added to the longitudinal model output yielding a joint safety utility, where the same conclusions were drawn for optimal dose as when only ≥Grade 3 neutropenia was considered. The RP3D had better predicted utility compared to the lower dose levels translating to a preferable benefit-risk profile. Across the range of weightings for safety and efficacy evaluated within the CUI sensitivity analysis, the higher dose was found to be optimal in all scenarios where efficacy was weighted higher than safety.
Conclusions: The CUI output successfully supported RP3D selection for an oncology compound. The ability to use longitudinal model outputs for the CUI enabled the inclusion of dose-dependent dose modifications and safety and efficacy predictions for intermediate doses not clinically evaluated. The quantitative assessment of safety and efficacy output from the adaptive dosing simulation platform can be utilized across other oncology compounds to select regimens for clinical testing or expansion and support dose optimization.
Citations: [1] Jermain et al. ACoP, 2022. [2] Zhu et al. CPT: PSP, 2019.
