Matteo Gazzin1, Pauline Bambury1, Niccolò Totis2, Giuseppe Pasculli1, Fianne Sips2, Mario Torchia1, Jane Knöchel1, Daniel Röshammar1, Jin Y Jin3, Brendan Bender3, Nastya Kassir3
1InSilicoTrials Technologies S.p.A., 2InSilicoTrials Technologies B.V., 3Genentech Inc.
Objectives: This work aims to illustrate the value of using InSilicoTrial’s clinical trial simulator for optimizing oncology study designs, using a combined workflow of previously developed PKPD models for trastuzumab emtansine (T-DM1) efficacy and safety in HER2-positive metastatic breast cancer patients. Introduction: InSilicoTrial’s drug development platform allows users to integrate data across multiple sources such as preclinical studies, clinical trials, publicly available information from scientific literature and real-world evidence with pharmaco-statistical models from sponsors, academic collaborators, or developed in-house resulting to guide the next phase of drug development. A comprehensive set of modeling methodologies are supported by this platform, including quantitative systems pharmacology (QSP), physiologically based pharmacokinetics (PBPK), pharmacokinetic/pharmacodynamic (PK/PD), dose-exposure-response, AI and machine learning (ML) models. When utilized for clinical trial simulation, the platform allows explorations of alternative dosing strategies, sample size calculations, as well as the impact of various patients’ characteristics and subgroups. The platform supports the generation of synthetic patient disease progression profiles to augment or replace traditional clinical trial arms as well as for enriched study designs. Methods: The capabilities of the platform are here demonstrated through the tool’s application in optimizing an adaptive oncology Phase 3 clinical study design. Longitudinal T-DM1 PK/PD models for drug exposure, tumor growth, progression-free survival and platelet counts (originally developed based on data in HER2-positive metastatic breast cancer patients) were implemented and connected on the platform [1]. Alternative T-DM1 intravenous infusion dosing regimens were compared, including weekly administration (q1w) of 1.2 or 2.4 mg/kg, as well as 3.6 mg/kg every 3 weeks (q3w). If patients experienced grade 3 or 4 thrombocytopenia (based on daily measurements), the next dose was delayed until platelet levels recovered to grade = 1. For those experiencing grade 4 thrombocytopenia, the subsequent doses were further reduced by 0.6 mg/kg for the q3w dosing and by 0.4 mg/kg for the q1w regimens. If more than 2 dose reductions were required, treatment was discontinued. Each treatment arm was simulated with 100 patients across 100 replicates. Results: Probability of success calculations showed that the 2.4 mg weekly dosing performed slightly better in terms of efficacy (progression-free survival) compared to the 3.6 mg q3w regimen with a median hazard ratio of 0.79 (0.75 ; 0.94). Simulations predicted that the 3.6 mg/kg q3w dosing results in a similar incidence of grade 3 and 4 thrombocytopenia (22.2% [18.7-24,8]) compared to the 2.4 mg/kg q1w schedule (16.9% [14-20.2]). However, dose delays and treatment discontinuations are predicted to be more frequent with the 2.4 mg/kg q1w schedule than with the 3.6 mg/kg q3w regimen. Key results (figures of efficacy and safety endpoint and tables of % patients predicted to meet target efficacy and safety criteria across dose levels) are presented in illustrative dashboards offering practical guidance for informed decision-making. Conclusions: In summary, this clinical trial simulator serves as a robust tool for informing go/no-go investment decision-making and dose selection. By combining general disease progression models with drug specific PK/PD models in automatized workflows, the platform offers a drug development tool both for technical users (such as statisticians and data scientists) and wider projects teams.
[1] Bender et al. PAGE 25 (2016) Abstr 5928 [www.page-meeting.org/?abstract=5928]
Reference: PAGE 33 (2025) Abstr 11442 [www.page-meeting.org/?abstract=11442]
Poster: Methodology - Study Design