(T-023) Optimizing Clinical Trial Design through In Silico Modeling, Virtual Populations, and Mathematical Optimization
Tuesday, October 21, 2025
7:00 AM - 1:45 PM MDT
Location: Colorado A
Evgueni Jacob – Modeling And Simulation – Nova In Silico; Riad Kahoul – Modeling And Simulation – Nova In Silico; Eliott Tixier – Scientific Software Engineering – Nova In Silico
Objectives: Randomized controlled trials (RCTs) aim to estimate treatment effects in a target population. However, selection based on investigator-defined eligibility criteria (ECs) can lead to limited generalizability and an increased risk of false-negative outcomes [1,2]. Precise optimization of ECs is therefore critical to enhance trial efficiency and statistical power. To address these challenges, we developed an integrated In Silico framework combining mechanistic disease modeling, virtual populations, and mathematical optimization to systematically evaluate and refine ECs, aiming to enhance treatment effect size, statistical power, and overall trial efficiency.
Methods: The framework necessitates: - a QSP disease model with both the natural disease progression and the treatment response [3] - a representative virtual population capturing clinical variability [4] and - a detailed trial protocol including treatment arms, randomization, dose administration, and scheduling. Clinical efficacy endpoints can include either continuous, rate-based, or time-to-event outcomes. Optimization algorithms—such as Sobol sequence [5] grid search with Monte Carlo refinement—are employed to explore EC combinations. A composite cost function balances clinical and operational objectives:
Cost function = (E × WE) – (SDE × WSD) – (SS × WSS) – (TD × WTD).
Where E is mean efficacy, SDE is efficacy standard deviation, SS is sample size, TD is trial duration, and the W terms are user-defined weights.
To demonstrate the framework, a synthetic two-arm study was simulated using a large cohort of virtual patients. The underlying physiologically based pharmacokinetic (PBPK) model comprised compartments for blood, lymph, peripheral tissues, and tumor. Eligibility criteria were applied based on selected baseline patient descriptors. Treatment was administered via subcutaneous injection. Efficacy was assessed as the relative change in tumor burden from baseline, defined as: %ΔTumor = 100 × (Initial – Final) / Initial volume (cm³).
Results: The optimization process identified combinations of eligibility criteria that substantially enhanced trial performance. Compared to a non-optimized design, the optimized scenario resulted in a 35% increase in predicted treatment efficacy and a reduction in required sample size from 548 to 251 patients per arm (–54%). These findings illustrate the efficiency gains achievable through model-based refinement of inclusion criteria.
Conclusions: Integrating QSP modeling with virtual populations and mathematical optimization offers a robust approach to refining clinical trial design. This methodology facilitates the identification of patient subgroups most likely to benefit from treatment, thereby improving trial efficiency and success rates. These In Silico methods are consistent with MIDD practices, which have shown measurable impact on reducing development timelines and clinical trial costs [6]. Future work will focus on validating this framework by comparing optimization results.
Citations: [1] Van Spall HGC, Toren A, Kiss A, Fowler RA. Eligibility criteria of randomized controlled trials published in high-impact general medical journals: a systematic sampling review. JAMA. 2007;297(11):1233–1240. doi:10.1001/jama.297.11.1233 [2] Kennedy-Martin T, Curtis S, Faries D, Robinson S, Johnston J. A literature review on the representativeness of randomized controlled trial samples and implications for the external validity of trial results. Trials. 2015;16:495. doi:10.1186/s13063-015-1023-4 [3] Gibbs, J. P., Kimko, H. C., & Gibiansky, E. (2022). Quantitative Systems Pharmacology: A Case for Disease Models in Drug Development. Clinical Pharmacology & Therapeutics, 111(1), 33–36. https://doi.org/10.1002/cpt.528 [4] Cheng Y, Straube R, Alnaif AE, Huang L, Leil TA, Schmidt BJ. Virtual Populations for Quantitative Systems Pharmacology Models. Methods Mol Biol. 2022;2486:129–179. doi:10.1007/978-1-0716-2251-4_6 [5] Bratley P, Fox BL. Algorithm 659: Implementing Sobol’s quasirandom sequence generator. ACM Trans Math Softw. 1988;14(1):88–100 [6] Viceconti M, Emili L, Afshari P, et al. Possible contexts of use for in silico trials methodologies: a consensus-based review. IEEE J Biomed Health Inform. 2021;25(10):3977–3982. doi:10.1109/JBHI.2021.3090469 [7] Musuamba FT, Manolis E, Holford NHG, et al. Impact of Model-Informed Drug Development on Drug Development Cycle Times and Clinical Trial Cost. Clin Pharmacol Ther. 2023;113(1):20–28. https://doi.org/10.1002/cpt.3636
