Clinical Trial Simulation and the Art of Adaptive Design Optimization
Rethinking Simulation—More Than Prediction
In statistics, the public often equates simulation with prediction. Sports forecasting, such as predicting the hurricane trajectory or predicting the results of the full NFL season, typify this viewpoint; simulation is used to answer “what will happen?” These predictive models have value, but in the context of adaptive clinical trial design, this perspective creates confusion.
For the creation of optimized clinical trials, simulation is not used for forecasting, instead, it serves as a practical instrument for developing and interrogating trial designs—well before any data is collected in the clinic. Clinical trial simulation allows scientists to stress-test every component: randomization approaches, interim analyses, arm dropping, sample size adjustments, and dose selection strategies, among others.
This method of simulating patient flow, data accrual, and decision-making creates a controlled environment for exposing strengths and vulnerabilities in a proposed design. Each synthetic “run” represents a possible real-world outcome—capturing operational complexity, uncertainties in patient response, variable accrual rates, and even drug supply logistics.
In Silico Design as Stepwise Refinement
The in-silico design process is defined by iterative refinement rather than rigid adherence to a single, scripted plan. It starts with a set of candidate trial designs—such as a phase 2 dose-finding study with multiple arms, or a seamless phase 2/3 trial. For every scenario, simulations are run across a wide range of assumptions—treatment effect distributions, sample sizes, biomarker prevalence, enrollment rates, and stopping rules.
Key elements of the process:
● Scenario Development: Stakeholders specify plausible scenarios, including both expected effects and challenging edge cases.
● Full Trial Emulation: Each simulation run executes the trial as written—enrolling, randomizing, performing interims, and recording “patient-level” outcomes.
● Decision Recording: Endpoints such as which dose advances, whether the trial stops early for futility or efficacy, correct dose selection, statistical significance, sample size utilization, and even realized revenue as an outcome are tracked for each trial.
What is powerful in clinical trial simulations is its commitment to stakeholder clarity. Instead of beginning with aggregate metrics (such as power or type 1 error alone), the team first shares example trials—walking clinicians and decision-makers through simulated patient data, interim analyses, and resulting adaptive decisions. This stepwise transparency builds mutual understanding, allowing all voices—statistical, clinical, operational, and regulatory—to calibrate their expectations to what “success” or “failure” truly look like for the study.
As simulation results are shared, feedback naturally drives changes. Stakeholders may request to alter allocation ratios, sample size, futility thresholds, or enrichment strategies. New scenarios are developed to probe “what if” situations. Each refinement round brings the design closer to optimizing stakeholder and scientific goals. The process is inherently iterative, often revealing issues or opportunities that would not have surfaced without trial simulations.
Infrastructure and Advanced Scenario Analysis
Efficient simulation demands more than coding proficiency. Iterative protocol evaluation requires rapid scenario generation, fast simulation turnaround, and immediate visualization of outcomes. FACTS software was developed to be a flexible and powerful advanced simulation platform.
FACTS enables:
● Rapid configuration of complex design elements (e.g. dose escalation methods, adaptive randomization, interim analyses)
● Comparison of multiple candidate designs within a unified framework
● Scenario libraries for stress-testing against key expectations or regulatory-relevant situations
● Counterfactual simulation, a critical feature: multiple trial designs are simulated on identical patient cohorts, elucidating the tangible impact of early stopping, adaptive rules, or alternative decision thresholds. This side-by-side design comparison is a powerful way to directly answer questions about tradeoffs in error rate, sample size, or efficiency between two designs under the same assumptions.
As simulation progresses, complex design behaviors are distilled into clear, actionable summaries for decision-makers: tiered operating characteristics, sample size distributions, probabilities of dose selection, type 1 error control under various nulls, and the potential impact of design innovations like response adaptive randomization.
When a design is finalized, a large-scale simulation run documents all relevant operating characteristics—supporting regulatory submissions, DSMB planning, and internal design sign-off. The path to that protocol, however, is paved with the learning and efficiency gains of iterative simulation-led redesign.
Simulation as the Empirical Backbone for Modern Trials
Clinical trial simulation is the backbone of the trial design process—a fully integrated practice of scenario exploration, iterative refinement, and unambiguous communication with all stakeholders. The discipline and openness demanded by this approach safeguard against blind spots and reduce the likelihood of unexpected or undesirable problems after launch. Clinical trial simulation ensures that every design advanced to the clinic is both scientifically robust and contextually fit for purpose, informed by millions of virtual runs and rigorously-tested and compared for complete optimization.