Planning for complex and innovative trial designs often require extensive simulations in order to optimize the design parameters (sample size, number and timing of interim looks, adaptive decision rules, futility and efficacy stopping rules, etc.) and to establish the operating characteristics (power, type-I error rate, etc.) of the design. Bayesian adaptive designs in general fall under simulated guided designs.

Even when design parameters and adjustments for adaptive changes can be handled with closed-form asymptotic expressions—for example, in designs with sample-size re-estimation at an interim look or those with group-sequential boundaries for early efficacy stopping—simulations add important insight. They reveal how the design behaves in practice and highlight opportunities for further optimization and robustness of the design under various scenarios. Such optimizations can reduce sample size, shorten trial duration, or lower the probability of incorrect interim decisions.

Simulations also help in aligning the trial design and objectives with respect to recruitment rates, lost-to-follow-up rates, disease prevalence or incidence rates as in a vaccine trial or a diagnostic device validation trial or prevalence of sub-populations in a study incorporating a stratified randomization.

For time-to-event trials, much of the asymptotic theory relies on assumptions such as constant (over time) and/or proportional hazards. Simulations allow us to assess the sensitivity of these assumptions under scenarios like delayed treatment effects or non-proportional hazards. This helps ensure that sample size and power calculations account for such variations and mitigate risks such as an underpowered trial.

Our statistical team routinely works on complex adaptive and innovative trial designs and are well equipped with agile programming skills to rapidly set up simulations to be able to finalize study designs in a timely fashion. Simulation results for complex adaptive designs are routinely requested by the regulatory bodies and we provide these as part of a statistical planning document which also provides a detailed description of the design and design parameters. This document can then feed into the design details in the protocol, statistical analysis plans, data monitoring plans and charters, and other study documents.