Estimation-Driven Adaptive Trials: A Framework for Balancing Design Efficiency with Reliable Estimation to Improve Clinical and Health Economic Decision-Making

Adaptive clinical trials are changing how we evaluate new treatments by allowing trial modifications based on accumulating data. Group sequential designs, for example, can stop a trial early if results suggest benefit or harm, improving efficiency and reducing patient exposure to ineffective treatments. However, these modifications affect the accuracy and uncertainty quantification of treatment effect estimates. Inflated or deflated estimates are then fed into benefit-risk assessment or cost-effectiveness models that often do not account for them, making treatments seem more (or less) beneficial than they actually are. As a result, healthcare systems may approve interventions that offer less value to patients than expected, or fail to adopt genuinely beneficial treatments. Current adaptive designs largely focus on controlling error rates, without explicitly incorporating estimation reliability as a design objective. Moreover, adaptive decisions and interim analyses are often incompletely reported, limiting transparency and making it difficult for downstream users to account for their impact on estimation. This project addresses both gaps, developing frequentist and Bayesian design strategies that balance estimation reliability with trial efficiency and promoting transparency and awareness of how adaptive features affect trial results. This work will ultimately improve the reliability and interpretability of evidence for clinical and policy decision-making.