Synthetic control method for clinical trials and precision medicine

The large data availability fuels research on "one patient, one treatment" personalized medicine methods in clinical studies, particularly impactful for rare diseases where randomized controlled trials are impractical. Existing approaches like "external controls" using data from similar patients can be problematic in cases where there are limited suitable controls. This work proposes a novel approach combining historical controls with an adaptive study design, allowing treatment assignment to adapt based on valid estimates of treatment effects obtained within patient subpopulations. The proposed method represents a novel use of the "econometric" synthetic control method within clinical studies. We use a Bayesian approach to inference, which makes it easy to quantify the uncertainty related to the estimates and propagate it to the next patient's treatment assignment. The novel combination of synthetic controls with adaptive designs leverages the flexibility of both methods for the effective handling of complex data patterns, improving treatment allocation efficiency while guaranteeing ethical studies. Simulation results support the promise of this approach for clinical studies.