A Practical Review of Adaptive Platform Trials

Traditional randomized controlled trials (RCTs) can provide rigorous evidence but are often slow and resource-intensive, requiring separate trials for each intervention. Adaptive platform trials (APTs) have been promoted as a solution, offering a framework that tests multiple therapies under a single protocol, with arms added or dropped as evidence accumulates. However, their advantages come with trade-offs that warrant scrutiny. In this review, we critically appraise 3 landmark APTs. The I-SPY2 trial accelerated Phase II oncology research by utilizing Bayesian adaptive randomization and surrogate endpoints; however, much of its efficiency stemmed from relying on intermediate outcomes, which may not reliably predict survival. RECOVERY demonstrated the power of scale on a pragmatic UK-wide platform, but its success reflected health system infrastructure, political leadership, and the unique circumstances of the COVID-19 pandemic as much as its design. REMAP-CAP, a perpetual platform trial for pneumonia, rapidly switched to pandemic mode in 2020 and tested COVID-19 therapies using Bayesian models and response-adaptive randomization (RAR); however, the RAR amplified random noise in some domains, exposing patients to interventions later shown to be ineffective. A recent systematic review confirmed wide heterogeneity in APTs and suboptimal reporting. APTs are not inherently better than classical RCTs. Gains in speed may depend on less rigorous endpoints, complex adaptive methods, or streamlined oversight, each of which introduces new risks of error. As APTs spread to new fields such as transfusion medicine, clinicians and researchers must learn to recognize both the potential benefits and the pitfalls of this design.