Real-time adaptive randomization of clinical trials

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Cardiovascular Research

Background:

• Randomized controlled trials (RCTs) are standard but can assign participants to inferior treatments.
• Delayed endpoints in RCTs pose challenges for real-time treatment adaptation.
• Balancing treatment allocation with learning about treatment efficacy is crucial.

Purpose of the Study:

• To evaluate real-time adaptive randomization (RTAR) for clinical trials with delayed endpoints.
• To assess the trade-offs of RTAR regarding mortality, confidence intervals (CIs), and statistical power.
• To compare RTAR with traditional RCTs and block-based adaptive designs.

Main Methods:

• Simulations using data from two large cardiovascular RCTs (GUSTO-1 and EUROPA).
• Resampling treatment assignments and endpoints to mimic real-time and block-based adaptive randomization.
• Evaluating RTAR's performance in optimizing treatment allocation and endpoint estimation.

Main Results:

• RTAR quickly identifies superior treatments, allocating more patients to them and reducing overall mortality.
• Endpoint rates and odds ratios remained within RCT confidence intervals, with tighter CIs for the superior arm.
• RTAR demonstrated improved adaptation to changing endpoint rates within trials.

Conclusions:

• RTAR enhances patient outcomes and narrows confidence intervals for superior treatments.
• Benefits of RTAR are balanced by wider CIs for inferior arms and odds ratios.
• Forward-looking adaptive variants offer intermediate benefits, and RTAR did not misidentify inferior arms as superior.