Augmented trial designs for evaluation of principal surrogates

Area of Science:

• Clinical Trials Methodology
• Causal Inference
• Biostatistics

Background:

• Evaluating intermediate responses as surrogates in randomized trials is challenging due to missing counterfactual data.
• Existing augmentation methods for counterfactual responses are primarily limited to vaccine trials.

Purpose of the Study:

• To extend existing augmentation procedures for obtaining counterfactual responses.
• To propose augmentations for three trial designs outside the vaccine trial setting.
• To enable the evaluation of intermediate responses as principal surrogates using standard methods.

Main Methods:

• Extension of existing augmentation procedures for counterfactual responses.
• Application of augmentation to crossover, individual stepped-wedge, and treatment run-in trial designs.
• Identification of causal estimands under augmented designs and application of likelihood-based methods.
• Development of a remediation analysis to address assumption violations.

Main Results:

• Augmented designs, including crossover and stepped-wedge, can improve efficiency by observing clinical endpoints under both treatment and control.
• The treatment run-in design allows for baseline measures as potential surrogates.
• Proposed methods facilitate the evaluation of intermediate responses as principal surrogates.

Conclusions:

• The study provides novel augmentation strategies for evaluating intermediate responses as surrogates in diverse clinical trial designs.
• The proposed methods, supported by a remediation analysis, enhance causal inference and efficiency in clinical trials.
• These advancements are applicable beyond vaccine trials, with an illustration in a drug-resistant tuberculosis trial.