Correction for non-compliance in equivalence trials

Area of Science:

• Biostatistics
• Clinical Trials
• Epidemiology

Background:

• Non-compliance in randomized trials complicates the interpretation of equivalence trial results.
• The intent-to-treat hypothesis may not align with the sharp null hypothesis when patients deviate from assigned treatment.

Purpose of the Study:

• To provide a unified overview of analytic methods for adjusting non-compliance in equivalence trials.
• To introduce and compare new structural (causal) models for handling non-compliance.
• To discuss sensitivity analyses for violations of the ignorable non-compliance assumption.

Main Methods:

• Comparative analysis of existing and novel analytic approaches for non-compliance.
• Introduction of coarse structural nested models, non-nested marginal structural models, and continuous-time structural nested models.
• Evaluation of methods based on assumption plausibility, robustness, computational burden, and sensitivity analysis.

Main Results:

• Non-compliance requires specific analytic adjustments in equivalence trials.
• New structural models offer alternative approaches to address non-compliance.
• Sensitivity analysis is essential to assess the impact of non-compliance assumption violations.

Conclusions:

• Accurate analysis of equivalence trials necessitates methods that adjust for non-compliance.
• The introduced structural models provide valuable tools for causal inference in the presence of non-compliance.
• These methods are also applicable to estimating time-varying treatment effects from observational data.