An overview of statistical methods for handling nonadherence to intervention protocol in randomized control trials: a...

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Epidemiology

Background:

• Nonadherence to interventions is common in randomized controlled trials (RCTs).
• Handling nonadherence appropriately is crucial for accurate treatment effect estimation.
• Naive per-protocol analysis is frequently used but can be biased.

Purpose of the Study:

• To conduct a methodological review of statistical approaches for managing intervention nonadherence in RCTs.
• To identify and evaluate the methods employed in published RCT literature.
• To assess the trends and challenges in statistical handling of nonadherence.

Main Methods:

• Systematic literature search of bibliographic databases.
• Inclusion of 58 articles published between 1991 and 2015.
• Categorization and analysis of 88 distinct methodological applications.

Main Results:

• 56% of identified studies were excluded for using naive per-protocol analysis.
• Complier average causal effect (56%) and instrumental variable (23%) were the most utilized advanced methods.
• Maximum likelihood estimators were predominantly used with these advanced methods.
• Advanced methods produced larger treatment effects than intention-to-treat, but without statistical significance due to larger standard errors.

Conclusions:

• A significant proportion of RCTs still rely on naive per-protocol analysis.
• There is a growing application of complier average causal effect and instrumental variable methods.
• These advanced methods are based on strong assumptions that may be violated.
• Further empirical research is needed to compare the performance and usability of different statistical methods for nonadherence in RCTs.