Application of the Marginal Structural Model to Account for Suboptimal Adherence in a Randomized Controlled Trial

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Epidemiology

Background:

• Suboptimal adherence in randomized controlled trials (RCTs) presents analytical challenges.
• Traditional methods like per-protocol analysis or censoring introduce biases.
• Causal modeling offers a novel approach to address adherence issues.

Purpose of the Study:

• To apply causal modeling, specifically marginal structural models with inverse-probability weighting, to adjust for adherence in RCTs.
• To predict intervention effects under conditions of full adherence.
• To gain insights into the intrinsic physiological effects of interventions by comparing with intention-to-treat results.

Main Methods:

• Implementation of a marginal structural model using inverse-probability weighting via a weighted generalized estimating equation model.
• Development of two ancillary models: stepwise linear regression for percent weight loss and stepwise logistic regression for early discontinuation.
• Calculation of participant- and time-specific weights based on ancillary models.

Main Results:

• The developed causal model can effectively predict intervention effects at full adherence.
• Comparison of model predictions with intention-to-treat results provides deeper understanding of intervention efficacy.
• The approach allows for nuanced analysis of adherence impact in clinical trials.

Conclusions:

• The marginal structural model with inverse-probability weighting is a complex but powerful tool for analyzing adherence in RCTs.
• Careful consideration of model details, including variable selection, interactions, and time-dependent covariates, is crucial for accurate application.
• This method enhances the ability to interpret intervention effects and understand physiological impacts, complementing traditional trial analyses.