Joint modeling compliance and outcome for causal analysis in longitudinal studies

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Causal Inference

Background:

• Noncompliance is common in longitudinal studies, complicating treatment effect estimation.
• Repeatedly measured outcomes and compliance data require specialized modeling techniques.
• Potential outcome frameworks are crucial for causal inference in observational and randomized settings.

Purpose of the Study:

• To develop a joint model for compliance and clinical outcomes in two-arm randomized longitudinal studies.
• To estimate treatment effects within principal strata defined by compliance behavior.
• To assess the impact of treatment on future compliance over time.

Main Methods:

• Proposed a Markov compliance and outcome model.
• Utilized the potential outcome framework to define pre-randomization principal strata.
• Employed Bayesian methods for parameter estimation.
• Illustrated results with a cognitive behavior therapy and depression study.
• Conducted a simulation study to evaluate model properties.

Main Results:

• The model successfully estimates treatment effects within defined compliance strata.
• It quantifies the influence of treatment on subsequent compliance behavior.
• Bayesian estimation provided robust parameter estimates.
• The cognitive behavior therapy example demonstrated practical application.

Conclusions:

• The joint modeling approach effectively addresses noncompliance in longitudinal studies.
• The model provides insights into both treatment efficacy and its effect on adherence.
• This methodology enhances causal inference in complex longitudinal trial data.