Beyond intent to treat (ITT): A complier average causal effect (CACE) estimation primer

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Psychiatric Epidemiology

Background:

• Randomized control trials (RCTs) are the gold standard for causal inference in treatment efficacy.
• Traditional RCT analyses often fail to account for imperfect participant compliance.
• New statistical methods are needed to address non-compliance in clinical research.

Purpose of the Study:

• To introduce and demonstrate the Complier Average Causal Effect (CACE) structural equation mixture model.
• To compare CACE with traditional RCT data analysis perspectives like intent-to-treat (ITT), per-protocol, and as-treated.
• To provide a practical guide for estimating treatment effects with imperfect compliance.

Main Methods:

• Utilized maximum likelihood parameter estimation and mixture modeling.
• Developed a CACE structural equation mixture model.
• Applied the model to simulated data from a cognitive-behavioral therapy trial for Juvenile Fibromyalgia.

Main Results:

• The CACE model effectively estimates treatment effects despite imperfect participant compliance.
• Demonstrated the model's assumptions, specification, estimation, and interpretation.
• Provided Mplus syntax and linear model equations for reproducibility.

Conclusions:

• The CACE model offers a statistically sound approach to analyzing RCT data with non-compliance.
• This method enhances causal inference in clinical trials where perfect adherence is rare.
• The study provides valuable tools for researchers investigating treatment efficacy in real-world settings.