Estimating the causal effect of compliance on binary outcome in randomized controlled trials

Area of Science:

• Biostatistics
• Epidemiology
• Clinical Trials

Background:

• Partial compliance in randomized trials complicates causal effect estimation.
• Existing methods using univariate compliance summaries can lead to boundary solutions and confounding.
• Temporal compliance patterns are crucial for accurate analysis.

Purpose of the Study:

• To develop and apply advanced likelihood-based methods for analyzing causal effects with partial compliance.
• To explicitly account for the temporal structure of compliance in randomized trials.
• To re-analyze a vitamin A trial using an extended methodology.

Main Methods:

• Applied Goetghebeur and Molenberghs' method for ordinal compliance and monotone dose response.
• Extended likelihood-based methodology to multivariate compliance with partial order structures.
• Incorporated Rubin's randomization assumptions and parametric modeling of potential outcomes and compliance.
• Analyzed a vitamin A trial data, considering temporal compliance patterns.

Main Results:

• The univariate compliance summary yielded meaningful estimators in a blood pressure trial but a boundary solution in the vitamin A trial.
• The extended method revealed a significant impact of vitamin A on mortality within the first 4 months.
• The greatest risk reduction was observed in children who received two doses of vitamin A.
• Findings align with previous vitamin A trials in Ghana and Nepal.

Conclusions:

• Accounting for temporal compliance structure improves causal effect estimation in randomized trials.
• Vitamin A supplementation, particularly with adequate dosage, can substantially reduce child mortality.
• The developed methodology offers a robust framework for analyzing complex compliance scenarios in clinical research.