Different ways to estimate treatment effects in randomised controlled trials

Area of Science:

• Biostatistics
• Clinical Trials
• Epidemiology

Background:

• Ongoing debate regarding baseline value adjustment in randomized controlled trial (RCT) data analysis.
• Misunderstanding exists on appropriate methods for adjusting baseline values.
• Need for clear guidance on analyzing RCT data.

Purpose of the Study:

• Explain various methods for estimating treatment effects in RCTs.
• Illustrate these methods using a real-world example.
• Provide recommendations for analyzing RCT data.

Main Methods:

• Theoretical explanation and application of longitudinal analysis of covariance (ANCOVA).
• Application of repeated measures analysis using baseline values as outcomes.
• Analysis of changes from baseline.
• All methods applied to a dataset on systolic blood pressure lowering treatment.

Main Results:

• Baseline differences between groups must be accounted for in RCT analysis.
• Standard repeated measures analysis and analysis of changes can yield biased treatment effect estimates.
• Real-life example demonstrated varying treatment effect estimates due to baseline differences.

Conclusions:

• Longitudinal analysis of covariance (ANCOVA) is advised for RCT data analysis.
• Alternatively, a repeated measures analysis including treatment-by-time interaction is recommended.
• These methods appropriately adjust for baseline differences, ensuring unbiased treatment effect estimation.