To condition or not condition? Analysing 'change' in longitudinal randomised controlled trials

Area of Science:

• Biostatistics
• Clinical Trial Analysis
• Longitudinal Data Modeling

Background:

• Traditional statistical methods like ANCOVA and LDA are common for analyzing longitudinal RCT data.
• Constrained longitudinal data analysis (cLDA) is an established technique that enforces baseline equality between study arms.
• Missing data can impact the reliability of statistical analyses in clinical trials.

Purpose of the Study:

• To illustrate the application and impact of missing data on ANCOVA, LDA, and cLDA methods.
• To compare the performance of cLDA against traditional methods using real-world RCT data.
• To demonstrate the theoretical concepts of these statistical techniques in a clinical context.

Main Methods:

• Analysis of fasting lipid profiles from the Group Medical Clinics (GMC) longitudinal RCT.
• Application of linear mixed models to both complete and all available data.
• Comparison of results from ANCOVA, LDA, and cLDA.

Main Results:

• With complete data, cLDA and ANCOVA showed a significant reduction in LDL cholesterol (11.2 mg/dL) for GMC compared to usual care.
• Using all available data, cLDA demonstrated a significant LDL improvement (8.9 mg/dL) in GMC.
• The LDA method, using all available data, yielded a non-significant LDL improvement (7.2 mg/dL).

Conclusions:

• cLDA offers efficient treatment effect estimates and robust inferential statistics, even with missing data.
• cLDA is a preferred method over ANCOVA and LDA for analyzing RCTs with potential missing data.
• The study highlights cLDA's utility in providing reliable results in the presence of incomplete longitudinal data.