Pitfalls of linear regression for estimating slopes over time and how to avoid them by using linear mixed-effects...

Area of Science:

• Clinical Epidemiology
• Nephrology
• Biostatistics

Background:

• Clinical epidemiological studies often investigate disease causes, such as the link between blood pressure and chronic kidney disease (CKD).
• Estimating kidney function trajectories over time is crucial for understanding disease progression, but patient heterogeneity in function, dropout rates, and data availability pose analytical challenges.
• Traditional methods like linear regression for individual slopes do not adequately address this heterogeneity.

Purpose of the Study:

• To highlight the challenges in estimating kidney function trajectories in clinical studies.
• To introduce and explain the utility of linear mixed-effects models (LMM) for analyzing such data.
• To provide a framework for modeling and interpreting LMM in the context of kidney function decline.

Main Methods:

• Comparison of linear regression with linear mixed-effects models (LMM) for analyzing longitudinal kidney function data.
• Demonstration using a clinical example to illustrate the application and interpretation of LMM.
• Development of a framework for modeling and interpreting LMM in epidemiological studies.

Main Results:

• Linear regression methods fail to account for significant heterogeneity in patient data, including varying kidney function levels, dropout times, and numbers of measurements.
• Linear mixed-effects models (LMM) can effectively incorporate this heterogeneity, retaining all available information and variability in the data.
• LMM provide a more robust and accurate estimation of kidney function trajectories over time.

Conclusions:

• Linear mixed-effects models (LMM) are superior to linear regression for analyzing kidney function trajectories due to their ability to handle patient heterogeneity and missing data.
• Understanding and applying LMM is essential for accurate epidemiological research on chronic kidney disease progression.
• The proposed framework facilitates the modeling and interpretation of LMM in clinical research settings.