On summary measure analysis of linear trend repeated measures data: performance comparison with two competing methods

Area of Science:

• Biostatistics
• Medical Research Methodology

Background:

• Repeated measures analysis is crucial in medical research, particularly with large datasets.
• The summary measure approach (SMA) offers a viable analytical tool when other methods are limited.

Purpose of the Study:

• To detail SMA techniques for linear trend repeated measures data.
• To compare the performance of SMA against linear mixed models (LMM) and unstructured multivariate approach (UMA).

Main Methods:

• Developed practical guidelines for SMA using least squares regression slope and mean response.
• Conducted Monte Carlo simulations to evaluate SMA, LMM, and UMA under various covariance structures.
• Applied all methods to two real-world medical data examples.

Main Results:

• SMA demonstrated superior performance over UMA and was comparable to LMM in testing time, group, and interaction effects.
• LMM showed a lack of robustness and produced unreliable results when covariance structures were misspecified.
• UMA yielded overly conservative inferences and is not recommended for this data type.

Conclusions:

• SMA is a simple, safe, and powerful analytical approach for linear trend data.
• The efficiency loss of SMA compared to optimal LMM is typically negligible.
• SMA is recommended as a primary method for analyzing linear trend data, especially with numerous measurements or limited sample sizes.