Appropriate within-subjects statistical models for the analysis of baroreflex sensitivity

Area of Science:

• Physiology
• Biostatistics

Background:

• Baroreflex sensitivity is commonly assessed by relating systolic blood pressure changes to pulse interval variations.
• Least squares regression (LSR) is frequently used to calculate individual baroreflex sensitivity slopes, pooling results for group analysis.

Purpose of the Study:

• To address the violation of independence assumptions in LSR for within-subject baroreflex sensitivity data.
• To introduce and advocate for the use of linear mixed models (LMM) as a more appropriate analytical approach.

Main Methods:

• Comparison of traditional LSR analysis with linear mixed models (LMM) using a representative dataset.
• Demonstration of how LMM accounts for the inherent correlation of repeated measures within individual subjects.

Main Results:

• LSR analysis of baroreflex sensitivity data violates independence assumptions, potentially leading to biased estimates and reduced statistical power.
• Linear mixed models provide a more statistically robust analysis by incorporating subject-level correlations.
• LMM analysis is more efficient, analyzing the entire dataset simultaneously, reducing bias, and enhancing statistical power.

Conclusions:

• Linear mixed models represent a superior analytical strategy for baroreflex sensitivity studies compared to conventional LSR.
• Adoption of LMM can lead to more reliable and powerful conclusions in baroreflex research.
• Researchers are encouraged to implement LMM for improved accuracy and efficiency in analyzing baroreflex sensitivity.