Objectives

Statistical techniques currently used in musculoskeletal research often inefficiently account for paired-limb measurements or the relationship between measurements taken from multiple regions within limbs. This study compared three commonly used analysis methods with a mixed-models approach that appropriately accounted for the association between limbs, regions, and trials and that utilised all information available from repeated trials.

Method

Four analysis were applied to an existing data set containing plantar pressure data, which was collected for seven masked regions on right and left feet, over three trials, across three participant groups. Methods 1–3 averaged data over trials and analysed right foot data (Method 1), data from a randomly selected foot (Method 2), and averaged right and left foot data (Method 3). Method 4 used all available data in a mixed-effects regression that accounted for repeated measures taken for each foot, foot region and trial. Confidence interval widths for the mean differences between groups for each foot region were used as a criterion for comparison of statistical efficiency.

Results

Mean differences in pressure between groups were similar across methods for each foot region, while the confidence interval widths were consistently smaller for Method 4. Method 4 also revealed significant between-group differences that were not detected by Methods 1–3.

Conclusion

A mixed effects linear model approach generates improved efficiency and power by producing more precise estimates compared to alternative approaches that discard information in the process of accounting for paired-limb measurements. This approach is recommended in generating more clinically sound and statistically efficient research outputs.