Principal component analysis of lifting waveforms

Abstract

One limiting factor in lifting research design has been the inability to effectively analyze waveform data, especially when differences in body mass, height, and load magnitude influence the derived kinetic variables. The purpose of this study was to demonstrate the sensitivity of principal component analysis to quantify clinically relevant differences in kinetic lifting waveforms over three load magnitudes and between two separate populations. Principal component analysis was applied to five kinetic lifting waveforms. The derived principal component scores were used as the dependent measures in a two-way (clinical status x load magnitude) MANOVA. Significant low back pain group differences (P<0.05) were found for three of the principal component scores on extension moment generation in the sacral and thoracic regions and for trunk compression. Significant differences were found for each variable with respect to the magnitude across the entire lift time between the three load conditions, as well as four significant differences related to inferred mechanical changes that resulted from lifting increasingly heavier loads. Principal component analysis of kinetic lifting waveforms was shown to be insensitive to a confounding factor of different load magnitudes when attempting to identify previously determined clinically relevant differences in the waveform trajectories. The analysis was able to partition the variability attributed to the direct influence of different external load magnitudes, versus those differences in spinal loading that arose from the variations in the lifting mechanics of increasing loads. The technique could be beneficial for other kinetic analyses where confounding magnitude modifiers like body size are present.