Assessing individual response to exercise training: Statistical considerations, analytical approaches, and moderators of response
Exercise , Individual response , Fitness
A growing body of research has classified “responders” or “non-responders” to exercise training and has assessed whether individuals respond differently to exercise training. In recent years, biostatisticians have raised concerns regarding the statistical rigour required to classify individual responses and determine whether interindividual variability can be attributed to differences in exercise trainability per se. This dissertation attempts to advance our understanding of these concerns by describing, critically evaluating, and demonstrating how to implement statistical approaches when classifying individual responses and estimating interindividual differences in trainability. The literature review first overviews the statistical principles relevant for examining individual responses to exercise training, and then reviews the purported biological mechanisms that explain why individuals respond differently to exercise training. The literature review also contains a systematic review with two novel analyses that: i) demonstrate the importance of considering error and meaningful thresholds when classifying individual responses, and ii) present evidence questioning the existence of interindividual differences in trainability. The following two chapters showcase and discuss the limitations with ignoring error and meaningful thresholds when classifying individual responses, and also demonstrate a new method for calculating response estimates – individual monoexponential regressions – that improves the accuracy in classifying individual VO2max responses. We then provide evidence supporting the hypothesis that higher response rates – the proportion of participants classified as “responders” – are explained by larger mean changes and not reduced interindividual variability. The remaining chapters focus on investigating the existence of interindividual differences in trainability by: i) critically evaluating the SDIR method, ii) presenting mixed evidence of interindividual differences in trainability across a range of morphological and molecular skeletal muscle outcomes following aerobic exercise training, and iii) revealing a lack of interindividual differences in trainability from an individual participant data meta-analyses that included over 1,500 participants from 8 RCTs. Collectively, this dissertation highlights the importance of adopting statistical methods when examining individual responses to exercise training.