Quantifying the Inter-Individual Variation in Response to Exercise Interventions

  • Philip Williamson

Student thesis: Doctoral Thesis


Interest in the concept of ‘precision’ or ‘personalized’ medicine has grown over the last three decades. While much of the literature published appears to support the notion that clinically-relevant individual response differences exist in phenotypes such as maximal aerobic capacity and weight loss, much of this research is based upon the observed response, as opposed to the ‘true’ inter-individual variation.
In this doctoral research programme, I investigated ‘true’ inter-individual variation in response to exercise interventions. The difference between observed and ‘true’ individual differences is that measurement error and other sources of random variation are fully considered in order to quantify ‘true’ individual differences. These were investigated due to the recent focus on ‘individual responses and precision and personalised approaches. This was achieved through a number of approaches, including a critical review of literature, a systematic review and meta-analysis, and both secondary analysis of randomised controlled trial (RCT) data and primary data collection through the novel use of a replicate crossover trial.
A critical review of the relevant literature on responses of maximal oxygen uptake to exercise training revealed that when the correct method for statistical analysis is utilised on data from published research claiming substantial inter-individual variability in response, it was actually observed that there was greater variability in the control sample versus the intervention sample. This finding implies that there is no substantial true individual training response variance, though the uncertainty in the estimate of true inter-individual variability in response is marked with the small sample sizes involved. The review also revealed that the vast majority of published research purporting to show individual variation in response does not utilise the most robust trial design (RCT) or statistical methods (comparison of the standard deviations of the changes in all groups).
A meta-analysis of supervised exercise RCT’s revealed that evidence is limited for clinically relevant ‘true’ inter-individual variation in weight change in response to an exercise intervention, once the random variability in weight over time in the control group is accounted for. This was the first systematic review and meta-analysis of individual response variance. The pooled mean weight loss (-1.4 kg) was much smaller than a conservative threshold for a clinically important change (2.5 kg), and inter-individual variation in weight change standard deviation (SD) was only 0.8 kg. A novel approach using a prediction interval revealed that in a future study in similar settings, the 95% plausible range for mean weight change vs. control would be -5.0 to 2.1 kg. The probability that the mean weight change in a future study would be clinically relevant was 26% (possibly clinically important). For the individual response variability, the prediction interval ranged from small negative to small positive, and the probability that the individual response variance was clinically relevant was 23% (unlikely).
A secondary analysis of data from dietary and lifestyle advice interventions (PREMIER trial) revealed substantial inter-individual variations in the body weight and blood pressure responses. Paradoxically, this response variance was not even partially accounted for by including a sex-by-treatment interaction term in the model, despite substantial sex differences in mean treatment effect. When analyses were stratified by sex, much larger true individual response variance for weight loss and blood pressure changes were observed in men compared to women, explaining the paradox. The observed effect in women is relatively consistent, whilst in men it is much more variable, reinforcing the requirement for thorough exploration of data prior to undertaking full analyses.
In a novel replicate crossover trial designed to properly partition variance and quantify ‘true’ inter-individual variation in response to acute high intensity aerobic exercise, results suggest the presence of substantial ‘true’ inter-individual variation in response. There were large sex differences in mean response, with greater blood pressure and heart rate response variables in females in comparison to males. This was the first replicate crossover designed and analysed in this way, using a specific model to elucidate the acute response to exercise.
Evidence from these studies indicates that, when quantified appropriately, chronic exercise interventions appear to elicit limited ‘true’ inter-individual variation in response in peak oxygen uptake and weight loss. Conversely, there appear to substantial inter-individual variations in blood pressure and heart rate responses to acute, high intensity aerobic bouts of exercise. Furthermore, with multicomponent interventions there appear to be substantial individual responses for blood pressure and weight loss in men, based on secondary analysis of existing trial data. It is clear that much of the research purporting to evidence individual variation in response is lacking a suitable control sample. To that end, in chronic exercise intervention trials, it is likely appropriate to focus upon the mean change, whilst for acute exercise interventions, further quantification of the magnitude of inter-individual variation in response may well be warranted.
Date of Award14 Dec 2018
Original languageEnglish
Awarding Institution
  • Teesside University
SupervisorAlan Batterham (Supervisor)

Cite this