Effect size thresholds to interpret comparisons with exercise interventions for tendinopathy

A systematic review with meta-analysis

Abstract

Introduction: The purpose of this systematic review with meta-analysis was to develop tendinopathy-specific effect size thresholds to aid interpretation of comparative effectiveness of exercise therapies.  

Methods: A comprehensive systematic literature search for studies comparing exercise, non-active controls, and non-exercise interventions for tendinopathy management was conducted. Trials with participants diagnosed with rotator cuff, lateral elbow, gluteal, patellar or Achilles tendinopathy of any severity or duration were included. Standardised mean difference comparative effect sizes were combined with Bayesian hierarchical models. Non symmetric (exercise vs. non-active), and symmetric (exercise vs. exercise, and exercise vs non-exercise) distributions centred on zero were constructed with small, medium, and large thresholds calculated to capture 25, 50, and 75% of the comparative effect-size distributions, respectively. Analyses were combined across all tendinopathy locations and separated according to outcome domains including disability, pain, physical function capacity and range of motion where sufficient data were available.

Results: Data were extracted from 96 studies and yielded 130 pairwise comparisons. When pooling data across all outcomes, 98 comparative effect sizes showed small: 0.11 [95%CrI:0.09-0.13], medium: 0.25 [95%CrI:0.23-0.27], and large: 0.46 [95%CrI:0.44-0.49] effect sizes for exercise therapy compared with non-active control. Symmetric distributions calculated from 636 effect sizes showed lower magnitude thresholds when comparing exercise therapies (small: 0.11 [95%CrI:0.09-0.13], medium: 0.25 [95%CrI:0.23-0.27], and large: 0.46 [95%CrI:0.44-0.49]) than comparing exercise therapies with non-exercise interventions (254 effect sizes; small: 0.17 [95%CrI:0.13-0.21], medium: 0.37 [95%CrI:0.33-0.41], and large: 0.70 [95%CrI:0.64-0.75]). Analyses showed greater magnitude effect sizes for pain and disability outcomes than physical function capacity and range of motion.

Conclusion: Exercise therapy generally results in improvements beyond natural healing processes and expectancy effects present in non-active controls. Comparative effect sizes between exercise therapies are, however, generally low in magnitude indicating therapy variations yield relatively minor incremental effects. These findings should inform the clinical interpretation of existing and future trials, and guide study design of future research – such as the necessity for larger samples than have been previously used –. It is recommended that small, medium, and large threshold values presented in this review be used instead of Cohen’s general values.