Introduction The purpose of this large meta-analysis was to synthesise intervention data to create empirically derived thresholds to interpret the effectiveness of exercise therapies in the management of tendinopathy.

Materials and Methods Data from exercise therapy interventions (rotator cuff, lateral elbow, patellar, Achilles and gluteal) across eight outcome domains (disability, physical function capacity (PFC), function, pain, painLoading, painTime, range of motion (ROM), and quality of life (QoL)) were included. Mean standardized effect sizes were calculated and meta-analysed using a five-level Bayesian hierarchical model with the 0.25-, 0.5- and 0.75-quantiles used to benchmark ‘small’, ‘medium’, and ‘large’ effects.

Results A total of 1521 effect sizes from 121 studies comprising 176 treatment arms and 4370 participants were included. Substantial differences in effect size distributions were identified across outcome domains. The small, medium, and large thresholds for patient reported subjective outcomes (disability, function, pain, painLoading, painTime) were typified by values of 0.6, 1.0 and 1.6, respectively. In contrast, lower effect sizes were obtained for QoL and objective outcomes (ROM, PFC), with small, medium, and large thresholds typified by values of 0.2, 0.4 and 0.7, respectively.

Conclusion Previous thresholds widely used to interpret small, medium, and large intervention effects (small = 0.2, medium = 0.5, large = 0.8) will tend to overestimate the effectiveness of patient reported subjective outcomes in the management of tendinopathy. The context specific benchmarks developed here should be used in future with the recommendation that evaluation of intervention effectiveness include both objective and subjective measurements.