Moderators of Exercise Effects on Cancer-related Fatigue: A Meta-analysis of Individual Patient Data

Abstract

Purpose Fatigue is a common and potentially disabling symptom in patients with cancer. It can often be effectively reduced by exercise. Yet, effects of exercise interventions might differ across subgroups. We conducted a meta-analysis using individual patient data of randomized controlled trials (RCT) to investigate moderators of exercise intervention effects on cancer-related fatigue. Methods We used individual patient data from 31 exercise RCT worldwide, representing 4366 patients, of whom 3846 had complete fatigue data. We performed a one-step individual patient data meta-analysis, using linear mixed-effect models to analyze the effects of exercise interventions on fatigue (z score) and to identify demographic, clinical, intervention- and exercise-related moderators. Models were adjusted for baseline fatigue and included a random intercept on study level to account for clustering of patients within studies. We identified potential moderators by testing their interaction with group allocation, using a likelihood ratio test. Results Exercise interventions had statistically significant beneficial effects on fatigue (β = −0.17; 95% confidence interval [CI], −0.22 to −0.12). There was no evidence of moderation by demographic or clinical characteristics. Supervised exercise interventions had significantly larger effects on fatigue than unsupervised exercise interventions (β_difference = −0.18; 95% CI −0.28 to −0.08). Supervised interventions with a duration ≤12 wk showed larger effects on fatigue (β = −0.29; 95% CI, −0.39 to −0.20) than supervised interventions with a longer duration. Conclusions In this individual patient data meta-analysis, we found statistically significant beneficial effects of exercise interventions on fatigue, irrespective of demographic and clinical characteristics. These findings support a role for exercise, preferably supervised exercise interventions, in clinical practice. Reasons for differential effects in duration require further exploration.