Abstract
OBJECTIVE: To evaluate the effectiveness of respiratory muscle training (RMT) on respiratory symptom severity and symptom duration in adults with chronic obstructive pulmonary disease (COPD) and to appraise the certainty of the evidence. METHODS: We searched seven databases and included parallel-group randomized controlled trials (RCTs) comparing RMT plus guideline-based care versus control. Meta-analyses were conducted in RevMan using fixed- or random-effects models as appropriate. Risk of bias was assessed using RoB 2. Certainty of evidence was assessed using the GRADE approach. Because multiple outcomes were analysed, we controlled the false discovery rate using the Benjamini-Hochberg (BH) procedure across prespecified meta-analysed outcomes and report both unadjusted and BH-adjusted p values. RESULTS: Seven RCTs (n = 1171) were included. No trial reported symptom duration in a way that matched our prespecified definition. Compared with control, RMT reduced dyspnoea (mMRC; unadjusted p = 0.003; BH-adjusted p = 0.007; moderate certainty) and improved health-related quality of life (SGRQ; unadjusted p < 0.00001; BH-adjusted p = 3.5e - 05; moderate certainty). Lung function effects were mixed: FVC improved (MD = 0.37 L, 95% CI 0.33 to 0.40; unadjusted p < 0.00001; BH-adjusted p = 3.5e - 05; low certainty) and FEV1/FVC improved (MD = 1.84%, 95% CI 0.29 to 3.39; unadjusted p = 0.02; BH-adjusted p = 0.035; low certainty), while FEV1 did not differ significantly (MD = 0.18 L, 95% CI -0.04 to 0.40; unadjusted p = 0.11; BH-adjusted p = 0.11; low certainty). Evidence for overall symptom burden and exercise capacity was uncertain: CAT (unadjusted p = 0.06; BH-adjusted p = 0.07; very low certainty) and 6MWD (unadjusted p = 0.05; BH-adjusted p = 0.07; very low certainty). CONCLUSION: In adults with COPD, adding RMT to guideline-based care probably reduces dyspnoea and improves health-related quality of life (moderate certainty). Evidence for benefits on overall symptom burden, lung function, and exercise capacity is low to very low, and the effect on symptom duration remains unknown due to the lack of reporting.