Study selection is summarized in the PRISMA flow diagram. After deduplication, 33 studies (n = 2538) met the inclusion criteria and 27 (n = 2 106) contributed data to the meta-analysis. The complete PRISMA flow diagram illustrating the process of study identification and selection through database searching is presented in Fig. 1.

A total of 5097 records were identified through the database search. After removing duplicates, 3293 titles and abstracts were screened, and 2309 records were excluded. The full text of 984 potentially relevant articles was reviewed, of which 950 were excluded. Ultimately, 33 studies were included in this review, and 27 studies were used in the meta-analysis.

Further details were included in Supplementary material B for a comprehensive overview of the 33 RCTs and their characteristics6,7,8,9,12,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48. Exercise capacity was assessed in 24 studies6,7,8,9,12,23,24,28,29,30,31,32,33,35,36,37,38,39,41,42,43,46,47,48, with 23 employing the 6-minute walk test (6MWT)6,10,11,12,15,26,27,31,32,34,35,36,38,39,40,41,42,44,45,46 and one utilising the endurance shuttle walk test (ESWT). Dyspnea severity was evaluated in nine studies6,7,8,22,29,32,36,40,42,46,47,48: eight administered the modified Medical Research Council (mMRC) questionnaire6,7,8,22,29,32,40,42 and one applied the London Chest Activity of Daily Living (LCADL) scale. HRQoL outcomes were reported across all 24 studies6,7,9,21,23,24,25,26,27,29,30,32,33,35,36,38,39,40,41,43,44,45,46,48, measured through heterogeneous instruments such as Chronic Obstructive Pulmonary Disease Assessment Test (CAT)6,7,24,27,29,39,44 and St George’s Respiratory Questionnaire (SGRQ)23,29,33,36,44,45.

A summary of the analysis results was presented in Figs. 2 and 3, taken usual care as the common control. No significant inconsistency was found in the NMA. Visual inspection of the funnel plot revealed symmetrical distribution of effect estimates, indicating a low risk of publication bias (Supplementary material H). Egger’s regression test demonstrated no statistically significant evidence of small-study effects (P > 0.05).

a Facility-based vs Non facility-based - 6MWT. b Facility-based vs Non facility-based -mMRC. c Four comparison of the four PR models - 6MWT. d Four comparison of the four PR models -mMRC.

Data from 23 studies6,7,8,9,