A literature search revealed 24,866 citations, and 740 full-text articles were assessed for inclusion eligibility (Figure 1). Review articles; retrospective studies; mixed population studies; and studies with interventions such as qigong, vibration training, neuromuscular electrical stimulation, and balance training, did not meet our inclusion criteria (eTable 3; Additional file 1 pp. 10–12). A total of 54 RCTs, involving 2,828 patients, were selected; these were categorized into three groups based on the timing of rehabilitation intervention (Figure 2): RCTs conducted during stable conditions (n = 36; eTable 4.1; Additional file 1 pp. 13–17), RCTs conducted after an acute exacerbation discharge (n = 8; eTable 4.4; Additional file 1 pp. 32–33), and RCTs conducted before an acute exacerbation discharge (n = 8; eTable 4.7; Additional file 1 pp. 38–39). The inclusion and exclusion criteria adopted in all RCTs are listed in eTables 4.2, 4.5, and 4.8 (Additional file 1 pp. 18–25, 34–35, and 42–43).
We classified all rehabilitation pulmonary programs into seven categories. The following four categories comprised single-component programs: (1) ET with cycle ergometry, treadmill training, or Nordic walking; (2) RT with elastic bands, self-gravity resistance, or weight training machines; (3) RMT with inspiratory or expiratory muscle trainers; and (4) usual care (eTable 4.10). The following three remaining categories involved multi-component programs that combined the aforementioned interventions: (5) ET+RT, (6) ET+RMT, and (7) ET+RT+RMT. The rehabilitation program details are provided in eTables 4.3, 4.6, 4.9, and 4.10 (Additional file 1 pp. 26–31, 36–37, 40–41, and 44–45).
During RoB assessment, studies conducted at the aforementioned three intervention time points were rated across five domains; “low” to “some concerns” of risk were noted (eFigures 5.1, 5.3, and 5.5; Additional file 1 pp. 53, 57, and 60). These ratings were based on an evaluation of each study's methodology and potential biases (eFigures 5.2, 5.4, and 5.6; Additional file 1 pp. 54–55, 58, and 61). The primary reasons for downgrading were inadequate randomization processes, missing outcome data, or insufficient detailing of information (eTable 5.1, 5.2, and 5.3; Additional file 1 pp. 56, 59, and 62). Across the seven rehabilitation categories, the average patient age, predicted FEV1, and body mass index ranged from 61.0 to 69.0 years, 43.8% to 50.7%, and 22.3 kg/m2 to 26.9 kg/m2, respectively. No significant differences were observed among the various interventions (eTable 4.12; Additional file 1 pp. 46-47).
Pulmonary rehabilitation for stable COPD
We analyzed 38 RCTs that focused on patients undergoing pulmonary rehabilitation for stable COPD without experiencing any acute exacerbation for more than 6 weeks (eTable 4.2; Additional file 1 pp. 18–25). The pulmonary rehabilitation interventions were supervised by specialists. In cases wherein supervision was not available, a physiotherapist routinely visited the patients, maintained an exercise diary, and conducted follow-up phone calls to ensure patient adherence to the program. The pulmonary rehabilitation programs were conducted for 30–120 min per session for 2–3 days weekly over 6–8 weeks (eTable 4.3; Additional file 1 pp. 26–31).
Overall, 29 RCTs with 1,675 participants provided 6 MWT data (Figure 2a). Groups undergoing single-component rehabilitation programs, including ET (mean difference [MD], 44.88; 95% confidence interval [CI], 25.40 to 64.36), RT (MD, 53.13; 95% CI, 16.37 to 89.88), and RMT (MD, 34.09; 95% CI, 17.09 to 51.10), as well as those undergoing multi-component rehabilitation programs, such as ET+RT (MD, 57.24; 95% CI, 40.87 to 73.60), ET+RMT (MD, 32.11; 95% CI, 5.82 to 58.39), and ET+RT+RMT (MD, 72.09m; 95% CI, 48.16 to 96.02), were found to be significantly superior to the control group in terms of the 6 MWT distance (Figure 3a). Among the aforementioned programs, the ET+RT+RMT combination demonstrated the highest improvement in terms of the SUCRA and relative ranking (Figure 4a, eTable 7.1, Additional file 1 pp. 73).
Overall, 14 RCTs with 909 participants provided data on the SGRQ scores (Figure 2a); only two intervention arms demonstrated significant improvements. The ET+RT+RMT combination achieved the most substantial improvement in the SGRQ scores (MD, -24.51; 95% CI, -45.26 to -3.75). The ET+RT combination also achieved significant improvements in the SGRQ scores (MD, -20.92; 95% CI, -35.06 to -6.79; Figure 3a). The ET+RT+RMT and ET+RT combinations emerged as the top-ranking options in terms of the SUCRA and relative ranking (eTable 7.2, Additional file 1 pp. 74).
Overall, 12 RCTs with 719 participants provided data on the Borg scale ratings. Among the arms evaluated, only those that underwent single-component rehabilitation programs with RMT achieved significantly decreased Borg scale ratings as compared to the scores of the control group (MD, -2.37; 95% CI, -3.58 to -1.15; eTable 7.3, Additional file 1 pp. 75).
The analysis for detecting publication bias, conducted using funnel plots and the Egger’s test, did not reveal any issues, except for a potential small-study effect observed in trials related to the 6 MWT (Additional file 1 section 8, pp. 86–90). To assess the impact of this small study, we conducted a sensitivity analysis by excluding studies that may have contributed to this effect[18]. Our findings remained consistent and were unaffected by the exclusion of this study, indicating that it did not influence the overall results. Furthermore, no loop or design inconsistencies were detected (Additional file 1 section 9, pp. 99). The level of evidence for confidence in the 6 MWT results ranged from “moderate” to “high.” However, in terms of the SGRQ scores, the evidence certainty was found to range from “very low” to “low” (Additional file 1 section 10, pp. 100–104).
Post-discharge pulmonary rehabilitation of patients with acute exacerbation
Seven RCTs were identified that specifically examined the impact of supervised post-discharge pulmonary rehabilitation on patients admitted for acute exacerbation of COPD (eTables 4.5; Additional file 1 pp. 34–35). The supervised pulmonary rehabilitation sessions typically lasted for 30–120 min and were mostly performed 7 days a week over 6–8 weeks (eTables 4.6; Additional file 1 pp. 36–37).
Four RCTs involving 308 participants evaluated mortality outcomes of different rehabilitation groups (Figure 2b). The group that underwent the ET+RT combination exhibited non-significant superiority over the control group in terms of survival (odds ratio [OR], 0.64; 95% CI, 0.19 to 2.10). Five RCTs involving 357 participants evaluated the outcomes of readmission due to acute exacerbation of COPD (Figure 2b); the arm that underwent the ET+RT combination showed significant superiority over the control group in terms of readmission rate (OR, 0.44; 95% CI, 0.21 to 0.91).
Four RCTs involving 269 participants evaluated the 6 MWT distance (Figure 2b). The arm that underwent the ET+RMT combination showed significant superiority over the control group (MD, 51.09; 95% CI, 5.42 to 96.76) but the arm that underwent ET alone did not (MD, 80.4; 95% CI, -10.43 to 171.3; Figure 3b). However, ET alone had the highest probability of performing best in terms of the SUCRA (Figure 4b) and relative ranking (eTable 7.8, Additional file 1 pp. 80).
Six RCTs with 352 participants provided data on the SGRQ scores (Figure 2b). Arms that underwent ET (MD, -13.63; 95% CI, -27.08 to -0.17) and the ET+RMT combination (MD, -11.30; 95% CI, -16.38 to -6.21) demonstrated significantly improved SGRQ scores as compared to the scores of the control group (eFigure 6.2, Additional file 1 pp. 70). ET appeared to be the most favorable option in terms of the SUCRA (Figure 4b), and the relative ranking is shown in eTable 7.9 (Additional file 1 pp. 81). No significant publication bias was detected based on the funnel plot and Egger's test results (Additional file 1 section 8, pp. 91–94). However, loop inconsistencies were identified in the analyses of the 6 MWT and SGRQ (Additional file 1 section 9, pp. 99). The confidence level of evidence ranged from “very low” to “low” (Figure 3b, Additional file 1 section 10, pp. 105).
Pre-discharge pulmonary rehabilitation of patients with acute exacerbation
Seven RCTs were identified that focused on patients admitted for acute exacerbation of COPD who underwent supervised pulmonary rehabilitation during their hospital stay, i.e., prior to discharge (eTable 4.8, Additional file 1 pp. 40–41). The supervised rehabilitation sessions typically lasted for 30–120 min and were mostly conducted for 7 days a week over a period of 6–8 weeks (eTable 4.9, Additional file 1 pp. 42–43).
Three RCTs involving 221 participants evaluated the mortality outcomes (Figure 2c). Neither the arms that underwent multi-component rehabilitation nor those that underwent single-component rehabilitation programs showed superiority over the control group in terms of survival. Four RCTs involving 253 participants evaluated the outcomes of readmission due to acute exacerbation of COPD (Figure 2c). The arm that underwent rehabilitation with ET alone showed significant superiority over the control group (OR, 0.09; 95% CI, 0.01 to 0.56).
Seven RCTs enrolled 380 participants and provided data on the 6 MWT (Figure 2c). Only the arm that underwent ET alone (MD, 167.69m; 95% CI, 81.23 to 254.15) demonstrated a significant superiority over the control group in terms of the 6 MWT distance (Figure 3c). SUCRA analysis identified ET as the best option in terms of the readmission rate and 6 MWT distance (Figure 4c, Additional file 1 p. 82–84). Additionally, five RCTs involving 263 participants provided data on the Borg scale ratings. The arms that underwent ET alone (MD, -0.90; 95% CI, -1.56 to -0.24) achieved significantly reduced Borg scale ratings as compared to the scores of the control group. SUCRA analysis also indicated that ET was the most favorable option (Figure 4, eTable 7.9, Additional file 1 p. 85). No significant publication bias was detected based on the funnel plot or Egger's test (Additional file 1 pp. 95–99). However, loop inconsistencies were observed in the 6 MWT analysis (Additional file 1 section 9, pp. 99). The confidence level of evidence was “very low” due to incoherence being the major concern in this section (Figure 3c, Additional file 1 section 10, pp. 106).