Search results

This process is illustrated in Fig. 1. A total of 1813 studies were initially obtained by searching the relevant databases, and 936 studies were obtained after eliminating duplicates using Endnote X9. Titles and abstracts were read to obtain 108 studies that met the requirements, and 57 studies were finally identified for inclusion after reading the full text.

Fig. 1

Flowchart of the study selection process

The basic characteristics of the included studies are presented in Table S3. A total of 57 studies were included, all RCTs in English and Chinese, including 4294 patients aged between 40 and 88 years. Among them, 19 studies reported Taichi, 5 studies reported Wuqinxi, 14 studies reported Liuzijue, and 19 studies reported Baduanjin. We have presented the references of the included literature in the supplementary materials.

Risk of bias assessment and GRADE

The results of the risk of bias assessment of the included studies are shown in Figures S1. Among them, most studies were assessed as having “some concerns or low risk,” with only six studies being at a high risk of bias. The main source of the high risk was potential bias due to the selection of the reported results. In addition, three studies had a high risk of bias owing to the randomization process. These issues may be partly due to the lack of published protocol analysis plans in the study or a lack of clarity in the randomized reporting methods. Other sources of concern were potential deviations from the intervention measures described and concerns about the risks in the outcome measures.

The GRADE results are presented in Table S4. For FEV1, the quality of evidence ranged from low to moderate. Regarding the FEV1%, the quality of evidence ranged from low to moderate. Regarding the FEV1/FVC ratio, the quality of evidence ranged from very low to moderate. Owing to the potential risk of bias in the inclusion of studies in terms of randomization, allocation concealment, and blinding methods, downgrading was achieved. Moreover, the sample size included in the study was small, and the confidence interval was large, leading to downgrading.

Evaluation of safety

Only one study reported that patients experienced mild chest tightness and shortness of breath during treatment, which was relieved after a few minutes of rest [16]. In addition, no adverse reactions were reported in other studies.

Inconsistency test and PSRF parameters

The outcome indicators of this study were based on the indirect comparison of conventional treatment groups with each exercise group, and the lack of direct comparison between the exercise groups did not constitute a closed loop; therefore, an inconsistency test was not required [17]. In addition, the PSRF parameter for each indicator was equal to or close to 1, indicating good convergence and high reliability of the study results.

Network meta-analysis

A network evidence diagram for each indicator is shown in Fig. 2. The results of the network meta-analysis are shown in Fig. 3. A total of 57 studies were included, with five interventions. The dots in the diagram represent interventions; larger dots represent larger sample sizes of the included studies; a solid line connecting two dots indicates a direct link between two interventions; the thicker solid line represents a more significant number of studies between the two interventions.

Fig. 2

Network evidence diagram. CT = conventional treatment

Fig. 3

Network meta-analysis results. CT = conventional treatment

FEV1

In terms of FEV1, 35 studies with 2769 patients were included. The results showed that Baduanjin was superior in improving FEV1 compared to the control group, and the difference was statistically significant (WMD = -0.27, 95% CI [0.40, 0.14], p < 0.05). Moreover, Baduanjin was superior to Taichi (WMD = -0.19, 95% CI[-0.36,-0.02], p < 0.05), Wuqinxi (WMD = -0.09, 95% CI [0.17, 0.33], p < 0.05), and Liuzijue (WMD = -0.03, 95% CI[-0.18,-0.23], p < 0.05), indicating that Baduanjin was better than the other interventions at improving FEV1.

FEV1%

In terms of FEV1%, 45 studies with 3371 patients were included. The results showed that Liuzijue was superior in improving FEV1% compared to the control group, and the difference was statistically significant ([WMD = -8.60, 95% CI [(-11.16, -5.92], p < 0.05). Moreover, Liuzijue was superior to Taichi (WMD = -6.07, 95% CI [-9.70, -2.38], p < 0.05), Wuqinxi (WMD = -5.73, 95% CI [-10.41, -0.72], p < 0.05), and Baduanjin (WMD = -3.56, 95% CI [-7.01, 0.03], p < 0.05), indicating that Liuzijue was better than the other interventions for improving FEV1%.

FEV1/FVC

In terms of FEV1/FVC, 41 studies with 2700 patients were included. The results showed that Liuzijue was superior in improving FEV1/FVC compared with that of the control group, and the difference was statistically significant (WMD = -7.35, 95% CI [-10.02, -4.51], p < 0.05). Moreover, Liuzijue was superior to Taichi (WMD = -3.89, 95% CI [-7.58, 0.02], p < 0.05), Wuqinxi (WMD = -0.96, 95% CI [-5.89, 4.39], p < 0.05), and Baduanjin (WMD = -3.42, 95% CI [-7.15, 0.45], p < 0.05, which indicated that Liuzijue was better than the other interventions for improving FEV1/FVC.

Cumulative probability ranking

The results of the cumulative probability ranking are shown in Fig. 4.

Fig. 4

Cumulative probability ranking. CT = conventional treatment

The results for FEV1 were ranked from highest to lowest and divided into five ranks: from 1 to 5. The ranking results were as follows: Baduanjin > Liuzijue > Wuqinxi > Taichi > conventional treatment.

Similarly, the FEV1% results were ranked from highest to lowest and categorized into five levels: from level 1 to 5. The specific results were as follows: Liuzijue > Baduanjin > Wuqinxi > Taichi > conventional treatment.

For FEV1/FVC, the results were ranked from highest to lowest and divided into five levels, from 1 to 5. The specific ranking results were as follows: Liuzijue > Wuqinxi > Baduanjin > Taichi > conventional treatment.

Subgroup analysis

Considering that differences in intervention time might have affected the study results, subgroup analyses were performed based on different intervention durations. Two subgroups were established: the duration of intervention < 6 months group and the duration of intervention ˃ 6 months group. The results are summarized in Supplementary Table S5. In the duration of intervention < 6 months group, Liuzijue showed superior results than the conventional treatment in improving FEV1. Moreover, Liuzijue was more effective than Baduanjin, Wuqinxi, and Taichi. In contrast, in the duration of intervention ≥ 6 months group, Baduanjin exhibited superior results than the conventional treatment in improving FEV1 and was more effective than Liuzijue, Wuqinxi, and Taichi. Based on the cumulative probability ranking results (Figure S2A, B), Liuzijue had better outcomes when the duration of intervention was < 6 months, whereas Baduanjin was considered superior for interventions lasting ≥ 6 months.In the duration of intervention < 6 months group, the results showed that improved FEV1% was observed with Liuzijue than the conventional treatment, and more effective Baduanjin, Wuqinxi, and Taichi. In addition, in the duration of intervention ≥ 6 months group, the results showed that Liuzijue was also better than conventional treatment and was superior to Baduanjin, Wuqinxi, and Taichi. The results of the cumulative probability ranking (Figure S2C, D) showed that Liuzijue was more effective in improving FEV1%. In the duration of intervention < 6 months group, the results showed that Liuzijue was superior to conventional treatment in improving FEV1/FVC, and was also superior to Baduanjin, Wuqinxi, and Taichi. In addition, in the duration of intervention ≥ 6 months group, the results showed that Liuzijue was also better than conventional treatment and was superior to Baduanjin, Wuqinxi, and Taichi. The results of the cumulative probability ranking (Figure S2E, F) showed that Liuzijue was the best option for improving FEV1/FVC.

Furthermore, subgroup analyses were performed according to the different baseline pulmonary functions (level of airway obstruction) of the COPD patients in the study, considering that the different baseline pulmonary functions could have a significant impact on the results of the study. Two subgroups were created based on the Global Strategy for Prevention, Diagnosis and Management of COPD: 2024 Report's classification of the severity of COPD patients: the moderate group and the severe group [18]. The results are summarized in Supplementary Table S6. In the moderate group, Liuzijue showed superior results than the conventional treatment in improving FEV1%. Moreover, Liuzijue was more effective than Baduanjin, Wuqinxi, and Taichi. In the severe group, the results showed that Liuzijue was also better than conventional treatment and was superior to Baduanjin, Wuqinxi, and Taichi. The results of the cumulative probability ranking (Table S7) showed that Liuzijue was more effective in improving FEV1%.

Finally, considering that the frequency of interventions might have affected the study results, subgroup analyses were performed based on the frequency of interventions. One subgroup was established: the frequency of intervention ≥ three times a week group. The results are summarized in Supplementary Table S8. In the frequency of intervention ≥ three times a week group, Liuzijue showed superior results than the conventional treatment in improving FEV1, FEV1% and FEV1/FVC. Moreover, Liuzijue was more effective than Baduanjin, Wuqinxi, and Taichi. The results of the cumulative probability ranking (Table S9) showed that Liuzijue was more effective in improving FEV1, FEV1% and FEV1/FVC.

Sensitivity analysis

Sensitivity analyses were performed by removing studies with sample sizes less than 30 and articles exhibiting a high risk of bias and lacking methodological rigor to verify the robustness of our results.

The results of the pairwise comparisons (Table S10, S11) and cumulative probability ranking (Figure S3, Table S12) were not significantly different from the overall results. Therefore, we conclude that our results are stable and reliable.

Publication bias

As shown in Fig. 5, owing to the small sample size and publication bias, the study findings need to be carefully explained.

Fig. 5