Comparative effectiveness of non-invasive interventions for stage I post-stroke shoulder–hand syndrome: a network meta-analysis of randomized controlled trials

Background To compare the effectiveness of non-invasive interventions for pain, upper-limb motor function, and activities of daily living in patients with stage I post-stroke shoulder–hand syndrome (SHS) using network meta-analysis. Methods Randomized controlled trials evaluating non-invasive interventions for stage I SHS were systematically retrieved. Outcomes included visual analogue scale (VAS), Fugl–Meyer Assessment–Upper Extremity (FMA-UE), and Barthel Index (BI). Mean differences (MDs) or standardized mean differences (SMDs) with 95% confidence intervals (CIs) were estimated, and surface under the cumulative ranking curve (SUCRA) values were calculated. Results Thirty-six RCTs involving 2,769 participants were included. The VAS, FMA-UE, and BI networks included 24, 28, and 18 RCTs, respectively. Compared with conventional treatment (CT), VAS scores were significantly reduced by kinesiology taping (KT) (MD = −1.95, 95% CI − 2.85 to −1.05), air pressure therapy (AT) (MD = −2.30, −4.31 to −0.29), Traditional Chinese Medicine External Herbal Therapy (TCM-EHT) (MD = −1.36, −2.08 to −0.64), Traditional Chinese Medicine External Physical Stimulation (TCM-EPS) (MD = −1.43, −2.36 to −0.50), and moxibustion (MD = −1.52, −2.53 to −0.51). FMA-UE improved with TCM-EPS (SMD = 2.12, 0.03 to 4.21), TCM-EHT (SMD = 2.07, 0.62 to 3.52), and KT (SMD = 2.32, 0.04 to 4.61). BI improved with TCM-EHT (SMD = 3.64, 1.76 to 5.51), TCM-EPS (SMD = 2.55, 0.60 to 4.50), and KT (SMD = 1.98, 0.27 to 3.70). However, 95% prediction intervals for all statistically significant comparisons crossed the null. AT, KT, and TCM-EHT ranked highest for VAS, FMA-UE, and BI, respectively. No statistically significant global inconsistency was detected, although localized disagreement was identified for CT versus TCM-EHT in the VAS network. Conclusion Non-invasive interventions showed outcome-specific potential benefits, but no intervention demonstrated consistent superiority across all outcomes. Although AT and KT ranked favorably for pain and KT, TCM-EPS, and TCM-EHT for functional outcomes, wide prediction intervals indicated substantial uncertainty. Given methodological limitations and imprecision, findings should be interpreted cautiously and confirmed by adequately powered multicentre RCTs. Systematic review registration The publicly accessible registration record is available through the PROSPERO database: https://www.crd.york.ac.uk/PROSPERO/view/CRD420251276361 . The systematic review was registered in PROSPERO (registration number: CRD420251276361).