Effects of traditional Chinese mind, body exercises on depressive symptoms in middle-aged and older adults: a multilevel meta-analysis with exploratory dose, response and machine learning analyses

ObjectiveThis study evaluated the effects of traditional Chinese mind, body exercises on depressive symptoms in middle-aged and older adults and examined potential dose-related and study-level moderators.MethodsPubMed, Web of Science, Embase, Cochrane Library, CNKI, VIP, and Wanfang were searched through May 3, 2026, for randomized controlled trials. A three-level random-effects model was used to account for dependent effect sizes. Subgroup analyses, meta-regression, exploratory dose, response analyses, and an XGBoost-based machine-learning extension were conducted. The risk of bias and the certainty of evidence were assessed using RoB 2 and GRADE.ResultsThirty randomized controlled trials were included. Traditional Chinese mind, body exercises were associated with a statistically significant reduction in depressive symptoms among middle-aged and older adults (SMD = −0.88; 95% CI, −1.18 to −0.58; p < 0.001). Sensitivity and supplementary analyses generally supported the robustness of the main findings; however, substantial heterogeneity and risk of bias contributed to low certainty of the evidence. Subgroup analyses indicated that most estimates favored traditional Chinese mind, body exercises, with statistically significant subgroup differences observed only by country. Meta-regression showed that older age was associated with a smaller intervention effect (β = 0.03, p = 0.025). Exploratory dose, response analyses did not identify a statistically supported optimal dose, although descriptive patterns were observed. XGBoost and SHAP analyses suggested that age, scale type, total dose, and session duration may be important study-level features, though these findings were exploratory.ConclusionTraditional Chinese mind, body exercises may reduce depressive symptoms in middle-aged and older adults; however, the certainty of the evidence is low because of substantial heterogeneity and risk of bias. Dose, response and machine-learning findings should be considered exploratory. Further high-quality randomized controlled trials are warranted to confirm these findings and refine intervention parameters.Systematic review registrationhttps://www.crd.york.ac.uk/PROSPERO/view/CRD420261388878, identifier PROSPERO (CRD420261388878).