Yonghua Liu, Haoran He, Jiajun Lan
Low-certainty evidence showed materially different estimates by comparator: Control-NoInt favored the practices, whereas the small Control-Int subgroup was inconclusive. The pooled SMD is a heterogeneous average and should not be interpreted as a uniform benefit or as superiority over usual care or another intervention control. High heterogeneity, study-level risk of bias, geographic concentration in China, and indirect dose modeling do not support a precise dose recommendation or routine implementation without confirmatory trials.
OBJECTIVE: To estimate the association between traditional Chinese mind-body practices and anxiety symptoms in university students and to explore, rather than prescribe, possible dose-response patterns.
METHODS: We searched PubMed, Web of Science, PsycINFO, the Cochrane Library, and China National Knowledge Infrastructure (CNKI) from inception through May 18, 2026. Randomized controlled trials of Taijiquan, Baduanjin, qigong, and related practices were synthesized as standardized mean differences (SMDs) in a three-level random-effects model. Heterogeneity, prespecified subgroup and meta-regression analyses, and an exploratory Bayesian dose-response model were examined.
RESULTS: Sixteen RCTs contributed 18 dependent effect estimates. Comparator subgroup estimates differed significantly (P-interaction = 0.001): Control-NoInt (waitlist or no intervention) yielded SMD = - 0.92 (95% CI - 1.28 to - 0.56), whereas Control-Int (usual care or another intervention control) yielded SMD = - 0.09 (95% CI - 0.38 to 0.56; p = 0.708). The Control-Int estimate was based on only six effect estimates and was underpowered, with a confidence interval compatible with benefit, no effect, or modest harm. Across all comparators, the pooled estimate was SMD = - 0.56 (95% CI - 0.92 to - 0.20; p = 0.002), with substantial heterogeneity (I² = 94.1%). The Bayesian curve had a model-estimated turning point near 20.7 cumulative hours; sparse dose coverage, wide uncertainty at the extremes, and between-study heterogeneity preclude treating this value as an optimal prescription.
CONCLUSIONS: Low-certainty evidence showed materially different estimates by comparator: Control-NoInt favored the practices, whereas the small Control-Int subgroup was inconclusive. The pooled SMD is a heterogeneous average and should not be interpreted as a uniform benefit or as superiority over usual care or another intervention control. High heterogeneity, study-level risk of bias, geographic concentration in China, and indirect dose modeling do not support a precise dose recommendation or routine implementation without confirmatory trials.