Comparative efficacy of various exercise types and doses for anxiety in older adults: A network and dose-response meta-analysis Authors: Haotian Jin, Yinan Zhang, Yi Sun, Yan Shao Published: 2026 ABSTRACT AbstractBackgroundAnxiety significantly impacts quality of life among older adults, particularly those with chronic conditions. Exercise has been established as an effective non-pharmacological intervention for anxiety, while the optimal type and dosage of exercise remain unclear. This study aimed to determine which exercise modality yields the largest efficacy for reducing anxiety and to model the optimal METs-min/week dose for anxiety reduction.MethodsThis systematic review and network meta-analysis searched PubMed, Web of Science, Embase, Medline, and Cochrane databases from database inception to 15 January 2025. We included randomized controlled trials (RCTs) investigating exercise interventions among older adults (≥ 60 years) and extracted data from dose metrics (METs-min/week), treatment regimens, intensity, and study duration.ResultsTwenty-three RCTs with a total of 2093 participants were included in the analysis. Tai Chi emerged as the most effective exercise for reducing anxiety compared with the control group (SMD = − 1.37, 95% CI: −2.03 to − 0.71). Resistance training (RT), Yoga, Qigong and aerobic exercise (AE) also demonstrated relative efficacy. However, mixed exercise (MIX) did not yield significant effects. Dose-response analysis indicated that Tai Chi demonstrated optimal efficacy for alleviating anxiety in older adults at a dose of 500 METs-min/week, with statistically significant effects observed at a relatively lower dose (340 METs-min/week).ConclusionOur findings indicate that Tai Chi is the most effective exercise for alleviating anxiety in older adults. In contrast, MIX showed no significant effects.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12877-026-07869-x.Keywords: Anxiety, Older adults, Exercise, Network meta-analysis, Dose-response, Tai Chi Introduction Anxiety is a prevalent condition among older adults, characterized by excessive worry, tension, and physiological changes that significantly affect quality of life and physical health [1–3]. Although drug therapy and psychological treatments are effective for anxiety management, they may be unsuitable for older adults due to adverse effects and accessibility issues [4, 5].Existing RCTs suggest that although pharmacological treatments (e.g., SSRIs) are effective in reducing anxiety in some populations, their efficacy is not universal [6, 7]. Therefore, non-pharmacological options like exercise have attracted attention, and previous researches have recognized exercise as an effective and beneficial adjunctive strategy [6, 8]. In recent years, various exercise modalities have demonstrated positive effects in improving physical and psychological health. Mind–body exercises (e.g., Yoga, Tai Chi) improve stress regulation, balance, and mental well-being [9]. RT has emerged as a key intervention, particularly for older adults, to counteract age-related declines in muscle strength and to improve functional performance, quality of life, motor control and overall health outcomes [10]. Higher levels of physical activity are associated with better mental health, and prior reports have demonstrated that exercise reduces anxiety through neurotrophic, inflammatory and cognitive-behavioral pathways [11, 12]. Mind-body exercises (MBEs), such as Yoga and Tai Chi, are thought to alleviate anxiety via multi-level biosemiotic synchronization. At the neurological level, MBEs promote coherence between prefrontal (cognitive) and sensory-motor cortical regions, improving information processing. Beyond the nervous system, this synergy engages endocrine and immune pathways to modulate stress physiology. Moreover, group-format MBEs elicit a “collective physiology” through which the synchronization of movement aligns participants’ biological rhythms, promoting social cohesion and enhancing collective resilience [9]. RT alleviates anxiety among older adults through integrated neurophysiological and psychosocial pathways. By enhancing motor-neuron excitability and augmenting neural drive, RT yields adaptations that underpin improved motor function. At the same time, RT directly enhances psychological well-being by attenuating symptoms of depression and anxiety, and the group-based format promotes social support that counters isolation and loneliness [10]. AE may influence the body’s principal stress-response systems via HPA Axis Adaptation, Autonomic Nervous System (ANS) Regulation and Neuroprotective Effects. These processes support brain health in the prefrontal cortex and hippocampus—regions critical to emotional regulation [13]. Although multiple meta-analyses have demonstrated the benefits of exercise in older adults, methodological limitations remain. Earlier systematic reviews were often restricted to pairwise meta-analysis, which could only compare exercise against control conditions without ranking the relative efficacy of different exercise modalities [14]. In addition, one investigation of exercise effects on anxiety did not specifically target older adults [15]. More recently, a pairwise and network meta-analysis in postmenopausal women identified mind-body exercise as the most effective exercise modality for reducing depression and anxiety. However, this study did not incorporate a dose-response analysis, leaving the minimal and optimal exercise dosage unclear [16]. To address these remaining evidence gaps, we undertook a comprehensive network and dose-response meta-analysis. This approach enables direct comparison and ranking of diverse exercise interventions to identify the most effective modality, while precisely quantifying the minimal and optimal dosages to support evidence-based exercise prescription. The primary aim of this study was to determine which exercise modality yields the largest efficacy for reducing anxiety and to model the optimal METs-min/week dose for anxiety reduction. Methods This meta-analysis adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for network meta-analyses (PRISMA-NMA) and the Cochrane Handbook. The review was registered (PROSPERO: CRD420251119717) and reported following the PRISMA checklist. Search strategy We conducted a thorough search in PubMed, Web of Science, Embase, Medline, and Cochrane databases from database inception to 15 January 2025. A combination of medical subject heading (MeSH) terms and relevant synonyms was used to develop the search strategy. The MeSH terms included in the search strategy were “exercise”, “anxiety” and “aged”. For each concept, synonyms were combined using the Boolean operators OR (within each concept) and AND (between concepts), such as (“Anxiety”[Mesh] OR Angst OR Nervousness) AND (“Aged”[Mesh] OR elderly). The complete search strategy is detailed in Supplementary 1.1. We screened the reference lists of relevant articles. Two investigators performed Title/abstract and full-text screening independently, and discrepancies were resolved by discussion with a third reviewer. Eligibility criteria Inclusion criteria followed the PICOS standard: (1) participants’ mean age ≥ 60 years; (2) healthy or with stable chronic conditions (e.g., hypertension, type 2 diabetes, osteoarthritis, stable coronary heart disease, early or middle stage Parkinson’s disease (Hoehn & Yahr I–III) or mild COPD); (3) exercises were divided into six types; (4) control groups included no intervention, usual care, wait-list, maintain daily activities and health education; (5) outcomes measured anxiety symptoms; (6) published RCTs (e.g., individual, cluster, or first-phase crossover designs); (7) clinical diagnosis or subclinical anxiety symptoms measured by validated scales. Exclusion criteria: (1) single-session (acute effect); (2) indistinct exercise types/doses; (3) no baseline anxiety assessment; (4) change scores for anxiety were not calculable; (5) participants’ mean age < 60 years; (6) pilot, feasibility, exploratory, preliminary, formative, proof-of-concept, or scoping studies. (7) severe or unstable somatic conditions, or major psychiatric conditions (e.g., active cancer under treatment, recent (< 6 months) stroke or myocardial infarction, severe heart failure (e.g., NYHA III–IV), severe COPD requiring long-term oxygen therapy, schizophrenia or advanced Parkinson’s disease (Hoehn & Yahr IV–V)). Data extraction Two researchers (HT.J and Y.S) independently extracted data from included studies. Any discrepancies were resolved with a third author (S.Y) by consultation. Extracted data included the following information: author information, publication year, detailed characteristics (age, gender, subject number, physical condition, baseline information and adherence), intervention characteristics (type, duration, sessions, frequency and intensity) and anxiety outcomes with cut-off scores. We verified that for all anxiety scales, including Self-Rating Anxiety Scale (SAS), Hospital Anxiety and Depression Scale-Anxiety Subscale (HADS-A), State-Trait Anxiety Inventory (STAI), Zhong Shan COPD Questionnaire-Anxiety Subscale (ZS-COPD-Q), Depression Anxiety Stress Scales (DASS), Anxiety and Depression Goldberg Scale (ADGS), Beck Anxiety Inventory (BAI), Profile of Mood States (POMS), Templer Death Anxiety Questionnaire (TDAQ), a higher score denotes a greater severity of anxiety. Intervention intensity and dosage calculation based on American College of Sports Medicine (ACSM) guidelines (e.g., low:9–11 Borg RPE, 57–64% HR max or 30–50% 1RM; moderate:12–13 Borg RPE, 64–76% HR max or 50–70% 1RM; vigorous: 14–17 Borg RPE, 76–96% HR max or 70–85% 1RM), and the METs value for relative intensity range was used to calculate weekly volume of exercise. Duration was extracted as the actual minutes per session reported in each study [17]. Weekly volume was calculated as METs value × minutes per session × weekly frequency. For exercises (e.g., Tai chi, Yoga, Qigong) where numerical intensity parameters are typically not reported, we directly used the standardized METs values assigned to these activities in the Compendium of Physical Activities [18]. SDs were derived from SEs, CIs, t-values, quartiles, ranges, or p-values if it was not given [19]. Corresponding authors were responsible for missing data; if there was no reply, the study would be excluded. Risk of bias and certainty of evidence Risk of bias was assessed by two reviewers (YN.Z and Y.S.) with Cochrane RoB 2 [20], which evaluated five perspectives (randomization process, deviations from the intended interventions, missing outcome data, measurement of the outcome and selection of the reported result) and overall bias. Disagreements were resolved by a third investigator and adjudicated by HT.J. The certainty of evidence was assessed according to the GRADE method using the CINeMA web application [21, 22]. Measures of treatment effect In order to assess changes of anxiety at the end of the intervention, the primary outcomes were derived as the mean (M), standard deviation (SD), and sample size (n). The assessment completed at the end of the intervention period would be evaluated when studies reported several different post-intervention anxiety outcomes (e.g., mid-treatment, follow-up, or secondary phases in crossover designs). Exercise categories were divided into six groups: AE (aerobic exercise), RT (resistance training), Tai Chi, Qigong, yoga, and MIX (a combination of aerobic, resistance, balance, coordination, or stretching exercises). Based on both type and dosage, each intervention was additionally classified. Dosage was defined by the total energy expenditure, which was measured in Metabolic Equivalent of Task (MET) and calculated on the basis of exercise duration, frequency, and intensity. The dose was quantified in METs-minutes per week [23, 24] and classified into levels of 0 (control group), 250, 500, 750, 1000, or 1200 for network connectivity [25]. Statistical analysis Frequentist NMA was conducted in Stata (version 17.0) utilizing standardized mean differences (SMDs) based on the change score (end-point minus baseline score) calculated as Hedges’ g with a small-sample correction. The random-effects models were estimated using restricted maximum likelihood (REML). We adopted both inverse-variance (i) and common-effect (c) models to calculate general effect estimates. For better visualization, forest plots for pairwise and network-level treatment effects were generated. A network plot was presented to visualize the geometry and connectivity of comparisons between different treatments. The ranking probability of interventions was estimated by a random-effects model, and the Surface Under the Cumulative Ranking (SUCRA) values were computed to summarize the ranking hierarchy of treatments. League table was drawn to illustrate relative effect estimates across all treatment pairs, and then it was sorted and exported to Excel. For local inconsistency, we used the node-splitting method, in which a p-value < 0.05 noted significant inconsistency between direct and indirect estimates. Besides, loop-specific inconsistency was assessed using the inconsistency factor plot. The 95% confidence interval of the inconsistency factor included zero, indicating that inconsistency was not significant. We used Egger’s test and comparison-adjusted funnel plots to evaluate publication bias and visualize it [26]. Trim-and-fill method was used to further assess the potential small-study effects and publication bias, ensuring the robustness of our findings. A Bayesian dose-response network meta-analysis (MBNMA) was performed through the MBNMAdose and multinma packages in R (version 4.5, R Foundation for Statistical Computing, Vienna, Austria). MBNMA combines the advantages of conventional network meta-analysis with the dose-response model and is capable of estimating treatment effects across different doses and agents. The connectivity and feasibility of the treatment network were visually screened at both treatment and agent levels. The network geometry was inspected to make sure all treatments and dose levels were properly connected. (Supplementary 2.1, 2.2) We fitted a random-effects network meta-analysis model and used SMD as the effect measure. For evaluation of local inconsistency, a node-splitting model was performed. The density plots (Supplementary 2.3) of the split comparisons were used to assess consistency, and Supplementary 2.4 demonstrated the different responses of different doses for different treatments, respectively. Model consistency and UME (unrelated mean effect) were assessed by comparing the residual deviance, DIC and other parameters between consistent and inconsistent models, and we drew a plot of posterior deviance distributions [27]. (Supplementary 2.5, 2.6) In order to investigate the dose-response relationship, restricted cubic splines were used, which allow flexible modeling of non-linear trends over different doses. Model fit was evaluated using predicted values and graphical summaries that were overlaid on observed data. We performed dose-response network meta-analysis by the MBNMAdose package in R. Model fitting was conducted through Markov Chain Monte Carlo (MCMC) methods carried out in JAGS (Just Another Gibbs Sampler), accessed via the rjags interface. Through visual inspection of trace plots and residual diagnostics, we assessed the convergence. The default prior distributions supported by MBNMAdose were used when no prior information was available. Sensitivity analysis To assess the robustness of the findings, we conducted a sensitivity analysis by excluding high-risk studies and reanalyzing the effect sizes for the outcome. Additionally, a sensitivity analysis of the dose-response relationship for exercise interventions on anxiety was performed. The results from these analyses were compared with the original analysis to evaluate the robustness and consistency of the results. Results Study selection and characteristics A total of 23 studies were included for analysis. The details of exercise types across these studies are listed below: 4 studies on Qigong, 3 studies on MIX, 5 studies on RT, 4 studies on Tai Chi, 6 studies on Yoga, and 5 studies on AE. Figure 1 demonstrates the flowchart of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) systematic review. After the first step of screening 19,531 studies, 3357 full-text articles were assessed. Finally, 23 RCTs including 6 types of exercise interventions were included in the analysis. Flow chart of the search process The included studies involved 2093 participants. The mean age of participants across all included studies was 68.38 years. The baseline mean age ranges from 60.00 to 84.26 years. The mean ages in different intervention groups were as follows: Control: 67.35 years; Tai Chi: 67.29 years; RT: 65.34 years; Yoga: 66.93 years; AE: 69.44 years; Qigong: 65.13 years; and MIX: 77.17 years. It should be noted that most studies did not adequately report or control for the use of antianxiety medications. The intervention duration ranged from 8 to 42 weeks, and the exercise frequency ranged from 1 to 7 times per week. The intensity of the exercises was tailored to the type of intervention. Further characteristic details for included studies were presented in Supplementary 1.2. Risk of bias and certainty of evidence Among the 23 included RCTs, 17(73.9%) demonstrated a low risk of bias. Two (8.6%) trials were not blinded to participants or personnel, and three (12.9%) studies deviated from the intended intervention due to the trial context. In one study (4.3%), outcome assessors were aware of the intervention received by study participants. Among all randomized clinical studies, the risks in terms of measurement of the outcome and missing outcome data were relatively minimal. Figure 2 shows the risk-of-bias graph, and we present details in Supplementary 1.3. Risk of bias graph Additionally, we graded the evidence in the network meta-analysis using CINeMA. For the main results, we assessed the credibility of 90.48% of the evidence as low. CINeMA confidence ratings are shown in Supplementary 1.4. Pairwise and network meta-analyses We conducted a frequentist network meta-analysis with a consistency model, which used a multivariate random-effects framework. Figure 3 shows the network plot. Network plot The results show that all exercise interventions, in addition to MIX, significantly reduced anxiety symptoms. To assess global inconsistency, we used the design-by-treatment interaction model, which showed no significant inconsistency between direct and indirect comparisons (χ²(4) = 4.24, p = 0.37). We performed node-splitting analyses to explore local inconsistencies between direct and indirect evidence. This enabled us to test inconsistency at the comparison level across treatment pairs in terms of both direct and indirect evidence. Node-splitting did not detect significant inconsistency between groups. Table 1 presents the league table visualizing the anxiety outcome of pairwise comparisons of SMD across different exercise interventions. The analysis indicated that Tai Chi emerged as the most effective anxiolytic exercise, with a statistically significant improvement compared with the control group (SMD = − 1.37, 95% CI: −2.03 to − 0.71). What’s more, Tai Chi also significantly outperformed other interventions, including AE and MIX, which suggested a relatively high grade of therapeutic benefit according to Table 1.The SUCRA for anxiety is presented in Fig. 4. Table 1League table of anxiety symptom. Boldface indicates statistically significant resultsTAICHI-0.61 (-1.47,0.24)RT-0.65 (-1.46,0.15)-0.04 (-0.63,0.55)YOGA-0.64 (-1.53,0.25)-0.03 (-0.83,0.78)0.01 (-0.74,0.77)QIGONG-0.84 (-1.63,-0.04)-0.22 (-0.90,0.46)-0.18 (-0.76,0.40)-0.20 (-0.94,0.55)AE-1.00 (-1.96,-0.04)-0.38 (-1.26,0.49)-0.34 (-1.16,0.48)-0.36 (-1.27,0.56)-0.16 (-0.90,0.59)MIX-1.37 (-2.03,-0.71)-0.75 (-1.29,-0.21)-0.71 (-1.17,-0.25)-0.73 (-1.32,-0.13)-0.53 (-0.97,-0.09)-0.37 (-1.06,0.33)CON League table of anxiety symptom. Boldface indicates statistically significant results SUCRA for anxiety symptom Note: SUCRA, the surface under the cumulative ranking curve RT also demonstrated a significant reduction in anxiety symptoms (SMD = − 0.75, 95% CI: −1.29 to − 0.21). Yoga, Qigong, and AE all alleviated anxiety symptoms: Yoga (SMD = − 0.71, 95% CI: −1.17 to − 0.25), Qigong (SMD = − 0.73, 95% CI:−1.32 to − 0.13), AE (SMD = − 0.53, 95% CI:−0.97 to − 0.09). In contrast, MIX showed only a low and statistically non-significant effect (SMD = − 0.37, 95% CI: −1.06 to 0.33), suggesting limited benefits when integrating various exercises. Overall, Tai Chi proved to be the most effective strategy for anxiety reduction. Mixed exercise did not yield significant effects on relieving anxiety symptoms. The detailed comparison of outcomes between groups is provided in Fig. 5. Pairwise comparison forest plot The comparison-adjusted funnel plot was highly symmetrical, and Egger’s test did not examine a statistically significant small-study effect (p = 0.087), which suggested a low risk of publication bias (Supplementary 3.1). The trim-and-fill method was employed to further detect potential small-study effects and publication bias. The results demonstrated that the funnel plot remained symmetric without imputing any studies, supporting the robustness of the findings. (Supplementary 3.2). We conducted loop-specific heterogeneity analysis to further test local inconsistencies within the treatment network. Three evaluated loops demonstrated good consistency between direct and indirect estimates. Loop-specific plot is presented in Supplementary 3.3. Dose response We detected a nonlinear dose–response relationship between total physical activity (PA) dose and anxiety symptoms among the older adults (Fig. 6). The most effective dose was observed at 550 METs-min/week (SMD: − 0.88; 95% CrI: − 1.29 to − 0.48). The dose range that showed significant decreases in anxiety symptoms was 170 to 860 METs-min/week, indicating a U-shaped response beyond moderate intensity. It should be noted that doses exceeding 870 METs-min/week did not show significant therapeutic benefits, and the effects were gradually diminishing. Posterior means of treatment effects by dose and intervention type can be found in Supplementary 3.4. Dose-response relationship between overall exercise dose and anxiety Different types of exercise resulted in different nonlinear dose relationships between exercise dose and anxiety symptoms in the older adults. Figure 7 shows that Tai Chi, RT, Yoga, Qigong and AE all presented a non-linear improvement in anxiety symptoms with increasing exercise dose. Tai Chi showed the optimal dose at 500 METs-min/week (SMD: -2.27; 95% CrI (-3.15 to -1.38)) in this study, and the minimum effective dose was 340 METs-min/week (SMD: -0.87; 95% CrI (-1.66 to -0.08)). In terms of AE, RT, Qigong and Yoga, the minimum effective doses were 460 METs-min/week, 400 METs-min/week, 700 METs-min/week and 360 METs-min/week. Dose-response relationship of different types of exercises Sensitivity analysis The results of excluding studies with a high risk of bias indicated that the effect size did not change significantly compared to the primary findings. This suggests that our main conclusions are robust. Moreover, the dose-response results demonstrated that the effective dose for improving anxiety was between 180 and 840 METs-min/week, with the optimal dose at 540 METs-min/week (SMD: − 0.85, 95% CrI: − 1.26 to − 0.40). The details of sensitivity analysis were provided in Supplementary 4.1. Discussion This network meta-analysis evaluates the effectiveness of different exercises and doses for relieving anxiety in older adults. Based on data from 23 RCTs, we compared six different types of exercise: (e.g., AE, RT, Tai Chi, Qigong, yoga, and MIX). Tai Chi emerged as the most effective intervention, especially at low-to-moderate doses, demonstrating its potential as a non-pharmacological method for alleviating anxiety in older adults. Tai Chi demonstrated superior efficacy for alleviating anxiety compared to AE, MIX, and control groups. Our findings are consistent with earlier research reporting greater improvements in anxiety for the Tai Chi group compared with the AE group [28]. Tai Chi improved both physical and psychological conditions, which might contribute to its anti-anxiety effects. A previous study has also found Tai Chi can consistently promote changes in heart rate variability (HRV) and perceived stress [29]. Another possible explanation for the superior effects of Tai Chi compared with AE is the higher and more consistent adherence. This factor may also explain its greater effect compared to MIX, as Tai Chi consists of a gentler, low-impact meditative sequence of movements that is sustainable in the long term [30]. Other exercise modalities, including AE, RT, Qigong and yoga, also showed statistically significant improvements in anxiety symptoms. AE also reduces anxiety through potential mechanisms such as increased neurogenesis, endorphin release, and improved cardiovascular function [31]. Yoga integrates controlled breathing, physical postures and meditation, therefore facilitating relaxation and reducing cortisol levels [32]. RT also reduces anxiety and depression through integrated neurophysiological and psychosocial pathways, such as enhanced self-efficacy and neurochemical modulation [10, 33]. Qigong, a traditional Chinese exercise similar to Tai Chi that involves slow movements and a strong focus on mindfulness and breath, showed a moderate effect [34]. In contrast, MIX, which combined various types of exercise (e.g., AE, RT, balance), did not show significant improvements in anxiety symptoms. This may be attributed to a lack of specificity or the decrease of psychological benefits when different exercise modalities are integrated without a targeted focus. We also found distinct dose–response relationships across exercise modalities. AE and RT showed efficacy within a range of 500–750 METs-min/week, while Qigong was also effective at 550–700 METs-min/week. Yoga was effective at a comparatively moderate dose (400–500 METs-min/week). It should be noted that Tai Chi was effective even at lower doses (350–500 METs-min/week, equivalent to 40 min per session, 3 times per week). These differences may reflect the physiological and psychological mechanisms underlying each exercise. AE and RT generally need higher energy expenditure and demonstrate a higher effective threshold. In contrast, Tai Chi and Yoga present a gentler and low-intensity character. Both practices involve mental concentration, control of body tension and breathing control, which may contribute to mental benefits. In summary, this study highlights Tai Chi as a particularly effective and accessible exercise modality for reducing anxiety in older adults. Other forms of exercise, such as AE, RT, Qigong and Yoga, also demonstrated benefits and should be adopted on the basis of individual preferences and capabilities. Limitations and future directions Although we applied stringent criteria to screen the included studies and data, some limitations remain. Firstly, some studies did not report or control for medication use, which may affect the accuracy of the outcomes. Future researchers should address this limitation and regulate medication use. Secondly, there is limited data on MIX, as few studies have investigated this modality. Finally, the calculation of intensity for mixed exercise is challenging since it integrates various exercise forms and often lacks intensity indicators (e.g., heart rate), thereby causing inaccuracy. Future research should establish standardized methods to quantify the intensity of MIX. Conclusion This systematic review and network meta-analysis suggested that Tai Chi is the most effective exercise modality for relieving anxiety in older adults, with optimal efficacy observed at low-to-moderate doses. Other exercise modalities, including AE, RT, Qigong and Yoga, also demonstrated beneficial effects, whereas MIX did not yield significant improvements. Furthermore, the dose-response relationship indicates that while low-to-moderate doses are effective for alleviating anxiety, exceeding an optimal threshold can have an adverse effect on therapeutic efficacy. Supplementary Information Supplementary Material 1 Supplementary Material 2 Acknowledgements The authors have no acknowledgments to report. Clinical trial number PROSPERO: CRD420251119717. Generative AI disclosure This article was written by the author independently, without using any AI tools or software to generate, edit or modify the content. Authors’ contributions HaoTian Jin and YiNan Zhang designed the study and developed the retrieval strategy. HaoTian Jin, Yi Sun and Yan Shao executed the systematic evaluation searching and screening the summaries and titles, assessing the inclusion and exclusion criteria, generating data collection forms and extracting data, and evaluating the quality of the study. HaoTian Jin and Yi Sun performed meta-analysis. HaoTian Jin drafted the article, which was reviewed and revised by YiNan Zhang. Funding This study was supported by "The Fundamental Research Funds for the Central Universities" (Grant number:3282026011) and "BESTI Teaching Cultivation and Incubation Program" (Grant number:jy202638). Data availability All data generated or analyzed during this study are included in this published article and its supplementary information files. Declarations Ethics approval and consent to participate This is a systematic review and meta-analysis, ethics approval and consent to participate are not applicable. Consent for publication Not applicable. The manuscript does not include the participant’s identification image or other personal or clinical details. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Data Availability Statement All data generated or analyzed during this study are included in this published article and its supplementary information files.