Introduction

Depression is a prevalent and debilitating condition that significantly impacts patients’ quality of life and overall health1,2. In 2017, the World Health Organization (WHO) reported that the global annual incidence of depression is approximately 4.4%, with over 300 million individuals affected by varying degrees of the condition3. The incidence of depression increased by approximately 27.6%4,5,6,7 during the COVID-19 pandemic and is projected to become the leading global disease burden by 20308. The primary treatments in clinical practice include antidepressant medications and psychotherapy; however, their limitations have become increasingly apparent9. Antidepressant medications are often associated with side effects and relapses, manifesting as withdrawal symptoms including nausea, vomiting, insomnia, seizures, and other adverse reactions. As depression severity increases, the risk of suicide attempts, significant impairment in social functioning, and a chronic disease course may escalate10. Although psychotherapy does not carry the adverse effects of medications, it is time-consuming, expensive, and yields a remission rate of only 50%11. Consequently, medical professionals and researchers must urgently explore alternative treatments that are simple, effective, cost-efficient, and safe, while also reducing the disease course.

The “Healthy China 2030” Plan outlines strategies to improve the physical fitness of the entire population through national fitness campaigns, integration of physical and medical health services, non-medical interventions, and promotion of physical activity among key demographic groups12. Exercise is recommended as an adjunctive therapy for depression in the WHO Guidelines on Physical Activity and Sedentary Behaviour13. Exercise therapy, a non-pharmacological treatment for depression, has gained widespread recognition among psychiatrists and researchers due to its low cost, ease of implementation, high adherence rates, and minimal adverse effects14. Research indicates that exercise therapy may influence neurogenesis and brain remodeling through the modulation of neurobiochemical levels and inhibition of inflammatory and apoptotic pathways, thereby producing antidepressant effects similar to those of pharmacological treatments14. Several meta-analyses have demonstrated the positive effects of various exercise modalities on alleviating depressive symptoms15-18.

The primary exercise modalities include aerobic exercise, mind-body exercise, and resistance training. Aerobic exercise, characterized by physical activity performed with adequate oxygen intake19engages large muscle groups and allows for the modulation of intensity based on the individual’s preferred pace. Common activities include walking, jogging, and running. Mind-body exercise, in contrast, is a distinct form of physical activity that emphasizes the dynamic interaction between the brain, mind, body, and behavior, highlighting the influence of psychological factors on physiological health. Notable examples of mind-body exercises include yoga, Pilates, Tai Chi, and qigong20which are characterized by relatively low intensity and MET values ranging from 1.5 to 421. Resistance training, in contrast, involves exercises performed against resistance, typically using equipment such as push-ups, dumbbells, and barbells22. Unlike aerobic and mind-body exercises, resistance training focuses on repeatedly overloading muscles through static, isometric, or dynamic contractions, establishing a stronger connection between muscle tissue and the nervous system. Most current studies on adult patients with depression have either conducted meta-analyses on the effects of aerobic exercise23mind-body exercise24and resistance training25,26or failed to differentiate between exercise types when analyzing their effects27. Few studies have directly compared the efficacy of aerobic exercise and resistance training in patients with depression18,28 making it challenging to distinguish the antidepressant effects of these exercise modalities. Additionally, a significant proportion of existing studies have conflated patients with depressive tendencies with those clinically diagnosed with depression, as defined by the Diagnostic and Statistical Manual of Mental Disorders (DSM) and the International Classification of Diseases (ICD). This methodological limitation may introduce bias, leading to inaccurate conclusions about the comparative efficacy and acceptability of these exercise modalities. Therefore, this study aims to perform a network meta-analysis to compare the antidepressant efficacy and acceptability of three common exercise modalities (aerobic exercise, mind-body exercise, and resistance training) in patients with clinically diagnosed depression.

Methods

Network meta-analysis is performed according to the preferred reporting items in the System Review and Metaanalysis (PRISMA) guide29. (PROSPERO: CRD42023424536).

Literature search strategy

A literature search team was first established. Subsequently, two researchers independently conducted searches in the China National Knowledge Infrastructure (CNKI), PubMed, Cochrane, Embase, Web of Science, PsycINFO, and Medline databases. The search focused on published Clinical Randomized Controlled Trials (RCTs) regarding exercise interventions for depression. The search period extended from the inception of each database until January 8, 2025. Search terms included variations of “Exercise,” “Walk,” “Jogging,” “Hiking,” “Swim,” “Aquatic,” “Cycling,” “Bicycling,” “Concurrent Training,” “Cross Training,” “Functional Training,” “Resistance Training,” “Aerobic,” “Qigong,” “Mind-Body Exercise,” “Tai Chi,” “Baduanjin,” “Wuqinxi,” “Yijinjing,” “Yoga,” “Pilates,” “Depression,” “Depressive Disorder,” and “Randomized Controlled Trial.” The search terms were adapted for each database using appropriate subject terms such as MeSH (PubMed) and Emtree (EMBASE), in addition to free-text terms. An example of the PubMed search strategy is provided in Table 1.

Table 1 Strategy for searching RCTs about depression treated using attention bias modification in PubMed.
Full size table

Inclusion and exclusion criteria

Inclusion criteria: (1)The study design must be a randomized controlled trial (RCT) with publicly available literature for complete data collection; (2)Participants must meet diagnostic criteria such as the Chinese Classification and Diagnostic Criteria of Mental Disorders (CCMD-3), Diagnostic and Statistical Manual of Mental Disorders (DSM-IV or DSM-5), International Classification of Diseases (ICD-10), or DSM-V, or be adults aged 18 or older who have been identified as having depressive symptoms above a certain threshold through validated screening tools (e.g., Beck Depression Inventory (BDI) or Hamilton Depression Scale (HAM-D); no restrictions based on nationality, race, gender, illness duration, or comorbid symptoms; (3)The intervention should involve exercise therapy, including aerobic exercise, resistance training, and mind-body exercises, with the control group receiving standard care, control conditions, or placed on a waitlist. Refer to Table 2 for precise definitions.

Exclusion criteria: (1)The study design was not a randomized controlled trial (RCT), case report, famous doctor experience, book report, animal test, or studies in which the intervention period was less than four weeks; (2)Subjects were minors (≤ 18 years old); (3)Patients lacked baseline clinical depression or were only possibly diagnosed with depression; (4)The intervention was a combined intervention or could not be fully classified as aerobic exercise, resistance training, or mind-body exercise; (5) The results of the study were unclear or the data were incomplete.

Table 2 Definition of exercise therapy and control measures.
Full size table

Literature screening and data extraction

The articles retrieved from various databases were imported into EndNote 20 software for duplicate title searches, and any duplicates were removed. Three researchers independently screened potential articles based on the title, abstract, and full text to identify those that met the inclusion criteria, forming a final list by consensus. If the full text of any article was unavailable, the corresponding author was contacted to provide it. Five researchers evaluated the articles for compliance, resolving any disagreements through discussion. Data extraction was performed using Excel, including: (1) Basic information (first author, year of publication, country); (2) Patient information (diagnostic criteria, age, gender, sample size); (3) Trial details (diagnostic tools, interventions, intervention duration); (4) Outcome measures (symptom relief score, acceptability); and (5) Information regarding the quality assessment of the included studies.

Quality assessment

The methodological quality of the literature was assessed according to the Cochrane Handbook for Systematic Reviews of Interventions version 5.130.The assessment included: (1)Random sequence generation; (2)Allocation concealment; (3) Blinding of participants and personnel; (4)Blinding of outcome assessors; (5)Selective reporting of outcomes; (6)Incomplete outcome data; (7)Other sources of bias. The risk of each bias item was assessed as low, unclear, or high risk based on the trial report. Two investigators independently assessed the risk of bias in each study and reached consensus through discussion when disagreements arose. If a dispute persisted, it was resolved through consultation with a third researcher.

Outcome indicators

(1)Symptom relief score (presented by the endpoint depression scale scores for depressive symptoms in response to the exercise intervention); (2)acceptability (using the number of participants who withdrew from the study before the end of the intervention period as a measure of acceptability).

Statistical analyses

In this study, the quality of the included studies was assessed using RevMan 5.4, and frequency-based reticulated meta-analysis was performed using Stata 17. Due to significant variations in evaluation tools (e.g., scale items, scoring criteria) across the individual studies, primary outcome results were pooled using the standardized mean difference (SMD), while secondary outcome results were pooled using the odds ratio (OR). In the reticulated meta-analysis, consistency was tested using the point score method. A P-value greater than 0.05 indicates that the difference between direct and indirect comparisons was not statistically significant. When a closed-loop structure exists between interventions, both global and local inconsistency tests were conducted. A lower limit of the 95% confidence interval (CI) for the inconsistency factor (IF) at or near 0 suggests strong agreement between direct and indirect comparison results31. The results of indirect comparisons in the reticulated meta-analysis were interpreted using effect sizes and 95% confidence intervals (95% CIs). This study ranks the likelihood of each intervention by integrating both the effect size and the Surface Under the Cumulative Ranking Curve (SUCRA) plot. The SUCRA value is a statistical measure used to compare the relative effectiveness of multiple interventions. The intervention effect is assessed by ranking the effects of different intervention groups and calculating the average percentage of the area under the cumulative ranking curve for each intervention group32. Publication bias was evaluated using funnel plots, with the Egger test conducted for additional validation. Evidence certainty in the network meta-analysis was appraised via the Confidence in Network Meta-Analysis (CINeMA) framework.

Results

Literature selection

By searching CNKI (n = 251), PsyclNFO (n = 728), PubMed (n = 1138), Embase (n = 3268), Cochrane (n = 2904), Web of Science (n = 2129), Medline (n = 1905), a total of 12,323 documents were obtained. After removing duplicates, 6692 pieces of eligible literature remained. Initial screening of titles and abstracts narrowed it down to 813 articles. After full-text assessment, 755 articles were excluded for not meeting the inclusion criteria, resulting in a final inclusion of 58 articles33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89. A visual representation of the search results can be found in the PRISMA fow chart in Fig. 1.

Fig. 1
figure 1

Preferred reporting items for systematic reviews and meta-analyses (PRISMA) fow chart of the study selection.

Full size image

Characteristics of the included studies

Four of the 58 studies were three-arm studies, while the remaining studies were two-arm studies, encompassing a total of 3751 participants, with 1790 in the experimental group and 1961 in the control group. Participants were assigned to experimental and control groups based on the principle of randomization. The experimental group included 62 distinct treatment protocols representing different exercise modalities (aerobic, mind-body, and resistance training), with 27 protocols for aerobic, 28 for mind-body, and 8 for resistance training. The control group consisted of 29 conventional control protocols, 1 control condition, and 10 waiting list protocols. The basic characteristics of the included studies are presented in Table 3.

Table 3 Basic characteristics of included studies.
Full size table

Result of assessment

A total of 58 studies were included in the analysis, all of which referenced the term “random”. Among these, one study employed the lottery method68while seven studies used the random number table method35,49,67,70,83,88. Study 16 employed a computerised randomisation process33,37,39,42,44,47,53,54,55,57,58,64,while four studies used a randomized block method36,40,51,63.The remaining studies did not specify the randomization method employed. The distribution of studies is as follows: 10 studies mention centre33,35,39,44,45,46,51,63,68,70while 4 studies mention opaque envelopes54,55,66,75.The remaining studies do not mention allocation concealment. Additionally, blinding was mentioned in only two studies, specifically regarding the outcome measure blinding37while the remaining studies did not address blinding of subjects, experimenters, or outcome measures. Regarding data completeness, two patients were lost to follow-up at a rate exceeding 20%73,75, while the remaining studies were within the acceptable range. The extent of selective reporting of study results, along with other potential sources of bias, remains unclear.

Fig. 2
figure 2

Schematic of cochrane bias risk assessment.

Full size image

Credibility assessment

The quality of evidence in the network meta-analysis was evaluated using the Confidence in Network Meta-Analysis (CINeMA) framework. This assessment covered six domains: within-study bias, reporting bias, indirectness, imprecision, heterogeneity, and incoherence. Evidence quality was categorized into four levels: high, moderate, low, or very low. For efficacy outcomes, evidence across intervention comparisons was judged moderate. Regarding acceptability, evidence quality ranged from high to low, depending on specific comparisons. Key factors influencing these ratings included: (1) inadequate reporting of randomization procedures, blinding, or allocation concealment in some studies, elevating bias risk; (2) variability in diagnostic criteria, treatment duration, and sample sizes across studies, potentially impacting result consistency and precision; and (3) moderate heterogeneity contributing to lower overall evidence quality assessments.

Results of reticulation Meta-analysis

Evidence networks and inconsistency analysis

The two closed rings in Fig. 2 represent efficacy (symptom relief score) and acceptability for the respective interventions. In the relationship between interventions, the dots represent the exercise intervention modality, the area of the dots indicates sample size, and the connecting lines between dots signify a direct comparison between interventions. A thicker connecting line indicates a larger number of studies. The absence of a connecting line between two interventions indicates no direct comparison; these interventions can be compared indirectly using network meta-analysis. Refer to Fig. 3 for a diagram illustrating the evidence network.

Fig. 3
figure 3

Evidence network for antidepressant efficacy and acceptability of different exercise modas.

Full size image

The overall inconsistency test indicated that both efficacy (symptom relief score) and acceptability for each outcome index had P-values > 0.05, with low heterogeneity, indicating good overall consistency. The consistency of each closed loop was further assessed. The IF values for each intervention ranged from 0.17 to 0.59, with the lower limit of the 95% CI including 0. The acceptability of each intervention showed IF values ranging from 0.05 to 1.33, with the lower limit of the 95% CI including 0, indicating good consistency for each closed loop, thereby justifying the use of the consistency model for analysis.

Comparison of efficacy between interventions

Sixty-two studies assessed the efficacy of each intervention. A comparison was conducted between the antidepressant efficacy of three exercise modalities (aerobic exercise, mind-body exercise, and resistance training) and the control group (conventional control, control condition, and waiting list) (Table 4). The antidepressant efficacy of the three exercise modalities was statistically significant (P < 0.05), suggesting a statistically significant difference (P < 0.05) in depression relief among patients. Aerobic exercise, mind-body exercise, and resistance training were significantly more effective than the control group. The difference between mind-body exercise (SMD: −0.43, 95% CI [−0.82, −0.04], P < 0.05) and aerobic exercise was statistically significant, suggesting that mind-body exercise was more effective in alleviating depression. No other pairwise comparisons were statistically significant. The cumulative probabilities from the network meta-analysis were as follows: mind-body exercise (94.1%), resistance training (81.0%), aerobic exercise (63.9%), control condition (30.5%), waiting list (23.9%), and conventional control (6.7%). A line chart illustrating the cumulative probabilities is shown in Fig. 4.

Table 4 Results of network meta-analysis of antidepressant effects of different exercise modes.
Full size table
Fig. 4
figure 4

Ranking of antidepression effects of different exercise methods.

Full size image

Comparison of acceptability between interventions

All 62 studies reported the acceptability of each intervention. The network meta-analysis revealed that the confidence interval for all pairwise comparisons of exercise methods included 1, indicating no statistically significant differences. The cumulative probabilities from the network meta-analysis were ranked as follows: aerobic exercise (67.1%), mind-body exercise (61.5%), resistance training (39.3%), control condition (20.5%), waiting list (65.5%), and conventional control (46.1%). A line chart illustrating the cumulative probabilities is presented in Fig. 5.

Fig. 5
figure 5

Ranking chart of antidepressant acceptability of different exercise modes.

Full size image

Subgroup analysis of the network meta-analysis

Subgroup analysis of the network meta-analysis was conducted based on the primary outcome, efficacy (symptom relief score). The 62 studies reported the intervention duration: 23 studies lasted between four and eight weeks, 29 studies between eight and 12 weeks, and 10 studies for more than 12 weeks. The subgroup analysis revealed that mind-body exercise was the most effective form of exercise intervention for depression, with efficacy percentages of 86.9% for 4–8 weeks, 83.2% for 9–12 weeks, and 98.5% for more than 12 weeks (Table 5).

Table 5 Results were sorted by subgroup analysis of different periods.
Full size table

Literature publication bias tests

LPublication bias in literature is typically assessed using comparison-corrected funnel plots when more than ten studies are included. The efficacy funnel plot revealed asymmetry, with several data points outside the inverted funnel, suggesting potential publication bias. In contrast, the acceptability funnel plot, after correction, showed a symmetrical distribution of studies around the midline, indicating a lower risk of publication bias. The dispersion of individual studies could be attributed to factors such as sample size and study quality (Fig. 6). As only the funnel plot method was employed, the analysis was relatively unidimensional, and the statistical effect was minimal. Therefore, the Egger test was conducted. The Egger test revealed significant publication bias (P = 0.000, p < 0.05). However, the Egger test for acceptability showed no significant publication bias (P = 0.243, p > 0.05).

Fig. 6
figure 6

Comparison-corrected funnel plot.

Full size image

Discussion

In recent years, exercise interventions have become increasingly prevalent in clinical settings for the prevention and treatment of depression. Aerobic exercise, mind-body exercise, and resistance training are three modalities recognized as effective adjuncts to non-pharmacological treatments. These approaches are valued for their low cost, convenience, high adherence rates, and minimal side effects14,90. Psychiatrists and researchers worldwide have increasingly acknowledged the benefits of these interventions. Mechanistic studies in both animals and humans have demonstrated that exercise can enhance the expression of PGC-1α in skeletal muscle, modulate peripheral kynurenine metabolism, regulate inflammatory responses, influence endocrine secretion, and affect neurotrophic factors, glucocorticoid hormone levels, CNS morphology, cytokine release, and hippocampal neurogenesis. These effects collectively stimulate the central nervous system (CNS) and alleviate depressive symptoms91,92. Such CNS stimulation parallels the biological actions of antidepressant medications14. Exercise may also provide neurogenic benefits that some pharmaceutical treatments do not, warranting further investigation. This study aimed to objectively assess the clinical efficacy and acceptability of various exercise modalities through a comprehensive meta-analysis of 65 clinical trials. The results show that exercise significantly reduces depression symptoms, with mind-body and physical exercises demonstrating the highest efficacy, while aerobic exercise has the greatest acceptability.

Antidepressant efficacy of different exercise modalities

Although exercise is a well-established treatment for depression, the comparative effectiveness of various exercise modalities remains contentious. This study aimed to evaluate the efficacy of different exercise modalities in managing depression, emphasizing the comparative advantages of mind-body exercises. The results demonstrated that mind-body exercises significantly improved adult depression symptoms, outperforming resistance training and aerobic exercise, which ranked subsequently. These effects are primarily explained by the unique physiological and psychological characteristics inherent to mind-body movements. A key attribute of mind-body exercise is its ability to induce relaxation, characterized by softness, slowness, and tranquility. This relaxation response has been shown to decrease sympathetic nervous system activity, facilitate cortical neuron rest, and protect against external stressors. Consequently, it alleviates internal anxiety and conflict, thereby reducing depressive symptoms93,94. The distinctive respiratory patterns inherent in mind-body exercises also warrant attention. Compared to other exercise forms, mind-body practices emphasize deep, slow, steady, and prolonged breathing. This breathing pattern has been demonstrated to improve circulation, enhance organ function, and modulate the autonomic nervous system95effectively alleviating emotional distress. Mind-body exercises also possess distinctive psychological attributes. Unlike other exercises, mind-body practices emphasize the integration of mental and physical processes through meditation, fostering holistic development96. This state is associated with increased alpha brain wave activity, which correlates with elevated endorphin and catecholamine levels, thereby mitigating depressive symptoms97,98. Notably, distinct mind-body exercises exhibit unique neurophysiological characteristics. For example, Tai Chi has been demonstrated to suppress left hemispheric activity while enhancing right hemispheric excitation, resulting in increased positive emotions and reduced depressive symptoms15.It is worth noting that this ranking both aligns with and diverges from findings reported in other studies. A 2019 network meta-analysis of elderly populations ranked exercise efficacy for depression improvement as mind-body exercise > aerobic exercise > resistance training99corroborating our findings that mind-body practices are superior in alleviating adult depression. Nevertheless, discrepancies exist between studies evaluating aerobic exercise and resistance training, likely stemming from subject heterogeneity across studies. For instance, Miller exclusively studied elderly participants, highlighting that most research accounts for the health status of older adults and special populations with muscle dysfunction. Conversely, among young and healthy individuals, low-intensity, short-duration resistance training shows poor compliance and challenges in sustaining regular adherence100,101. Consequently, resistance training appears less effective than aerobic exercise in alleviating depression. A network meta-analysis examining resistance, aerobic, and combined exercises in depressed patients found resistance training superior to aerobic exercise for young adults102consistent with our results. Subgroup analysis further identified mind-body exercise as the most effective intervention for depression, showing efficacy across durations > 4 to ≤ 8 weeks (86.9%), > 9 to ≤ 12 weeks (83.2%), and > 12 weeks (98.5%). These findings further substantiate the therapeutic advantages of mind-body exercises in treating adult depression.

Antidepressant acceptability of different exercise modalities

Aerobic exercise was found to be more acceptable than other exercise modalities, reflecting its superior preference among participants. This higher level of acceptability may be attributed to the simplicity of activities such as walking, running, and cycling, which comprised 72.4% of the studies examined in this review. Existing research suggests that initiating exercise with less complex activities can encourage initially reluctant individuals with depression to engage, thereby improving the effectiveness of interventions91. Nevertheless, this study focuses solely on acceptability and does not explore the specific factors that might explain the preference for one exercise modality over another. Furthermore, acceptability is likely influenced by factors such as exercise intensity, duration, and workload, highlighting the need for further research to confirm the study’s findings.

Limitations of this study

The application of network meta-analysis to evaluate the efficacy and acceptability of various exercise interventions for patients with depression is subject to several limitations: (1) In terms of study characteristics, more than 50% of participants in this review had comorbid somatic disorders, which introduced complexity due to variations in factors such as disease severity, gender, age, and clinical history (e.g., first vs. recurrent episodes), potentially leading to study heterogeneity; (2) With regard to study design, discrepancies exist in the assessment tools for depression, intervention settings, duration, intensity, frequency, and other parameters, which could affect the accuracy of the results; (3) During literature screening, this study exclusively searched electronic databases and did not include grey literature or sources other than Chinese and English publications. This limitation may have introduced database and language biases. Additionally, the study excluded combined interventions and non-randomized controlled trials, which may have further restricted the findings. Moreover, the inability to access the full text of some studies may have compromised the reliability of the results.

Summary

Limited evidence indicates that mind-body exercises may have superior antidepressant efficacy compared to other modalities, while aerobic exercises generally demonstrate higher acceptability. Given patient preferences and medical recommendations, interventions with proven efficacy should be prioritized in clinical practice. However, due to the limitations of current studies, future research should rigorously stratify subjects based on disease status and conduct robust clinical trials for validation.