Association between physical activity and risk of anxiety: a dose-response meta-analysis of 11 international cohorts

Summary

Background

Anxiety disorders are the world’s most common mental disorders and the risk of anxiety may be reduced by physical activity. However, the dose-response relationship between physical activity and the risk of anxiety remains uncertain.

Methods

Following Meta-analysis of Observational Studies in Epidemiology (MOOSE) reporting guidelines, PubMed, Embase, Web of Science, and PsycINFO databases were searched for all prospective cohort studies on physical activity and the risk of anxiety up to 7 February 2025, with no language restrictions. We included prospective cohort studies reporting physical activity at three or more exposure levels and risk estimates for anxiety. Data extraction and article quality assessments were completed independently by two extractors. Data were synthesized using random-effects models, and nonlinear dose-response meta-analyses were performed using restricted cubic spline models. We used the Newcastle–Ottawa Scale (NOS) to assess the risk of bias for the included articles. Egger’s test and Begg’s test were used to evaluate the publication bias.

Findings

11 studies (I² = 71.4%; P < 0.0001) including 295,738 participants and 1,906,409 person-years were included. For all studied population, this study found a L-shaped dose-response relationship between physical activity and anxiety risk. The maximum beneficial dose for physical activity is 30 metabolic equivalent task hours per week (MET-h/week), and the anxiety risk is reduced by 16% [95% confidence intervals (CI): 13%–20%]. Anxiety risk slightly increases when the dose exceeds 50 MET-h/week. In subgroup analysis, studies with follow-up time ≤5 years (I² = 68.4%; P = 0.023) exhibited a more typical J-shaped pattern, with a smaller maximum benefit dose of 12.5 MET-h/week and a greater reduction in anxiety risk, up to 49% (95% CI: 35%–59%); anxiety risk significantly increases above 23.5 MET-h/week. The results for studies with follow-up time>5 years are similar to those for the all studied population. Studies on participants over 50 years old (I² = 0.9%; P = 0.32) found an inversed-line relationship between physical activity and anxiety risk. And physical activity levels less than 14.5 MET-h/week were not statistically significant in reducing anxiety risk [CI of relative risks (RR) included 1.00]. The NOS assessment showed low risk of bias for all 11 studies. Egger’s test and Begg’s test found no publication bias in the included studies (P = 0.73).

Interpretation

Our research confirmed a dose-response relationship between physical activity and anxiety risk, and increasing physical activity within the range recommended by the World Health Organization (WHO) (10–20 MET-h/week) can significantly reduce anxiety risk. Subgroup analysis showed that this protective effect was most significant during the 5-year follow-up period, with a maximum anxiety risk reduction of 49%. Exceeding the recommended dose, there may be additional smaller benefits between doses 20–30 MET-h/week, and exceeding 50 MET-h/week may become a risk factor. In addition, the protective effect decreases with time and age, and the recommended dose by WHO (10–14.5 MET-h/week) has no statistically significant protection for the elderly, nor has the maximum beneficial dose been found, indicating the need for more relevant research on the elderly in the future.

Funding

The National Traditional Chinese Medicine innovation team and talent support projects fund this study.

Research in context.

Evidence before this study

We conducted a systematic search of the PubMed, Embase, Web of Science, and PsycINFO databases (from their inception to November 20, 2023, with no language restrictions) using the keywords related to “risk of anxiety”, “physical activity”, “dose” or “intensity”. Existing studies include two meta-analyses on physical activity and anxiety risk. One of the studies categorised Post-Traumatic Stress Disorder (PTSD) as an anxiety disorder, and the results of another meta-analysis suggested that physical activity was associated with a reduced risk of anxiety, but as with the previous study, no quantitative studies were conducted using dose-response. Other studies reported epidemiological cohort studies examining the association between physical activity and anxiety risk. However, the results of cohort studies were not entirely consistent, particularly the lack of dose-response meta-analyses for physical activity and anxiety risk.

Added value of this study

Our research found a dose-response relationship between physical activity and anxiety risk. All included studies analysis showed an increase in physical activity and a decrease in anxiety risk. The maximum benefit dose was 30 metabolic equivalent task hours per week (MET-h/week), and the anxiety risk was reduced by 16%. Among them, studies with follow-up time ≤5 years showed more significant dose-response, with a smaller maximum benefit dose of 12.5 MET-h/week, and a greater reduction in anxiety risk, up to 49%. The risk of anxiety increases significantly when exceeding 23.5 MET-h/week. Studies with participants over 50 years old found that physical activity levels less than 14.5 MET-h/week were not statistically significant in reducing anxiety risk, indicating that the sensitivity of the body to physical activity may decrease with age. Therefore, people need physical activity slightly above the recommended level to achieve long-term mental health benefits.

Implications of all the available evidence

Available evidence has confirmed a clear dose-response relationship between physical activity and anxiety risk. Achieving the World Health Organization (WHO) recommended level of 10–20 MET-h/week for physical activity can yield significant anxiety reduction benefits, especially during the five-year follow-up period when a smaller dose increase can lead to a greater risk reduction. However, as age increases, the magnitude of this benefit may decline over time; therefore, the maximum recommended physical activity volume for long-term adherence or older adults could be adjusted up to 30 MET-hours/week. These findings can provide a basis for developing evidence-based personalized strategies for preventing anxiety disorders according to short-term or long-term goals or different ages.

Introduction

Anxiety disorders are the world’s most common mental disorders. Approximately 4% of the global population suffers from anxiety,¹ which ranks as the sixth largest contributor to non-fatal health loss globally and appears in the top 10 causes of Years Lived with Disability (YLD) in all WHO Regions.² Findings from the Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) show that the prevalence of anxiety disorders in 204 countries and territories worldwide increased from 194.9 (165.1–231.2) per million people to 301.4 (252.6–356.0) per million people between 1990 and 2019.³ In 2019, 301 million people worldwide suffered from anxiety disorders, and anxiety disorders (22.9% [18.6–27.5]) accounted for the second-largest proportion of mental disorder Disability-Adjusted Life-Years (DALY), following depression (37.3% [32.3–43.0]).³ Results by type of mental disorder and region show that anxiety disorders had the highest age-standardized DALY rate in Tropical Latin America and the lowest in Central Asia in 2019,³ however, even in Central Asia, anxiety disorders had the second-highest age-standardized DALY rate for mental disorders.⁴ People with anxiety have a significantly higher risk of dying from natural and unnatural causes compared to the general population,⁵ and as a result, anxiety disorders represent an economic burden and a social stressor in many countries and regions around the globe.6, 7, 8

Multiple studies have shown that increased physical activity is an important means of improving public health and that high levels of physical activity are effective in preventing and improving physical⁹ and mental health¹⁰ problems in a cost-effective manner. On the other hand, physical activity can be detrimental to health,^11,12 especially at high levels. Excessive physical activity not only increases the load on the skeletal muscles, cardiovascular system, and respiratory system, leading to sports injuries, but can also negatively impact mental health.¹² Therefore, there is a need to conduct research on the relationship between physical activity and health risks and to provide scientific evidence to raise awareness of the importance of physical activity and to rationalize the improvement of sedentary behaviors and physical inactivity habits.

The World Health Organization (WHO) defines physical activity as any bodily movement produced by skeletal muscles that requires energy expenditure.¹³ Physical activity encompasses all forms of movement, including activities during leisure time, for transportation to and from places, or as part of one’s work or domestic activities. Both moderate- and vigorous-intensity physical activity contribute to health improvement. The WHO recommends that all adults engage in regular physical activity. Specifically, adults should perform at least 150–300 min of moderate-intensity aerobic physical activity; or at least 75–150 min of vigorous intensity aerobic physical activity; or an equivalent combination of moderate- and vigorous-intensity activity throughout the week, for substantial health benefits.¹⁴ This is the physical activity equivalent to 10–20 metabolic equivalent task hours per week (MET-h/week).

A significant amount of international research has investigated the relationship between physical activity and the risk of anxiety, including randomized controlled trials, cohort studies, case-control studies, and meta-analyses. However, the relationship remains controversial based on the results of prospective studies. The vast majority of prospective studies have demonstrated that physical activity reduces the risk of anxiety in healthy populations. For example, two cohort studies by McDowell et al.^15,16 suggest that high levels of physical activity can reduce the risk of anxiety in individuals aged ≥50 years. Hallgren et al.¹⁷ followed 24,480 adults aged ≥18 years in Sweden for 13 years and concluded that engaging in leisure-time physical activity at levels recommended for general health may reduce the risk of elevated anxiety symptoms. Physical activity not only reduces the risk of anxiety in healthy individuals, but also improves the mental health status of those already suffering from anxiety disorders. For instance, Boschloo et al.¹⁸ followed 1275 depressed and anxious patients aged 18–65 years in the Netherlands for two years and found that, compared to patients with low levels of physical activity, patients with high levels of physical activity had significantly lower persistence of the illness at two years, and physical activity levels greater than 50 MET-h/week were found to have a therapeutic effect on anxiety. However, some studies differ from the above findings. For example, Harvey’s analysis¹⁹ of 33,908 adults followed prospectively for 11 years showed that physical activity prevented the onset of depression but not anxiety. The main reason for the divergent findings may be that different intensities of physical activity have varying effects on preventing the onset of anxiety disorders.^20,21 It may also be influenced by long follow-up periods, which limit reverse causality and give rise to risk of bias, but may also be subject to regression dilution bias due to exposure measurement error caused by changes in participants’ physical activity levels during the follow-up period.²² In addition, the results of two meta-analyses that did not use dose-response relationships did not clarify the benefits of physical activity. Schuch²³ showed that high levels of physical activity can provide a protective effect against anxiety. However, this meta-analysis included studies on post-traumatic stress disorder (PTSD),^24,25 in 2013, Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5)²⁶ no longer categorizes PTSD as anxiety symptoms, so more rigorous meta-analyses are needed to clarify the relationship between physical activity and anxiety risk. Another meta-analysis by McDowell²⁷ showed while current evidence demonstrates that physical activity prevents the onset of anxiety, there are some noteworthy challenges to the current evidence base, such as issues with exposure factors and outcome measures. Traditional meta-analyses analyze the relationship by binary categorization of exposure factors, for example, the intensity of physical activity into active and inactive. This approach leads to a loss of information and does not clarify the change in the risk of anxiety when the intensity of physical activity changes. It is necessary to characterize trends in the relationship between physical activity intensity and the risk of anxiety by conducting dose-response meta-analyses using harmonized exposure estimates. Dose-response meta-analysis techniques have been used to study the relationship between physical activity and a variety of disorders,28, 29, 30 but not yet been applied to the relationship between physical activity and the risk of anxiety.

The dose-response relationship between physical activity and anxiety risk is also controversial. A study by Sun-Young Kim²⁰ suggests that there may be a U- or J-shaped dose-response curve between physical activity and anxiety risk. However, the results of two recent studies^31,32 based on the UK BioBank indicate an L-shaped dose-response curve. There are three potential reasons for this discrepancy. First, Kim’s study²⁰ was a cross-sectional study, whereas the other two studies^31,32 were based on the same cohort data from the UK Biobank. Second, participants in Kim’s study²⁰ were adults aged 18–64, while the other two studies^31,32 focused on middle-aged and older adults aged 40–69. Third, Kim’s study²⁰ measured physical activity levels using the International Physical Activity Questionnaire-Short Form (IPAQ-SF), whereas physical activity levels in the UK Biobank studies^31,32 were obtained through device measurements at baseline and were only measured for seven days. Therefore, further studies are needed to clarify the curves of the dose-response relationship.

None of the available meta-analyses have further examined the association between physical activity and the risk of anxiety using dose-response meta-analysis techniques. The aims of this study were to explore the dose-response relationship between physical activity and anxiety risk, to investigate whether this association is influenced by the follow-up duration, and to examine whether the relationship between physical activity and the risk of anxiety is different in older adults, thereby providing theoretical support for physical activity recommendations for the prevention of anxiety disorders.

Methods

Search strategy and selection criteria

Our study followed Meta-analysis of Observational Studies in Epidemiology (MOOSE) reporting guidelines.³³ Four databases (PubMed, Embase, Web of Science, and PsycINFO) were searched, along with lists of systematic reviews references searched through the databases from databases inception to 7 February 2025. There were no language limitations. Search strings are available in the Supplementary material S1. Search terms were related to physical activity and different forms of physical activity such as swimming, running, and others. Eligibility criteria included prospective cohort studies, and that the study reported an association between physical activity and anxiety. Studies that reported any dimension of physical activity at three or more exposure levels were included and studies were excluded if participants mixed with elevated anxiety symptoms or comorbid other psychiatric disorders at baseline or if the presence of a diagnosis at the outcome that could not be strictly defined as an anxiety disorder, such as PTSD.²⁶ The outcome of interest was anxiety, including (1) registry data, or diagnostic interviews and (2) elevated anxiety symptoms established using validated cutoffs for an anxiety screening instrument (e.g., World Mental Health Composite International Diagnostic Interview). Titles and abstracts as well as subsequent full-text screening were reviewed independently by 2 reviewers, with a third reviewer deciding on inclusion or exclusion in case of disagreement. Screening compliance for both authors: Kappa = 0.85.

Data extraction

Each of the two researchers extracted data independently using a standardized extraction form, which included: authors, study name, year of publication, study site, sex, age at baseline, duration of follow-up, sample size, type of physical activity (leisure time physical activity, home or occupational physical activity), measure of physical activity, dose of physical activity (e.g., MET-hours/week, minutes or times per week), number of participants in each dose category, number of cases, number of person-years, multivariable-adjusted risk estimates of relative risks (RR), hazard ratios (HR), incidence rates (IR), or odds ratios (OR) and their corresponding 95% confidence intervals (CI). A third researcher verified the information extracted. Characteristic tables of the study were produced and comparing against the planned groups for each synthesis. If the required data were not reported in the original article and could not be obtained through other publications in the same cohort, we contacted the authors to obtain key information if necessary.

Assessment of risk of bias and quality of evidence

Risk of bias and quality of evidence in cohort studies was assessed by Newcastle–Ottawa Scale (NOS),^34,35 two independent reviewers conducted a quality assessment of the included studies to identify potential evidence of bias. If disagreements arose during the assessment process, they were resolved through negotiated agreement. Egger’s test and Begg’s test were used to analyze publication bias.

Data harmonization

Physical activity levels reported in all studies were harmonized to a common unit of measure: MET-h/week, using the method of data conversion that depended on the content of the study’s original report. The quantification of physical activity levels (frequency and duration) drew on relevant studies.^28,36 Physical activity levels were taken from the median of the exposure level interval, and if there was no maximum value for the described maximum physical activity exposure level, the minimum value of the maximum activity exposure level plus 0.5 times the width of the interval of the proximity exposure level was taken as the value of the maximum exposure level.^37,38 Light physical activity is usually considered to be less than 3 metabolic equivalent task (MET), moderate is 3–6 MET; vigorous activity is greater than 6 MET,³⁹ therefore it is considered to be 1.5 MET for light activity, 3.5 MET for moderate activity, and 7.0 MET for vigorous activity.²⁸ In some articles, physical activity was reported as inactive, minimally active and very active. Since inactive activity is defined as between 0.5 and 2 MET,²⁸ which is 1.25 MET after taking the median. Minimally active and mildly active intensities are 1.5 MET, and very active levels of physical activity are between moderate and vigorous (3.5–7 MET), which is taken as a median of 5.25 MET. If the original article did not provide a duration, it was assumed that the duration of physical activity per day was 45 min²⁸ and then multiplied the duration by the intensity value.

Statistical analysis

A narrative synthesis was performed for each primary outcome. When sufficient data were available for dose-response meta-analyses (more than 2 studies per outcome of interest), dose-response meta-analyses were performed using STATA version 18 software, along with visualisation of the results. The extracted data was processed statistically by the two researchers separately and if the results obtained were inconsistent then the previous step was repeated until the results were consistent. We chose RR and 95% CI to be used as effect size for the meta-analysis, and the reported OR, HR, was by default comparable to the RR.⁴⁰ When the reported effect size was IR, we converted the RR value based on the information of the data available in the original article. When there were no directly reported levels of physical activity or number of cases of anxiety disorders, or number of participants, or number of person-years of follow-up, we estimated them based on the information that could be extracted for each exposure category and calculated RR values and 95% CI. Effects were estimated using the most adjusted model. If the effect sizes of the original article were not based on the lowest physical activity level as the reference group, we converted the reference group to the lowest dose group and calculated the RR.⁴¹

If gender-specific results were reported, gender-specific results were included in the analysis as separate studies. A gender-stratified meta-analysis was not conducted because only one study⁴² reported gender-stratified results. If there were multiple studies with participants from the same database and overlapping baseline times, we included only one of these studies (choosing the one with the highest NOS score) for dose-response analyses. Data were synthesized using random-effects models. Constrained cubic spline models were used to perform a nonlinear dose-response meta-analysis of studies.

Sensitivity analysis by different methods of data standardization

We performed two sensitivity analyses. First, when physical activity intensity had to be assumed for the physical activity exposure metrics, 4.5 MET, 7.5 MET and 6 MET were set as moderate, vigorous and moderate-to-vigorous activity intensity, respectively. Second, when the study needed to assume physical activity duration, we specified either a 30-min or 1-h duration.

Subgroup analysis

Four types of subgroup analyses were conducted: Studies with ≤5 years of follow-up (n = 4) and studies with >5 years of follow-up (n = 7); studies of individuals aged ≥ 18 (n = 9) and aged ≥50 years (n = 2).

Heterogeneity analysis

Heterogeneity across studies was assessed using the Cochran Q test and the I² statistic. When heterogeneity was found, a heterogeneity sensitivity analysis was performed.

Sensitivity analysis considering participant characteristics

If the population included in a study might be different from the general population, we excluded the study to re-run the dose-response analysis.

Ethics approval

As this study utilized publicly available data, ethical approval and participant informed consent were not required.

Role of funding source

JL is funded by the National Traditional Chinese Medicine (TCM) innovation team and talent support projects (ZYYCXTD-C-202210). However, the funder did not ‌intervene in‌ the study design, data collection, analyses, interpretation, manuscript writing nor in submission choices.

Results

As shown in Fig. 1, the systematic literature search yielded 16,726 results after the removal of duplicates. After excluding 16,632 records based on title and abstract screening, 94 full-text articles were reviewed. Full-text screening identified 13 eligible articles (the characteristics can be found in the Supplementary material S1). Considering that the participants in both studies of McDowell^15,16 and the study of Herring²¹ originated from the same database and overlapped at baseline time, we included only the study of Herring²¹ (with the highest NOS score) when pooling the data from all studies to avoid double counting. The results of 11 cohort studies involving 295,738 participants, including 1,906,409 person-years. The total number of anxiety cases was 13,341. Approximately 49% of the participants in the studies were female. The 11 studies included in the meta-analysis included 9 studies from Europe,^17,19,21,31,42, 43, 44, 45, 46 one from South Korea,⁴⁷ and one from Iran⁴⁸ (characteristics are described in Table 1). All but one of the 11 included studies were from high-income countries. The relevant data and statistics for each study are provided in the Supplementary material S2.

Fig. 1.

Study selection.

Table 1.

Study characteristics and exposure harmonization.

Study (country)	Study name	Participants (case)a	Sex	Age group at baseline	Follow-up (years)	Baseline time	Anxiety disorder definition	Exposure harmonizationb
Ströhle et al.43 2007 (Germany)	Early Developmental Stages of Psychopathology (EDSP)	2099 (228)	Both (49.1% women)	14–24	4	1995	CIDI	Article reported physical activity as no activity, non-regular activity and regular activity. Multiplied by intensity (MET) for no activity (1.25), non-regular activity (1.5), regular activity (5.25).
Bäckmand et al.46 2009 (Finland)	Personality and Mood of Former Elite Male Athletes	776 (51)	Men	52–100	6	1995	BSI-53c	Used volume category midpoints.
Jonsdottiret al.44 2010 (Sweden)	none	2553 (42)	Both (86.5% women)	47 ± 9.9	2	2004	HAD > 10	Article reported physical activity as Sedentary,LPA, MVPA. Multiplied by intensity (MET) for Sedentary (1.25), LPA (1.5), MVPA (5.25).
Ten Have et al.45 2011 (The Netherlands)	Netherlands Mental Health Survey and Incidence Study (NEMESIS)	3815 (218)	Both (49.4% women)	18–64	3	1996	CIDI	Used duration category midpoints.Multiplied by intensity (MET) for moderate to vigorous (5.25)
Brunes et al.42 2015 (Norway)	Health Survey in Nord-Trøndelag (HUNT)	23,991 (1861)	Both (53.3% women)	19–85	13	1995–1997	HADS-A ≥ 8	Article reported physical activity as Low, moderate, high. Multiplied by intensity (MET) for Low (1.5), moderate (3.5), hige (7).
Harvey et al.19 2018 (Norway)	Health Study of Nord-Trøndelag County (HUNT)	22,564 (1972)	Both (49.5% women)	≥20	11	1984–1986	HADS-A ≥ 8	Used duration category midpoints. Multiplied by intensity (MET) for moderate to vigorous (5.25).
Hallgren et al.17 2019 (Sweden)	Swedish National March Cohort study	24,480 (198)	Both (63.10% women)	≥18	13	1997	Registryd	Article reported physical activity as MVPA min/week. Multiplied by intensity (MET) for MVPA (5.25).
Sun-Young Kim et al.47 2020 (Korean)	Kangbuk Samsung Health Study	134,957 (5086)	Both (34.51% women)	18–64	6	2012	BAI ≥ 19	Used volume category midpoints.
Najafipour et al.48 2021 (Iran)	Kerman Coronary Artery Disease Risk Factors Study (KERCADRS)	2813 (97)	Both (59.40% women)	15–80	5	2014	BAI ≥ 15	Used volume category midpoints.
Herring et al.21 2024 (Ireland)	The Irish Longitudinal Study on Ageing (TILDA)	6134 (1079)	Both (53.60% women)	≥50	10	2009	PSWQ-A ≥ 23, CIDI-SF	Used volume category midpoints.
Kai Yu et al.31 2024 (UK)	The UK Biobank	71,556 (2509)	Both (45.50%)	40–69	8	2013	ICD-10e	Article reported physical activity as MVPA min/week.Used midpoints.Multiplied by intensity (MET) for MVPA (5.25).

UK, United Kingdom; CIDI, World Mental Health Composite International Diagnostic Interview; CIDI-SF, World Mental Health Composite International Diagnostic Interview- Short Form; HAD, Hospital Anxiety and Depression scale; HADS-A, Hospital Anxiety and Depression Scale anxiety subscale; BAI, Beck Anxiety Inventory; BSI-53, Brief Symptom Inventory-53; PSWQ-A, Penn State Worry Questionnaire; ICD-10 International Classification of Diseases, Tenth Revision; MVPA, moderate-to-vigorous physical activity; LPA, light physical activity; MET, metabolic equivalent task; MET-min/week, metabolic equivalent task minutes per week.

^aCases of incident anxiety.

^bVolume category refers to the description of physical activity volume in the article, e.g., <600, 600–6000, and >6000 MET-min/week. Duration category refers to the description of physical activity duration in the article, e.g., 0, 1–3, and >3 h/week. These ranges had to be converted to point estimates for the meta-analysis.

^cMood was defined by two partial scales of the short stress symptom survey extracted from the Brief Symptom Inventory-53 symptom survey.

^dAscertained through linkages to existing nationwide, complete and continuously updated specialist medical registers, including inpatient and outpatient records.

^eAnxiety was defined as ICD-10 F40 (phobic anxiety disorders) and F41 (other anxiety disorders).

Pooling the results of 11 studies^17,19,21,31,42, 43, 44, 45, 46, 47, 48 showed that 68% of participants reported physical activity levels below 20 MET-h/week. Among participants who experienced anxiety at follow-up, 78% had physical activity levels below 30 MET-h/week. Physical activity levels were lower in older adults compared to the general population, with approximately 58% of participants aged 50 years or older reporting physical activity levels below 10 MET-h/week.

As shown in Fig. 2-A, a nonlinear dose-response meta-analysis of the 11 included studies (I² = 71.4%; P < 0.0001) revealed a L-shaped dose-response relationship between physical activity and the risk of anxiety. A steep gradient of association was observed in the 0–30 MET-h/week interval, with a mean reduction in the risk of anxiety of 2.7% for each 5 MET-h/week increase in physical activity. However, physical activity above 50 MET-h/week no longer significantly reduces the risk of anxiety. Compared to individuals who did not report any physical activity, those who met the WHO-recommended physical activity level (10–20 MET-h/week) had a 8% (95% CI: 6%–10%) to 14% (95% CI:10%–17%) lower risk of anxiety. At approximately 30 MET-h/week, the dose-response relationship between physical activity and anxiety risk showed an asymptotic relationship. The maximum benefit was obtained for physical activity up to 30 MET-h/week, with a 16% (95% CI: 13%–20%) reduction in the risk of anxiety, and the additional potential benefit was less and more uncertain above 50 MET-h/week.

Fig. 2.

Dose-response relationship between physical activity and Anxiety risk. Abbreviations: MET-h/wk, metabolic equivalent task hours per week. Dark line represents the meta-analytical dose-response curve. The light line indicates the 95% Confidence Intervals (CI). A: Dose-response relationship between physical activity and anxiety risk for the combined 11 studies. B: Dose-response relationship between physical activity and anxiety risk for studies with ≤5 years of follow-up. C: Dose-response relationship between physical activity and anxiety risk in studies with >5 years of follow-up. D: Dose-response relationship between physical activity and anxiety risk in participants aged ≥50 years.

Studies analyzing adults aged >18 years (I² = 68.5%; P = 0.00077) found the dose-response relationship was similar to that observed in the 11 included studies. Detailed findings are provided in the Supplementary material S1.

As shown in Fig. 2-B, analysis of studies with ≤5 years of follow-up (I² = 68.4%; P = 0.023) revealed a typical J-shaped dose-response relationship between physical activity and the risk of anxiety. Compared to individuals with no physical activity at all, those meeting the WHO recommended physical activity level (10–20 MET-h/week) had a 42% (95% CI: 30%–51%) to 47% (95% CI: 35%–58%) lower risk of anxiety. The greatest benefit was observed at much lower dose of 12.5 MET-h/week, with a very high, up to 49% (95% CI: 35%–59%) reduction in the risk of anxiety. Within the 0–12.5 MET-h/week range, each 5 MET-h/week increase in physical activity associated with an average 19.6% reduction in the risk of anxiety. However, physical activity above 23.5 MET-h/week resulted in a significant reduction in benefit, and above 39 MET-h/week no longer significantly reduced the risk of anxiety (CI of RR included 1.00).

As shown in Fig. 2-C, analysis of studies with >5 years of follow-up (I² = 73.1%; P = 0.00052) found that, compared to individuals with no activity at all, physical activity at the WHO recommended level (10–20 MET-h/week) was associated with only an 8% (95%CI:5%–10%) to 13% (95%CI:9%–16%) reduction in the risk of anxiety. At 30 MET-h/week, the dose-response relationship exhibited an asymptotic relationship, and physical activity at a level of 30 MET-h/week resulted in the greatest benefit, with a 15% (95%CI:11%–18%) reduction in the risk of anxiety. Within the 0–30 MET-h/week range, each 5 MET-h/week increase in physical activity was associated with an average 2.5% reduction in the risk of anxiety.

As shown in Fig. 2-D, studies analyzing older adults aged >50 years (I² = 0.9%; P = 0.32) found an inversed-line relationship between physical activity and risk of anxiety, in which each 5 MET-h/week increase in physical activity was associated with a mean 2.6% reduction in the risk of anxiety. Physical activity levels of 10–14.5 MET-h/week were associated with a 10% (95% CI: 0%–18%) to 14% (95% CI: 0%–26%) reduction in the risk of anxiety; however, the benefits of physical activity at such low levels were uncertain (CI of RR included 1.00). Physical activity levels higher than 14.5 MET-h/week associated with a statistically significant decrease in the risk of anxiety. The highest recommended physical activity dose by the WHO (20 MET-h/week) was associated with a 18% (95% CI: 3%–30%) reduction in the risk of anxiety. No specific physical activity dose was found to provide the greatest benefit in older adults.

We performed two sensitivity analyses using different methods of data standardization and reran the dose-response analyses excluding the study in which participants were former sports athletes,⁴⁶ neither of which revealed substantial alterations in dose-response associations (the maximum benefit dose was consistently close to 30 MET-h/week). Detailed findings are provided in the Supplementary material S1.

The Q-test analyzed 11 studies and found greater heterogeneity (I² = 71.4%, P < 0.0001) (Fig. 3). After removing one study⁴³ involving participants aged 14–24 years, one cohort²¹ of participants aged ≥50 years followed for 10 years, and one study³¹ that used devices to measure physical activity, Q-tests of the remaining 10 studies revealed no significant heterogeneity (I² = 17.7%, P = 0.29). Dose-response analyses did not reveal substantial differences (the maximum benefit dose was consistently about 30 MET-h/week). Detailed findings are provided in the Supplementary material S1.

Fig. 3.

Q-test analyzed 11 studies. Abbreviations: CI, Confidence Intervals; DL, Dersimonian-Laird.

The 11 included studies had NOS scores ranging 7–9 (detailed ratings are in the Supplementary material S1), all of which were high-quality evidence with low risk of bias. No publication bias was detected using Egger’s test and Begg’s test analyses (P = 0.73).

Discussion

To our knowledge, this study is the first dose-response meta-analysis of the relationship between physical activity and anxiety. We included a total of 11 studies, all of which were shown to be of high quality and low risk of bias by NOS results. Our findings show an inverse curvilinear relationship between anxiety and low-dose physical activity in the general population. Increasing physical activity can reduce the risk of anxiety within a range of doses (0–30 MET-h/week), and the greatest benefits of physical activity for anxiety are achieved at lower doses. On average, for every 5 MET-h/week increase in physical activity, the risk of anxiety was reduced by 2.7%; maximum benefit was obtained for physical activity up to 30 MET-h/week, which was associated with a 16% reduction in the risk of anxiety. When physical activity exceeded 30 MET-h/week, the reduction for anxiety was no longer significant, and greater benefits were no longer obtained by increasing physical activity. The dose-response relationship also appears to show an asymptotic relationship at 30 MET-h/week in heterogeneity and sensitivity analyses.

Subgroup analyses showed that this relationship of physical activity reducing anxiety risk was more pronounced within 5 years followup than over the long term, and analyses of studies with ≤5 years of follow-up found that larger benefits were realized when going from inactivity to at least some activity. The equivalent of a cumulative 1 h of brisk walking per week was associated with a 21% reduction in the risk of anxiety over 5 years. But the recommended dose of physical activity (10–20 MET-h/week) may have more significant short-term than long-term efficacy in preventing anxiety (e.g., the same level of physical activity of 12.5 MET-h/week was associated with a 49% and 9% reduction in short- and long-term risk of anxiety, respectively). Therefore, public health recommendations for physical activity levels to prevent anxiety may be up to 30 MET-h/week, equivalent to 1 h of brisk walking per day.

Previous meta-analyses showed that people with high levels of physical activity had lower risks of developing anxiety disorders than those with low levels of physical activity. But one study²³ included PTSD in the meta-analysis. Schuch’s²³ meta-analysis, published in 2019, showed that adults with high levels of physical activity had a lower risk of developing an anxiety disorder than those with low levels of physical activity. However, Schuch’s²³ study included PTSD in the meta-analysis. PTSD was considered an anxiety disorder prior to the DSM-5,²⁶ the most recent guidelines have redefined anxiety disorders, and it is possible that the results of this meta-analysis are no longer applicable at the present time due to the inclusion of two PTSD-related studies.^24,25 Following Schuch’s²³ publication of a meta-analysis, McDowell et al.²⁷ also conducted a systematic review and meta-analysis of longitudinal associations between physical activity and anxiety in 2019, and while current evidence demonstrates that physical activity prevents the onset of anxiety, there are some noteworthy challenges to the current evidence-base, such as issues with exposure factors and outcome measures. None of the available meta-analyses further analyzed the association using dose-response meta-analysis. Our study quantifies the differences in risk of anxiety associated with different doses of physical activity. Therefore, our findings have important new implications for health practitioners in making lifestyle recommendations to people.

Sun-Young Kim²⁰ conducted a cross-sectional study of 124,434 participants in Seoul, South Korea, from January 1, 2012, to December 31, 2016, to analyze the dose-response associations between physical activity and anxiety as defined by the Beck Anxiety Inventory (BAI), and found there was either a U-shaped or J-shaped association between physical activity and anxiety risk, and physical activity up to 30 MET-h/week had the lowest risk of developing an anxiety disorder. However, these findings contradict the results (L-shaped) from studies using accelerometer-measured baseline physical activity.^31,32 All three associations, U-, J-, and L-shaped, indicate that greatest benefits can be obtained from low doses of physical activity, and additional benefits are no longer obtained above the maximum beneficial dose. The difference between the U- or J-shaped and the L-shaped associations is the trend of the curve when physical activity exceeds the maximum beneficial dose (U- or J-shaped curve shows an upward trend, L-shaped curve shows a horizontal trend). Our study found a J-shaped dose-response relationship for studies with less than 5 years of follow-up and an L-shaped relationship for those with more than 5 years of follow-up.

Possibly influenced by publication bias, most cohort studies reported results showing that physical activity was indeed associated with a reduced risk of anxiety, with only Harvey’s¹⁹ findings suggesting that physical activity levels were not associated with a reduced risk of anxiety. The results of our meta-analysis may explain the contradictions between the results of prospective cohort studies. Harvey’s¹⁹ study was an 11-year prospective follow-up of 33,908 adults with a follow-up time more than 10 years. We conducted separate independent dose-response meta-analyses of studies with a follow-up time of ≤5 years and those with a follow-up time of >5 years, found that the short-term benefits of increased levels of physical activity may be greater than the long-term benefits. In addition, we analyzed that the risk of bias may be due to the longer follow-up period limiting the risk of bias due to reverse causation, or regression dilution bias may occur due to exposure measurement errors caused by real changes in physical activity behaviors during the follow-up period.²² For example, in Lee’s⁴⁹ study, it was shown that the association between physical activity and mortality estimated at 28 years of follow-up from a single physical activity measurement at baseline was 2–3 fold weaker than the mortality association estimated using repeated measurements.

Authoritative guidelines^13,50 recommend adults aged 18–64 years should engage in at least 150–300 min of moderate-intensity physical activity or 75–150 min of vigorous-intensity aerobic physical activity per week, which is equivalent to a physical activity level of 10–20 MET/week. Our analysis of studies with a follow-up time of ≤5 years found that, relative to people with no physical activity at all, physical activity levels up to the minimum of the recommended dose, 10 MET-h/week, were associated with a 42% reduction in the risk of anxiety; maximum benefit was obtained with physical activity levels up to 12.5 MET-h/week, with a 49% reduction in the risk of anxiety. Analysis of studies with a follow-up > 5 years and with a follow-up of 2–13 years both showed that the greatest benefit was obtained from physical activity up to 30 MET-h/week. The volume of physical activity that maximized the benefits of the former (≤5 years of follow-up) was lower than the latter (>5 years of follow-up). This means that if a person maintains 2 h of brisk walking per week (10 MET-h/week) for 2–5 years, the risk of anxiety can be reduced by 42%; however, the same 2 h of brisk walking per week for 6–13 years only reduces the risk of anxiety by 8%, which requires an appropriate increase in the dose of physical activity in order to ensure a long-term state of mental health. Because the physical activity levels used in our dose-response analyses were all collected at baseline and may have been subject to regression dilution bias due to exposure measurement error caused by changes in physical activity levels during follow-up, studies with a combined follow-up time of ≤5 years showed a stronger association between the recommended activity volume and reduced risk of anxiety. Whereas subgroup analyses with a follow-up time >5 years showed a weaker association between the recommended volume and reduced risk of anxiety, an alternative explanation may be related to ageing, a speculation corroborated by our subgroup analysis of older adults.

In subgroup analyses, we included two cohort studies^21,46 of older adults with participants ≥50 years of age, and preliminary findings suggest that higher doses of physical activity may be needed in the older population to achieve a reduction in the risk of anxiety. Our subgroup analyses showed that older adults had higher uncertainty (wider confidence intervals) about potential benefits at physical activity below 14.5 MET-h/week, but less uncertainty (narrower and more stable confidence intervals) about potential benefits as physical activity levels increased above 20 MET-h/w. Because the results of our subgroup analyses were limited by the number of included studies (only two studies in older adults), we did not find a dose of physical activity that would be of maximal benefit in older adults. However, findings from subgroup analyses suggest that older adults may be less sensitive to physical activity.

Negative emotions can imbalance the hypothalamic-pituitary-adrenal axis (HPA axis), which increases levels of the stress hormone cortisol. Physical activity can reduce stress hormone levels and promote mental health in many ways. For example, physical activity can promote activation of the endocannabinoid (eCB) system.⁵¹ Physical activity also promotes the release of endorphins from the brain, exerting a pain-relieving and mood-enhancing effect. In addition, physical activity boosts dopamine and 5-hydroxytryptamine production, helping to reduce anxiety and depression. A recent study⁵² have shown that lactate produced by muscles during physical activity can enter the brain and modulate synaptic function by enhancing lactylation of multiple synaptic proteins, helping the brain to better handle stress and relieve anxiety. Our results found that older adults may require slightly higher levels of physical activity than the general population to achieve significant benefits. This may be related to the fact that older people have lower basal metabolism and neuroendocrine levels than younger people,⁵³ reduced physical functioning, and less effective activation of the eCB system, or reduced release of substances such as endorphins and dopamine by physical activity. It may also be associated with psychosocial factors, loneliness, frailty, ageing, and chronic illness in older adults. More cohort studies are needed in the future to determine the relationship between physical activity and the risk of anxiety in older adults in order to give more rational recommendations for physical activity in older adults.

This study covered prospective cohort studies from the time of construction of each database. The included studies covered participants of different genders and ages. We conducted the first dose-response meta-analysis of physical activity and the risk of anxiety, a pooled analysis of studies with 2–13 years of follow-up, and separate subgroup analyses of studies with ≤5 years and 5–13 years of follow-up, as well as subgroup analyses of studies of participants ≥50 years of age. Our study identified the mental health benefits of physical activity and found that the recommended level of physical activity for older adults may be slightly higher than the WHO recommendation. Our study explains the disagreement between cohort studies on the association between physical activity and the risk of anxiety and fills a research gap. In addition, another major strength of this meta-analysis is our extensive coordination of exposure levels to maximise the inclusion of available evidence.

Our study found a dose-response relationship between physical activity and the risk of anxiety. Substantial mental health benefits can be obtained by reaching the WHO recommended dose within five years. But the benefits of recommended dose may diminish over time and with age, which requires an appropriate increase in the level of physical activity in order to ensure a long term state of mental health. The public health guidelines may suggest people engaging in up to 30 MET-hours/week of physical activity to help prevent anxiety, which is equivalent to brisk walking for 1 h a day. It is recommended that academics and relevant organizations conduct more research on older people to explore the levels of physical activity that maximize the benefits for older people. Public health practitioners should take measures to encourage people to participate in a variety of physical activities to promote their mental health.

We did not include results using more than one measure of physical activity because no studies were found that met our inclusion criteria. With the exception of one study,³¹ we included studies that used only self-reported measures of physical activity, which are susceptible to recall and social desirability bias, and we included studies that reported physical activity levels only at baseline. Data for higher physical activity doses were relatively limited in our analysis. Modern technological tools such as artificial intelligence⁵⁴ or mobile health apps⁵⁵ have been widely used in the research, prevention and treatment of a variety of important diseases, and wearable devices and mobile apps⁵⁶ have begun to be used to guide and monitor physical activity levels. Therefore, future studies are necessary to use more modern technologies such as digital health products and device-based activity measures to capture a larger range of exposures with repeated measurements during follow-up. There were also not enough studies that analyzed stratified subgroups for outcomes such as gender, age or geography, and the included studies were underrepresented in low- and middle-income countries.

We also included a study⁴⁶ in which the participants were former elite athletes, and although these individuals may be different from the general population, the participants in this study were already 52–100 years old (no longer practising for about 20 years or more) and the main difference from the general population may only be the higher levels of physical activity, which is the key exposure factor of interest to us. Moreover, the number of studies in older cohorts that we included was too small, and in order to understand the relationship between physical activity exposure and anxiety risk in people of different ages, we eventually included the study and performed sensitivity analyses, and the inclusion of this study did not substantially change our dose-response results. Future research involving distinct populations and follow-up durations is essential to validate the conclusions of this study.

Heterogeneity implies that conclusions may be inconclusive, so we analyzed heterogeneity. After effective exposure coordination, we included studies with greater heterogeneity (I² = 71.4%, P < 0.0001) due to the inclusion of three studies: a study⁴³ of participants aged 14–24 years, a study of older adults with longer follow-up times (10 years),²¹ and a study³¹ used devices to measure 7-day habitual physical activity. First, physical activity levels in younger adults differed from those in the general population, and the dose-response relationship also varies. Second, the risk of anxiety in older adults may have differed from that in the general population, and with the long follow-up period of the study, the dose-response association may have become weaker.²² Third, in the device-based measurement study, participants wore accelerometers on their dominant wrist for seven days to measure physical activity, while other studies used questionnaires to measure baseline physical activity. Thus creating heterogeneity. Q-tests on the remaining 8 studies after removing the three studies did not reveal heterogeneity (I² = 17.7%,P = 0.29). We ultimately included all 11 studies in our analysis, as our respective dose-response analyses consistently indicated that the maximum benefit dose was 30 MET-h/week, with no substantial differences observed. Additionally, we wanted to ensure that no valuable information was overlooked.

This study has several limitations. First, this study systematically synthesizes various studies featuring diverse populations and follow-up durations, which may introduce methodological inconsistencies and lead to less precise estimates. To address these issues, we performed thorough subgroup analyses, sensitivity analyses, and heterogeneity assessments to ensure the analytical rigor of our findings. Second, for nonlinear dose-response analyses, we included only studies with at least three levels of physical activity exposure, a requirement that limited the number of eligible findings compared with previous meta-analyses. Physical activity exposures in the study were categorized using multiple methods and represented in different forms, which may have led to the exclusion of additional studies. Third, the included studies used different anxiety assessment methods (or tools) with different thresholds, which may introduce heterogeneity and affect the internal validity of the meta-analysis. It is true that different countries and doctors may have different diagnoses for different populations, especially for anxiety, a mental illness. However, for the large-scale population studies included in our study, we can assume that the tools and methods used by researchers for anxiety assessment should be the most suitable, reliable, and therefore comparable, just like the WHO statistics of diseases across countries around the world. In addition, the differences in assessment tools between studies are listed in Table 1 for comparison.

Contributors

Conceptualization: Xuesi Li, Weicong Chen, Zijiao Zhang, Zheng Li, Tongyu Ma, Shunjing Ding, Fangyu Yang, Jianbo Lei.

Data curation: Xuesi Li, Zijiao Zhang, Zheng Li.

Formal analysis: Xuesi Li, Zijiao Zhang, Zheng Li, Xianming Fan.

Investigation: Xuesi Li, Zheng Li.

Methodology: Xuesi Li, Weicong Chen, Zheng Li, Xianming Fan, Jianbo Lei.

Supervision: Tongyu Ma, Shunjing Ding, Fangyu Yang, Jianbo Lei.

Validation: Xuesi Li, Zijiao Zhang, Zheng Li, Jianbo Lei.

Writing—original draft: Xuesi Li.

Writing—review & editing: Xuesi Li, Jianbo Lei.

All authors read and approved the final version of the manuscript. Xuesi Li and Jianbo Lei accessed and verified the data.

Data sharing statement

The underlying data and associated codes for this study can be viewed in the Supplementary materials, and other data can be shared by contacting the corresponding author upon reasonable request for academic and research purposes.

Declaration of interests

All authors declare that they have no conflict of interest.