Article Text

Exercise for anxiety disorders: systematic review
  1. Kaushadh Jayakody1,
  2. Shalmini Gunadasa2,
  3. Christian Hosker3
  1. 1Academic Unit of Psychiatry and Behavioural Sciences, Leeds Institute of Health Sciences, University of Leeds, Leeds, UK
  2. 2Department of Health Sciences, The University of York, York, UK
  3. 3Department of Liaison Psychiatry, Leeds Partnerships NHS Foundation Trust, Leeds, UK
  1. Correspondence to Dr Kaushadh Jayakody, Division of Applied Medicine (Psychiatry), University of Aberdeen, Royal Cornhill Hospital, Aberdeen, AB25 2ZH, Scotland; jarkjayakody{at}


Background Anxiety disorders are commonly treated with antidepressants and psychological treatments. Some patients may prefer alternative approaches such as exercise.

Objective To investigate the treatment effects of exercise compared with other treatments for anxiety disorders.

Data sources Randomised controlled trials (RCTs) of exercise interventions for anxiety disorders were identified by searching six online databases (July 2011). A number of journals were also hand searched.

Main results Eight RCTs were included. For panic disorder: exercise appears to reduce anxiety symptoms but it is less effective than antidepressant medication (1 RCT); exercise combined with antidepressant medication improves the Clinical Global Impression outcomes (1 RCT, p<0.05); exercise combined with occupational therapy and lifestyle changes reduces Beck Anxiety Inventory outcomes (1 RCT, p=0.0002). For social phobias, added benefits of exercise when combined with group cognitive behavioural therapy (CBT) were shown (p<0.05). There was no significant difference between aerobic and anaerobic exercise groups (1 RCT, p>0.1) with both seeming to reduce anxiety symptoms (1 RCT, p<0.001). It remains unclear as to which type of exercise; moderate to hard or very light to light, is more effective in anxiety reduction (2 RCTs).

Conclusions Exercise seems to be effective as an adjunctive treatment for anxiety disorders but it is less effective compared with antidepressant treatment. Both aerobic and non-aerobic exercise seems to reduce anxiety symptoms. Social phobics may benefit from exercise when combined with group CBT. Further well-conducted RCTs are needed.

  • Exercise
  • Physical activity and exercise methodology
  • Psychology

Statistics from


Anxiety disorders are a common cause of morbidity and mortality. These disorders manifest as a range of cognitive, behavioural and physical symptoms. Prevalence varies among types of anxiety disorder,1 but worldwide lifetime prevalence of anxiety disorders is generally estimated to be 16.6%, with considerable heterogeneity between studies.2

Anxiety disorders are commonly treated with antidepressants and psychological treatment methods.1 However, concordance with antidepressant medication can be poor and it is associated with side effects. Patients may also not wish to follow psychological treatments due to the commitment of adhering to therapy and issues of stigma. There is preliminary evidence that an acute bout of exercise has an antipanic activity in healthy subjects.3 Smits and Zvolensky4 reported that physical inactivity is significantly associated with greater levels of anxiety sensitivity and can affect the severity of panic disorder. Further, adults who are more likely to have a stressful lifestyle benefit more from the exercise training than those who do not.5 There have been a few hypothesised mechanisms of anxiety reduction following exercise; enhanced self-efficacy, experiences of mastery, distraction from anxiety-provoking stimuli, a method of exposure therapy, neurotransmitter changes, peptide changes and changes of self-concept have been proposed.6 ,7 Exercise might also provide exposure therapy by exposing agoraphobics to unsafe environments and may also produce more monoamine neurotransmitters leading to anxiety reduction.8

Petruzzello et al,9 conducted three separate meta-analyses and the results substantiate the claim that exercise is associated with reductions in anxiety. Meta-analysis conducted by Wipfli et al10 also showed larger reductions in anxiety among exercise groups than no-treatment control groups. Exercise training also seems to reduce anxiety symptoms among sedentary patients who have a chronic illness.11

To our knowledge, there have been several formally conducted systematic reviews and meta-analyses published with regard to exercise effects on anxieties in healthy individuals5 ,9 ,10 ,12 but none on diagnosed clinical anxiety disorders alone. Our objectives are to examine the treatment effect of exercise on clinically diagnosed anxiety disorders where the diagnosis has been made following a formal assessment.


Inclusion and exclusion criteria

Types of studies

A systematic review was conducted using all relevant published randomised controlled trials (RCTs) in which exercise was being used as an intervention for diagnosed clinical anxiety disorders.

Types of participants

Subjects were adult men and women with a clinical diagnosis of anxiety disorder according to International Classification of Disease,13 Diagnostic Statistical Manual,14 diagnostic tools, diagnostic questionnaires, research diagnostic criteria or any other validated criteria. Studies investigating mainly depressive disorder were also excluded.

Types of interventions

RCTs with following interventions were considered:

  1. Exercise versus no intervention

  2. Exercise versus other interventions

    • Exercise versus placebo pills

    • Exercise versus medication

    • Exercise versus cognitive behavioural therapy (CBT)

  3. Exercise in combination with treatments versus other interventions

    • Exercise+occupational therapy+life style changes versus standard general practitioner care

    • Exercise+group CBT versus relaxation+group CBT

    • Exercise+antidepressant medication versus relaxation+placebo

  4. Comparing different forms of exercise

    • Aerobic versus non-aerobic

    • Hard or strenuous or moderate to hard exercise versus light or very light exercise

Types of outcome measures

Main primary outcomes are

  1. Clinically significant changes in anxiety symptoms—as define by each of the studies

  2. Clinically relevant outcome/s such as ‘important improvement’ in general mental state

  3. Clinically important improvement in quality of life

  4. Relapse

  5. Compliance with exercise treatment

Secondary outcomes include

  1. Adverse effects associated with exercise treatment

  2. Treatment satisfaction

  3. Economic outcomes

Search methods for identification of studies

We searched MEDLINE, EMBASE, PsycINFO, CINAHL, AMED and Cochrane database of systematic reviews (July 2011). Six journals were hand searched; British Journal of Sports Medicine, British Medical Journal, The Psychiatrist, British Journal of Psychiatry, American Journal of Psychiatry and Archives of General Psychiatry. For additional references, bibliographies of included trials and relevant review articles were examined. Attempts were also made to contact some authors for further clarification. Search terms are described in see online supplementary appendix 1.

Data collection and analysis

Selection of studies

KJ and SG inspected all the studies from the search and identified relevant abstracts. Any article considered relevant was accessed fully. Above inclusion and exclusion criteria were used to select suitable studies. Any discrepancies were resolved through a discussion with the other reviewer (figure 1).

Assessment of risk of bias in included studies

The Cochrane Collaboration Handbook criteria15 and the Cochrane Collaboration on Depression Anxiety and Neurosis (CCDAN) Quality Rating System16 ,17 were used to assess the methodological quality of included studies.

Data extraction and analysis

Data from included studies were extracted using a standardised extraction sheet. Cochrane Collaboration's Review manager 5.1 was used to enter data using duplicate data entry facility. Attempts were made to calculate the weighted mean difference. Random effects model was used in all the analysis.

For continuous data, mean difference, 95% CI and probability value (p) were calculated when means and SDs were presented. We used the generic-inverse-variance method18 when means and SDs were not presented but the mean differences or standard mean differences were presented instead. In Merom et al,19 data were presented using effect size and CIs. Here we assumed ‘effect size=standard mean difference’ according to the Cochrane Hand Book of Systematic Reviews on Interventions.20 For Binary data, the ORs and 95% CIs were estimated with the p value.

When continues data were not normally distributed, we applied the following two standards to all data before inclusion:

  1. SDs and the means were reported or were obtained from authors

  2. The SD when multiplied by two was less than mean (otherwise, mean is unlikely to be appropriate measure of the distribution).21 Data not meeting this standard were not entered to Review Manager for graphical presentation as this assumes normal distribution. These data were entered under the ‘other data’ tables using Other Data Entry Facility instead

The above two standards were applied to avoid the pitfall of applying parametric tests to non-parametric data.

Attempts were also made to do a subgroup analysis where data permitted. These included

  1. Disease severity groups according to mild, moderate and severe

  2. Comorbid physical illnesses

No subgroup analysis was done due to limited data availability. Attempts were also made to do a sensitivity analysis in order to examine the robustness of the observed findings but given limited data availability, this was not possible. We also used the GRADE approach to interpret findings22 and used the GRADE profiler V.3.5 (GRADE PRO) to import data from Review Manager 5.1 to create ‘Summary of findings’ tables.


Search results

The initial search found 4134 records. The process of study selection is shown in figure 1.

Characteristics of included studies

Eight studies met the eligibility criteria and were included: Broocks et al,8 Esquivela et al,23 Lambert et al,24 Martinsen et al,6 Merom et al,19 Oeland et al,25 Sexton et al26 and Wedekind et al.27 Study participants included patients diagnosed with an anxiety disorder from both sexes, with ages ranging from 18 to 65 years. Characteristics of the included studies are shown in table 1.

Table 1

Characteristics of included studies

Excluded studies and studies awaiting assessment

Most studies were excluded on the basis of study design as they were not RCTs or failed to evaluate clinically diagnosed anxiety disorders.51–65 O'Connor60 compared exercise with no intervention for patients with panic disorder. It was not clear as to whether its control group had an adequate wash-out period and was excluded. One study,62 evaluating the effects of exercise on post-traumatic stress disorder was excluded due to non-randomisation. Three studies assessing obsessive compulsive disorder63–65 were excluded because no control arm was found. One study66 is awaiting assessment.

CCDAN quality rating scores for included studies (table 2) are as follows:

  • Broocks et al8—27/46

  • Esquivela et al23—22/46

  • Lambert et al24—37/46

  • Martinsen et al6—26/46

  • Merom et al19—29/46

  • Oeland et al25—33/46

  • Sexton et al26—27/46

  • Wedekind et al27—38/46

Table 2

Cochrane Collaboration on Depression Anxiety and Neurosis (CCDAN) quality rating scores for included studies

Main results

Exercise versus no intervention

One study25 evaluated the effects of exercise verses no intervention. It compared exercise (both aerobic and non-aerobic) versus no exercise for patients with panic disorder and generalised anxiety disorder. Quality of life was increased in both exercise and no exercise groups from baseline at 20 and 32 weeks but the improvement was more prominent in exercise group, although the EuroQol 5D results were not significant (table 3).

Table 3

Results summary

Exercise versus placebo pill

One study8 found beneficial effects of structured exercise (running) including anxiety reduction compared with placebo pill in panic disorder with or without agoraphobia in all outcome scales. The results were statistically significant in six of nine outcome scales (table 3).

Exercise versus antidepressant medication

One RCT8 compared structured exercise (running) with antidepressant medication (clomipramine 37.5–112.5 mg/day) for patients with panic disorder with or without agoraphobia. Beneficial effects of this medication including a significant anxiety reduction were observed in five of nine outcome scales (table 3).

Exercise+antidepressant medication versus relaxation+placebo

One study27 examined the effect of combined exercise plus antidepressant medication (paroxetine 40 mg/day) against relaxation plus placebo pill for patients with panic disorder with or without agoraphobia (table 3). Here, the outcome was assessed using Clinical Global Impression scale at 10 weeks. There was a significant improvement in Clinical Global Impression in the exercise plus antidepressant (paroxetine) group (F value=8.61, p<0.05).

This study27 also compared exercise plus placebo pill against relaxation plus placebo pill using the same outcome scale (Clinical Global Impression scale). There was a greater non-significant improvement in Clinical Global Impression in the exercise plus placebo pill group (F value=3.7, p=0.06).

Exercise versus CBT

None of the studies compared exercise with CBT. Heiden et al61 compared CBT with exercise (included in physical activity programme). It broadly looked at patients diagnosed with a stress-related illness including anxiety but not anxiety disorders.

Exercise+group CBT versus educational sessions+group CBT

For generalised anxiety disorder, social phobia and panic disorder, one RCT19 showed a non-significant reduction in anxiety symptoms following exercise plus group CBT compared with educational sessions plus group CBT (table 3), and there was a significant reduction in depressive symptoms in the Depression and Anxiety Stress Scale: Depression. Added benefits of exercise when combined with group CBT were shown for social phobics (p<0.05) when considering the results of regression coefficients of analysis of covariance (ANCOVA).

Exercise+occupational therapy+life style chages versus standard general practitioner (GP) care

One study24 compared exercise in combination with occupational therapy (including life style changes) against standard general practitioner care for patients with panic disorder with or without agoraphobia (figure 2). To measure the panic attack frequency, severity and symptomatology, it also used Anxiety Disorder Interview Schedule-IV.67 At 20 weeks in Beck Anxiety Inventory (Supplementary File for editors only figure 3), it showed a significant reduction in anxiety in the exercise arm (which included occupational therapy and life style changes) compared with standard general practitioner care (table 3). Even at 10 months, exercise arm patients remained improved compared with general practitioner care (OR 0.10, 95% CI 0.01 to 0.84, p=0.03). At 20 weeks, exercise arm patients had been more panic free compared with those receiving general practitioner care (OR 0.38, 95% CI 0.15 to 0.96, p=0.04). At 10 months, again exercise arm patients remained more panic free non-significantly (OR 0.43, 95% CI 0.16 to 1.16, p=0.09). Its GRADE PRO summary of findings is shown in table 4.

Table 4

Summary of findings table; exercise+occupational therapy+life style changes versus standard general practitioner care

Figure 2

Exercise+occupational therapy+life style changes versus standard general practitioner care.

Moderate, strenuous or hard exercise versus very light or light exercise

Two studies26 ,23 compared moderate, hard or strenuous exercise versus very light or light exercise (table 3). These two studies used different exercise programmes and outcome scales causing significant heterogeneity, and were not combined for meta-analysis. Nonetheless the data were extracted and were entered in to the Review Manager.

Sexton et al26 compared light exercise (walking) against strenuous or hard exercise (jogging) for patients with an anxiety diagnosis (figure 3). A greater non-significant reduction in anxiety symptoms were observed in the jogging group compared with the walking group in the State—Trait Anxiety Inventory—Trait scale (mean difference 5, 95% CI −2.27 to 12.27, p=0.18) and the State—Trait Anxiety Inventory—State scale (mean difference 5.20, 95% CI −3.56 to 13.96, p=0.24). Similarly, there was no significant outcome difference between these two groups in the Brief Psychiatric Rating Scale (mean difference 0.29, 95% CI −2.79 to 3.19, p=0.90). To the contrary, the Global Assessment Scale outcome results non-significantly favoured walking (mean difference 5, 95% CI 14.02 to 4.02, p=0.28). The GRADE PRO summary of findings table comparing moderate, hard or strenuous exercise against light or very light exercise is shown in online supplementary table S4.

Figure 3

Moderate, strenuous or hard exercise (jogging) versus very light or light exercise (walking): continuous data.

Esquivela et al26 compared moderate with hard exercise with very light exercise by having patients exercised on a bicycle ergometer with different workloads, for patient with panic disorder with or without agoraphobia. Outcomes were assessed using the Panic Symptom List (PSL-IIIR) consisting of 13 symptoms of a panic attack (as described by DSM III-R) on a scale from 0 to 4.35 Also, the frequency and the severity of panic attacks were assessed using Visual Analogue Anxiety Scale after provoking panic attacks using 35% carbon dioxide panic provocation challenge.36

The Visual Analogue Anxiety Scale showed a significant reduction in anxiety symptoms in moderate to hard exercise compared with very light exercise (p<0.01). The panic rate (Supplementary File for editors only figure 5) was also lower significantly in moderate to hard exercise group than the light exercise group (OR 0.07, 95% CI 0.01 to 0.82, p=0.03).

Aerobic versus non-aerobic exercise

One study6 compared the effects of aerobic and non-aerobic exercise on panic disorder, agoraphobia without panic attacks, social phobia and generalised anxiety disorder (table 3). At the end of the study, both groups had achieved significant reductions in scores compared with admission values in the Comprehensive Psycho-pathological Rating Scale and Phobic Avoidance Rating Scale (p<0.001). The difference between aerobic and non-aerobic groups was small and non-significant (p>0.1).

Exercise for obsessive compulsive disorder and post-traumatic stress disorder

Our search identified a protocol for a one RCT evaluating the effects of exercise on post traumatic stress disorder.68 No RCTs were found evaluating the effects of exercise on obsessive compulsive disorder. However, there were other study designs evaluating the effect of exercise on obsessive compulsive disorder63–65 and post-traumatic stress disorder.62

Exercise for adjustment disorder and acute stress disorder

This review did not find any RCTs specifically evaluating the effects of exercise for adjustment disorder or acute stress disorder. Again, we found other study designs evaluating this effect.69


The strengths of this review include: thorough searching for evidence; systematic appraisal of the quality of included studies using the CCDAN quality assessment tool; and grading of evidence (GRADE PRO). Furthermore, attempts were made to cover a wide range of anxiety disorders and a total of 563 subjects were included.

There was a significant reduction in anxiety symptoms following structured exercise (running) for patients with panic disorder with or without agoraphobia in comparison with placebo pill treatment.8 However, some limitations were identified in this study. It can be argued that running exercise induces prolonged exposure to feared agoraphobic situations which might act as a graded exposure technique. As the authors note, ‘running exposes patients to the internal feared cues of palpitations, sweating, rapid breathing, and the like that are induced by exercise’.70 On the other hand, it could also be argued that running exercise may not induce exposure to ‘all of the cues that panic disorder or agoraphobic sufferers fear’—and in that sense it may have additional benefits than just a graded exposure technique. Marks70 commenting on this study8 reported, ‘The study may have achieved even more improvement had its exposure been tailored to involve all of the patients’ feared cues systematically rather than just incidental to the exercise schedule’.70

One study24 concluded that exercise and ‘occupational therapy led life style treatment’ may provide a clinically effective intervention at least as effective as routine general practice care for patients with panic disorder. It is not clear about the type of treatment patients had as routine general practitioner care which would have affected the outcomes.

One study19 suggests that there is no benefit of exercise plus group CBT upon educational sessions plus group CBT for social phobia, panic disorder and generalised anxiety disorder. Possible explanations for this could be that the strategies used with group CBT may also be applied to exercise training. The authors state, “these may include situational analysis, goal setting, self monitoring, home work activities, and supportive follow up.”71 However, when considering the results of regression coefficients of ANCOVA, added benefits of exercise when combined with group CBT were shown for social phobics (p<0.05).19 In this study, attendance rates were high for patients with social phobia compared to the other two diagnoses. Here, the effects of unmeasured confounders were not accounted for analysis.19 “The study had a limited statistical power to demonstrate significant effects.” As a positive note, assessors had used audio-taped recordings to assess the inter-rater reliability.

It remains uncertain as to how effective exercise is when it comes to hard or light exercise. Our review found two studies evaluating this:23 ,26 one23 highlighted the beneficial effects of moderate to hard exercise for patients with panic disorder in the Visual Analogue Anxiety scale, Panic Symptom List and Panic Rate (statistically significant result). One of the strengths in this study23 was that the effects of comorbid disorders across the two treatment arms (confounders) were minimised after effective randomisation. However, in the second study, Sexton et al26 failed to show a significant difference among light exercise (walking) and hard exercise (Jogging). In fact, all the outcome scales used in this study delivered a non-significant result. The authors commented: “There was an un-equal distribution of confounders at the baseline when substance misuse was taken in to consideration. There was also a high dropout rate in jogging group.”26 It remains unclear as to why these two studies contradict each other. Possible explanations include that the two studies used different patient groups, different exercising methods and different outcome scales. For instance, Esquivela et al23 used patients with panic disorder while Sexton et al26 used patients with an anxiety disorder; also Esquivela et al23 used exercising on a bicycle ergometer while Sexton et al26 used walking or jogging. In contrast to the aforementioned findings, Dishman and Buckwort72 reported that moderate intensity activities such as walking are more successful than vigorous physical activity programmes for anxiety and depression.

In terms of aerobic and non-aerobic exercise, one study6 indicated that both aerobic and non-aerobic groups achieve similar and significant reductions in anxiety scores. The patients in this study6 were hospitalised, and had received traditional psychiatric treatment in addition to exercise, which could have affected the outcomes. Carrera et al73 reported that people with panic disorder rate their quality of life as lower than that of healthy people. Our review found one RCT25 showing a non-significant improvement in quality of life in the exercise group (at 20 and 32 weeks) compared with a ‘no exercise’ group for patients with panic disorder and generalised anxiety disorder. When exercise combined with an antidepressant for patients with panic disorder,27 combined group (exercise+antidepressant) showed a greater improvement in Clinical Global Impression outcomes compared with relaxation plus placebo (p<0.05). In CCDAN quality rating this study scored 38 of 46 but it should also be noted that Wedekind et al27 were unable to establish true double-blind conditions.


We considered English language reported articles only. There were very few studies to enter into a meaningful funnel graph to assess publication bias. Most of the included studies were of shorter duration with methodological limitations. It can be argued that studying treatment effects of anxiety disorders brings unique difficulties; distressed patient experiencing anxiety symptoms may not wait till the end of exercise treatment programme and thus drop out. Patients may request medication or other psychological treatment options to deal with their acutely distressing symptoms at an earlier stage rather than waiting to see exercise effects.

Clinical implications

Considering the results from above eight included studies, it seems that exercise shows a treatment effect beyond the placebo effect. Although it appears that the antianxiety effects of exercise are lesser than antidepressants for clinical anxiety disorders, it can still be beneficial as an adjunctive treatment. Medical clearance is needed for these patients to proceed with exercise programme.7 The Physical Activity Readiness Questionnaire is a simple screening instrument commonly used in preparticipation screening for moderate intensity physical activity programme.74 Exercise prescription or motivational messages in printed form or by computer seem to be more effective than face-to-face counselling alone.72 Further, interventions that target specific groups or are tailored to the individual seem more effective than more generic interventions for anxiety and depression.75–78

What this study adds

  • Treatment effects of exercise for patients with clinical anxiety disorders.

  • Identification of less-researched areas such as; treatment effects of exercise on different types of clinical anxiety disorders; comparing treatment effects exercise versus psychological treatment methods for clinical anxiety disorders.


The authors would like to thank Clive Adams, Mahesh Jayaram, Samantha Roberts and Claire Irving from the Cochrane Schizophrenia Group, Alistair Cardno and John Holmes form the University of Leeds, Ian Reid and Mariesha Jaffray from the University of Aberdeen and Karen Sinclair from Grampian Health Board.


View Abstract

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Contributors KJ participated in protocol development, searching, study selection, data extraction, data analysis and report writing. SG participated in protocol development, data analysis, statistical expertise, searching, study selection and data extraction. CH participated in protocol development, searching, study selection and report writing.

  • Funding None.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.