Is Exercise a Viable Therapy for Anxiety? Systematic Review of Recent Literature and Critical Analysis

Abstract

Objectives:

Exercise has been promoted as a treatment for a variety of psychiatric conditions. The benefits of exercise for depression are widely recognized, but the benefits of exercise for anxiety are uncertain. Although several reviews promoted exercise as a treatment for anxiety, concerns about the quality of studies prompted us to provide a critical review of the recent literature to re-assess the value of exercise for treating anxiety.

Methods and Materials:

We conducted a systematic review of all peer-reviewed randomized clinical trials (RCTs) among adults, published between January 2014 and December 2021, with an exercise intervention and anxiety as the a priori primary outcome. Two reviewers independently extracted data from studies meeting inclusion criteria, including sample characteristics, exercise intervention, control conditions, primary anxiety measure, relevant findings, and methodological quality quantified by PEDro scores.

Results:

7240 published studies from CINAHL, EMBASE, MEDLINE, and PsycINFO were screened in April 2022, with 1831 participants across 25 eligible RCTs, of which 13 included elevated anxiety at study entry as an eligibility criterion. Only two of these 13 studies, and five of 12 studies of non-anxious individuals, found anxiety to be reduced unequivocally with exercise. Most studies suffered from significant methodological limitations including concurrent therapies and lack of intention-to-treat analyses.

Conclusion:

There remains considerable uncertainty about the value of exercise in reducing symptoms of anxiety, particularly among anxious individuals. The paucity of methodologically sound studies of patients with anxiety represents a significant gap in our knowledge and calls for more research in the area.

1. Introduction

Anxiety disorders are among the most commonly diagnosed forms of mental illness in the United States and are responsible for one-third of the total expenditures of the federal government for mental illness [1–3]. Anxiety also is highly prevalent in cardiac populations, with estimates ranging from 25% to 44% [4–6]. In 2013, Tully and Cosh [6] reported an 11–14% prevalence of Generalized Anxiety Disorder (GAD) across 12 studies and a pooled lifetime prevalence of 26%. Anxiety also has important prognostic significance, especially for patients with cardiovascular disease (CVD). In a study of 808 post-MI patients, Frasure-Smith and colleagues [7] reported that more than 40% of patients had elevated symptoms of anxiety and that anxiety was associated with an odds ratio of 1.67 for major adverse CVD events. Strik and colleagues [8] found that while symptoms of depression and anxiety both were associated with increased CVD events, anxiety was an independent predictor of adverse events and higher health care costs, and further noted that anxiety accounted for the relationship between depressive symptoms and worse prognosis.

The widespread prevalence of anxiety disorders, along with the increased risk of adverse events, particularly in patients with CVD, underscore the need to identify effective treatments for anxiety in healthy persons and patients with CVD. There is reason to believe that exercise might be one such approach.

“Lifestyle Psychiatry” is an emerging field that recognizes the important contribution of a variety of health behaviors on mental health, including nutrition, smoking, sleep, and physical activity [9]. Physical exercise is arguably the most established health behavior that is widely recognized as having important psychological as well as physical health benefits. Observational and interventional studies have shown that exercise can impact a wide range of psychiatric conditions, ranging from Alzheimer’s disease [10] to schizophrenia [11]. Studies reporting the benefits of exercise on depression are especially well-documented and have been the subject of numerous systematic reviews and meta-analyses [12–15]. Indeed, recent international guidelines now recommend exercise as a first-line treatment for mild/moderate depression and as an adjunctive treatment for severe mental illness [16].

Although there is substantial literature on exercise and depression, the effect of exercise on anxiety has received comparatively less attention, and results have been less consistent and there is a paucity of exercise studies in patients with CVD. In one early review, Petruzzello and colleagues [17] performed a meta-analysis of the chronic and acute effects of exercise on anxiety. More than 100 studies were identified, with 408 effect sizes and 3,048 participants; the authors concluded that no matter how anxiety was assessed (i.e., state anxiety, trait anxiety, or a psychophysiological measure), aerobic exercise, but not anaerobic exercise, was associated with reduced anxiety. In a subsequent meta-analysis of 36 randomized controlled trials (RCTs) of acute bouts of exercise, Ensari and colleagues [18] reported a small but statistically significant effect of acute exercise in reducing anxiety. However, they did not address the issue of the benefits of regular (i.e., chronic) exercise on measures of anxiety.

Since the initial Petruzzello review, there have been a number of meta-analyses of the effects of regular exercise on anxiety, with disparate conclusions [19–23], in part because participants without elevated anxiety or with no anxiety diagnosis were included in these studies. For example, Conn [20] only included healthy participants without elevated anxiety and found significant heterogeneity in study designs and outcomes. Bartley and colleagues [19] examined seven exercise RCTs in 407 individuals with any diagnosed anxiety disorder and reported no significant difference between exercise and control conditions for anxiety outcomes (effect size = 0.02). However, individuals with anxiety symptoms, but without a psychiatric diagnosis, were not included. Similarly, systematic reviews of RCTs have been conducted among individuals without elevated anxiety or anxiety disorders, including studies examining older adults without anxiety diagnoses [24] and studies comparing results of mindful and non-mindful exercise approaches [25]. However, the clinical significance of reduced anxiety among persons who are not anxious at study entry is not clear. In a review of studies of resistance training, Gordon and colleagues [26] reported that exercise significantly relieved anxiety across 16 trials including 922 participants. However, the review did not distinguish between studies of individuals with and without elevated anxiety at baseline, and the analysis grouped anxiety disorders with other psychiatric conditions, including substance use disorder.

Previously Stonerock and colleagues [27] examined publications from inception through July 2014 for RCTs in which participants were pre-selected on the basis of either a diagnosis of an anxiety disorder or elevated symptoms of anxiety and then randomized to treatment with exercise as one of the treatment arms of the trial. The review identified 12 RCTs and five meta-analyses; of the 12 RCTs, three were considered to show positive evidence of exercise reducing anxiety, whereas four were considered not to show a positive relationship and five were deemed to have mixed results. Importantly, most trials were considered to suffer from significant methodological limitations that left the issue of the use of exercise to treat anxiety unresolved. Unlike previous reviews, the authors concluded that the existing body of evidence was not of sufficient scientific rigor to recommend exercise as a treatment for individuals with clinically significant anxiety.

Because of the inconsistencies in recommendation for exercise as a treatment for anxiety, the present review sought to provide an update of the literature since the Stonerock and other more recent reviews [23, 27], to a) extend the earlier reviews by examining all studies published between January 2014 and December 2021 that assessed the effects of chronic exercise on anxiety outcomes among individuals with an anxiety disorder or with elevated symptoms of anxiety, and b) determine if there is now sufficient evidence to conclude that exercise can be considered efficacious in reducing anxiety among anxious individuals.

2. Methods

2.1. Search Strategy

In April 2022, we conducted a systematic search of MEDLINE, CINAHL, EMBASE, and PsycINFO for peer-reviewed RCTs in English published between January 2014 and December 2021, in which measurements were obtained from participants for either symptoms of an anxiety-related disorder or elevated symptoms of anxiety. This systematic review was not pre-registered but followed Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines for reporting [28].

Articles were identified with titles or abstracts that included at least one match from all three of the following groups of search terms: (1) physical activity OR exercise OR exercise training OR aerobic exercise OR resistance exercise OR strength training OR resistance training OR weight-bearing exercise OR exercise therapy; AND (2) anxiety OR stress OR anxiety disorder OR Obsessive-Compulsive Disorder OR OCD OR Posttraumatic Stress Disorder OR PTSD OR Generalized Anxiety Disorder OR Social Anxiety Disorder OR Panic; AND (3) randomly OR randomized OR randomized controlled trial OR randomized clinical trial. In addition, the literature was searched for reviews or meta-analyses of the area, to identify any additional RCTs that were not included in the primary search strategy. Our search strategy aimed to include all relevant empirical research, without assuming any specific mechanism by which exercise would reduce anxiety or requiring that a study assess mediators or moderators of an exercise-anxiety relationship.

Inclusion Criteria

All RCTs that met the following criteria were included: 1. articles published in a peer-reviewed journal in English; 2. articles published between January 2014 and December 2021; 3. participants were at least 18 years of age; 4. a minimum of 30 participants were randomized; 5. exercise interventions consisted of > 1 bout of aerobic exercise or resistance training; 6. anxiety was a primary outcome measure; and 7. elevated anxiety symptoms were measured using a validated assessment instrument, or participants were diagnosed with an anxiety disorder by an established procedure such as a psychiatric interview. These criteria were selected to replicate those included in the Stonerock review [27]. Studies that examined the effects of a single bout of exercise on anxiety have been reviewed previously [29–33] and were not included in the present analysis.

In the course of title and abstract review, we also identified several reviews and meta-analyses from the search that addressed the topic of exercise as an intervention for anxiety. All articles cited in these meta-analyses were reviewed for eligibility criteria and inclusion, independent of the database search described above.

2.2. Procedures

Every unique article collected was examined to confirm that each inclusion criterion was met (e.g., study included only human participants, article published no earlier than 2014, at least 30 randomized participants, etc.). All of the articles identified for full-text review were reviewed by two raters; discrepancies were discussed, and a third reviewer reconciled any differences if needed. Studies that were found to have met all inclusion criteria after the full-text review underwent data extraction. Information gathered included study inclusion criteria, participant demographics, exercise intervention, control condition, percent attrition, primary outcome of the trial, anxiety measures, and anxiety-related findings, including a determination as to whether exercise was associated with reduced anxiety. Extracted data were reviewed by all authors to resolve discrepancies, and descriptive statistics were tabulated.

Additionally, study quality and relative risk of bias from design of each RCT were assessed using the Physiotherapy Evidence Database (PEDro) scale, a widely used instrument to rate the overall quality of RCTs [34]. PEDro scores were tabulated independently by one author and reviewed by the other authors, with discrepancies reviewed and resolved through re-review of the published full text. PEDro scores were summed across domains; those with higher scores (range 0–10) are considered to have better quality. After the first PEDro item (‘eligibility criteria specified’), the following ten criteria were scored (1 or 0); random allocation of participants to groups, allocation concealment, similarity of groups at baseline, blinding of participants, blinding of interventionists, blinding of assessors of at least one key outcome, inclusion of one key outcome variable from 85 % or more participants, intention-to-treat (ITT) analysis, results reported for between-group differences on at least one key outcome, and point measurements and measurements of variability reported for at least one key outcome. PEDro scores have been shown to be reliable [35], and these quantify the methodological quality, but not necessarily the validity or importance, of the published findings.

3. Results

The selection process for the present review is depicted in Figure 1. The initial electronic database search yielded 7240 records, with 1952 duplicate entries from separate databases, leaving 5288 unique articles. Of these, the vast majority (n = 5218, 98.7%) were excluded after review of the title and abstract. The most common reasons for exclusion were that no exercise intervention was present, anxiety was not a primary outcome measure, or the article was not an empirical research study.

Figure 1:

PRISMA 2020 flow diagram of the selection process for RCTs eligible for full review

In the course of the systematic search, four meta-analyses of exercise and anxiety published between 2014 and 2021 were found [23, 36–38]. The publications identified in these reviews were examined to determine whether any additional relevant studies had been missed or misidentified. Of the 33 studies identified in these four publications, four additional studies had publication dates and sample sizes that appeared to meet our inclusion criteria, and a full-text review was conducted. Two of those four studies were found to have met inclusion criteria [39, 40] and were included in the present review. Of note, the study of United States veterans with Posttraumatic Stress Disorder (PTSD) by Hall and colleagues [39] was initially not included in our review, as this report characterized anxiety scores as a secondary outcome, with ‘intervention feasibility’ as the primary outcome. However, upon examination of the study design publication [41], it was determined that anxiety was initially characterized as a primary study outcome. Because all other inclusion criteria were met, this RCT was included in the review. One of the four studies [42] was excluded because some participants were under 18 years of age, and another [43] was excluded because all participants completed the same exercise intervention.

In total, 70 studies were examined at full-text level, resulting in the exclusion of an additional 44 studies. Inclusion decisions required a third rater for adjudication on two of 70 studies. The most common reason for exclusion (19/44) was that anxiety was not a primary study outcome, and was either a secondary outcome or only briefly mentioned. Additional reasons for exclusion included, but were not limited to, no non-exercise control group, insufficient sample size; exercise intervention was not aerobic or resistance training (e.g., Tai Chi or Pilates); exercise regimen was identical for all groups; and randomization was not specified. Additionally, two publications reported results from the same sample [44, 45] and these records were consolidated during data extraction. As a result, a total of 25 RCTs met our eligibility criteria and are included in the present review.

3.1. Sample Characteristics

The RCTs that met the inclusion criteria were conducted in a number of countries across five continents. Of the 25 studies identified, eight were completed in the United States, three each in Brazil and Canada, two each in Australia and Taiwan, and one each for France, Germany, Hong Kong, Ireland, South Korea, Spain, and the United Kingdom. In total, 1831 participants were included across all of the studies. The median sample size was 56, with 11 of 25 studies enrolling fewer than 50 participants. The studies varied considerably in the proportion of participants’ sex (range: 0–100% female) and average age (range: 19.5 – 67.4 years).

Several studies recruited from select populations. For example, Legrand et al. [46] recruited only prisoners. Other studies restricted participant recruitment to private psychiatric hospitals [47] or psychiatric clinics [48]; two studies included only university students [49, 50]; and one study recruited United States military veterans [39]. Several studies included individuals with specific medical conditions, including Parkinson’s disease [51, 52], stroke [53], and coronary heart disease [54].

Diagnoses of anxiety disorders and measures of anxiety symptoms also varied considerably across studies. Thirteen studies had anxiety as a prerequisite to participate (see Table 1), with eight of these studies selecting patients with specific anxiety diagnoses (i.e., Panic Disorder [PD], PTSD, Obsessive-Compulsive Disorder [OCD], GAD, Social Anxiety Disorder [SAD]) [39, 44, 45, 47, 48, 54–57]. Of note, OCD and PTSD, which are not considered among the group of anxiety disorders in the current Diagnostic and Statistical Manual of Mental Disorders (DSM-5), were included in this review because these diagnoses were grouped among the anxiety disorders in DSM-IV-TR [58], which was current at the time the trials were conducted. The remaining five studies included individuals with elevated anxiety symptoms by self-report questionnaire rather than interview-determined psychiatric diagnosis. Three RCTs included those with questionnaire scores above a specific cutoff; two [46, 59] used the Spielberger State-Trait Anxiety Inventory (STAI) [60] but used different cutoff values, and one [61] used the Anxiety Sensitivity Index (ASI) [62]. Another RCT [63] included individuals with elevated, but subclinical, levels of GAD symptoms assessed by questionnaire, and the final RCT [40] included individuals with positive screens for PTSD by questionnaire. The number of anxious participants in study samples ranged from 30 to 128, with 788 unique participants.

Table 1:

Summary of Randomized Controlled Trials in which Anxiety is a Prerequisite for Eligibility

Author (Year), Country, N	Inclusion Criteria	% Female	Age (mean yrs)	Exercise Intervention	Control Condition	% Attrition	Primary Outcome	Anxiety Measures	Anxiety Findings	Mediator/Moderator Reported?	Does Exercise Reduce Anxiety?	Limitations
Abrantes, Brown, et al. (2017) Abrantes, Farris, et al. (2019) United States N = 56	Treatment-resistant OCD diagnosis; Y-BOCS score of 16+; physically inactive prior to study (<60 min/week aerobic exercise)	64%	38.8	Aerobic; 12 weekly sessions, increasing from 20 to 35–40 min; moderate, treadmill or elliptical, with discussion sessions on increasing physical activity, plus instructed to exercise on their own, 2–4 days/week	Health education	19.6 % (post-test)	Negative affect (OCD symptoms, anxiety, mood)	Y-BOCS (2017) Y-BOCS BAI NIMH-SR (2019)	Y-BOCS, BAI: Scores decreased from baseline to posttest for both groups. No effect of group. NIMH-SR: Anxiety decreased from pre- to post-session (p = .05), and decrease was larger in exercise group (p = .001), but no time x group interaction. At posttest, exercise group, but not control group, showed reduced anxiety (p = .014)	No	Equivocal	-Participants varied in exercise time/intensity -Lack of blinding of assessors -Inadequate control for expectations and attention -Confounded intervention (exercise participants were receiving concurrent treatment)
Blumenthal et al. (2021) United States N = 128	Coronary heart disease patients; DSM-5 diagnosis of anxiety disorder or HADS ≥ 8	29%	64.6	Aerobic: 12 weeks, 3 days/week, 30–45 min	Two groups: Placebo pill, and up to 20 mg equivalent escitalopram	3.9% after 12 weeks	HADS	HADS STAI-State STAI-Trait HAM-A GAD-7 ASI	All groups improved at posttest on composite of anxiety measures (p < .001). Exercise and escitalopram groups had greater reductions compared to placebo in HADS-Anxiety (p = .03) and STAI-State (p = .01). Post hoc contrasts of exercise and placebo showed no difference for HADS-Anxiety. Planned contrasts showed greater reductions for escitalopram compared to exercise for HAM-A (p = .036), STAI-Trait (p = .042), and GAD-7 (p = .022), but not ASI (p = .129).	Mediator: Aerobic fitness - Not significant.	Equivocal	-Participants in ‘pill’ condition were blinded to escitalopram or placebo but not to exercise
Gaudlitz et al. (2014) Germany N = 58	Diagnosis of PD, with no, mild, or moderate agoraphobia; ages 18–70	55%	35.8	Aerobic; 8 weeks, 3 days/week, 30 min walking on treadmill	Movement; light stretching for equivalent times	19% at month 1 34% at month 7	Anxiety	HAM-A BAI PAS	HAM-A: Aerobic group saw greater reduction in scores at 7-month follow-up (group x time interaction, p = .030), but not at posttest. No main effect for group or time. BAI, PAS: Not significant.	No	Equivocal	-Confounded intervention (exercisers received additional therapies) -Did not follow ITT principle (only those who completed month 1, 47/58, were kept for analyses)
Goldste in et al. (2018) United States N = 47	Veterans who met DSM-IV criteria for diagnosis of PTSD or partial PTSD	19.1%	46.8	Aerobic and resistance: 12 weeks of ‘integrative exercise’ including three (or more) 1 hr/wk sessions of aerobic exercise, strength training, yoga, mindfulness	Waitlist	19.1%	PTSD symptom severity and quality of life	CAPS	CAPS: Greater reductions in PTSD symptoms in exercise group compared to waitlist controls (p = .038, d = −.90)	No	Equivocal	-No ITT analysis -Confounded intervention (exercisers received additional therapies) -Inadequate control for expectations and attention -Non-PTSD anxiety symptoms not reported -Small sample
Gordon et al. (2021) Ireland N = 49	Subclinical GAD as determined by PDSQ-GAD subscale >=6 and PSWQ > 45	66.7%	26.6	Resistance exercise training; 8 weeks, 2 days/week, approx. 25 min, 8–12 repetitions of 8 exercises	Waitlist	44.9%	Subclinical GAD status (using cut-scores for the PDSQ -GAD and PSWQ)	PDSQ-GAD subscale, PSWQ, STAI-Trait	For PDSQ-GAD <6 OR PSWQ <45, 6 exercisers achieved remission (33%) compared to 2 controls (6%). For PDSQ-GAD <6 AND PSWQ <45, 2 exercisers achieved remission (11%) versus 1 control (3%). Significant group x time interactions for PSWQ worry, worry engagement, and STAI. Exercisers had lower worry (p≤0.04) and anxiety symptoms (p≤0.001), but not worry engagement	No	Equivocal	-Inadequate control for attention -Some participants also received concurrent therapies -Uneven randomization (18 to exercise, 31 to waitlist) - ITT not primary analysis -Sample size and remission rates too small to provide meaningful conclusions
Hall et al. (2020) United States N = 54	Known clinician diagnosis of PTSD prior to recruitment	9.3%	67.4	Combined resistance, flexibility, and aerobic: 12 weeks, 3 days/week, target 150 min/week of moderate to vigorous physical activity	Usual care	11%	PTSD symptoms	PCL-5	PCL-5 and subscales p>0.05. Authors note average change score among exercisers of 7 points considered a “reliable change ” and 42% of exercisers showed 10+ point reduction in PCL-5 scores.	No	No	-No ITT analysis reported; authors state a post hoc, all-participant analysis was done and “very similar,” but main analyses do not include all participants
Lanoye et al., 2021 United States N = 45	18–35 years of age and elevated anxiety sensitivity (ASI-3 score ≥23)	90.9%	26.1	Aerobic: 2 weeks, 3 days/week, 20 min/day treadmill walking	No intervention (assessments only)	26.7%	Anxiety sensitivity	ASI-3	No differences between exercise condition and controls at follow-up intervals of 4, 6 and 8 weeks after start of exercise	No	No	-Exercise intensity and adherence unclear -No equipoise in conditions -No ITT analysis -Point and variability measures for anxiety not reported -Small sample
LeBouthillier et al. (2017) Canada N = 48	DSM-5 anxiety diagnosis (PD, SAD, GAD, PTSD); ages 18–65; <150 min/week of moderate to vigorous physical activity, no other treatment aside from medication (stable, not benzodia zepine or antipsychotic)	77%	29.6	Aerobic exercise OR resistance training (two groups); 4 weeks, 3 days/week, 40 min on spin cycle or resistance machines	Waitlist (re-randomized to one of two exercise conditions after 1-week post-treatment)	14%	Presence of anxiety diagnosis; disorder-specific symptoms	SCID-5-RV diagnosis change T-scores for disorder-specific outcome measures: (OCI-R, PDSS -SR, PSWQ, SIPS, SMAA, SMSP-A) PCL-5 DASS-21	SCID-5-RV: Disorder status improved at posttest for intervention groups versus control (p = 0.001; 100% of resistance group; 56% of aerobic group; 9% of waitlist group). Disorder-specific symptoms: no change over time for waitlist or aerobic group. Resistance training displayed 7.6T reduction from baseline at posttest (Cohen’s d = −0.39), but no change at 1-week and 1-month follow-up. DASS-21 anxiety: Aerobic group improved relative to zero change (p = 0.012), no change for waitlist and resistance groups.	No	Equivocal	-Did not follow ITT principle -Nonadherence: one third of patients randomized did not complete the interventions -Waitlist patients were re-randomized into exercise groups but treated as equivalent to first exercise group in analyses -Study registration and data analysis were modified after data were collected -Small sample
Legrand et al. (2020) France N = 37	Prisoners, first time incarceration; STAI-State > 40; Less than 20 min/day moderate-to-vigorous physical activity/day in prior year	0%	34.8	HIIT; 6 weeks, 3 days/week, 40 min, “some what hard” full-body exertion	Waitlist	0%	State anxiety	STAI-State	STAI-State: Group x time interaction was significant (p = 0.044). In 2 × 2 ANOVA post-hoc comparisons, only exercise group showed decreased anxiety at posttest (p = .002; Cohen’s d = −0.71).	No	Yes	-Inadequate control for expectations and attention -No covariate adjustment in analyses -Small sample
Ma et al. (2017) Taiwan N = 86	Patients with a DSM-IV anxiety disorder	67.5%	40.1	Aerobic; 3 months; at-home exercise program with video instructions for aerobic activity, Tai Chi, and walking; instructed to exercise 30 min/day, 5 days/week	Usual care (received home exercise program after follow-up completed)	3.5%	State and trait anxiety levels	STAI	STAI: State and trait anxiety levels significantly dropped from baselin e to 3-month follow-up in exercise group, though differences were not found at posttest.	No	Equivocal	-Control group with lower STAI at baseline -No ITT analysis
Rosenbaum et al. (2014) Australia N = 81	DSM-IV-TR primary diagnosis of PTSD, recruited during inpatient treatment	16.0%	47.8	Walking plus resistance bands; 12 weeks, 3 days/week (one with supervision), plus usual psychiatric care (medication, individual and group therapy)	Usual care	29% for self-report measures 54% for physical measures	PTSD symptoms	PCL-Civilian; DASS-21	PCL-Civilian: Exercise group had significantly lower scores at posttest than controls, with moderate effect size (p = 0.04). DASS-21: Scores significantly improved for exercise over usual care (p = 0.003).	No	Yes	-High attrition -Nonadherence (exercisers completed only 58% of scheduled sessions) -Baseline tests of group equivalence not reported; control group older and more women
Whitworth et al. (2019) United States N = 30	Exposure to recent traumatic event and positive screen for PTSD on PDS5 (28+)	73.3%	29.1	Resistance, 3 weeks, 3 days/week, 30 min each: squat, bench press, pulldown, overhead press, and bicepscurl	Educational videos, time-matched for contact time with exercise group	20%	PTSD symptoms	PDS5, avoidance and hyper -arousal symptoms STAI	PDS5: Exercise group reported lower symptoms, but no group differences after adjustment for multiple comparisons. STAI: Not significant.	No	No	-No ITT analysis -Nonadherence: Participants completed an average of 6–7 of 9 sessions -Authors refer to “large, non-significant” findings of lower PTSD symptoms -Small sample
Zheng et al. (2018) Australia N = 69	Age 18–60, healthy, STAI > 50	78% at end	33.9	12 weeks, 2–5 hours/week of unspecified exercise at a fitness center	Two groups: Tai Chi (considered a separate intervention arm), 6 weeks of group and 6 weeks at home (total 12), and waitlist	27.5%	Anxiety	STAI	STAI-State anxiety decreased at 6 weeks for exercise and Tai Chi groups (p<0.001), but not waitlist. STAI-Trait anxiety decreased at weeks 6 and 12 for exercise and Tai Chi groups (p < 0.001) but not waitlist. No difference between exercise and waitlist in post hoc analysis, but Tai Chi reported superior to waitlist at week 12 (p = 0.010).	No	Equivocal	-Baseline comparisons not reported for several tests -Exercise and Tai Chi groups are not compared with post hoc tests; authors state Tai Chi “not inferior” to exercise without reporting tests -Exercise intervention is inadequately described (intensity, duration, modality) and adherence is not reported

Note: ANOVA - analysis of variance; ASI - Anxiety Sensitivity Index; BAI - Beck Anxiety Inventory; CAPS - Clinician-Administered PTSD Scale; DASS - Depression Anxiety Stress Scales; DSM - Diagnostic and Statistical Manual of Mental Disorders; GAD - Generalized Anxiety Disorder; HADS - Hospital Anxiety and Depression Scale; HAM-A - Hamilton Anxiety Rating Scale; ITT - intention-to-treat; NIMH-SR - National Institute of Mental Health Self-Rating Scale; OCD - Obsessive-Compulsive Disorder; OCI-R - Obsessive-Compulsive Inventory-Revised; PAS - Panic and Agoraphobia Scale; PCL - Posttraumatic Stress Disorder Checklist; PD - Panic Disorder; PDS5 - Posttraumatic Stress Disorder Scale for DSM-5; PDSQ - Psychiatric Diagnostic Screening Questionnaire; PDSS-SR - Panic Disorder Severity Scale-Self Report; PEDro - Physiotherapy Evidence Database; PSWQ - Penn State Worry Questionnaire; PTSD - Posttraumatic Stress Disorder; SAD - Social Anxiety Disorder; SCID - Structured Clinical Interview for DSM (-IV or −5, RV - Research Version); SIPS - Social Interaction Phobia Scale; SMAA - Severity Measure for Agoraphobia; SMSP-A - Severity Measure for Specific Phobia-Adult; STAI - Spielberger State-Trait Anxiety Scale; Y-BOCS - Yale-Brown Obsessive Compulsive Scale.

The remaining 12 studies [49–53, 64–70] included anxiety as an outcome, but did not require an anxiety disorder or heightened anxiety for eligibility (see Table 2). A total of 1043 individuals were randomized in these studies, with sample sizes ranging from 32 to 252. Of these studies, nine of 12 studies selected participants with other psychiatric or medical comorbidities such as lung cancer [64], Parkinson’s disease [51, 52], methamphetamine dependence [68], stroke [53], fibromyalgia [66], and schizophrenia [69].

Table 2:

Summary of Randomized Controlled Trials where Anxiety is NOT a Prerequisite for Eligibility

Author (Year), Country, N	Inclusion Criteria	% Female	Age (mean yrs)	Exercise Intervention	Control Condition	% Attrition	Primary Outcome	Anxiety Measures	Anxiety Findings	Mediator/Moderator Reported?	Does Exercise Reduce Anxiety?	Limitations
Aidar et al. (2018) Brazil N = 43	Ischemic stroke >1 year prior to study entry	47.2% (at end of study)	52.2	Aerobic; 12 weeks, 2 days/week, 45–60 min of walking, pedaling, and swimming	Waitlist	16.3%	Depression and anxiety	STAI	STAI: Exercise group showed lower anxiety at posttest with at least medium effect size (two-way ANOVA with Bonferroni corrections: trait: p = 0.002, Cohen f2 = 0.358; state: p = 0.003, Cohen f2 = 0.223). Control group unchanged.	No	Yes	-No baseline group comparisons -No ITT analysis -Authors state perceived exertion was measured but do not report results or targets -Inadequate control for attention -Small sample
Chen et al. (2015) Taiwan N = 116	Primary lung cancer; age 18+; moderate physical activity <150 min/week; no cognitive impairment; no behavioral therapy in prior 6 months; no repeated onset of depression	53. 4%	64.2	Home-based walking regimen; 12 weeks, 3 days/week, 40 min, with weekly phone calls	Usual care	13% (post-test) 21.6% (3-month follow-up)	Anxiety and depression	HADS	HADS: Exercise group anxiety scores were unchanged from baseline to posttest or 3-month follow-up, whereas control group showed an increase from baseline anxiety. However, exercise group showed greater decrease in anxiety over time at posttest (p = 0.001) and follow-up (p = 0.042). Significant decrease in number of participants at/above clinical cutoff for symptoms only for exercise group (p = 0.049).	No	Equivocal	-Nonadherence and attrition; only 44.8% of the exercise group completed the program -Inadequate control for expectations and attention -Blinding of assessors insufficient
Ferreira et al. (2018) Brazil N = 35	Parkinson’s disease, with stable medication use; ages 60+; no exercise protocol in prior 3 months; intact mental status, able to walk independently	Not stated	65.8	Resistance training program: 6 months, 2 days/week, 30–40 min	Usual care	0%	Anxiety	BAI	BAI: Scores decreased at posttest in exercise group (t-test, effect size = 0.415, p = .0001), but not control. 2 × 2 ANOVA is stated to have been done, but results not reported.	No	Yes	-Inadequate description of patient sample (e.g., concurrent medications) -Incomplete data analysis -Inadequate control for expectations and attention -Small sample
Gokal et al. (2016) United Kingdom N = 50	Women aged 18–75 with stage I-III breast cancer, relatively inactive physically, with no psychiatric illness to hinder participation	100%	52	Aerobic; 12 weeks, self-managed, moderate-intensity walking, recommended up to 30 minutes 5 times/week	Usual care	16% (study period) 0% (post test)	Psychosocial functioning (anxiety, mood, self-esteem)	HADS	HADS: Anxiety improved over time for both groups (p < 0.01), without a significant effect of group (p = 0.06) or time x group (p = 0.35)	No	No	-Inadequate control for attention -Sampling bias (dropouts less educated, less likely to be employed) -Small sample
Izquierdo-Alventosa et al. (2020) Spain N = 32	Women between 30–70 years old with diagnosed fibromyalgia, with no clinical improvement after 3+ months of medication treatment	100%	54.1	Both aerobic and resistance; 8 weeks, 2 days/week, walking plus lifting	No intervention, maintain baseline physical activity and complete logs	0%	Pain catastro phizing	PCS HADS	PCS and HADS-Anxiety: Exercise group lower than control at posttest, p < 0.05 (Bonferroni-corrected multivariate ANOVA post hoc tests)	No	Yes	-Baseline differences not reported for anxiety, only select variables -Inadequate control for attention -Small sample
Kwok et al. (2019) Hong Kong N = 138	Clinical diagnosis of idiopathic Parkinson’s disease, age 18+, no medication or surgical treatment for psychiatric disorders	52. 9%	63.7	Resistance training and stretching: 8 weeks, 60 min/week, encouraged to complete 2 more independent sessions of 20 min/week	Yoga and mindfulness program: 8 weekly sessions of 90 min Note exercise condition was considered the control group.	18.8%	Anxiety and depression	HADS	No change over time for exercise group. Yoga group had greater reductions in anxiety from pre- to posttest (group x time interaction, p = .001), including posttest and 12-week follow-up.	No	No	-Convenience sampling -Limited generalizability -Treatment nonadherence; 22% of exercise group attended no sessions, compared to 0% for mindfulness yoga group
Loh et al. (2019) United States N = 252	Primary diagnosis of cancer other than leukemia; age 60+; undergoing first-time chemotherapy without radiation	91.7%	67 (median - did not report mean)	Home-based walking and resistance exercise, low to moderate intensity, 6 weeks. 3 days/week, 40 min, with weekly phone calls	Usual care; participants recorded steps via a pedometer for 4 days at baseline and 4 days before the end of the study	Not reported	Anxiety, mood, and social and emotional well-being	STAI	Exercise group participants with higher anxiety had greater improvement in STAI scores over time, (baseline anxiety x treatment interaction, p = .001). However, individuals below the 75th percentile did not show significant improvements.	Moderator: Baseline anxiety - Greater reduction in anxiety for exercise participants with higher baseline anxiety. Mediator: Daily steps - not significant.	Equivocal	-Secondary analysis of subset of participants (age 60+) from a larger trial -Intervention not standardized -Treatment nonadherence; 71% of exercise group reported doing resistance exercise, but change in steps/day was no different across groups
Lucibello et al. (2019) Canada N = 50	University students, ages 18–30, no more than 60 min/week of moderate to vigorous physical activity for prior 6 months. Participants were divided post hoc into high and low anxiety severity subgroups.	82.2%	19.4	Aerobic; 9 weeks, 3 days/week, 30 min of moderate-intensity cycling	No exercise training. Directed to continue physical inactivity for 9 weeks	16%	Anxiety	STAI-State, 6-item version, emailed twice per week, spaced 40 minutes apart	Main effects found for group, anxiety severity, group x time anxiety severity x time (all p = .006 or lower). In those with high baseline anxiety, state anxiety decreased more over the 9-week exercise intervention (p=.029). Scores not significant for the low anxiety exercise group or either control group. No differences found in the size of acute reductions in state anxiety over 9 weeks.	Moderator: Baseline anxiety - Greater reduction in anxiety for exercise participants with higher baseline anxiety.	Equivocal	-Determination of sample size not reported -No ITT analysis -Only measured acute effects of exercise on state anxiety -Ns not reported for subgroups -No anxiety descriptive statistics - No rationale given for dividing non-clinical sample into low vs. high anxiety post-hoc -Small sample
Lucibello et al. (2020) Canada N = 60	University students ages 18–30 engaging in no more than 1 hour/day of moderate to vigorous physical activity prior to study	63%	19.8	HIIT; 9 weeks, 3 days/week, 20 minutes of stationary cycle	Placebo; patients told study was of “acute” exercise, referring to pre- and posttest exercise tasks, and that they should not exercise between sessions	23.3%	Anxiety, depression, and pro-inflammatory cytokines	BAI	Anxiety decreased for both HIIT and placebo groups, with no between-group differences.	No	No	-No ITT analysis -Not true ‘placebo’ group -Unclear control for expectations of benefit (not measured) -Unclear if control group performed similar amount of exercise at pre- and posttest
Raws on et al (2015) United States N = 135	Methamp hetamine dependence, newly abstinent, in residential treatment, ages 18–45 (male) or 18–55 (female)	29.6%	31.7	Combined aerobic and resistance exercise; 8 weeks, 3 days/week, 60 min	Health education	2.2%	Anxiety and depression	BAI	Both groups demonstrated significant reductions in anxiety at posttest, but exercise group showed greater reductions in anxiety compared to education controls (p = 0.001).	Moderator: Attendance of exercise sessions - More sessions associated with lower anxiety at posttest.	Yes	-Limited generalizability due to unique sample -No ITT analysis -Did not adjust for baseline differences in levels of anxiety and depression
Ryu et al. (2020) South Korea N = 60	Adults with schizophrenia; ages 18–65; stable dose of anti-psychotic medication for 4+ months	45%	38.8	Aerobic; 16 weeks, once per week, group-based outdoor cycling; 15 min goal setting plus education, 10 min warm-up, 40 min of bike training at moderate intensity, 10 min cool down, and 15 min discussion	Occupational therapy for 90 min once per week	16.6%	Mental health (psychiatric symptoms, depression, anxiety, global function); cognition; physical activity; adherence	STAI	STAI: Outdoor cycling for 16 weeks significantly reduced anxiety compared to occupational therapy; significant group x time interactions for state (p = .031) and trait (p = .002) anxiety	No	Yes	-No ITT analysis -Multiple “primary” outcomes including anxiety -Adherence to the exercise intervention not well described -Exercise intervention confounded with more than 30 min of social interaction and counseling
Vancini et al. (2017) Brazil N = 72	Age 18–66, BMI > 25; at least 6 months ‘without practicing physical activity’	60.0%	46.6	Home-based walking regimen; 12 weeks, 3 days/week, 40 min, with weekly phone calls	2 groups: No exercise (control) and Pilates	12.5%	Quality of life, anxiety and depression	STAI	Both walking and Pilates groups showed decreased trait anxiety (p = 0.06) per post hoc analysis, and walking group also showed decreased state anxiety (p = 0.01). No between group comparisons reported.	No	Equivocal	-No-exercise control group and randomization not described --No baseline group tests of equivalence -No ITT analysis

Note: ANOVA - analysis of variance; BAI - Beck Anxiety Inventory; BMI - body mass index; HADS - Hospital Anxiety and Depression Scale; HIIT - high-intensity interval training; ITT - intention-to-treat; PCS - Pain Catastrophizing Scale; PEDro - Physiotherapy Evidence Database; STAI - Spielberger Stait-Trait Anxiety Inventory.

3.2. Exercise Interventions

Aerobic exercise alone was the primary intervention in 12 of the 25 studies. The most common modalities of aerobic exercise were walking and cycling, including one study using a high intensity interval training (HIIT) cycling program [50]. Resistance training was the primary intervention in four studies [40, 51, 52, 63]. Resistance and aerobic exercise were combined in six studies [39, 47, 56, 66–68].

Of the remaining three studies, one [46] had participants engage in both aerobic and anaerobic exercises that were described as HIIT, with examples of activities including abdominal crunches, pushups, and squat jump exercises. Another [57] included two separate intervention groups, one using aerobic exercise and the other using resistance exercise. Finally, in a study by Zheng and colleagues [59], participants in the exercise group were directed to do 2–5 hours per week of exercise for 12 weeks at a fitness center and provided a membership and classes. Of note, the authors did not specify the mode of exercise or the content of the exercise classes, and the exercise group was considered a comparison group, in addition to a waitlist control group, for a Tai Chi intervention.

Among those studies that involved aerobic exercise, six of 20 reported exercise intensity, which ranged from 60% to 85% of maximum heart rate [49, 54, 55, 61, 68] with the exception of the sprint intervals of 95% of maximum in a trial of HIIT cycling [50]. Eight of 20 reported ratings of perceived exertion [39, 46, 47, 49, 53, 64, 66, 67]. In the other aerobic exercise studies, the intensity of exercise was not reported [45, 48, 56, 57, 65, 69].

Exercise supervision was provided in 21 of the 25 studies. In three separate studies [48, 65, 67], participants engaged in an unsupervised, home-based exercise regimen for approximately 12 weeks, such that adherence to the exercise prescription was not reported and could not be confirmed. An additional study failed to describe the exercise intervention in sufficient detail to determine if any supervision was provided [59].

Length of interventions varied from 2 to 26 weeks, with the modal intervention lasting 12 weeks (11/25 studies). Exercise frequency ranged from one to five sessions per week. Five studies [52, 54, 55, 61, 64] provided analyses of follow-up measurements of anxiety beyond the post-testing immediately following the completion of the intervention.

3.3. Control Conditions

A majority of the RCTs (16 of 25) compared the exercise intervention condition to a single, non-exercise control group without any expectation of benefit, such as a waitlist, usual care, or no intervention at all. Of these, one study [67] also instructed participants to count their steps for four days at the start and at the end of the study period. Another study [49] of inactive individuals instructed the control group to maintain their physical inactivity for the duration of the study period. In one attempt to provide equipoise, Lucibello and colleagues [50] conducted an exercise bout at baseline with all participants, but non-exercise control participants were told that the study concerned the effects of a single bout of acute exercise over a follow-up period and had no further contact prior to posttest. In a partial crossover design, LeBouthillier and colleagues [57] used a waitlist control condition and re-randomized participants from the waitlist group into either aerobic or resistance exercise, such that all participants received exercise training. However, no separate analyses were reported comparing these participants to those who exercised without first being on a waitlist. Also of note, one trial [48] reported that controls received usual care during the study period, and then were given written materials for home-based, self-directed exercise after the trial was completed. However, no data were collected from this home-based intervention.

Non-exercise ‘movement’ was a component of the control group protocol for five of the remaining trials. Two studies included both a waitlist control group and a comparison group that carried out structured movements, such as Pilates or Tai Chi [59, 70]. A third study [52] compared individuals participating in mindfulness yoga exercise (poses, controlled breathing, and meditation) to individuals participating in stretching and resistance exercise, which Kwok and colleagues considered the control condition. In another study, participants carried out light stretching for the same duration as that of the exercise intervention group [55]. Finally, one trial assigned individuals with schizophrenia in the control group to weekly occupational therapy for 90 minutes [69].

Education was employed as a control in three trials [40, 44, 45, 68], including one in which the time that control participants spent interacting with the study team was directly matched with that of exercisers [40].

Finally, Blumenthal and colleagues [54] included a placebo pill as a control condition, as well as a third arm of the trial involving an established anxiolytic medication, escitalopram.

3.4. Anxiety Measures

Nineteen of the 25 studies used only self-report questionnaires to measure anxiety prior to delivery of the intervention, including forms of the STAI [60], the anxiety subscale of the Hospital Anxiety and Depression Scale (HADS) [71], Beck Anxiety Inventory (BAI) [72], Anxiety Sensitivity Index (ASI) [62], Posttraumatic Stress Disorder Checklist (PCL) [73], and Psychiatric Diagnostic Screening Questionnaire (PDSQ) [74]. Four studies assessed baseline anxiety via structured interviews, including the Clinician-Administered PTSD Scale (CAPS) [75], Hamilton Anxiety Rating Scale (HAM-A) [76] and the Structured Clinical Interview for the DSM that was current at the time of the study (e.g., SCID-IV-TR, SCID-5-RV) [77]. The remaining study [57] utilized both a self-report measure and structured interviews.

Anxiety outcome measures varied widely across studies. The most commonly used instruments were the STAI, BAI, and the HADS. The STAI, either in English or translated versions, was used in 10 of the 25 studies. Five studies included English or translated versions of the BAI. The HADS also was used in five RCTs. Other measurement instruments included the ASI [62], CAPS [75], Depression Anxiety Stress Scales (DASS) [78], Generalized Anxiety Disorder Assessment (GAD-7) [79], National Institute of Mental Health Self-Rating Scale (NIMH-SR) [80], Obsessive-Compulsive Inventory-Revised (OCI-R) [81], Pain Catastrophizing Scale (PCS) [82], Panic and Agoraphobia Scale (PAS) [83], Posttraumatic Stress Diagnostic Scale for DSM-5 (PDS5) [84], Panic Disorder Severity Scale-Self Report (PDSS-SR) [85], Penn State Worry Questionnaire (PSWQ) [86], Social Interaction Phobia Scale (SIPS) [87], Severity Measure for Agoraphobia (SMAA) [88], Severity Measure for Specific Phobia-Adult (SMSP-A) [89], and Yale-Brown Obsessive Compulsive Scale (Y-BOCS) [90]. Six studies assessed outcomes using multiple measures of anxiety.

3.5. Outcomes of the Exercise Interventions

Seven of the 25 studies offered clear evidence that the exercise intervention groups showed a greater improvement in anxiety symptoms [46, 51, 53, 66, 68, 69] or disorder status [47] compared to controls. Only two of 13 studies among anxious individuals showed unequivocal reductions in anxiety for the exercise group compared to controls [46, 47].

Six of 25 studies showed minimal or no differences in anxiety scores for exercisers compared to controls. Finally, we considered findings to be equivocal or inconclusive in the remaining 12 studies (48% of sample), including eight of the 13 trials conducted among anxious individuals. In most of these studies, findings were considered ‘equivocal’ because results showed a decrease in anxiety for only a subset of measures or assessment points. We note that in three studies [56, 59, 63] the authors reported unequivocally positive findings, but we determined that whether exercise reduced anxiety was equivocal, due to significant methodological limitations (see Discussion). Of the 25 studies, only four report any analysis of potential mediators or moderators of an anxiety-exercise relationship. For moderation, three studies reported that reduced anxiety was moderated by either the number of exercise sessions attended (more exercise associated with greater anxiety reduction) or the level of anxiety at baseline (higher baseline associated with greater anxiety reduction). Two studies assessed potential mediators, specifically participants’ daily steps or increased levels of aerobic fitness. Neither of these analyses yielded a significant result.

3.6. Methodological Quality Assessment

Assessment of all 10 PEDro criteria has been detailed in Table 3. Among the 25 studies, no study satisfied all criteria, and scores ranged from 2 to 8. All of the included studies met the criterion for randomizing participants to groups appropriately. No study met the criterion that all therapists who administered the therapy (i.e., exercise protocol) were blinded. We note that satisfying this criterion is typically not possible given the nature of exercise interventions. Only one study [48] blinded participants to their group allocation; this criterion was considered met through a waitlist condition, as the individuals who were monitored for 3 months of usual care then received the same home-based exercise materials as the intervention group. However, it was unclear whether participants were informed that there were two arms to the trial and that their exercise intervention was delayed. Another group [50] attempted to blind participants to their allocation by characterizing the no-intervention control group as an exercise condition. That said, no manipulation check was reported for this approach. In another study [54], participants knew that they were in an exercise or ‘pill’ condition, but participants in the ‘pill’ condition (and the treating psychiatrist) were unaware of whether they were receiving escitalopram or placebo.

Table 3:

PEDro Criteria for Included RCTs

Author (Year)	Elevated Baseline Anxiety	Random allocation to groups	Allocation concealed	Similar baseline groups	Blinding: participants	Blinding: interventionists	Blinding: assessors for ≥1 key outcome	≥1 key outcome reported for ≥85% participants	ITT analysis	Between-groups result for ≥1 key outcome	Point and variability measures for ≥1 key outcome	PEDro rating (total)
Abrantes et al. (2017, 2019)	Yes	Y	-	-	-	-	-	Y	Y	Y	Y	5
Aidar et al. (2018)	-	Y	Y	Y	-	-	-	-	-	Y	Y	5
Blumenthal et al. (2021)	Yes	Y	Y	Y	-	-/Y (pill condition)	Y	Y	Y	Y	Y	8
Chen et al. (2015)	-	Y	Y	Y	-	-	-	Y	Y	Y	Y	7
Ferreira et al. (2018)	-	Y	Y	Y	-	-	Y	Y	Y	-	Y	7
Gaudlitz et al. (2014)	Yes	Y	-	Y	-	-	Y	-	-	Y	Y	5
Gokal et al. (2016)	-	Y	Y	Y	-	-	-	Y	Y	Y	Y	7
Goldstein et al. (2018)	Yes	Y	-	-	Y	-	-	-	-	Y	Y	4
Gordon et al. (2021)	Yes	Y	-	Y	-	-	-	-	Y	-	Y	4
Hall et al. (2020)	Yes	Y	Y	Y	-	-	-	Y	Y	Y	Y	7
Izquierdo-Alventosa et al. (2020)	-	Y	Y	Y	-	-	-	Y	Y	Y	Y	7
Kwok et al (2019)	-	Y	Y	Y	-	-	Y	-	Y	Y	Y	7
Lanoye et al., 2021	Yes	Y	-	-	-	-	-	-	Y	-	Y	3
LeBouthillier et al. (2017)	Yes	Y	-	-	-	-	-	Y	-	-	-	2
Legrand et al. (2020)	Yes	Y	Y	Y	-	-	-	Y	Y	Y	Y	7
Loh et al. (2019)	-	Y	Y	Y	-	-	-	-	Y	Y	-	5
Lucibello et al. (2019)	-	Y	Y	Y	-	-	-	Y	Y	Y	-	6
Lucibello et al. (2020)	-	Y	Y	Y	-	-	-	-	-	Y	Y	5
Ma et al. (2017)	Yes	Y	-	Y	Y	-	-	Y	-	-	-	4
Rawson et al. (2015)	-	Y	Y	Y	-	-	-	Y	-	Y	Y	6
Rosenbaum et al. (2014)	Yes	Y	Y	Y	-	-	Y	-	-	Y	Y	6
Ryu et al. (2020)	-	Y	Y	Y	Y	-	-	Y	-	Y	Y	7
Vancini et al. (2017)	-	Y	-	-	-	-	Y	Y	-	Y	Y	5
Whitworth et al. (2019)	Yes	Y	Y	Y	-	-	-	-	-	Y	Y	5
Zheng et al. (2018)	Yes	Y	Y	Y	-	-	-	-	Y	Y	Y	6

The remaining PEDro criteria that most commonly were not satisfied were non-blinding of assessors (not met in 68% of studies), no ITT analysis (52% of studies), missing data for key outcomes in >15% of sample (44% of studies), and lack of allocation concealment (40% of studies).

3.7. Review of Recent Reviews and Meta-Analyses

Our search identified several new exercise and anxiety systematic reviews and meta-analyses published since 2014. In a review by Aylett and colleagues [38], 15 studies were identified with a total of 675 patients. Nine trials had participants with diagnosed anxiety disorders and six trials had participants with raised anxiety. Aerobic exercise was effective in the treatment of raised anxiety compared to waiting list control groups (effect size − 0.41, 95% CI = − 0.70 to − 0.12), with high intensity exercise showing greater effects compared to low intensity exercise. A meta-analysis by Stubbs et al. [23] covered articles published through December 2015, screening 3999 entries, reviewing 62 full-text publications, and identifying six qualifying RCTs including 262 adults. Exercise was reported to significantly decrease anxiety symptoms more than control conditions, with a moderate effect size (standardized mean difference [SMD] = −0.582). In a follow-up meta-analysis to the Stubbs review, Ramos-Sanchez and colleagues [37] identified 13 RCTs that met their eligibility criteria, comprising 731 adult participants. Exercise was considered to have ‘small/bordering medium anxiolytic effects in persons with anxiety and related disorders.’ Finally, the meta-analysis by Jacinto et al. [36] examined physical activity as a method to reduce anxiety in individuals with disabilities. A total of 330 studies from January 2001 to June 2021 were initially identified, but only four met the inclusion criteria, three of which involved individuals with intellectual disabilities. Of note, their search identified no study conducted in English-speaking countries, and only two studies had what the authors considered ‘good’ methodologies.

4. Discussion

The present systematic review identified 25 new RCTs since January 2014 that evaluated exercise as an intervention for anxiety as a primary outcome, including 13 studies among anxious individuals. Search results indicated that the number of RCTs meeting inclusion criteria has more than doubled since the Stonerock 2015 review [27]. The increase in the number of published studies in this area could be considered an encouraging opportunity for more definitive conclusions regarding the merits of exercise as a treatment for anxiety. However, examination of study methodologies and results suggest that there remains considerable uncertainty as to the efficacy of exercise as a treatment for anxiety. Indeed, 18 of 25 RCTs either had equivocal findings or found that exercise did not improve anxiety, whereas only seven studies reported positive findings.

For most of the seven positive trials, the control group was either a waitlist or usual care. Of the two positive RCTs conducted with anxious individuals, one study showed moderate improvements in PTSD symptoms and anxiety among psychiatric inpatients diagnosed with PTSD who were randomized to exercise with both walking and resistance bands rather than usual care [47]. In the other, first-time prisoners who were physically inactive in the prior year completed 6 weeks of HIIT training and reported lower state anxiety at posttest than those on a waitlist [46].

The remaining five positive studies were conducted among non-anxious individuals. In one, a group of ischemic stroke survivors who completed 12 weeks of various aerobic exercises reported lower state and trait anxiety at posttest versus waitlist controls [53]. Among a group of women with fibromyalgia who did not benefit from medication, lower pain catastrophizing and self-reported anxiety were reported after 8 weeks of walking and weight lifting at a low intensity compared to no intervention [66]. Resistance training for 6 months was associated with lower BAI scores compared to usual care in a RCT of 35 individuals with Parkinson’s disease [51]. In addition, individuals with schizophrenia who were randomized to a 16-week group cycling intervention reported lower anxiety ratings compared to those spending equivalent time in occupational therapy [69]. It is noteworthy that these authors identified anxiety as one of a number of ‘primary’ outcomes, including global functioning and psychiatric symptoms related to schizophrenia, but a separate, validated measure of anxiety, the STAI, was included. Finally, in a study of 135 individuals in residential treatment for methamphetamine dependence, Rawson and colleagues [68] reported that those participants who engaged in a combined aerobic and resistance exercise program reported lower anxiety at posttest compared to health education controls.

Upon in-depth review, we considered three purportedly ‘positive’ trials to, in actuality, have equivocal results, despite the authors’ claim of unequivocal benefit of exercise for anxious individuals. Goldstein and colleagues [56] described a 12-week “integrative” exercise regimen, combining aerobic and resistance exercise, conducted among military veterans with diagnosed PTSD or partial PTSD. They reported a significant reduction in symptoms for the exercise group at posttest compared to waitlist controls. However, their integrative intervention included other treatment modalities, such as yoga, mindfulness, and psychotropic medications, not available to the control group, and did not control for attention or expectation of benefit among controls. Thus, we deemed the study inconclusive as to whether exercise resulted in a reduction in anxiety.

In another study, 8 weeks of resistance exercise was associated with greater remission rates of subclinical GAD versus waitlist [63]. Statistically significant interactions were reported for some, but not all, worry symptoms, as well as for self-reported anxiety on the STAI, indicating lower symptoms in the exercise group. However, remission rate differences were substantially different depending on the criterion for remission used, with the most stringent criteria showing remission in only two exercise participants compared to one waitlist participant; thus, remission rates were too small to draw meaningful conclusions. Additionally, this trial had an attrition rate over 40% and conducted ITT analyses only as sensitivity analyses, raising concerns about the positive interpretation of remission rates.

Finally, Zheng and colleagues [59] compared three intervention arms among individuals with high self-reported anxiety symptoms. This group considered the exercise arm, and a waitlist, as control groups to compare against a program of 12 weeks of Tai Chi (group and home). Those assigned to exercise were directed to complete 2–5 hours per week of exercise at a fitness center, including certain offered classes. Although the exercise group had better outcomes compared to waitlist controls at many time points, no basic information about exercise modality, duration, intensity, or adherence was provided, leaving insufficient information to conclude that exercise reduced anxiety.

In other cases, RCTs were considered equivocal due to a mixture of positive and null findings. Eight studies among anxious participants were considered to have equivocal findings. In a study of 56 individuals with treatment-resistant OCD [44], those randomized to a moderate-intensity walking intervention showed a marginal trend for more ‘treatment responders’ in OCD symptoms compared to health education controls (30.4% versus 7.7%), but post-treatment group differences on the BAI and Y-BOC-S were not significant. In a second publication from the same study, Abrantes et al. [45] reported a significant decrease in acute anxiety measured by the NIMH-SR from the start to end of each weekly session, with a larger decrease for exercisers compared to education controls. However, exercisers had higher scores at baseline, suggesting possible regression to the mean. Further complicating the interpretation of results was that exercisers, but not controls, also received 20 minute ‘discussion’ sessions on topics related to increasing physical activity, prior to each session, which the health education controls did not receive. Because exercise was confounded with these discussion sessions, the effects of exercise independent of these discussion sessions cannot be determined.

Gaudlitz and colleagues [55] compared treadmill walking for 8 weeks, 3 days per week to light movement among individuals with PD without severe agoraphobia. Only one of the three self-report measures indicated a decrease in anxiety for the aerobic exercise group. Additionally, the groups did not differ significantly at posttest, but only at 7-month follow-up. In what was described as a ‘transdiagnostic’ approach, LeBouthillier and colleagues [57] randomly assigned individuals with one of several DSM-5 anxiety diagnoses to either aerobic or resistance exercise, some after having completed measures in a waitlist control group. The number of individuals meeting diagnostic criteria improved at posttest in the two exercise groups compared to waitlist controls. However, measures of symptom relief did not consistently improve for exercisers. In another study of 86 individuals with any diagnosed anxiety disorder, those who participated in a 3-month home-based aerobic program reported no difference from usual care at posttest but improved state and trait anxiety at 3-month follow-up [48].

There have been few studies of anxiety in patients with CVD. In 2004, Lavie and Milani [4] reported more than a 69% reduction in anxiety among highly anxious participants in an exercise-based cardiac rehabilitation program; however, there was no control group, and exercise was only one component of the intervention, leaving uncertainty about the benefits of exercise in treating anxiety in this patient population. In one of the few RCTs of anxious CVD patients, Blumenthal and colleagues [54] reported greater reductions in state, but not trait, anxiety, among exercisers compared to controls, after 12 weeks of treatment. Escitalopram, but not exercise, resulted in greater reductions in HADS anxiety scores, the primary outcome measure, as well as scores on other trait anxiety measures compared to placebo controls. Moreover, the benefit of exercise compared to placebo controls was marginal, showing greater improvement on state anxiety, but not on multiple measures of trait anxiety. Follow-ups at 6-months [91] and, more recently at 12-months [92], showed that the pattern of results persisted. Participants in the escitalopram group reported lower anxiety compared to placebo controls and exercisers, despite exercisers achieving greater functional capacity and reporting more physical activity during the follow-up.

Four trials with equivocal findings were conducted among non-anxious individuals. Two trials were conducted in patients with cancer, comparing usual care to a home-based walking program. In a sample of 116 adult lung cancer patients in Taiwan, those who completed a 12-week walking regimen reported no statistically significant change from baseline to posttest, whereas those in the usual care group reported an increase in baseline anxiety over the same interval [64]. Loh and colleagues [67] described a similar, 6-week intervention that included resistance exercise of low to moderate intensity, among 252 individuals undergoing first-time chemotherapy for cancer other than leukemia. Only those who had the highest anxiety at baseline showed significant improvements in anxiety over time, when compared with usual care. Of note, step counts per day were not different between the exercise and usual care groups, though a majority of exercise group participants reported doing resistance exercise. A sample of 72 physically inactive adults in Brazil participated in one of three treatment arms: a 12-week, home-based walking regimen, a Pilates program of equivalent length, and a no-exercise control group [70]. Both exercise and Pilates were associated with lower trait anxiety at posttest compared with the no-exercise control group, but post hoc significance tests showed differences between exercise and Pilates only for state anxiety. Finally, Lucibello and colleagues [49] divided a group of physically inactive college students into high or low (but not clinically elevated) baseline anxiety groups, and randomly assigned them to either 9 weeks of HIIT cycling training or continued physical inactivity. Decreases in state anxiety over 9 weeks were reported only for the higher-anxiety HIIT group, not for all exercise groups relative to the control group. No rationale for partitioning the participants was provided.

Negative results were observed in six RCTs, indicating no difference in anxiety between exercise and control groups. Kwok and colleagues [52] reported that participants with Parkinson’s disease who engaged in strength and resistance exercises over eight weeks displayed less improvement in anxiety symptoms compared to participants who engaged in a mindfulness yoga intervention. Of note, in this study, resistance exercise was considered the “control” condition, and participants were recruited from a convenience sample. Additional studies of individuals with positive PTSD screenings [40] or clinician diagnoses of PTSD prior to enrollment [39] showed non-significant differences between the exercise group and individuals receiving usual care or time-matched study contact controls without an intervention. In a study of 44 individuals with high anxiety sensitivity, change in anxiety did not differ between participants who completed 2 weeks of treadmill exercise as an interceptive exposure and no-intervention controls [61]. A 12-week, home-based walking intervention did not produce a greater change in HADS anxiety scores, relative to usual care, in a sample of 50 breast cancer patients receiving chemotherapy [65]. Finally, a study of HIIT among non-anxious college students indicated that both controls and those who participated in a HIIT intervention had similar changes in anxiety [50].

Many studies in this review had what we considered serious methodological problems. For studies of participants in whom anxiety was an eligibility criterion, only three studies obtained PEDro scores of 7 or greater and only one study [54] achieved a score of 8. For studies in which anxiety was not a prerequisite, no study obtained a PEDro score of 8 or higher, but six studies achieved scores of 7; three were positive, two were negative, and one was equivocal. Overall, 16 of 25 studies had a PEDro score of ≤6. Among the methodologic shortcomings that were considered important, failure to adhere to the ITT principle, confounding of exercise with socialization or concurrent therapies, and lack of blinding of assessors were common.

In addition to standards of methodological quality addressed with PEDro ratings, a number of further methodologic concerns were sufficiently common to render available evidence as inconclusive. A common source of bias was that few studies attempted to provide equal attention to participants in both exercise and control conditions and expectations for benefit were not equivalent. Indeed, 16 of 25 studies compared the exercise intervention group to a single control group with either usual care, waitlist, or no intervention. This lack of equipoise constitutes a source of bias that leaves open whether differential expectations or attention in the groups could explain group differences. Additionally, studies with a group meeting component for exercisers only (e.g., [51, 59, 70]) would introduce the possibility that social interaction contributed to group differences. Although some studies mitigated this type of design issue with control groups such as group-based health education, in one case time-matched to the exercise intervention [40], most did not. Additionally, we identified several studies in which the intervention either was not described with sufficient detail or was not adequately standardized or monitored, leaving unclear as to what ‘dose’ of exercise participants received. As an example, several interventions (e.g., [48, 52, 59]) provided exercise instructions or recommendations for unsupervised exercise, such as a home-based walking regimen and educational materials for how to exercise, but did not actually determine whether participants engaged in exercise. In one case, the authors report that participants were expected to exercise at a fitness center and provided access, but no effort was made to monitor their participation or to report adherence [59]. Although such designs offer the potential for greater ecological validity outside of a supervised research setting, any advantage is offset by the uncertainty regarding adherence to the exercise intervention.

Similarly, several RCTs reported low adherence to the intervention, which compromised the internal validity of the studies. In one study [52], 22% of those participants assigned to exercise completed zero scheduled sessions. In one study of a home-based walking regimen among physically inactive lung cancer patients [64], less than half of participants assigned to exercise completed the study. One study of a home-based walking regimen used pedometers to assess adherence and found no significant difference between the exercise group and the no-intervention control group [67].

Several RCTs did not include ITT analyses to account for missing data and participant attrition. Such ‘completers’ analyses provide inconclusive results when evaluating the efficacy of a new treatment. In addition, several studies interpreted findings as significant despite the fact that analyses approached, but did not achieve, statistical significance. Some reports [46, 47, 57] did provide effect sizes to help with interpretation of such findings, which is important particularly in studies with small samples. In a number of cases, although a statistically significant result was found and reported, the effect size was considered small, which raises questions about real-world or clinical utility of the intervention. We also identified one study [49] in which the authors only provided summary information about participant dropout across study groups, rather than providing full information in the form of a Consolidated Standard of Reporting Trials (CONSORT) flowchart for participants [93].

Some studies included an exercise intervention combined with an additional treatment, such as pharmacotherapy or cognitive-behavioral therapy for anxiety. In some cases, all members of the exercise group received an additional treatment, such as psychotherapy [55] or mindfulness training [56], and in others, patients were already receiving treatments for anxiety, such as medications [47, 57]. The designs of these RCTs do not allow for analyses of the unique contribution of exercise to any observed benefits. This source of bias could be remedied by separating those who receive exercise alone from those who also receive other interventions. Alternatively, RCTs assessing for a unique benefit of exercise could exclude participants currently receiving any concurrent treatment. For example, the Understanding the Benefits of Exercise and Escitalopram in Anxious Patients With Coronary Heart Disease (UNWIND) trial by Blumenthal et al. [54] excluded patients receiving any supplemental treatment, which allowed for a more direct comparison of two active treatments (i.e., exercise or escitalopram) to a true control condition (i.e., a placebo pill). This trial showed clear superiority of escitalopram compared to both exercise and placebo. Results of exercise provided inconsistent results, however, in that exercise produced greater reductions in state, but not trait, anxiety compared to placebo controls.

Only five studies included a follow-up assessment of anxiety beyond a posttest at the end of the intervention period [48, 55, 61, 64, 91]. Thus, the longer-term benefits of exercise are uncertain. Of note, it is not clear if exercise was maintained during the follow-up interval and all of these studies had equivocal or negative findings as to whether exercise reduced anxiety, and only one of these [64] reported improvements to anxiety that remained significant at follow-up.

4.1. Summary

Taken together, available evidence does not offer conclusive support for the use of exercise to treat anxiety. Over 70% of RCTs reviewed provided equivocal or negative findings. Broadly, this is consistent with the prior review by Stonerock et al. [27], in which only 3 of 12 RCTs meeting inclusion criteria were considered to have unequivocal positive findings. Positive findings were more frequent in RCTs of individuals who were not anxious at baseline, whereas findings are more frequently equivocal or negative in the studies of anxious participants. Moreover, most studies suffered from significant methodologic shortcomings, as over half of the RCTs identified, 16 of 25, had PEDro scores of 6 or lower. Studies with greater methodological quality did not show more positive results. Among the 9 studies with PEDro ratings of 7 or higher, five studies had findings that were equivocal or negative compared to only four studies that were positive. Only one of the three studies of individuals with elevated anxiety, the trial by Legrand and colleagues [46], had both a PEDro score above 6 and a positive result for exercise reducing anxiety.

This conclusion stands in contrast to recent meta-analyses identified during our search. Stubbs et al. [23] concluded that exercise was effective in improving anxiety symptoms in people with a current diagnosis of anxiety and / or stress-related disorders. However, Stubbs also noted that five of six identified RCTs were of low quality, and most were considered to have a ‘high risk of bias.’ In addition, OCD and PTSD were no longer included in the anxiety disorder category according to DSM-5 [58]. When these studies were removed from the analysis, the SMD was reduced to −0.48, which was no longer statistically significant. In the review by Aylett and colleagues [38], exercise was considered a viable treatment option for anxiety, although the authors noted that “conclusions were limited by the small number of studies and wide variation in the delivery of exercise interventions.” In the review by Ramos-Sanchez et al. [37], the calculated effect size was smaller compared to the Stubbs analysis (SMD = −0.425 vs 0.58). However, of the 13 identified studies, only four studies (31%) reported a statistically significant difference between exercisers and controls. Additionally, Ramos-Sanchez advised caution in interpreting their results, identifying multiple serious methodological concerns across studies to include randomization problems, selection of outcome measures, high degree of missing data, high attrition, and substantial risk of bias due to missing data in half of the studies. Lastly, in the review of exercise in adults with disabilities by Jacinto et al. [36], all four included studies had significant confounds (e.g., combining exercise with socialization and other treatments), only two were considered to have “good” methodology by the authors, and all of the studies were thought to have limited generalizability. Despite the fact that all reviews highlighted a paucity of methodologically sound trials, authors all independently concluded that exercise was highly beneficial.

4.2. Limitations

We were only able to search in English and include studies with published full text in English. Fortunately, the original research includes individuals from many non-English speaking locations, and the frequency of keywords and titles/abstracts given in both English and another language further mitigates the possibility of missing a study that would otherwise qualify for this review. Our review focused on the benefits of multiple modalities of exercise, primarily aerobic exercise, and did not specifically focus on resistance training, which has been reviewed recently [26]. Other reviews have attempted to quantify the benefits of exercise using a meta-analytic approach. However, because of the high degree of sample heterogeneity, variability in exercise prescriptions and outcome measures, and the limited number of studies with sound methodologies, we did not perform a meta-analysis. We rather aimed to provide a critical review of the literature, so that the reader is better able to judge the quality of evidence for exercise as a treatment for individuals with heightened anxiety. Because of limited sample sizes and significant methodological weaknesses, calculation of ‘effect sizes’ was not considered appropriate.

5. Conclusion

The present review sought to update prior reviews on the potential benefits of exercise in reducing anxiety among individuals with anxiety diagnoses or elevated anxiety symptoms. Although 25 new RCTs were identified since 2014, including 13 conducted among anxious individuals, the paucity of high-quality research and negative or equivocal findings for more than half of the published studies precludes any definitive conclusions about the presumptive value of treating patients for anxiety with exercise. To our knowledge, there was only one RCT targeting CVD patients with heightened anxiety. Additional research is needed to better determine the benefits of exercise in anxious individuals (see Table 4 for recommendations). Clinicians recommending exercise to treat anxiety in their practice should make sure that their patients can exercise safely, provide an explicit exercise prescription (including 4 key elements: mode, frequency, intensity, and duration), set modest goals at first to maximize the likelihood of success, and monitor their anxiety and refer to a mental health specialist if symptoms persist or worsen.

Table 4.

Recommendations for Future Research

• Enrollment of participants with significant anxiety • Inclusion of participants from diverse backgrounds and underrepresented groups • Designation of appropriate control or comparison conditions • Adequately powered, multi-center trials • Careful assessment of anxiety using established instruments • Clear specification of exercise prescription, including mode, intensity, frequency and duration • Allocation concealment • Blinding of assessors to treatment group •Assessment of aerobic capacity or fitness as manipulation check of aerobic exercise intervention (or muscle strength if resistance training) • Documentation of adherence • Intention-to-treat as primary analytic approach • Follow-up after completion of intervention • Consideration of potential moderators and mediators

Key Points.

• Twenty-five randomized clinical trials of exercise interventions from January 2014 to December 2021 were identified, with 13 trials enrolling participants with heightened anxiety at baseline and 12 trials in which high anxiety was not a prerequisite but served as a primary outcome

• The majority of studies were either negative or inconclusive, and because many suffered from serious methodological shortcomings exercise should not be considered an efficacious intervention for anxiety disorders at this time

Alphabetical List of Abbreviations:

ANOVA

analysis of variance

ASI

Anxiety Sensitivity Index

BAI

Beck Anxiety Inventory

BMI

body mass index

CAPS

Clinician-Administered Posttraumatic Stress Disorder Scale

CONSORT

Consolidated Standard of Reporting Trials

CVD

cardiovascular disease

DASS

Depression Anxiety Stress Scales

DSM

Diagnostic and Statistical Manual of Mental Disorders

GAD

Generalized Anxiety Disorder

HADS

Hospital Anxiety and Depression Scale

HAM-A

Hamilton Anxiety Rating Scale

HIIT

high-intensity interval training

ITT

intention-to-treat

NIMH-SR

National Institute of Mental Health Self-Rating Scale

OCD

Obsessive-Compulsive Disorder

OCI-R

Obsessive-Compulsive Inventory-Revised

PAS

Panic and Agoraphobia Scale

PCL

Posttraumatic Stress Disorder Checklist

PCS

Pain Catastrophizing Scale

PD

Panic Disorder

PDS5

Posttraumatic Stress Disorder Scale for DSM-5

PDSQ

Psychiatric Diagnostic Screening Questionnaire

PDSS-SR

Panic Disorder Severity Scale-Self Report

PEDro

Physiotherapy Evidence Database

PRISMA

Preferred Reporting Items for Systematic Reviews and Meta-Analyses

PSWQ

Penn State Worry Questionnaire

PTSD

Posttraumatic Stress Disorder

RCT

randomized controlled trial

SAD

Social Anxiety Disorder

SCID

Structured Clinical Interview for DSM (-IV or −5, RV - Research Version)

SIPS

Social Interaction Phobia Scale

SMAA

Severity Measure for Agoraphobia

SMD

standardized mean difference

SMSP-A

Severity Measure for Specific Phobia-Adult

STAI

Spielberger State-Trait Anxiety Scale

Y-BOCS

Yale-Brown Obsessive Compulsive Scale

Data availability

The datasets generated during and/or analyzed during the current study are available from the first author upon reasonable request.