Anxiety disorders are the sixth leading cause of global years lived with disability [1]. Anxiety symptoms are associated with negative effects on the individual, including a lower quality of life and comorbid mental illness [2,3]. Furthermore, economic and social burdens related to anxiety disorders are enormous across the developed and developing world [4].

According to various guidelines for treating anxiety disorders, selective serotonin reuptake inhibitors (SSRI) and serotonin-norepinephrine reuptake inhibitors (SNRI) are first-line drugs, and cognitive behavioral therapy (CBT) is regarded as the psychotherapy with the highest level of evidence [5-7]. However, about a third of patients do not respond to antidepressants or CBT [8,9]. Given these criticisms, increasing level of physical activity has been regarded as a new alternative and self-help for relieving anxiety symptoms.

Numerous meta-analyses and systematic review studies have concluded that exercise can significantly decrease anxiety symptoms with a moderate effect size [10-12]. However, studies on the dose-response relationship between physical activity and anxiety symptoms are rare. According to a recent metaanalysis of the association between effect size and level of physical activity using twelve randomized controlled trials, the effect size increased when exercise approached a dose of 12.5 kcal·kg−1·week−1. It then began to decrease when the exercise dose was further increased [13]. Unfortunately, the sample size of that study for the dose-response examination was too small.

Several observational studies have also been conducted to address this issue, and the results are equivocal. Stubbs et al. [14] reported that low physical activity levels are associated with increased prevalence of anxiety by analyzing 237,964 individuals from 47 countries. However, they used a cross-sectional design, categorized levels of physical activity into only two groups, and did not identify sex-specific effect. Brunes et al. [15] found sex differences in the relationship between amounts of physical activity and anxiety symptoms; less than two hours weekly of physical activity with moderate-high intensity was associated with less likelihood of developing anxiety symptoms in healthy men, but not in women. However, only leisuretime physical activity was included as the quantity of exercise, non-leisure time physical activity such as one’s regular occupation, housework, or mode of transportation was disregarded.

Given these aforementioned limitations, our objective was to examine the dose-response relationship between physical activity and anxiety symptoms and find out what amount of physical activity was beneficial for improving anxiety. Additionally, considering sex differences, data were stratified by sex to analyze the effect of sex on the association between physical activity and anxiety symptoms.

Table 1 displays group-specific sociodemographic characteristics. Non-anxious subjects were more likely to be married with a higher level of education than anxious subjects. In comparison with non-anxious subjects, anxiety symptoms were more prevalent among those who were females, unemployed, of lower income, and with a higher level of AUDIT and total CESD score.

Table 2 shows the results of logistic regression analysis with BAI-defined anxiety as the main outcome variable and the level of physical activity as the principal predictor after adjusting for age, sex, location of the health-screening center, marital status, final degree, employment, income level, alcohol consumption, smoking history, BMI, and CESD score. Compared with the sedentary group (0-600 METs-min/wk), participants achieving 600-6,000 METs-min/wk had a significantly lower risk of case-level anxiety symptoms, with a Ushaped relationship [600-1,200 METs-min/wk: OR, 0.85 (95% CI, 0.77-0.94); 1,200-1,800 METs-min/wk: OR, 0.77 (95% CI, 0.68-0.88); 1,800-3,000 METs-min/wk: OR, 0.65 (95% CI, 0.56-0.74); and 3,000-6,000 METs-min/wk: OR, 0.69 (95% CI, 0.58-0.81)]. We observed that engaging in more than 6,000 METs-min/wk was not associated with the risk of anxiety symptoms. Visual representations of a U-shaped dose-response relationship between total physical activity and the odds ratio of case-level anxiety are shown in Figure 2.

Given sex differences in physical activity and anxiety, we stratified data by sex. Table 3 shows adjusted relationships between physical activity and anxiety symptoms for men, and Table 4 shows these data for women. Compared with the sedentary group (0-600 METs-min/wk), participants performing 600-9,000 METs-min/wk had a significantly lower risk of case-level anxiety symptoms for men [600-1,200 METs-min/ wk: OR, 0.84 (95% CI, 0.73-0.96); 1,200-1,800 METs-min/ wk: OR, 0.78 (95% CI, 0.66-0.91); 1,800-3,000 METs-min/ wk: OR, 0.68 (95% CI, 0.57-0.80); 3,000-6,000 METs-min/ wk: OR, 0.64 (95% CI, 0.52-0.80); 6,000-9,000 METs-min/ wk: OR, 0.61 (95% CI, 0.38-0.98)]. For women, we observed a lower risk of case-level anxiety among individuals performing 1,200-3,000 METs-min/wk [1,200-1,800 METs-min/wk: OR, 0.79 (95% CI, 0.64-0.97); 1,800-3,000 METs-min/wk: OR, 0.60 (95% CI, 0.47-0.76)]. The most beneficial levels of physical activity for lowering the risk of case-level anxiety symptoms were higher for men than for women [men with 6,000-9,000 METs-min/wk: OR, 0.61 (95% CI, 0.38-0.98); women with 1,800-3,000 METs-min/wk: OR, 0.60 (95% CI, 0.47-0.76)] (Figure 2).

We found that the relationship between physical activity and anxiety symptoms was curvilinear, suggesting that there were both the most efficacious levels (1,800-3,000 METs-min/week) and upper limits (≥6,000 METs-min/week) of physical activity for decreasing anxiety symptoms. There are potential neurobiological and psychological mechanisms that might explain the anxiolytic effect of physical activity. One mechanism to account for such a benefit is through the action of a brain-derived neurotrophic factor (BDNF). BDNF activates signaling pathways that can result in cognitive improvement and relieve depression and anxiety [28]. Other potential mediations may involve endocannabinoids (N-arachidonoylglycerol and 2-arachidonolylglycerol) [29]. Previous studies have reported that endogenous arachidonate-based lipids can reduce arousal and induce euphoria. Other biological mechanisms for the anxiolytic effects of physical activity might be through regulation of autonomic system functioning [30]. People with anxiety disorders have reduced heart-rate variability (HRV), which can aggravate anxiety symptoms by increased sympathetic and decreased parasympathetic activity. Because exercise regulates autonomic system functioning, anxiety symptoms can be alleviated. There are also psychological theories, including increases in self-esteem and self-efficacy [31]. In addition, exercise can reduce anxiety sensitivity, providing frequent exposure to the anxious situation in the exercise context [32].

For the upper limit of physical activity that provides no significant reduction of anxiety symptoms, numerous hypotheses have been suggested. In terms of biologically detrimental effects of excessive physical activity, the cytokines, hypothalamic, and glycogen hypotheses are most strongly supported. When excessive exercise that can cause tissue microtrauma is prolonged without adequate rest, it may result in chronic inflammation. Furthermore, inflammation is augmented by dysregulation of the stress axis after excessive exercise [33,34]. Pro-inflammatory cytokines, such as TNF-α, IL-1, and IL-6, can increase anxiety symptoms by having negative effects on the brain regions critical for regulation of fear and anxiety (prefrontal cortex, insula, amygdala, and hippocampus) [35]. The other mechanism explaining detrimental effects of excessive exercise is the glycogen-deficit hypothesis. Decreased glycogen causes increased oxidation and decreased levels of branchedchain amino acids (BCAA) known to be involved in central neurotransmitter synthesis as a target for anxiolytics [36]. For the psychologically negative effects of excessive physical activity, lots of studies have suggested that exercise addiction is a behavior addiction, congruous with the work-craving model [37]. The cognitive component of exercise addiction is a negative perfectionism, which is characterized as unrealistic self-expectations, a fear of not trying hard enough, and an implied desire to avoid disapproval that comes from not meeting the expectations of others [38]. Its behavioral component is an underlying obsessive-compulsive drive that make excessive exercise a habitual behavior [39,40]. The hedonic component of exercise addiction is its reduction of negative affect. Because negative affectivity is pervasive when one is unable to exercise, vulnerable people tend to obsess about engaging in vigorous physical activity [37]. These psychological components of exercise addiction, such as negative perfectionism, obsessive-compulsive drive, and negative affectivity, are crucial risk factors for developing anxiety symptoms. To put the risk-to-benefit ratio of exercise into perspective, extreme physical activity could not have positive effects on relieving anxiety.

We found that the most beneficial levels of physical activity for relieving anxiety symptoms were higher for men than for women. According to recent systematic reviews and metaanalyses, intense exercise can result in a significant decrease of both total estradiol and free estradiol, independent of a menopausal state or weight loss [41,42]. Sex hormones contribute to endogenous anxiolytics, such as serotonin and allopregnanolon, thus promoting adaptive responses to stress [43]. Because the reproductive system of women is sensitive to physiological stress such as strenuous exercise, the optimal volume of physical activity might be lower in women than in men.

From the perspective of public health, the recommended level of physical activity for reducing anxiety symptoms can be suggested in real-world terms as follows. In terms of walking at an average pace, 3-45.4 hours per week were efficacious for men and 6-15 hours per week were efficacious for women. In terms of moderate intensity activities, such as carrying light loads, bicycling at a regular pace, and playing tennis doubles, the efficacious amount of physical activity for relieving anxiety symptoms was 2.5-37.5 hours per week for men and 5-12.5 hours per week for women. In terms of vigorous intensity activities, such as heavy lifting, digging, aerobics, and fast bicycling, the appropriate level of physical activity for reducing anxiety symptoms was 1.25-18.75 hours per week in men and 2.5-6.25 hours per week in women.

Our study had several limitations. First, since we measured both exposure and outcome simultaneously, there might be some reverse causation (anxiety leading to a reduced level of exercise). Second, although we attempted to account for several confounding variables, several crucial potential confounders such as diet, personality, and attitude toward health, could not be measured. Third, physical activity and depression were measured by means of self-report screening tools with a risk of reporting bias. Fourth, we defined anxiety as having a BAI score of 19 and above. The operationalization of anxiety symptoms via a self-rating scale cannot be considered equivalent to a clinical diagnosis. Thus, a risk of misclassification remains.

Our study also had several strengths. First, using a large sample, we identified the dose-response relationship between physical activity and anxiety symptoms. Second, considering the correlation between depression and anxiety, we adjusted for depression to clarify the association between anxiety and physical activity. Third, given the effect of sex differences on the relationship between anxiety symptoms and physical activity, we stratified data by sex and identified efficacious sex-specific levels of physical activity for reducing anxiety symptoms.