Association of physical activity and the risk of COVID-19 hospitalization: a dose–response meta-analysis

1 / 14 1Association of physical activity and the risk of COVID-19 hospitalization: 2a dose–response meta-analysis 3Dan Li 1☯, Shengzhen Jin 2,1☯ and Songtao Lu 1,3* 41 School of Sports, Wuhan University of Science and Technology, Wuhan 430081, China.; [email protected] 62 Tennis College, Wuhan sports university, Wuhan 430079, China; [email protected] 73 School of Physical Education and Sports, Central China Normal University, Wuhan 430079, China. 8☯ These authors contributed equally to this work. 9* [email protected] 10Abstract 11Background 12Many people have experienced a high burden due to the spread of the coronavirus disease (COVID-19) and its serious 13consequences for health and everyday life. Prior studies have reported that physical activity (PA) may lower the risk of COVID- 1419 hospitalization. The present meta-analysis (PROSPERO registration number: CRD42022339672) explored the dose– 15response relationship between PA and the risk of COVID-19 hospitalization. 16Methods 17Epidemiological observational studies on the relationship between PA and the risk of COVID-19 hospitalization were included. 18Categorical dose–response relationships between PA and the risk of COVID-19 hospitalization were assessed using random 19effect models. Robust error meta-regression models assessed the continuous relationship between PA (metabolic equivalent 20[MET]–h/week) and COVID-19 hospitalization risk across studies reporting quantitative PA estimates. 21Results 22Seventeen observational studies (cohort\case–control\cross-section) met the criteria for inclusion in the meta-analysis. 23Categorical dose-relationship analysis showed a 40% (risk ratio (RR) 0.60, 95% confidence intervals (CI): 0.48–0.71) reduction 24in the risk of COVID-19 hospitalization compared to the lowest dose of PA. The results of the continuous dose–response . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprintthis version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.22.22276789doi: medRxiv preprint NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice. 2 / 14 25relationship showed a non-linear inverse relationship (Pnon-linearity < 0.05) between PA and the risk of COVID-19 26hospitalization. When total PA was less than or greater than 10 Met-h/week, an increase of 4 Met-h/week was associated with 27a 14% (RR = 0.83, 95%CI: 0.85–0.87) and 11% (RR = 0.89, 95%CI: 0.87–0.90) reduction in the risk of COVID-19 28hospitalization, respectively. 29Conclusions 30There was an inverse non-linear dose–response relationship between PA level and the risk of COVID-19 hospitalization. 31Doses of the guideline-recommended minimum PA levels by WTO may be required for more substantial reductions in the 32COVID-19 hospitalization risk. 33 341. Introduction 35The coronavirus disease 2019 (COVID-19) outbreak continues worldwide. As of 7 May 2022, COVID-19 has caused 51,587,3758 36infections and 6,272,357 deaths worldwide [1]. It is essential to identify high-risk groups that require special attention under these 37conditions [2]. For non-communicable disease outcomes, lifestyle risk factors have been consistently associated with morbidity, 38mortality, and loss of disease-free life [3,4]. For example, physical inactivity and smoking appear to be independently associated 39with a higher risk of community-acquired pneumonia morbidity and mortality [5,6]. 40It is also well established that the risk of developing respiratory disease is much higher in people with low physical activity (PA), 41whereas COVID-19 patients with a physically inactive lifestyle (e.g., sedentary behavior) are more likely to be hospitalized and 42have a greater likelihood of poor clinical outcomes [7]. Moreover, it has previously been shown that regular physical activity and 43higher physical fitness levels enhance immune function and, therefore, might reduce susceptibility to COVID-19 infection and 44infection severity [8,9]. Recent studies retrospectively evaluating cohorts of COVID-19 positive adults have described the benefit 45of regular physical activity in decreasing the incidence of adverse outcomes in confirmed cases of COVID-19 [10,11,12]. 46However, research on such topics is just emerging, and the impact of PA on the infectious and clinical outcomes of COVID-19 47remains unclear. The protective effects of different levels of physical activity against COVID-19 are controversial. Rahmati et al. 48[13] conducted a meta-analysis on this topic, which did not address the controversy regarding the protective effects of different 49levels of physical activity. In addition, Rahmati et al. [13] classified the case–control group as a cross-sectional study. In a meta- 50analysis, they assumed cardiopulmonary function as physical activity, which inevitably led to unconvincing results. Finally, the 51study only used the binary variables of physical activity included in the literature to analyze the outcome variables, ignoring the 52moderate dose in the multi-level doses, making it difficult to explain the heterogeneity generated by the meta-analysis studies. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprintthis version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.22.22276789doi: medRxiv preprint 3 / 14 53Furthermore, no systematic review or meta-analysis has reported the exact dose–response relationship between pre-diagnosis 54PA and COVID-19 hospitalization. Consequently, there is still substantial uncertainty regarding the association between pre- 55diagnosis PA levels and hospitalization due to COVID-19 among the general population. To precisely quantify the association 56between pre-diagnosis PA and the risk of COVID-19 hospitalization, we conducted a systematic review and meta-analysis of 57observational studies published up to May 8, 2022. 582. Materials and Methods 592.1. Inclusion Criteria 60The criteria for inclusion, and each article determined for inclusion, were discussed by three authors, and the discussions on 61inclusion and exclusion occurred more than three times. The inclusion criteria were as follows: (1) studies published as 62epidemiological observational cohort studies, case controls, and cross-sectional design investigation studies; (2) studies providing 63at least an odds ratio, relative risk (RR or HR), and 95% confidence interval (95%CI) between the level of physical activity and the 64risk of hospitalization for COVID-19, or raw data provided to calculate these indicators. The repeated literature was excluded. Only 65the latest studies were selected if they were conducted at different time points in the same cohort. Additionally, if multiple articles 66were published in the same group, we chose articles where subjects were followed for a longer time or with a larger sample size. 672.2. Search Strategy 68We searched PubMed (1980 to the present) and the Web of Science database (1980 to the present) for literature on the 69relationship between physical activity and the risk of COVID-19 hospitalization. The search strategy used keywords such as 70"exercise or physical activity or sport or walking or motor activity", "COVID-19 or SARS-CoV-2", and "severe or hospitalization". 71These searches were screened for cohort studies, case controls, and cross-sectional design studies. The latest search date was 72April 2022, and there was no language limit. The reference lists of the selected and related review articles were screened to 73identify potentially relevant studies. All searches were conducted independently by two authors, and the differences were 74resolved by group discussion. 752.3. Quality Assessment 76The Newcastle–Ottawa Scale (NOS) was used to evaluate literature quality, and scores of 0–3, 4–6, and 7–9 were determined as 77low, moderate, and high quality, respectively [14]. Each article was evaluated independently by two authors and cross-checked. 78In the group meeting, the results were publicized, and the reasons for the score of each item were specified. If the evaluation of 79the literature quality was inconsistent, the group focused on solving the final score of its quality. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprintthis version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.22.22276789doi: medRxiv preprint 4 / 14 802.4. Synthesis Methods 81Stata16.0 software was used for the meta-analysis. The p-value was set at p < 0.05, and all tests were double-sided. The effect 82value-adjusted risk ratio (RR) and 95% confidence interval (CI) of the group with the highest dose of PA (physical activity) compared 83to the control group with the lowest dose of PA in each study were combined. The combined effect values were calculated using 84a random-effects model. Heterogeneity was assessed and described using I2 statistics as the percentage of variation in the study; 85I2 values of 25%, 50%, and 75% represented low, moderate, and high levels of heterogeneity, respectively [15]. Egger and Begg 86tests were used to determine publication bias. During the sensitivity analysis, each study was deleted one by one to check whether 87the combined effect of the remaining studies had changed [15]. The subgroup meta-analysis was conducted according to PA 88intensity classification (LPA-light intensity physical activity, VPA-vigorous-intensity physical activity, MVPA-moderate-to-vigorous 89physical activity, MPA-moderate-intensity physical activity), sex, age, study area, study quality, and adjustment for confounding 90factors. Meta-regression was used to examine the heterogeneity among studies. 91According to categorical and continuous dose PA, this study analyzed the dose–response relationship with the risk of COVID-19 92hospitalization. Categorical doses were divided into dichotomous and multi-classified doses shown in the study, and the combined 93effect value RR was generated by comparing the highest and lowest doses. To analyze the continuous dose–response relationship, 94we calculated the total weekly dose of PA for each effect value RR based on the PA intensity, duration, and weekly frequency of 95the baseline survey provided in the literature. Furthermore, we assumed that the dose remained at this level during the follow- 96up survey. To determine the exposure value of the included dose, the median was set as the determined dose. If the development 97interval was < 0.5, it was set to 0.25. If the upper open interval was ≥ 1, the difference between the intermediate dose intervals 98was 0.25, and the exposure value was set to 1.25 [16]. Met-h/week was considered as the final unit of analysis. These are combined 99absolute indices of intensity, duration, and frequency used to calculate exposures to Met units not directly reported in the 100literature. Met, a physiological measure of PA energy, is defined as energy expenditure per kilogram of body weight per hour: 1 101Met = 1 kcal/kg ∗h. To address the differences in PA units in different studies, we used Ainsworth et al. 's classification [16], 102classifying PA into low-intensity LPA (3 Mets, such as walking exercise), moderate-intensity MPA(4 Mets), and high-intensity VPA 103(8 Mets). We then converted the duration of a particular PA intensity (h/week) to Met-h/week in combination with the frequency 104of the week [17]. 105To establish the dose–response relationship between PA and the risk of hospitalization for COVID-19, robust error meta- 106regression (REMR) was used for model fitting [18]. The REMR approach is based on a "one-stage" framework that treats each 107study as a cluster and fits the revised regression to the average PA dose across the entire dataset. In addition, the method also 108uses the inverse variance method to weight each dose-specific effect in the data and balances heteroscedasticity in the REMR . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprintthis version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.22.22276789doi: medRxiv preprint 5 / 14 109model to ensure unbiased estimation of parameters. Finally, we used restricted cubic splines as connection functions to fit the 110linear and non-linear dose–response models. Based on the dose-centralization treatment, the independent variable PA dose of 111the model was set as three nodes (0, 6.75, and 21), including two regression splines. The χ2 test was used to test the hypothesis 112that the regression coefficient of the second regression spline is significant (p < 0.05), indicated by a linear or non-linear dose– 113response relationship. A dose–response relationship curve was drawn using the Stata software XBLC command [16]. 1143. Results 1153.1. Study Selection and Characteristics 116In total, 170 articles were preliminarily identified. According to the literature inclusion and exclusion criteria formulated in this 117study, 17 studies, including seven cohort studies, five case–control studies, and five cross-sectional design studies, were finally 118included. There were 1,038,768 subjects and about 3022 hospitalized COVID-19 cases (some studies had not reported the number 119of cases). The steps for retrieval and inclusion are shown in Fig 1. The characteristics of the literature are listed in Table 1. Among 120the 17 studies, 3 were from North America [10,19,20], 7 were from Asia [11,12,21,22,23,24,25], 6 were from Europe 121[26,27,28,29,30,31], and 1 was from Oceania [32]. The NOS was used to score the included studies. Eight studies were considered 122high quality, with a score greater than or equal to seven, and the other nine were considered medium quality. 123 124Fig 1. Flow diagram of studies considered for inclusion in the systematic review. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprintthis version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.22.22276789doi: medRxiv preprint 6 / 14 125Table 1. Summary of the extracted studies. 126 Author（year） Country Study type Case\total Age（SD） female % Measurement and Categories of PA adjustment * QA AlKetbi（2021） AUH Cohort 135\641 44(NP) 36% Self-report PA: five Categories(times\w) NP* 5 Brandenburg（2021） CA Case-control NP\263 86%<65 57% PA-R: No, Moderate, 1 h of vigorous 2 5 6 8 9 12 15 7 Bielik(2021） SLK Cohort 104\2343 18 to 65 49% physically active(≥3 times\w), cold-water swim NP 5 Ekblom-Bak(2021） SWE Case-control 172/407131 49.9(NP) 30% Never/irregular, 1–2 times/w, ≥ 3 times/w 1 2 5 16 7 Katsoulis（2021） UK Cohort NP\ 85308 18-69(NP) NP Self-report PA: Low, Gentle, Moderate, Vigorous 1 2 7 9 7 Hamer(2020) UK Cohort 760\387109 56.2± 8.0 55.1% IPAQ: Sufficient\ Insufficient\ None 1 2 6 8 9 7 Halabchi（2021） Iran. Cross-Section 60\4694 36.45±9.77 55% Regular sports participation (yes/no) NP 6 Hamdan(2021） PLE Cross-Section 59\300 30.5±12.2 55.0% PA questionnaire: yes\no NP 5 Lee (2021) KR Cohort 277\118768 NP 51.2% Self-report PA: None, Gentle, Moderate, Vigorous 1 2 5-9 14 16 Latorre-Roman(2021) al.,2021 Spanish Cross-Section NP\420 33 (20–54) 52.6% IPAQ: Moderate PA>150 min\w, 30–150 min\w, None 1 6 Malisoux(2022） LUX Cohort 106\452 42 (31–51) 48% PA questionnaire: yes\no 1 2 3 5 16 18 19 7 Maltagliati(2021) France Case-control 66\3139 69.3 ± 8.5 53% four-point PA scale ranging (>1, 1; 150 min/w(moderate), Insufficient 1 2 19 7 Sallis（2021） USA Cohort 1199\2970 47±16.97 61.9% UPAG: active, inactive, some activity 1-15 9 Steenkamp（2021） ZA Cohort NP\65361 41±12.1 48.2% Low activity, Moderate activity, High activity 1 2 6 8 9 14 15 8 Tavakol(2021) Iran Cross-Section 64\188 18-75(NP) 52.7% GPAQ: Low, Moderate to high NP 6 Yuan(2021) CN Case-control 29\164 61.8 ± 13.6 48.8% Self-report PA: Inactivity, activity NP 6 127Case \ total: Number of cases and total sample size; Age characteristics: Single value indicates mean age, others are age range.* Adjustment factors: 1 age, 2 gender, 3 socioeconomic status, 4 race, 5BMI, 6 cardiovascular disease, 7 cancer, 1288 diabetes, 9 hypertension, 10 use of antihypertensive drugs, 11 corticosteroids, 12 chronic lung / respiratory disease, 13 liver disease, 14 HIV, 15 end-stage renal disease and immune disease, 16 smoking, 17 alcohol,18 sedentary behaviour 129, 19 comorbidities.NP *: Not reported. AUH: AUH , ZA: South Africa, UK: UK, AUH: UAE, CA: Canada, SWE: Switzerland, Iran: Iran, KR: Korea, ESP: Spain, USA: US, ZA: New Zealand, CN: China, SLK: Slovak, LUX: Luxembourg, PLE: Palestine. 130IPAQ: International Physical Activity Questionnaire. PA-R: physical activity rating questionnaire. UPAG: US Physical Activity Guidelines. GPAQ: Global physical activity questionnaire. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprintthis version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.22.22276789doi: medRxiv preprint 7 / 14 1313.2. Categorical analysis between PA and COVID-19 hospitalization 132Compared with the lowest PA dose, the highest PA in the included studies reduced the risk of COVID-19 hospitalization by 40% 133(RR = 0.60, 95%CI: 0.48, 0.71). The heterogeneity test result was I2 = 66.22% (p < 0.01), indicating significant heterogeneity of the 134study results (Fig 2). The pooled effect size results did not change significantly after excluding each study from the sensitivity 135analysis (Supporting information S1 Fig). Published bias analysis with Begg’s test p = 0.54 > 0.05, Egger test p = 1.330 > 0.05, and 136funnel diagram also showed no significant published bias (Supporting information S2 Fig). The effect values of the cohort study, 137case–control study, and cross-sectional design study were 0.63 (95%CI:0.54,0.71), 0.59 (95%CI: 026, 0.91), and 0.58 (95%CI: 0.42, 1380.74), respectively. 139Fig 2. Forest plot showing categorical analysis between PA and COVID-19 hospitalization 140As for the source of heterogeneity (presented in Table 2), between-group heterogeneity only appeared in the comparative 141analysis of the relationship between PA at different dose levels and the risk of hospitalization for COVID-19 (P b < 0.01), indicating 142that PA at different doses significantly reduces the risk of hospitalization. Within-group heterogeneity appeared in the multi-dose 143PA, case–control study, high quality, European, adjustment for age, sex, adjustment for high blood pressure, adjustment for 144diabetes, adjustment for cancer, and cardiovascular diseases subgroup, illustrating that the subgroup study effect value of the 145results may not be stable. The results may be affected by other factors. 146 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprintthis version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.22.22276789doi: medRxiv preprint 8 / 14 147 148 149Table 2. Subgroup analysis 150 151Multi-dose comparison *: The highest dose is higher than the 10 h-met / week and the lowest dose, and the moderate dose is lower than 10 h-met / week. The binary dose 152is generally expressed as exercise and non-exercise; * multiple dose grouping represents the highest, relative and lowest dose comparison,at least three lever; P a and P b 153represent heterogeneity within and between subgroups, respectively. 1543.3. Continuous dose-response relationship between PA and COVID-19 hospitalization 155Fig 3 shows the continuous dose–response relationship. The results showed a negative non-linear relationship between PA and 156the risk of hospitalization for COVID-19 (Pnon-linearity < 0.01). When PA < 10 Met-h/week, an increase of 4 Met-h/week (1 h of Subgroup N RR(95% CI) I2 (%) pa* Pb* Total effect size (highest vs. lowest) 17 0.60（0.48,0.71） 66.22 0.01 binary does * 7 0.51(0.33,0.69) 45.43 0.09 0.24PA Categories multi-dose * 10 0.65(0.51.0.79) 68.34 0.02 highest vs. lowest 12 0.59(0.55,0.63) 0 0.86multi-class dose comparison* moderate vs. lowest 13 0.75(0.65,0.85) 26.88 0.33 0.01< cohort 7 0.63(0.54,0.71) 20.83 0.53 0.85 Case-control 5 0.59(026,0.91) 82.07 0.00 Study type Cross-section 5 0.58(0.42,0.74) 2.05 0.28 Study quality 》7 8 0.67(0.53,0.81) 70.32 0.02 0.1 7< 9 0.49(0.32,0.66) 38.93 0.15 Europe 6 0.68(0.43,0.94) 68.68 0.02 0.46 Asia 7 0.49(0.30,0.68) 46.31 0.11 Different continent others (America) 4 0.59(0.54,0.64) 0 0.96 Adjusted confunding factor yes 10 0.63(0.51,0.75) 66.73 0.01 0.36age no 7 0.51(0.29,0.74) 43.20 0.14 yes 10 0.65(0.54,0.78) 59.93 0.03 0.15sex no 7 0.47(0.25,0.69) 44.89 0.12 yes 7 0.66(0.48,0.84) 67.76 0.02 0.29BMI no 10 0.54(0.40,0.87) 43.12 0.15 Adjusted baseline disease yes 8 0.58(0.31,0.85) 10.28 0.57 0.83hypertension no 9 0.61(0.55,0.67) 71.22 0.00 diabetes yes 8 0.61(0.55,0.87) 11.67 0.49 0.91 no 9 0.59(0.34,0.84) 67.09 0.00 cardiovascular yes 8 0.61(0.55,0.87) 11.67 0.49 0.91 no 9 0.59(0.34,0.84) 67.09 0.00 cancer yes 6 0.57(0.45,0.69) 0 0.70 0.75 no 11 0.61(0.43,0.79) 80.95 0.00 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprintthis version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.22.22276789doi: medRxiv preprint 9 / 14 157moderate-intensity or 1/2 h of high-intensity) was associated with a 14% reduction in the risk of hospitalization for COVID-19 (p 10 Met-h/week, the risk of hospitalization for COVID-19 decreased by 11% for each 159additional 4 Met-h/week (p < 0.01, RR = 0.89, 95%CI: 0.87–0.90). 160Fig 3 Continuous dose-response relationship between PA and COVID-19 hospitalization 1614. Discussion 162This study is the first dose–response meta-analysis of the relationship between PA and the risk of COVID-19 hospitalization. The 163literature included observational studies on the relationship between PA and the risk of COVID-19 hospitalization. Through a meta- 164analysis of the categorical dose, our main conclusion is that the risk of COVID-19 hospitalization is reduced by 40% compared with 165the lowest dose of PA. For continuous dose–response analysis, we confirmed that the relationship between PA and the risk of 166COVID-19 hospitalization is non-linear and inverse. For every four Met-h\week PA increase, the risk of COVID-19 hospitalization 167decreased by 11–14%. Sensitivity and published bias analyses further support these results, and these main quantitative features 168have important clinical significance. 169The dose–response association between PA and the risk of COVID-19 hospitalization has been previously reported. As for the 170categorical analysis, Rahmati et al. observed that PA was significantly associated with a reduction in COVID-19 hospitalization 171compared with control (RR = 0.58) by meta-analysis [13]. In the present study, we observed a similar RR of 0.60. However, we 172included multiple categorical analysis data with different PA dose levels to analyze the relationship between PA and the risk of 173COVID-19 hospitalization. We observed significantly different protective effects of varying PA levels on the hospitalization risk of 174COVID-19 by heterogeneity analysis. Compared to the results obtained by Rahmati et al., our observations are more abundant. 175As for the continuous analysis, Malisoux et al. observed an inverse dose–response association between PA and the risk of 176moderate COVID-19 illness, and increased PA was associated with a slightly lower risk of moderate illness (OR:0.99) [25]. In the 177present study, we observed a similar inverse dose–response association between PA and the risk of COVID-19 hospitalization. An 178increase of 4 Met-h/week PA was associated with a 12–17% (RR = 0.83–0.88) reduction in the risk of COVID-19 hospitalization. A 1794 Met-h/week PA is equivalent to one hour MPA or half an hour high VPA; our results are more specific and close to a practical 180exercise. Otherwise, the observed dose–response relationship between PA and the risk of COVID-19 hospitalization was non-linear. 181We assumed that after 10 Met-h/week, the degree of enhancement in lowering the risk of moderate COVID-19 illness when PA 182increase is weakened. However, the threshold is 30 Met-h/week according to the j-shaped association of PA and risk of moderate 183illness by Malisoux et al.. [25]. Meanwhile, our meta-analysis findings are more robust and specific, and we observed increasing 184PA benefits for the inpatient burden due to COVID-19. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprintthis version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.22.22276789doi: medRxiv preprint 10 / 14 185This difference in the magnitude of risk reduction for COVID-19 hospitalization could be related to differences in the mechanism 186through which PA modifies the risk of respiratory viral infections. This is supported by previous studies that showed a stronger 187association between maximal fitness exercise capacity and the risk of hospitalization due to COVID-19 [33]. Exercise capacity is an 188important index for measuring overall health and the ability of the body to cope with external stressors. More specifically, it is an 189important index to bear the burden on the heart and lungs [33]. However, PA greatly influences exercise capacity; more specifically, 190regular moderate-intensity to vigorous-intensity aerobic exercise daily can improve exercise capacity. In addition, the beneficial 191effects of regular PA on the immune system may involve several mechanisms, including enhanced immunosurveillance, reduced 192systemic inflammation, improved regulation of the immune system, and delayed onset of immunosenescence [34,35]. A recent 193meta-analysis investigated the effects of regular PA on the immune system [36]. This study showed that moderate to vigorous 194intensity exercise (e.g., walking, running, cycling) is overall beneficial with a lower concentration of neutrophils and a higher 195concentration of CD4 T helper cells and salivary IgA. These biochemical indicator changes may be the critical mechanism for regular 196PA to lower the risk of hospitalization due to COVID-19. 197As for the analysis of the confounding factors, results on between-subgroup heterogeneity showed significant heterogeneity in 198all subgroups adjusted for underlying disease. Thus, confounding factors of underlying disease significantly affected the 199association between PA and the risk of hospitalization. However, contrary to expectations, heterogeneity was observed in the 200within-subgroup analysis adjusted for age, sex, and BMI. This indicated that our data failed to demonstrate significant differences 201in the impact of PA on the risk of hospitalization for COVID-19 according to age, sex, and body weight. This may be because the 202overall heterogeneity of this meta-analysis was precisely concentrated in these subgroups. The other possible reason is that the 203effect of PA on reducing the risk of COVID-19 hospitalization is very stable among these different demographic characteristics. 204The WHO global recommendation on the health benefits of PA states that adults have at least 150 minutes of moderate- 205intensity aerobic PA per week, at least 75 minutes of high-intensity aerobic PA per week, or a combination of moderate and high- 206intensity activities; this is equivalent to 10 met hour/week [1]. Our analyses also show that when the PA level is >10 met 207hours/week, the risk of hospitalization for COVID-19 is reduced, but the degree of reduction becomes significantly smaller. The 208practical significance of this conclusion is that PA or exercise within 10 met hours/week has the most apparent effect on reducing 209the risk of hospitalization for COVID-19, with an additional benefit for reducing the risk when PA level exceeds 10 met hours/week. 210Increased hospitalization due to COVID-19 is a threat to health and a heavy disease burden on all aspects, such as individuals 211and the country. We found that increased physical activity significantly reduces the hospitalization risk of COVID-19, and physical 212activity should be a positive factor in decreasing the COVID-19 disease burden. However, the Global Burden of Diseases (GBD) 2132019 ranked low physical activity 19th out of 20 risk factors in terms of disability-adjusted life years, down from 10th in the . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprintthis version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.22.22276789doi: medRxiv preprint 11 / 14 214equivalent 2010 GBD publication [37,38]. Moreover, PA decreased in all age groups, independent of sex, due to the COVID-19 215pandemic, according to a recent meta-analysis [39]. In the face of the global spread of COVID-19, we must regain the vital role of 216physical activity in reducing the burden of the disease. The conclusions of our study will undoubtedly have significant implications 217for public health. 218Finally, our study is the first meta-analysis on the dose–response relationship between PA and the hospitalization risk of COVID- 21919. The results of this study are based on a large sample cohort study, case–control group study, and the advantages of a long 220period of follow-up investigation by cohort study; therefore, the results are relatively stable. However, this study may have 221limitations. First, the literature we included may be insufficient in the continuous dose–response analysis. One possible reason for 222this is that we set strict inclusion criteria. In addition, cohort studies require long-term follow-up surveys and extensive sample 223data. Therefore, few cohort studies have been conducted on related topics. Second, the methods of PA evaluation included in the 224literature of this study are subjective measurements, which may lead to inaccurate doses, and different measurement and 225observation methods of physical activities and hospitalized cases may have a significant impact on the research conclusion. In 226addition, exercise habit is based on the assumption that there is no change in the long-term follow-up; that is, the dose of PA is 227constant in the long-term observational study. This assumption may make the results inaccurate. Moreover, the results of some 228cohort studies included were not adjusted by the confounding factors such as sex, age, and other concomitant medical conditions, 229which should be paid attention to in future studies. 2305. Conclusions 231There was an inverse non-linear dose–response relationship between PA levels and the risk of COVID-19 hospitalization. An 232increase in the physical activity dose significantly reduced the hospitalization risk of COVID-19. The degree of risk reduction is 233weakened when PA is higher than 10 Met-h/week. Doses of the guideline-recommended minimum PA levels by the WTO may be 234required for more substantial reductions in the COVID-19 hospitalization risk. Future studies with different doses of PA or exercise 235interventions are needed to determine the optimum PA dose required for COVID-19 prevention. 236 237Supporting information 238S1 Checklist. PRISMA checklist of the meta-analysis.(DOC) 239S1 Fig. Sensitivity analysis result. (TIF) 240S2 Fig. Published bias funnel plot. (TIF) . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprintthis version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.22.22276789doi: medRxiv preprint 12 / 14 241Acknowledgements 242We would like to acknowledge the assistance of Chang Xu in helping us to develop an dose-response model with 243REMR approach. 244Author Contributions 245Conceptualization: Dan Li, Songtao Lu. 246Data curation: Dan Li, Shenzhen Jin, Songtao Lu. 247Formal analysis: Shenzhen Jin, Songtao Lu. 248Methodology: Dan Li, Shenzhen Jin. 249Software: Dan Li, Shenzhen Jin. 250Supervision: Shenzhen Jin, Songtao Lu. 251Visualization: Dan Li, Shenzhen Jin. 252Writing – original draft: Dan Li, Shenzhen Jin, Songtao Lu. 253Writing – review & editing: Shenzhen Jin, Songtao Lu. 254References 1. World Health Organization, Physical Activity Guidelines. 2021. Available from: https://www.who.int/emergencies/diseases/novel-coronavirus-2019. 2. Chen J, Qi T, Liu L, Ling Y, Qian Z, Li T, et al. Clinical progression of patients with COVID-19 in Shanghai, China. The Journal of infection. 2020;80(5):e1-e6. doi: 10.1016/j.jinf.2020.03.004. PubMed PMID: 32171869; PubMed Central PMCID: PMC7102530. 3. Colpani V, Baena CP, Jaspers L, van Dijk GM, Farajzadegan Z, Dhana K, et al. Lifestyle factors, cardiovascular disease and all-cause mortality in middle-aged and elderly women: a systematic review and meta-analysis. Eur J Epidemiol. 2018;33(9):831-45. doi: 10.1007/s10654-018-0374-z. PubMed PMID: 29524110. 4. Schlesinger S, Neuenschwander M, Ballon A, Nothlings U, Barbaresko J. Adherence to healthy lifestyles and incidence of diabetes and mortality among individuals with diabetes: a systematic review and meta-analysis of prospective studies. Journal of epidemiology and community health. 2020;74(5):481-7. doi: 10.1136/jech-2019-213415. PubMed PMID: 32075860. 5. Baik I, Curhan GC, Rimm EB, Bendich A, Willett WC, Fawzi WW. A prospective study of age and lifestyle factors in relation to community-acquired pneumonia in US men and women. Arch Intern Med. 2000;160(20):3082-8. doi: 10.1001/archinte.160.20.3082. PubMed PMID: 11074737. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprintthis version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.22.22276789doi: medRxiv preprint 13 / 14 6. Inoue Y, Koizumi A, Wada Y, Iso H, Watanabe Y, Date C, et al. Risk and protective factors related to mortality from pneumonia among middleaged and elderly community residents: the JACC Study. J Epidemiol. 2007;17(6):194-202. doi: 10.2188/jea.17.194. PubMed PMID: 18094518; PubMed Central PMCID: PMC7058467. 7. Team CC-R. Severe Outcomes Among Patients with Coronavirus Disease 2019 (COVID-19) - United States, February 12-March 16, 2020. MMWR Morb Mortal Wkly Rep. 2020;69(12):343-6. doi: 10.15585/mmwr.mm6912e2. PubMed PMID: 32214079; PubMed Central PMCID: PMC7725513 Journal Editors form for disclosure of potential conflicts of interest. No potential conflicts of interest were disclosed. 8. Duggal NA, Niemiro G, Harridge SDR, Simpson RJ, Lord JM. Can physical activity ameliorate immunosenescence and thereby reduce age-related multi-morbidity? Nature reviews Immunology. 2019;19(9):563-72. doi: 10.1038/s41577-019-0177-9. PubMed PMID: 31175337. 9. Simpson RJ, Kunz H, Agha N, Graff R. Exercise and the Regulation of Immune Functions. Progress in molecular biology and translational science. 2015;135:355-80. doi: 10.1016/bs.pmbts.2015.08.001. PubMed PMID: 26477922. 10. Sallis R, Young DR, Tartof SY, Sallis JF, Sall J, Li Q, et al. Physical inactivity is associated with a higher risk for severe COVID-19 outcomes: a study in 48 440 adult patients. Br J Sports Med. 2021;55(19):1099-105. doi: 10.1136/bjsports-2021-104080. PubMed PMID: 33849909; PubMed Central PMCID: PMC8050880. 11. Lee SW, Lee J, Moon SY, Jin HY, Yang JM, Ogino S, et al. Physical activity and the risk of SARS-CoV-2 infection, severe COVID-19 illness and COVID-19 related mortality in South Korea: a nationwide cohort study. Br J Sports Med. 2021. doi: 10.1136/bjsports- 2021-104203. PubMed PMID: 34301715; PubMed Central PMCID: PMC8300550. 12. Tavakol Z, Ghannadi S, Tabesh MR, Halabchi F, Noormohammadpour P, Akbarpour S, et al. Relationship between physical activity, healthy lifestyle and COVID-19 disease severity; a cross-sectional study. Zeitschrift fur Gesundheitswissenschaften = Journal of public health. 2021:1-9. doi: 10.1007/s10389-020-01468-9. PubMed PMID: 33558839; PubMed Central PMCID: PMC7858040. 13. Rahmati M, Shamsi MM, Khoramipour K, Malakoutinia F, Woo W, Park S, et al. Baseline physical activity is associated with reduced mortality and disease outcomes in COVID-19: A systematic review and meta-analysis. Reviews in medical virology. 2022:e2349. doi: 10.1002/rmv.2349. PubMed PMID: 35416354. 14. Wells GA, Peterson J, Welch V, et al. The Newcastle-Ottawa scale (NOS) for assessing the quality in nonrandomized studies in meta-analyses, 2009, Ottawa, University of Ottawa. Available from: http://www. ohri. ca/ programs/ clinical_ epidemiology/ oxford. Asp 15. Higgins J, Thompson S. Quantifying heterogeneity in a meta-analysis. Statistics in medicine. 2002;21(11):1539-58. doi: 10.1002/sim.1186. PubMed PMID: 12111919. 16. Zhang Tiansong,Dong Shengjie,Zhou Zhirui et al. The Advanced meta-Analysis Method-Based on stata. Fudan University Press 2015;220-232 17. Ainsworth B, Haskell W, Herrmann S, Meckes N, Bassett D, Tudor-Locke C, et al. 2011 Compendium of Physical Activities: a second update of codes and MET values. Medicine and science in sports and exercise. 2011;43(8):1575-81. doi: 10.1249/MSS.0b013e31821ece12. PubMed PMID: 21681120. 18. Xu C, Doi S. The robust error meta-regression method for dose-response meta-analysis. International journal of evidence-based healthcare. 2018;16(3):138-44. doi: 10.1097/xeb.0000000000000132. PubMed PMID: 29251651. 19. Brandenburg JP, Lesser IA, Thomson CJ, Giles LV. Does Higher Self-Reported Cardiorespiratory Fitness Reduce the Odds of Hospitalization From COVID-19? J Phys Act Health. 2021;18(7):782-8. doi: 10.1123/jpah.2020-0817. PubMed PMID: 33984837. 20. de Souza FR, Motta-Santos D, Dos Santos Soares D, de Lima JB, Cardozo GG, Guimaraes LSP, et al. Association of physical activity levels and the prevalence of COVID-19-associated hospitalization. Journal of science and medicine in sport. 2021;24(9):913-8. doi: 10.1016/j.jsams.2021.05.011. PubMed PMID: 34090826; PubMed Central PMCID: PMC8141261. 21. Yuan Q, Huang HY, Chen XL, Chen RH, Zhang Y, Pan XB, et al. Does pre-existent physical inactivity have a role in the severity of COVID-19? Therapeutic advances in respiratory disease. 2021;15:17534666211025221. doi: 10.1177/17534666211025221. PubMed PMID: 34148444; PubMed Central PMCID: PMC8221695. 22. Baynouna AlKetbi LM, Nagelkerke N, Abdelbaqi H, F AL, AlSaedi M, Almansoori S, et al. Risk Factors for SARS-CoV-2 Infection Severity in Abu Dhabi. Journal of epidemiology and global health. 2021;11(4):344-53. doi: 10.1007/s44197-021-00006-4. PubMed PMID: 34734381; PubMed Central PMCID: PMC8381146. 23. Halabchi F, Mazaheri R, Sabeti K, Yunesian M, Alizadeh Z, Ahmadinejad Z, et al. Regular Sports Participation as a Potential Predictor of Better Clinical Outcome in Adult Patients With COVID-19: A Large Cross-Sectional Study. J Phys Act Health. 2021;18(1):8-12. doi: 10.1123/jpah.2020-0392. PubMed PMID: 33260140. 24. Hamdan M, Badrasawi M, Zidan S, Sayarah A, Zahra LA, Dana S, et al. Risk factors associated with hospitalization owing to COVID- 19: a cross-sectional study in Palestine. The Journal of international medical research. 2021;49(12):3000605211064405. doi: 10.1177/03000605211064405. PubMed PMID: 34939466; PubMed Central PMCID: PMC8721739. 25. Malisoux L, Backes A, Fischer A, Aguayo G, Ollert M, Fagherazzi G. Associations between physical activity prior to infection and COVID-19 disease severity and symptoms: results from the prospective Predi-COVID cohort study. BMJ open. 2022;12(4):e057863. doi: 10.1136/bmjopen-2021-057863. PubMed PMID: 35487745. 26. Hamer M, Kivimaki M, Gale CR, Batty GD. Lifestyle risk factors, inflammatory mechanisms, and COVID-19 hospitalization: A community-based cohort study of 387,109 adults in UK. Brain, behavior, and immunity. 2020;87:184-7. doi: 10.1016/j.bbi.2020.05.059. PubMed PMID: 32454138; PubMed Central PMCID: PMC7245300. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprintthis version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.22.22276789doi: medRxiv preprint 14 / 14 27. Maltagliati S, Sieber S, Sarrazin P, Cullati S, Chalabaev A, Millet GP, et al. Muscle Strength Explains the Protective Effect of Physical Activity against COVID-19 Hospitalization among Adults aged 50 Years and Older. journal of sports medicile medRxiv. 2021. doi: 10.1101/2021.02.25.21252451. PubMed PMID: 33688683; PubMed Central PMCID: PMC7941660. 28. Bielik V, Grendar M, Kolisek M. A Possible Preventive Role of Physically Active Lifestyle during the SARS-CoV-2 Pandemic; Might Regular Cold-Water Swimming and Exercise Reduce the Symptom Severity of COVID-19? Int J Environ Res Public Health. 2021;18(13). doi: 10.3390/ijerph18137158. PubMed PMID: 34281096; PubMed Central PMCID: PMC8297290. 29. Latorre-Roman PA, Guzman-Guzman IP, Delgado-Floody P, Herrador Sanchez J, Aragon-Vela J, Garcia Pinillos F, et al. Protective role of physical activity patterns prior to COVID-19 confinement with the severity/duration of respiratory pathologies consistent with COVID-19 symptoms in Spanish populations. Research in sports medicine. 2021:1-12. doi: 10.1080/15438627.2021.1937166. PubMed PMID: 34128446. 30. Katsoulis M, Pasea L, Lai AG, Dobson RJB, Denaxas S, Hemingway H, et al. Obesity during the COVID-19 pandemic: both cause of high risk and potential effect of lockdown? A population-based electronic health record study. Public health. 2021;191:41-7. doi: 10.1016/j.puhe.2020.12.003. PubMed PMID: 33497994; PubMed Central PMCID: PMC7832229. 31. Ekblom-Bak E, Vaisanen D, Ekblom B, Blom V, Kallings LV, Hemmingsson E, et al. Cardiorespiratory fitness and lifestyle on severe COVID-19 risk in 279,455 adults: a case control study. The international journal of behavioral nutrition and physical activity. 2021;18(1):135. doi: 10.1186/s12966-021-01198-5. PubMed PMID: 34666788; PubMed Central PMCID: PMC8524225. 32. Steenkamp L, Saggers R, Bandini R, Stranges S, Choi Y, Thornton J, et al. Small steps, strong shield: directly measured, moderate physical activity in 65 361 adults is associated with significant protective effects from severe COVID-19 outcomes. British journal of sports medicine. 2022. doi: 10.1136/bjsports-2021-105159. PubMed PMID: 35140062. 33. Brawner CA, Ehrman JK, Bole S, Kerrigan DJ, Parikh SS, Lewis BK, et al. Inverse Relationship of Maximal Exercise Capacity to Hospitalization Secondary to Coronavirus Disease 2019. Mayo Clin Proc. 2021;96(1):32-9. doi: 10.1016/j.mayocp.2020.10.003. PubMed PMID: 33413833; PubMed Central PMCID: PMC7547590. 34. Nieman DC, Wentz LM. The compelling link between physical activity and the body's defense system. J Sport Health Sci. 2019;8(3):201-17. Epub 2019/06/14. doi: 10.1016/j.jshs.2018.09.009. PubMed PMID: 31193280; PubMed Central PMCID: PMCPMC6523821. 35. Shephard R, Shek P. Potential impact of physical activity and sport on the immune system--a brief review. British journal of sports medicine. 1994;28(4):247-55. doi: 10.1136/bjsm.28.4.247. PubMed PMID: 7894956. 36. Chastin SFM, Abaraogu U, Bourgois JG, Dall PM, Darnborough J, Duncan E, et al. Effects of Regular Physical Activity on the Immune System, Vaccination and Risk of Community-Acquired Infectious Disease in the General Population: Systematic Review and Meta-Analysis. Sports medicine. 2021;51(8):1673-86. doi: 10.1007/s40279-021-01466-1. PubMed PMID: 33877614; PubMed Central PMCID: PMC8056368. 37. Collaborators GRF. Global burden of 87 risk factors in 204 countries and territories, 1990-2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet (London, England). 2020;396(10258):1223-49. doi: 10.1016/s0140-6736(20)30752- 2. PubMed PMID: 33069327. 38. Stamatakis E, Ding D, Ekelund U, Bauman A. Sliding down the risk factor rankings: reasons for and consequences of the dramatic downgrading of physical activity in the Global Burden of Disease 2019. British journal of sports medicine. 2021;55(21):1222-3. doi: 10.1136/bjsports-2021-104064. PubMed PMID: 33926966. 39. Wunsch K, Kienberger K, Niessner C. Changes in Physical Activity Patterns Due to the Covid-19 Pandemic: A Systematic Review and Meta-Analysis. Int J Environ Res Public Health. 2022;19(4). doi: 10.3390/ijerph19042250. PubMed PMID: 35206434; PubMed Central PMCID: PMC8871718. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprintthis version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.22.22276789doi: medRxiv preprint . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprintthis version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.22.22276789doi: medRxiv preprint . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprintthis version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.22.22276789doi: medRxiv preprint . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprintthis version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.22.22276789doi: medRxiv preprint