Physical exercise frequency and cognition: a multicenter cross-sectional cohort study

Background and aims Dementia imposes a heavy burden on society and families, therefore, effective drug treatments, exploring and preventing factors associated with dementia, are paramount. To provide reference points for the best frequency of physical exercise (physical exercise), we investigated the association between frequency of PE and cognition in Chinese old adults. Methods 16,181 Chinese participants aged 65 years or older were included in this study. Associations between PE and cognition were estimated multivariate logistic and linear regression analyses. Associations were further investigated across dementia subtypes (Alzheimer dementia, vascular dementia, and other types of dementia). Subgroup analyses were performed in different age groups, in populations with and without stroke, and those with and without hypertension. Results PE associated with dementia after adjusting for full covariates (OR: 0.5414, 95% CI: 0.4536–0.6491, p < 0.001). Exercise performed at ≥3 times/week associated with lower risk of dementia (OR: 0.4794–0.6619, all p value <0.001). PE was associated with improved cognition (β: 12851, p < 0.001), and any PE frequency contributed to cognitive improvement (p values for exercise performed ≥1 time/week were <0.001). Similar conclusions were identified when we repeated analyses in different dementia subtypes and age groups. Subgroup analyses suggested that the cognition of individuals without hypertension also benefitted from exercising 1–2 times/week (OR: 0.6168, 95% CI: 0.4379–0.8668, p = 0.005). Conclusion The best exercise frequency is exercising ≥3 times/week for individuals from different dementia subtypes and age groups. While for those without hypertension, PE at 1–2 times /week is also beneficial.


Introduction
Dementia leads to a loss of independence thereby affecting families and the economy.In global terms, China has the largest population of individuals with dementia (Jia et al., 2020).In populations aged ≥65 years, the prevalence of all-cause dementia is 9.11%, while this prevalence is higher in rural areas when compared with urban areas (Hu et al., 2022).In 2050, the annual total cost of dementia will be approximately $1.89 trillion (Jia et al., 2020).Currently, no disease-modifying treatments are available for dementia.Therefore, exploring dementia prevention mechanisms and risk reduction approaches is paramount in China (Gao and Jia, 2023).Previous studies have shown that physical exercise (PE) is a potential cognition protective factor for individuals in early dementia stages (Liu et al., 2022), such as subjective cognitive decline (Wen et al., 2021) and mild cognitive impairment (Law et al., 2020).These findings suggest that PE interventions can affect cognition at an earlier stage than previously thought.
Physical inactivity is a modifiable risk factor associated with the reduced age-specific incidence of dementia (Tarassova et al., 2020;Alshagrawi and Abidi, 2023).Data from 7,000 individuals over a 2 years follow-up period showed that PE prevented or delayed cognitive impairment progression (He et al., 2021).PE increases cerebral blood flow (CBF) and nervous system plasticity (Farì and Lunetti, 2021).PE also reduces the neuroinflammation, oxidative stress, and amyloid β-protein (Aβ) deposition (Zhang et al., 2019).However, some studies have also shown inconsistent results; after a 5 years exercise intervention, older individuals showed no significant improvements in cognition (Zotcheva et al., 2022).Meanwhile, the most effective PE modalities for different population subgroups remain limited (Bull et al., 2020).Thus, there is a need for large sample studies to confirm such associations and provide evidence for the best PE intervention modality (Cha, 2022).To address this, we assessed the effects of PE on dementia and provided evidence showing the best PE interventions in China.

Participants
This study is our second multicenter, cross-sectional epidemiological survey, from April to October 2019, of dementia in elderly Chinese participants aged 65 years or older.
We collected data from 13 provinces, metropolitan areas, and autonomous areas which represented different geographical regions, urbanization levels, and economic development status in China.These areas included: Beijing, Tianjin, Chongqing, Fujian, Guizhou, Heilongjiang, Hubei, Hebei, Henan, Hunan, Liaoning, Shanxi, and Xinjiang.The detailed multistage, stratified cluster-sampling procedure was described in our previous study (Chen et al., 2022).A total of 23,382 individuals were interviewed, while only 21,745 individuals were eligible.All eligible individuals were aged 65 years or older and had lived in the same community or village for at least 1 year preceding the survey date.Among the eligible population, 301 refused to participate; 353 were untraceable; 72 had life-threatening illness; 78 were severe disability; 86 were deceased.Therefore, information from 20,855 individuals were collected.After excluding individuals with incomplete data and hearing or vision deficit, we included 16,181 individuals in our study (Figure 1).

Screening interview
This cross-sectional, door-to-door, questionnaire-based survey was conducted by senior neurologists and medical staff.All interviewers and experts received the same training on collecting information, neuropsychological assessments, and diagnosis, and retrained every 2 months.Participants' findings were recorded from physical and neurological examinations.
During interviews, participants completed a self-designed questionnaire using help from reliable informants (participant's spouse, children, other relatives or close friends, in descending order).Informants provided information if participants were unable to do so.The average interview lasted 30 min.Information collected from questionnaire included demographic factors (age, sex, and education years), lifestyle factors (smoking, drinking, PE status and PE frequency), and comorbidities (hypertension, diabetes, cerebrovascular disease etc).PE was defined as performing physical exercise that lasted 30 min or more and was evaluated with reference to a health survey (Kurtze et al., 2008).Participants were thought to perform PE if they answered "yes" for the question "Do you perform physical exercise that lasted 30 min or more?"PE frequency was collected by asking "How often do you perform physical exercise?"The answer was selected from the following choice:: 0 times/week (never), 1-2 times/week, 3 times/week, 4-5 times/week, and >5 times/ week.The answers would be confirmed by reliable informants.
Comorbidities, including stroke, hypertension, diabetes mellitus (DM), and coronary heart disease (CHD) history, were recorded from medical registers, and then confirmed with senior neurologists and medical staff to ensure accuracy.Stroke was defined as having a diagnosed or a known history of hemorrhagic or ischemic stroke.Hypertension was defined as having an average systolic blood pressure ≥ 140 mmHg or an average diastolic blood pressure ≥ 90 mmHg on ≥ three occasions or patients taking antihypertensive drugs.DM was defined as having a fasting serum glucose level ≥ 7 mmol/L, a non-fasting serum glucose level ≥ 11.1 mmol/L, or using hypoglycemic agents.CHD was defined as coronary atherosclerotic heart disease, which meant heart disease caused by coronary artery stenosis or occlusion.-Rodriguez et al., 2015), the Clinical Dementia Rating (CDR) scale (Morris, 1993), and Activities of Daily Living (ADL) scale (Eto et al., 1992;Chen et al., 1995) were administered by qualified and experienced specialists in neurology.Interviewers at each site included four junior neurologists and four neurologists from the local cooperative hospital.An expert panel and interviewers reviewed all the gathered information, and primary diagnoses were made at the end of each workday.If consensus was not reached, an expert returned to the participant's residence the following day to reexamine and reevaluate the participant and provide a final, definitive diagnosis.Data were stored on a secure server accessible by authorized personnel only.
In our survey, a non-dementia status was assigned when participants scored 0 on global CDR and ≥27 on the C-MMSE.When the C-MMSE test score was ≤ the cutoff point (≤17 for illiterate persons, ≤20 for persons with 1-6 years of education, and ≤24 for persons with ≥7 years of education), dementia was defined based on clinical criteria from the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV edition).DSM-IV criteria for dementia required   et al., 2011).We used diagnostic criteria for vascular dementia (VD) as proposed by the Neuroepidemiology Branch of the National Institute of Neurological Disorders and Stroke (Román et al., 1993).

Ethical considerations
The study was approved by the ethics committee of Tianjin Huanhu Hospital (ID:2019-40).Written informed consent was obtained from participants or their guardians.The procedures were performed in accordance with the ethical standards of the Committee on Human Experimentation.Study data were anonymous.

Statistical analysis
We used Kolmogorov-Smirnov normality tests and Quantile-Quantile plots to assess data normality.Variables were transformed using the "car" package in R software (Xu et al., 2020) to generate approximate normal distributions (Supplementary Figure S1).Statistical analyses were conducted on transformed values.Differences in categorical variables between two groups were analyzed using Chi-square tests, and numerical variables were analyzed using Wilcoxon tests.Difference comparisons between two groups involved multiple comparisons, which may have generated uncontrolled type I error rates (the rate of rejecting the null hypothesis when it should not be rejected) (Cabral, 2008).We used the false discovery rate (FDR) to adjust for multiple comparisons (threshold q < 0.05).

Model building and covariate selection
Statistically significant indicators in univariate analysis (age, sex, education years, and stroke, Supplementary Tables S1, S2) were included in multivariate analysis.Although no significance was identified for hypertension, DM, CHD, smoking, and drinking, they were previously considered risk factors for dementia (Baumgart et al., 2015) and were also included in multivariate analysis and categorized.M2 was adjusted for demographic factors (age, sex, and education years).M3 was additionally adjusted for comorbidities and lifestyle indicators (stroke, hypertension, DM, CHD, smoking, and drinking).
First, associations between PE (status and frequency) and dementia were estimated using a univariate logistic regression model (M1).Then, associations were further confirmed using a multivariate logistic regression model (M2).Finally, indicators were added to the multivariate logistic regression model as covariates to assess the robustness of results (M3).
Considering the fact that PE could affect cognition, we investigated if PE (status and frequency) was associated with C-MMSE scores.Associations between PE and cognition were estimated using linear regression models M1-M3, which adjusted for the same aforementioned factors.We also conducted an association study in different dementia subtypes, in participants with AD, VD, and other dementia types.
Variables in multivariate regression analyses were selected for interaction analyses.Accordingly, we conducted subgroup analyses stratified by age (65-74, or ≥75), hypertension (yes or no), and stroke (yes or no).Variance inflation factors were used to assess multicollinearity, which we found no evidence of in our analyses.
Two-tailed p < 0.05 values were considered statistically significant.Analyses were conducted in R software (version 3.6.1).

FIGURE 3
Association between physical exercise and dementia subtype.M1: the univariate regression analysis.M2: the multivariate analysis adjusted for age, sex, and education years.M3: the multivariate analysis adjusted for age, sex and education years, hypertension, DM, CHD, stroke, smoking and drinking.

Subgroup analyses
Interaction analyses demonstrated that associations between PE frequency and dementia were possibly affected by age, hypertension status, and stroke status (Supplementary Table S8).Therefore, subgroup analyses were conducted at different ages (65-74, and ≥75 years, Supplementary Table S9), stroke status (yes and no, Supplementary Table S10) and hypertension status (yes and no, Supplementary Table S11).

Discussion
In a large sample cohort from China, we assessed associations between PE frequency and dementia.PE exerted positive effects on cognition in our cohort.When compared with inactivity and having a lower PE frequency (1-2 times/week), performing PE ≥ 3 times/ week was better for improving cognition.Further findings, based on stratified analyses by age, hypertension, and stroke status, consistently and significantly showed the protective role of PE and performing PE ≥ 3 times/week.Of note, participants without hypertension might benefit from performing PE 1-2 times/week.Our findings strengthen the evidence showing the protective effects of PE frequency on dementia, and provide insights on PE for dementia prevention.
In our study, PE was a protective factor for dementia, regardless of age and comorbidity.These observations were consistent with previous studies; in a 10 years follow-up study, PE was inversely associated with cognitive impairment onset (Jedrziewski et al., 2010).Similarly, a systematic review involving 5,606 individuals from 73 articles concluded that all types of PE protected individuals from decreased global cognition (Huang et al., 2022).However, recent studies have also reported inconsistent results.A multicenter trial randomized 494 individuals, followed them for 4 months, and showed that moderate to high-intensity PE did not slow cognitive impairment in individuals with mild to moderate dementia (Lamb et al., 2018).A recent systematic review reported that neither a combination of strength and aerobic exercise, nor aerobic exercise alone, exerted Logistic regression analyses of subgroup.M1: the univariate logistic regression analysis.M2: the multivariate logistic regression analysis adjusted for age, sex and education years.M3: the multivariate logistic regression analysis adjusted for age, sex and education years, hypertension, DM, CHD, stroke, smoking and drinking.Wen et al. 10.3389/fnagi.2024.1381692Frontiers in Aging Neuroscience 08 frontiersin.orgbeneficial effects toward cognition and dementia (Steichele et al., 2022).PE exposure and sample size heterogeneity could account for some of these inconsistencies.Other covariates potentially affecting cognition, such as lifestyle factors and comorbidity, may also contribute to this heterogeneity.Different PE levels can generate different effects, but few studies have quantified PE interventions for dementia (López-Ortiz et al., 2021;Liu et al., 2022).Previous studies also confirmed a doseresponse relationship between exercise and cognition (Gallardo-Gómez et al., 2022).Higher PE frequency appeared to contribute to better cognitive results, while specific PE thresholds require clarification (Jia et al., 2019).Our analyses indicated that performing PE ≥ 3 times/week was associated with dementia.Of note, performing PE 1-2 times/week was a protective factor for dementia in individuals without hypertension.This finding supported a previous study showing that low-frequency PE exerted positive effects on cognitive function in individuals with chronic diseases (Cai et al., 2017).
Although recommendations for PE levels and types for particular groups have been reported, the evidence for the best PE modality for individuals with different cognitive states remains uncertain (Ding et al., 2020;Steichele et al., 2022).In our study, we analyzed different dementia types, including AD, VD, and other dementias.Consistent with previous studies, we confirmed that PE was a protective factor against AD and benefited cognition (Norton et al., 2014).Regular PE protects non-dementia individuals from AD pathological changes in cerebrospinal fluid (CSF) (Zhong et al., 2022).Also, PE was associated with VD in our study.This finding supports previous studies showing that PE exerted beneficial effects toward VD and putatively prevented VD development (Aarsland et al., 2010).This association may relate to the fact that PE increases oligodendrocyte precursor cell populations in the sub ventricular zone of the brain (Ohtomo et al., 2020).
Animal model and human studies have explored underlying mechanisms at multiple levels (Stillman et al., 2020;Tarassova et al., 2020).Firstly, PE may increase cerebral perfusion by increasing cerebral blood flow (CBF) (Huang et al., 2022).Such increases during PE could meet the energy demands related to cognition in the brain and improve cognitive function (Buxton, 2021;Yamada et al., 2021).Results of the researches on the association between PE and cognition were inconsistent.Some studies have pointed out that high-intensity exercises were associated with hyperventilation and hypoxia, which constricted blood vessels and reduced CBF (Verges et al., 2012).This observation may explain a decline in cognition after high-intensity PE (Gallardo-Gómez et al., 2022).However, a study has pointed out that cognitive impairment caused by high-intensity exercise was not related to CBF (Komiyama et al., 2020).Our study explored the relationship between PE frequency and cognition in a large sample size.Secondly, PE improved cognition by promoting neuroplasticity and neuroprotection (Soshi et al., 1991;Farì and Lunetti, 2021).These processes were directly mediated by increased brain-derived neurotrophic factor (BDNF) levels induced by PE (Wheeler et al., 2020).BDNF activates multiple intracellular signaling pathways, including phospholipase C-γ1/protein kinase C, Ras-mitogenactivated protein kinases, and phosphoinositide 3-kinase/seronine protein kinase, to regulate cerebral cortex thickness and synaptic density, thus increasing brain plasticity (Wang and Holsinger, 2018).At peripheral levels, PE promoted fibronectin type III domaincontaining 5 cleavage into irisin, which may have activated the brain cyclic adenosine phosphate/Protein Kinase A/cAMP-response element binding protein and enhanced BDNF levels (Madhu et al., 2022).PE also stimulated the ketone body D-β-hydroxybutyrate and cathepsin B to activate BDNF expression.Also, serotonin (Pietrelli et al., 2018) and several growth factors, including insulin growth factor-1 and vascular endothelial growth factor, were also induced by PE and exerted synergistic effects with BDNF in terms of neuroplasticity and neuroprotection (Jachim et al., 2020).Finally, neuroinflammation increases with aging and contributes to cognition decline, whereas PE was shown to attenuate this process (Huang et al., 2021).PE appeared to regulate micro-RNA expression (Hu et al., 2015), and significantly decreased pro-inflammatory markers, including interleukin-1β (IL-1β), IL-6, and tumor necrosis factor-α (TNF-α) (Qin et al., 2022).Proinflammatory microglia and astrocytes were suppressed by PE (Nakanishi et al., 2021).Additionally, PE was positively associated with CSF Aβ42 levels (Zhong et al., 2022), which may be mediated by activating lysosomal function (Wang et al., 2022) and promoting microglial Aβ clearance (Liang et al., 2022).
Our subgroup analysis showed that performing PE 1-2 times/ week was associated with a lower risk of dementia in patients without hypertension.In participants with hypertension, performing PE ≥3 times/week was beneficial.Chronic hypertension had detrimental effects on cognition via mechanisms underpinning cerebral small vessel disease, reduced white matter integrity, and impaired autoregulation in the brain (Claassen et al., 2021;Triposkiadis and Xanthopoulos, 2023).Thus, hypertension may reduce or negate the benefits of low-frequency PE.
A major advantage of our study is its large sample size which provides considerable data reliability and robustness.Also, cognitionrelated covariate adjustments highlighted the independence of PE as a protective factor for dementia and cognition.Our dementia subtype analyses could help us understand the effects of PE on different dementia types.However, our study had notable limitations.
As a cross-sectional study, we did not provide causation information similar to other prospective cohorts.Objective measurements for PE, such as pulse oximetry or the calculation of "mets, " and some PE-related indicators (types, intensity, or duration) were unavailable.Information about the consistent length of comorbidities was not collected.Additionally, participants were primarily elderly Chinese individuals, thus diversity across ages, races, and regions was not confirmed, and so populations should be expanded to identify more generalizable findings.Finally, our study was limited to dementia population and cognitively normal population.More studies should be performed in individuals at pre-clinical dementia stages so appropriate PE interventions can be implemented for these individuals.

Conclusion
In conclusion, PE was associated with cognitive decline when different adjustments were applied, regardless of dementia subtype.Performing PE ≥ 3 times/week was most effective in preventing dementia, whereas cognition appeared to benefit from any PE frequency.Moreover, the protective effects of PE were consistently observed in subgroups.Our findings underscore the importance of PE as a non-pharmaceutical therapy for delaying cognitive decline and preventing dementia in China.

FIGURE 4
FIGURE 4 a gradual onset and progressive course, and not due to any other process.An Alzheimer's disease (AD) diagnosis was based on The National Institute of Neurological and Communicative Disorders-AD and Related Disorders Association criteria (McKhann

TABLE 1
Characteristic of participants.
Continuous variables were represented as mean ± SD.Categorical variables were represented as numbers (proportion).Differences in categorical variables between two groups were analyzed using Chi-square tests, and numerical variables were analyzed using Wilcoxon tests.q value: significance after false discovery rate (FDR) correction.C-MMSE, Chinese Mini-Mental State Examination; DM, diabetes mellitus; CHD, coronary heart disease.