Effects of physical exercise on cognitively impaired older adults: a systematic review and meta-analysis of randomized control trials

Objectives: The main aim of this study was to compare the effects of different physical activities on cognitive functions in older adults divided according to cognitive impairment levels. Methods: We searched Web of Science, Scopus, and PubMed for randomized control trials (RCT). A standardized mean difference (SMD) of the pre-post intervention score of global cognitive function tests was calculated by the random model in the Cochrane meta-analyses for people with cognitive impairment generally and across three levels - borderline intact, mild, and moderate cognitive impairment separately. Additionally, an unstandardized coecient beta (B) was calculated in generalized linear models to estimate the effects of exercise, cognitive impairment severity, age, female ratio, duration and frequency of exercise program on the global cognitive function. Results: Data from 40 studies involving 1,780 participants from intervention groups and 1,508 participants from control groups were analyzed. After sensitivity analysis, physical exercise had a positive effect on cognitive functions in people across all levels of cognitive impairments, SMD (95 % condence interval [CI]) = 0.41 (0.29 - 1.54). All the activities were signicantly associated with better results in global cognitive functions when compared to active control (B = 0.538 in aerobic, 0.999 in resistance, 0.640 in combined exercise and 0.746 in Tai Chi). Age was signicantly associated with global cognitive functions decreasing and a higher number of female participants in intervention groups had a statistically signicant effect on the global cognitive function (B = 0.021). Conclusions: Physical exercise was associated with cognitive function improvement in older people with cognitive impairments. Cognitive impairment severity was not associated with cognitive functions changes after exercise interventions. Additionally, we calculated SMD for intervention as well as control groups separately. Then generalized linear models were used to estimate the inuences of selected factors and covariates to the SMD as the continuous dependent variable. We calculated an unstandardized coecient beta (B), standard error (SE) and 95% CI. B represents the amount by which dependent variable changes if we change the independent variable by one unit, keeping other independent variables constant. If 95 % CI does not cross the 0, then the result is statistically signicant. Statistics were calculated using RevMan 5.3 and IBM SPSS Statistics 24.


Background
The number of older adults with dementia is on the rise due to a global population ageing. Current estimates suggest that more than 131.5 million people will be affected by dementia by the year 2050 [1]. Dementia is generally characterized by a progressive decline in cognitive and physical functions, often leading to a loss of independence, and institutionalization in some cases [2]. Thus, dementia impacts not only daily lives of individuals diagnosed with the condition but also their families and broader society.
During the past two decades, epidemiological research has highlighted the link between modi able lifestyle factors and cognitive functions. For example, current evidence has demonstrated that a physically active lifestyle may help to delay the onset of cognitive decline and to slow down disease progression [3]. Also, physically active individuals have been shown to have a smaller risk of developing dementia or mild cognitive impairment than those who do not take part in any regular physical activity [4]. Moreover, results from several prospective studies have shown that exercise and physical tness seem to have a positive effect on brain health [5,6]. In particular, it has been demonstrated that regular physical activity in mid-life is associated with a lower risk of dementia in later life [7], as well as that one of the most effective protections against neurodegenerative or vascular dementia is to be su ciently physically active from mid-life [3]. In addition, it is now well known that exercise interventions increase the functional performance and activities of daily living in patients with cognitive impairments [8,9,10,11,12]. A positive effect of physical exercise on global cognition in individuals with mild cognitive impairments was partly con rmed [13,14,15,16,17,18]. Nevertheless, the effects of exercise on global cognitive function in people taking into account the level of cognitive impairment has still not been clearly elucidated. Likewise, the effects of aerobic and resistance exercise require further investigations too. Therefore, the main aim of this study was to generally analyze the effects of exercise on cognitive functions in older adults divided according to cognitive impairment severity, taking into consideration the effects of resistance exercise and aerobic exercise separately. Additionally, we aimed to investigate the association between selected factors including the passive or active control, cognitive impairment severity, age, sex, frequency and duration of exercise program. We hypothesized that there is a difference between aerobic and resistance exercise in terms of the effect on cognitive functions and that the effect might vary across different levels of cognitive impairment. We also hypothesized that different activity programs in control groups might in uence the results. For example, a social program without physical activities may be bene cial for older adults with cognitive impairment. We also assumed that social or education activities in control groups might be more helpful against the cognitive decline rather than inactivity in passive control groups.

Methods
This study assessed the effects of physical exercise programs on people with cognitive impairment. It is reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [19]. A compiled PRISMA checklist is included in Table 1.
The PICO (population, intervention, comparisons, and outcomes) framework was used for framing the inclusion criteria (see below).
Participants: people with a cognitive impairment being diagnosed with one of the standardized tools with a closed scale Intervention: activities requiring increased energy output excluding interventions combining both physical and cognitive training Comparisons: active or passive controls according to reported activities Outcomes: cognitive performance; focus on: type of exercise or control group activities, age and sex of participants, exercise program duration, frequency of exercise, and severity of impairment Inclusion criteria for this study Based on the above-mentioned PICO framework, the following inclusion criteria were applied: only data from randomized trials (RCT) the participants had to be diagnosed with a cognitive impairment according to one of the standardized tools written in the English language Concerning the exercise programs, only exercise activities that required increased energy output were included. All intervention groups from studies with a combination of physical exercise and cognitive training were excluded.

Exercise intervention and control group classi cation
We divided interventions into four categories accordingly: 1. Aerobic exercise -walking, stretching, ergometer cycling, as well as activities generally speci ed as "aerobic exercise" 2. Resistance exercise -strengthening exercise with elastic bands, balls, ankle weights, own body weight, dumbbells, resistance machines training, strength training, and activities generally speci ed as "resistance exercise" 3. Combined exercise -if the intervention program included both aerobic and resistance exercise, then music-based dance therapy, kinesiotherapeutic exercise, and handball training 4. Tai Chi -for signi cant similarity baduanjin was also included According to activities that were prescribed, we have also divided control groups into two categories -active and passive control groups. All control groups where extra activities that could have potentially been bene cial for cognitive functions (for example, attention-control educational programs, social visits, or recreational activities such as card playing or home craftwork), were categorized as "active control groups". Control groups asked to maintain their usual activities were categorized as "passive control groups".
We also analyzed the duration of exercise program, frequency of exercise, and female ratios. For exercise program duration and frequency of exercise per week, we used the same classi cation as Forbes at al (2013) in Cochrane systematic review -"up to three times per week" or "more than three times per week" and "up to 12 weeks or "more than 12 weeks" [25].

Search strategy
The analysis was conducted by identifying relevant papers referenced in the Web of Science, Scopus, and PubMed. Search terms used in all databases are presented in Table 2.
Data extraction and quality assessment All potential papers were rst downloaded into EndNote. Then, our three reviewing authors (LS, AT, and MS) deleted all the duplicates and scanned the titles and abstracts of the papers in order to identify studies that had the potential to meet the eligibility criteria. Full texts were subsequently assessed for eligibility. Any disagreements among the reviewers (KD, and IH) were resolved through discussions. We used the Physiotherapy Evidence Database (PEDro) scale to assess the methodological quality of the included studies [26].
We collected the following data for both exercise groups and control groups separately: baselines and after intervention means with 95% con dence interval (CI) and/or standard deviation (SD); and if described, also the mean of post-pre intervention score with SD or 95% CI were collected. Additionally, for factors or covariates for general linear models, we collected information about the type of exercise or control group activities, age of participants, female ratio, exercise program duration, and frequency of exercise.

Cognitive impairment classi cation
We divided the participants according to the level of their cognitive impairment into three categories -borderline intact, mild, and moderate cognitive impairment. In the classi cation, we used a mean of the baseline using the standard classi cation of each diagnostic tools.

Data analysis
The standardized mean difference (SMD) was calculated from the sample size, mean post-pre intervention score with SD from intervention and control groups. The random effect models were used for all the analyses [27]. If there were no available data, we calculated the mean of postpre intervention score as a mean of the post-intervention score -a mean of the pre-intervention score. The SD we estimated as: See formula 1 in the supplementary les.
We used Corr = 0.8 based on the assumption of a relatively high correlation between pre and post-measurements.
To assess the heterogeneity, I 2 was considered. A rough guide to the interpretation of I 2 is as follows: 0 to 40% might not be important, 30% to 60% may represent moderate heterogeneity, 50% to 90% may represent considerable heterogeneity, and 75% to 100% represents substantial heterogeneity [28]. We made a sensitivity analysis using funnel plots to eliminate heterogeneity. Additionally, we calculated SMD for intervention as well as control groups separately. Then generalized linear models were used to estimate the in uences of selected factors and covariates to the SMD as the continuous dependent variable. We calculated an unstandardized coe cient beta (B), standard error (SE) and 95% CI. B represents the amount by which dependent variable changes if we change the independent variable by one unit, keeping other independent variables constant. If 95 % CI does not cross the 0, then the result is statistically signi cant. Statistics were calculated using RevMan 5.3 and IBM SPSS Statistics 24.

Results
We included 40 RCT in the nal analysis out of the 1,258 publications resulting from the database search. These were controlled trials on physical activity and its effect on cognitive functions in people with cognitive impairments. Figure 1 shows the PRISMA ow diagram. Across the studies, we extracted data from 3,288 participants, all being over 50 years of age (69.1 % females). The majority of interventions was aerobic exercise (18 out of 46) and the frequency of exercise varied between two and seven sessions per week. The shortest duration of the exercise program was 6 weeks, and the longest was 60 weeks. As the main outcome, the following were used: In general, and after sensitivity analysis, physical exercise had a positive effect on cognitive functions in people with cognitive impairment SMD (95 % CI) = 0.41 (0.29 -0.54); heterogeneity was low I 2 = 24 %. In separate groups and according to the cognitive impairment severity, the exercise had statistically signi cant effect SMD (95 % CI) = 0.32 (0.18 -0.47) I 2 = 0% in borderline intact and 0.64 (0.44 -0.84) I 2 = 0% in mild cognitive impairment respective. Nevertheless, there was not signi cant effect in moderate cognitive impairment SMD (95 % CI) = 0.20 (-0.18 -0.58) I 2 = 53%. A forest plot with a graphical representation of individual effects is presented in Figure 2.
In the generalized linear models, when we used the active control groups as a reference category, its change on any type of exercise caused a signi cant increase in SMD estimate. Aerobic exercise B = 0.538, resistance exercise B = 0.999, combined exercise B = 0.640, and Tai Chi B = 0.746. When comparing the passive groups to the active groups, no signi cant effect was found B = 0.038. Age was signi cantly negatively associated with global cognitive functions B = -0.026. Cognitive impairment level, as well as the duration of exercise program, were not signi cantly associated with SMD estimate. The result of the generalized linear model of intervention and control groups is presented in Table  6.
For the intervention groups, there was not any signi cant association between SMD estimate and type of exercise, cognitive impairment, as well as the frequency of exercise. Age played a signi cant negative role (B = -0.048), and a higher number of women in intervention groups had a statistically signi cant positive effect (B = 0.021). The result of the generalized linear model of intervention groups separately is presented in Table 7.

Discussion
It is well-established that cognitive functions decline gradually over time as part of the natural ageing process [69]. The overall results of this meta-analysis indicate that physical exercise and speci cally aerobic exercise may have the power to mitigate cognitive decline process even in people with cognitive impairment.
According to our results, aerobic exercise had a twice higher impact on cognitive functions than resistance exercise when compared to active controls, which was four times higher when only the interventions groups were compared. Previous studies partly con rm these results demonstrating a positive effect of physical exercise on executive functions [14,15], and on global cognition [16,17,18] in individuals with mild cognitive impairments. However, we found that aerobic exercise also had a statistically signi cant positive effect in moderate to severe cognitively impaired people. Probably the positive effect of aerobic exercise on brain health seems to lie in the proposed mechanisms behind aerobic exercise such as neovascularization, synaptogenesis and angiogenesis, hippocampal high-a nity choline uptake and upregulation of muscarinic receptor density, increasing of mitochondrial volume in Purkinje cells, inhibition of the apoptotic biochemical cascades, identi ed primarily through animal research [70,71,72,73].
Moreover, a higher number of female participants in intervention groups had a positive effect on global cognitive function. This result could be explained by both different cognitive responses to exercise between men and women as well as by the different ratios in elderly females suffering dementia. As described by Baker et al. (2010), aerobic exercise improved performance on multiple tests of executive function, increased glucose disposal during the metabolic clamp, and reduced fasting plasma levels of insulin, cortisol, and brain-derived neurotrophic factor in women but not in men [74]. They also found that peak oxygen consumption was associated with improved executive function in women. It turns out that gender differences in cognitive functions can be related to the metabolic effects of physical activity. However, there are several other reasons that sex may in uence trial results. For instance, women have a higher lifetime risk of dementia [75], greater vulnerability to certain risk factors such as sex-speci c chromosomes, APOE ε4, sex differences in hormone levels etc. [76], and they demonstrate higher differential associations between biomarkers and cognitive impairment than men [77]. Moreover, there was a higher percentage of female participants in the intervention studies (40 of 46 intervention groups had a majority of female participants). One reason for this fact could be higher life expectancy in females [78] although the gender gap has been narrowing in Europe recently [79]. Another explanation could be greater adherence to health-related exercise programs in older women [80]. Thus, it would be of interest to explain which of the above-mentioned proposed factors is responsible for gender differences.
Studies included in this meta-analysis varied in terms of duration of exercise programs. In twenty seven studies, the duration of interventions was less than half a year, and in another nineteen, the duration of the interventions was for more than or equal to half a year. According to our analysis, it seems that the duration of the exercise program was associated with cognitive decline, which may be caused by the natural cognitive decline during ageing. Surprisingly, the frequency of exercise per week did not play any signi cant role in global cognition.
It should be noted that several limitations are involved in this study. Before the sensitivity analysis, there was considerable heterogeneity in all the analyses. In fact, heterogeneity is a common problem when conducting meta-analyses on this topic [14,18]. Nevertheless, using general linear models involved some limitations too. For example, we used only individual SMD and not the total amplitude, such as 95% CI. Therefore, the statistical signi cance of individual studies could not be drawn. Moreover, it was almost impossible to create a category with similar cognitive impairment because it varied considerably among the studies so the classi cation has some limitations, because if the variability was high then we could not be sure that all the participants were allocated rightly. The same is true for exercise interventions because the interventions included many different activities with different durations and intensities.

Conclusion
Despite the numerous limitations mentioned above, this study has shown that physical exercise and especially aerobic exercise may have the power to in uence cognitive functions in people with cognitive impairment. Such ndings could have practical implications such as to recommend physical activity as a nonpharmacologic treatment to combat the progression of cognitive decline in patients with dementia. Future research based on longitudinal epidemiological studies is needed to con rm such ndings further.

Declarations
Our results have not been published previously and are not under submission elsewhere. Co-authors are cognizant of the submitted text and agree to its publication in BMC Public Health.
Ethics approval and consent to participate

Availability of supporting data and material
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.

Funding
This research was supported by the Alzheimer Endowment Fund -AVAST, the project Q41, the AZV research project NV18-09-00587 of the Ministry of Health and project SVV 260466.
The funding agencies played no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.   -Eligibility criteria 6 Specify study characteristics (e.g., PICOS, length of follow-up) and report characteristics (e.g., years considered, language, publication status) used as criteria for eligibility, giving rationale.

5-7
Information sources 7 Describe all information sources (e.g., databases with dates of coverage, contact with study authors to identify additional studies) in the search and date last searched.
7, Table 2 Search 8 Present full electronic search strategy for at least one database, including any limits used, such that it could be repeated. Table 2 Study selection 9 State the process for selecting studies (i.e., screening, eligibility, included in systematic review, and, if applicable, included in the metaanalysis).

7, Fig 1
Data collection process 10 Describe method of data extraction from reports (e.g., piloted forms, independently, in duplicate) and any processes for obtaining and confirming data from investigators.

7-8
Data items 11 List and define all variables for which data were sought (e.g., PICOS, funding sources) and any assumptions and simplifications made.
-Risk of bias in individual studies 12 Describe methods used for assessing risk of bias of individual studies (including specification of whether this was done at the study or outcome level), and how this information is to be used in any data synthesis.
7 Table 4 Summary measures 13 State the principal summary measures (e.g., risk ratio, difference in means).

8
Synthesis of results 14 Describe the methods of handling 8-9 data and combining results of studies, if done, including measures of consistency (e.g., I 2 ) for each meta-analysis.
Risk of bias across studies 15 Specify any assessment of risk of bias that may affect the cumulative evidence (e.g., publication bias, selective reporting within studies).
-Additional analyses 16 Describe methods of additional analyses (e.g., sensitivity or subgroup analyses, metaregression), if done, indicating which were pre-specified.

RESULTS
Study selection 17 Give numbers of studies screened, assessed for eligibility, and included in the review, with reasons for exclusions at each stage, ideally with a flow diagram. 9, Figure 1 Study characteristics 18 For each study, present characteristics for which data were extracted (e.g., study size, PICOS, follow-up period) and provide the citations. Table 3 Risk of bias within studies 19 Present data on risk of bias of each study and, if available, any outcome level assessment (see item 12).

9
Results of individual studies 20 For all outcomes considered (benefits or harms), present, for each study: (a) simple summary data for each intervention group (b) effect estimates and confidence intervals, ideally with a forest plot.  Table 5 Synthesis of results 21 Present results of each metaanalysis done, including confidence intervals and measures of consistency.

Figure 2
Risk of bias across studies 22 Present results of any assessment of risk of bias across studies (see Item 15).

DISCUSSION
Summary of evidence 24 Summarize the main findings including the strength of evidence for each main outcome; consider their relevance to key groups (e.g., healthcare providers, users, and policy makers).

11-13
Limitations 25 Discuss limitations at study and outcome level (e.g., risk of bias), and at review-level (e.g., incomplete retrieval of identified research, reporting bias).         Present results of each meta-analysis done, including con dence intervals and measures of consistency.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. formula.docx