EFFECT OF EXERCISE ON COGNITIVE FUNCTION IN PERSONS WITH DEMENTIA: A SYSTEMATIC REVIEW AND META-ANALYSIS

As the population ages, the number of people suffering from dementia will rise significantly. Current estimates of total societal cost of dementia exceed $8 billion dollars (US). Epidemiological studies have shown that increased lifetime engagement in exercise reduces cognitive decline and the incidence of dementia in normal older adults. While existing research suggests that lifelong exercise may be preferable, the adoption of exercise at any age and stage of dementia-onset to delay or reverse cognitive decline is worthwhile given the prevalence of sedentary lifestyle, and the increasing proportion of older adults and dementia incidence. Recently, trials have started to explore the impact of exercise on cognitive symptoms in individuals diagnosed with dementia. These studies are reporting promising findings, which call for further meta-analytical review. The primary objective of this meta-analysis is to examine the effects of exercise interventions on cognitive function compared to standard care in older persons with dementia. The secondary objectives are to identify covariates and/or moderators that affect the effectiveness of these exercise programs. The trials included in this meta-analytic review were identified from systematic searches of major medical and psychological databases, including EMBASE, PubMed/MEDLINE, PsycARTICLES, and Cochrane Central Register of Controlled Trials (1966-2018) on 30 April 2018 using the concepts of dementia, cognitive function, and exercise. Trials were selected in which older people, diagnosed with dementia, were allocated either to exercise programs or to control groups (standard care) with the aim of improving cognition. One rater retrieved the articles, assessed for inclusion and methodological quality, and extracted the data. Data was analyzed for summary effects using mean difference (MD) and standardized mean difference (SMD). Data was synthesized for each outcome using a random-effects model. Exploration of heterogeneity was planned in relation to type, frequency and duration of exercise program. The collected data were analyzed by Review Manager (5.3). This review evaluated the results of 21 trials, including 1548 participants, that tested whether exercise programs could improve cognition (which includes such things as memory, learning, attention) in older people with dementia. The included trials were heterogeneous in terms of participant characteristics, and type, duration, and frequency of exercise. Meta-analysis demonstrated positive effects of exercise on cognitive function in older adults with dementia (SMD = 0.49, 95% CI [0.24 0.75], P = 0.0002). Fifteen trials demonstrated that exercise improves cognitive function for individuals with dementia, while the remaining six studies did not display a beneficial effect of exercise on cognitive function. This analysis revealed substantial heterogeneity (I2 = 79%), most of which could not be explained, and the quality of evidence was rated as low. Thus, findings should be interpreted with caution. This meta-analysis and systematic review revealed some evidence supporting the benefit of exercise programs in improving cognitive ability or slowing the decline of cognition in people with dementia. Future welldesigned RCTs with clear intervention criteria, large samples, and long-term follow-up are needed to enhance the quality of such a review by assessing the exercise programs that are best for people with various types and severity of dementia and by addressing additional outcomes (e.g., mortality, quality of life, healthcare service use, expenditures).

x LIST OF TABLES   Dementia is a clinical syndrome characterized by cognitive decline, motor deficits and/or behavioral problems, resulting in a decline in daily functioning (Scott & Barrett, 2007). With advancing age as the main risk factor, and an aging population worldwide, the prevalence of dementia worldwide is expected to double from approximately 50 million cases in 2019 to 65.7 million in 2030 (Prince et al., 2013). The total global societal cost of dementia was estimated to be US$ 818 billion in 2015, equivalent to 1.1% of global gross domestic product (GDP). This expected increase in prevalence of dementia will thus have increasingly dramatic social and financial consequences. WHO has denoted dementia as a public health priority (Wortmann, 2012). No disease modifying drugs for dementia are currently available, and pharmacological treatment is restricted to medications that may alleviate symptoms. As a first approach, best practice guidelines currently recommend that behavioral and psychological interventions be explored before initiating pharmacological interventions, due to the limited benefits of pharmacological treatments in reducing functional decline and their possible side effects (Hogan et al., 2008). Thus, exercise treatment may be an appealing alternative or adjunct to medication (Deslandes et al., 2009). Furthermore, in addition to the preventative benefits exercise has on dementia, exercise is among the potential protective lifestyle factors identified as a strategy for treating the symptoms of dementia or delaying its progression (Lautenschlager et al., 2010).

JUSTIFICATION FOR AND SIGNIFICANCE OF THE STUDY
1. Exercise for the prevention and treatment of chronic disease.
The World Health Organization (WHO) reported that physical inactivity is the fourth leading risk factor for global mortality, accounting for 6% of deaths globally (World Health Organization, 2010). Moreover, a 25% decrease in physical inactivity could prevent 1.3 million deaths annually , which could contribute to dramatic reductions in healthcare costs. Such findings have led to the development of randomized controlled trials (RCTs) assessing exercise's preventative, and more recently, treatment benefits in those at risk or already diagnosed with chronic disease (e.g., Belardinelli et al., 2012;Zwisler et al., 2008).

Multifaceted benefits of exercise.
Despite the variability of exercise effects in preventing, slowing the progress of, or improving various chronic diseases, it is important to note that the biomedical model of disease does not account for the complex factors associated with living with chronic diseases (e.g. medication side-effects, costs). This demonstrates the need for a broader more integrative framework in the treatment of chronic health conditions that includes lifestyle interventions, like exercise, to decrease disease-related morbidity, improve quality of life and physical and psychological health outcomes. Population-level exercise interventions have a multiplicity of benefits across a variety of populations, diseases, and domains (i.e. physical and mental health) while being relatively low cost compared to the healthcare and medical expenses in treating these diseases. Exercise and cognitive functioning in older adults.
In recent years, exercise programs for healthy older adults without dementia have demonstrated improvement in cognitive function and prevention of the onset of dementia (Erickson et al., 2011;Tseng, Gau, & Lou, 2011). Many such studies used a 60-minute exercise regimen scheduled three times per week for 24 weeks (Tseng, Gau, Lou, 2011).
For example, Hamer and Chida (2009)  Further, other research shows that mid-life exercise may contribute to maintenance of cognitive function and may reduce or delay the risk of late-life dementia (Chang et al., 2010). Exercise is one lifestyle factor that has recently been identified as a potential means of reducing or slowing the progression of dementia symptoms, including cognitive performance. Animal model studies demonstrate that exercise may slow the progress of existing dementia, and an increasing amount of trials are examining the impact of exercise on individuals with dementia and are reporting promising findings.
For example, such as work by Intlekofer and Cotman (2012) suggests that evidence is starting to emerge that exercise supports brain health, even when initiated after the appearance of Alzheimer's disease pathology.
Exercise and cognitive functioning in persons with dementia.
The existing RCTs on exercise treatment for dementia are mixed with some demonstrating enhanced brain vitality (Farina et al., 2014;Hess et al., 2014), while others do not (Forbes et al., 2015). For example, Groot et al. (2015) performed a meta-analysis of RCTs investigating effects of physical activity on cognitive function in patients with dementia. The findings suggest that exercise interventions led to improved cognitive function in patients with dementia. This effect was found across various types of dementia and different exercise frequencies. Conversely, a Cochrane systematic review of exercise RCTs for dementia treatment failed to find exercise benefits in cognition, neuropsychiatric symptoms, and depression (Forbes et al., 2015). This study included 11 RCTs examining cognitive function. Given the growing research base and increasing 5 interests in lifestyle interventions for chronic disease prevention and treatment, it would be reasonable to expect that there are a number of additional RCTs from the last four years that should be accounted for in these collective analyses.
With a burgeoning scientific literature on exercise and dementia, further justified by the recent release of the WHO Guidelines on risk reduction of cognitive decline and dementia (2019), a mechanism is needed to objectively synthesize data across the accumulating research. Fortunately, systematic review and meta-analysis can operate as this mechanism.
Furthermore, while exercise interventions for the treatment of dementia are becoming more prevalent, evidence varies across some effects and outcomes, and the mechanisms of exercise for the treatment of dementia remain inconclusive. For example, why do certain exercise interventions demonstrate beneficial effects on cognitive function while others fail to do so? What are the common factors underlying these exercise interventions that are effective in treating dementia?
There is a scarcity of review and meta-analytic studies that try to answer these questions. Many meta-analyses are looking at more global cognitive effects of exercise on dementia and other chronic diseases. However, the field now needs to look at the variables (e.g. recruitment, duration, exercise type) producing differential effect sizes to examine where exercise is making a difference in cognitive outcomes in people with dementia. Furthermore, progress in policy is hindered by the absence of clarity on actions most likely to make exercise effective and feasible in a given context (e.g. health status).

How the intervention might work.
Exercise mechanisms.

6
Physical activity refers to "body movement that is produced by the contraction of skeletal muscles and that increases energy expenditure" (Chodzko-Zajko et al., 2009).
Exercise refers to "planned, structured, and repetitive movement to improve or maintain one or more components of physical fitness" (Chodzko-Zajko et al., 2009). Detailed exploration and explanations of the potential mechanisms of physical activity and exercise is beyond the scope of this paper. For additional information, the reader is directed to two recent reviews by Erickson, Weinstein, and Lopez (2012) and Davenport et al. (2012). Briefly, exercise improves vascular health by decreasing blood pressure, arterial stiffness, oxidative stress, inflammation, and strengthens endothelial function (Fleg, 2012;Ghisi et al., 2010), all of which are associated in the maintenance of cerebral perfusion (Davenport et al., 2012). Recent evidence has shown a strong association between cerebral perfusion (i.e. balance between the supply and demand of nutrients to the brain), cognitive function, and fitness in older healthy adults (Brown et al., 2010).
Glucose intolerance or insulin resistance is linked with the formation of amyloid plaque (Watson 2003), which is a notable feature of Alzheimer's Disease. Exercise may also keep neuronal structure intact, as well as promote neurogenesis, synaptogenesis, and capillarization (formation of nerve cells, the gaps between them, and blood vessels, respectively, which may be associated with exercise-induced increases in brain-derived neurotrophic factor (BDNF), and insulin-like growth factors (Colcombe et al., 2003).
Studies suggest that BDNF supports the health and growth of neurons and may regulate neuroplasticity (adaptability of the brain) through aging (Cheng, 2003). More recently, Intlekofer et al. (2012) reported that exercise reinforces hippocampal function, which is responsible for memory, by strengthening the expression of BDNF to promote 7 neurogenesis, the formation of blood vessels, and synaptic plasticity. In summary, animal and human studies indicate that exercise can facilitate a process counteracting the progressive memory loss in older age and Alzheimer's Disease (Erickson et al., 2012).

Justification for review and meta-analysis.
One in four American adults has multiple chronic conditions (e.g. depression, cancer, type 2 diabetes) that last one year or more and require ongoing medical attention  (Di Raimondo et al., 2016). For example, a multitude of meta-analyses and systematic reviews have examined exercise effects for the prevention and treatment of common chronic diseases, such as depression (Cooney et al., 2013), kidney disease (Heiwe & Jacobson, 2011), and type 2 diabetes (Thomas et al., 2006).
However, unlike other highly researched chronic diseases, much less is known about the relationship between exercise and dementia in regards to slowing cognitive decline. A more detailed exploration into the optimal "prescription" of exercise for cognitive health is needed for those suffering from dementia.

Current review.
Aim and scope.
This study aims to assess the effects of exercise interventions for the treatment of cognitive symptoms in dementia, one of the most pervasive and increasingly burdensome 8 chronic diseases. The study will compare the exercise effects across a variety of cognitive outcomes and exercise interventions among individuals with dementia. Evaluating the effectiveness of exercise interventions for the treatment of dementia may not only reveal if exercise is having an effect, but also reveal when exercise is having an effect. That is, research has yet to determine how exercise intensity, duration, frequency or mode influence exercise effects on cognitive outcomes. Findings may contribute in developing more effective and affordable evidence-based interventions for treating, managing and delaying the often devastating effects of dementia. Published meta-analyses in this area have often targeted specific modes of exercise (e.g., aerobic exercise), restricted their inclusion criteria to a single cognitive outcome (e.g., MMSE), or investigated participants with either mild cognitive impairment (MCI) or dementia (Farina, Rusted, & Tabet, 2014;Song et al., 2018;Young, et al., 2015;Zheng, et al., 2016), which tends to restrict their meta-analyses to a handful of studies rendering incomplete summaries of the available evidence on exercise effects on cognitive functioning specifically in individuals with dementia. Many reviews do not examine or account for exercise prescription variables, though a more thoughtful examining of these variables may help clinicians and policymakers establish and disseminate more thoughtful guidelines on the type, intensity, duration and frequency of exercise that is recommended for dementia patients.

Research objectives and hypotheses.
Objective 1) Examine the effects of exercise interventions on cognitive functioning in individuals with dementia. Exercise effects on cognitive functioning from 9 the included trials will be pooled across all trials to calculate a pooled mean effect size using a common effect size (SMD).
Hypothesis 1) Based on the existing literature exploring exercise for healthy older adults (Erickson et al., 2011;Northey et al.., 2018;Tseng, Gau, & Lou, 2011) and adults with MCI (e.g., Hess et al., 2014;Song, Yi, Li, & Lei, 2018) it is hypothesized that exercise interventions will slow the progression of cognitive decline relative to the cognitive decline in control groups.
Objective 2) Examine the effects of exercise interventions within specific cognitive assessment measures/instruments. Mean exercise effects will be calculated within each specific cognitive outcome measure represented in the included studies.
Hypothesis 2) It is hypothesized that exercise interventions will slow cognitive decline relative to control groups within each cognitive outcome measure.
Objective 3 Hypothesis 4) It is hypothesized that exercise interventions with greater frequency, intensity, and duration will be most effective in slowing cognitive decline.
Objective 5) Examine the impact of study design (e.g., dementia type, dementia severity at baseline).
Hypothesis 5) It is hypothesized that exercise interventions with participants with less severe types and forms of dementia will experience greater improvement (or slowed progression) in cognitive function.
Such analyses have the potential to determine which variables are likely to predict successful treatment and possibly help to further elucidate the mechanisms of exercise interventions. These findings may identify ways to tailor the exercise intervention delivery to maximize effects for different groups. Building upon existing findings from 1) meta-analyses exploring the role of exercise in preventing dementia in those without dementia, and 2) newer and novel RCTs exploring exercise effects in treating dementia, this study will examine the collective effectiveness of exercise interventions in the treatment of cognitive symptoms in individuals with dementia, while also identifying what type of exercise interventions are most effective in treating dementia.
To examine mechanisms and/or predictors of successful exercise interventions, RCT variables can be coded to determine covariates within each RCT and across the included RCTs. Meta-analysis can be used to analyze which intervention variables 11 influence effect size for a specific intervention and compare how intervention variables affect effect size. One such meta-analysis assessed mean effect sizes across health behavior RCTs and moderator variables, such as recruitment type (Noar, Benac, & Harris, 2007

Design
Meta-analysis and systematic review will be used to: 1) calculate the pooled effect of all included exercise interventions for the treatment of cognitive symptoms in individuals with dementia; 2) calculate the pooled exercise intervention effect within each cognitive outcome measure; 3) calculate the pooled exercise intervention effect within each domain of cognition; and 4) identify any exercise intervention characteristics and study design characteristics that are moderators affecting intervention effectiveness.

Types of studies.
Only studies examining the effects of an exercise intervention in the treatment/management of dementia will be included. Older people (60+) diagnosed with dementia must be allocated to either an exercise program or a control program (usual care or social contact/activities). Only studies providing pre-and post-intervention primary outcome measures will be included.
Cross-over trials were eligible for inclusion, but only data from the first treatment phase (prior to the cross-over) was considered for inclusion. Non-blinded trials were included, as it was unrealistic to expect blinding of the participants and those who conducted the exercise interventions (for further details, see Assessment of risk of bias in included studies). Outcome assessors were expected to be blinded to treatment allocation, however, studies without blinding of outcome assessors were not excluded. Studies were rated for blinding. 13 Types of participants.  (McKhann, 1984), ICD-10 (World Health Organization, 1992), or CERAD-K (Hwang, 2010).

Types of interventions.
In this review, RCTs in which older people diagnosed with dementia were allocated to either an exercise program or a control group (usual care, standard care, or social contact/activities). Searches will be limited to trails with physical activity interventions, defined by the American College of Sports Medicine as "body movement that is produced by the contraction of skeletal muscles and that increases energy expenditure" (Chodzko-Zajko et al., 2009).
Exercise refers to "planned, structured, and repetitive movement to improve or maintain one or more components of physical fitness" (Chodzko-Zajko et al., 2009).
Searches were limited to interventions lasting 2 weeks or more with the aim of improving cognitive or neuropsychiatric symptoms in older people with dementia. Included interventions were both supervised and unsupervised programs.
Trials will be included where the only difference between groups was the exercise intervention, and the types, frequencies, intensities, duration, and settings of the exercise programs were described. In order to isolate exercise treatment effects, an experimental arm had to involve an intervention that solely delivers exercise (in addition to The methodology for data extraction and analysis was based in the Cochrane Handbook of Systematic Reviews of Interventions (Higgins & Green, 2011). Once the full text of selected RCTs were obtained, information was extracted using the predefined data extraction form (Appendix B). This form summarized key information, including details of the participant population (e.g., severity of dementia), sample size, the intervention(s) including intervention details (e.g., exercise type, duration), the comparison group (e.g., standard care), and cognitive outcome measures and relevant data. If a study included multiple cognitive outcome measures, the form also included the study's "primary "cognitive outcome measure of interest if identified in the study. If studies included multiple measures of cognition, the author decided a priori to include the study's "primary" cognitive outcome in the main analysis across all studies. If a primary outcome of interest was not identified, the most commonly used measure was used to maximize comparability between studies.
The mean change from baseline to measurement at the conclusion of the intervention, and the number of participants for each group was extracted. Data is presented in a series of summary tables and figures.

Assessment of risk of bias in included studies.
The methodological quality of the evidence in included RCTs was assessed based on the Cochrane collaboration tool for assessing risk of bias and any summary of findings tables within them using a criteria list (Higgins & Green, 2011) (see Appendix C. 'Cochrane risk of bias tool for randomized controlled trials'). Criteria for judging risk of bias were based on the Cochrane Handbook for Systematic Reviews of Interventions (Higgins & Green, 2011). This assessment tool was used to determine whether there was a low, high, or unclear risk of bias for each factor (Higgins & Green, 2011). The identity of the publication and author information for each trial report was not masked. If the description of a process or outcome was unclear or missing, the original author of the trial was contacted in an attempt to retrieve the required information. The following criteria were assessed: 1. Selection bias -systematic differences between baseline characteristics of the groups being compared, including: i) random sequence generation; ii) allocation concealment.
2. Performance bias -systematic differences between groups in the care that is provided, or in exposure to factors other than the interventions of interest, this includes: i) Blinding of participants and personnel.
3. Detection bias -systematic differences between groups in how outcomes are determined, this includes: i) blinding of outcome assessments.
4. Attrition bias -systematic differences between groups in withdrawals from a study, this includes: i) incomplete outcome data.

Reporting bias -systematic differences between reported
and unreported findings, that is: i) outcome reporting bias.
6. Other bias (i.e., bias due to other problems) Of note, blinding participants and personnel (performance bias) is inherently a problem when conducting rehabilitation trials, such as exercise interventions. Blinding of participants and personnel to the exercise intervention is not possible due to the nature of exercise interventions (e.g., exercises, devices, manual therapy), blinding for physiotherapists and personnel and patients may be challenging to impossible. Blinding for health care providers, patients, and outcome assessors is less frequently reported in trials involving nonpharmacological interventions (Boutron, Tubach, Giraudeau, et al., 2003). Thus, a high risk of bias assessment on this domain was not weighed in for final judgment for each study's total risk of bias.

Measures of treatment effect.
Effect sizes standardize findings across RCTs for direct comparison. Effect sizes represent the magnitude and direction of the exercise intervention effects within and across RCTs. Summary statistics were required for each included trial and outcome. For continuous data, the mean difference (MD) was used when the pooled trials used the same rating scale or test to assess an outcome. Standardized mean difference (SMD) was calculated and reported to compare the effectiveness across different outcomes. Both effect sizes were calculated and provided when possible.
SMD was used when the pooled trials used different rating scales or tests for continuous outcomes that were conceptually the same but measured in different ways.
The particular definition of SMD used in Cochrane reviews and implemented in Review Manager 5.3 software is the effect size known in social science as Hedges' (adjusted) g, which is very similar to Cohen's d, but includes an adjustment for small sample bias Post-intervention effect sizes were collected from each study. SMD was computed using Review Manager 5.3 meta-analysis software using the statistical data provided, such as mean, standard deviation, F-, or t-tests statistics. Based on the magnitude of effect sizes found in population-based health RCTs, 0.15, 0.20 and 0.25 represented small, medium, and large effects (Rossi, 2013).
An SMD of zero means that the new treatment and the control have equivalent effects. If improvement is associated with higher scores on the outcome measure, SMDs greater than zero indicate the degree to which treatment is more efficacious than control, and SMDs less than zero indicate the degree to which treatment is less efficacious than control. If improvement is associated with lower scores on the outcome measure, SMDs lower than zero indicate the degree to which treatment is more efficacious than control and SMDs greater than zero show the degree to which treatment is less efficacious than control.
The inverse variance method was used in this meta-analysis. All outcomes were reported using 95% confidence intervals (CI). None of the trials included in the review reported dichotomous data of interest to this review.
Dealing with missing data.
Reasons for missing data were extracted from original studies. Many types of information could not be identified from the published articles, such as descriptions of the 20 process of randomization, blinding of outcome assessors, attrition and adherence to the exercise intervention, reasons for withdrawing, and statistical data (i.e., means, SDs).
Author contacts were emailed and asked to provide the missing data. The possible impact of the missing data on the results depended on the extent of missing data, the pooled estimate of the treatment effect, and the variability of the outcomes. The variation in the degree of missing data was also considered as a potential source of heterogeneity. If available, intention-to-treat (ITT) data was used to best account for noncompliance, protocol deviations, withdrawal, and anything that happens after randomization. If these data were not available, only the reported completers' data was used in the analyses.

Assessment of heterogeneity.
Homogeneity analysis was used to test the assumption that all of the effect sizes are estimating the same population mean. Clinical heterogeneity could be assessed by inspecting the type of participants, details of the interventions, and outcomes within each study. Trials demonstrating clinical homogeneity, such that they tested an exercise intervention and examined similar cognitive outcome measures, were considered as potentials for this meta-analysis. Heterogeneity was initially explored through visual exploration of the forest plots. A test for statistical heterogeneity (a consequence of clinical or methodological diversity, or both, among trials) using the Chi 2 test (with a P < 0.10 indicating significance) and I 2 analysis was then performed. The I 2 analysis (Cochrane's Q test) is a useful statistic for quantifying inconsistency (I 2 = [(Q -df )/Q] x 100%, where Q is the Chi 2 statistic and df is its degrees of freedom (Higgins & Thompson, 2002;Higgins et al., 2003). This describes the percentage of variability in effect estimates that is due to heterogeneity rather than sampling error (i.e., chance).

21
Values greater than 50% are considered to represent substantial heterogeneity. If there was evidence of heterogeneity of the population or treatment effect, or both, between trials, random-effects model was used, for which the confidence intervals are broader than those of a fixed-effect model (Higgins & Green, 2011). If the value was less than 30%, the overall estimate would be presented as a fixed-effect model. Funnel plots were used to examine heterogeneity of effect sizes.

Assessment of reporting biases.
Funnel plots were examined to look for indications of publication bias. To investigate reporting biases within the included studies, the outcomes listed in the methods sections were compared with reported results.

Data synthesis.
A random-effects model was used when the I 2 measure of heterogeneity was greater than 30% indicating significant diversity between studies in participants or Data is presented as a meta-analysis and systematic review with tabulated data of the statistical outcomes reported in the original RCTs. Comparisons were determined by the data in the original RCTs. Where possible, data was grouped by specific cognitive outcome measure (e.g. MMSE), and also by specific cognitive domain (e.g., attention).
Outcomes of interest were continuous. The PRISMA (Liberati et al., 2009) statement was used as a framework to guide the selection of RCTs for the meta-analysis and systematic review. Another list was constructed with excluded studies and reasons for exclusion.

Subgroup analyses and investigation of heterogeneity.
If there was sufficient data, it was decided a priori that the following subgroup analyses would be conducted to explore possible causes of heterogeneity. In accordance with Objective 4 of this review and meta-analysis, the characteristics of the exercise intervention were coded into categorical variables. These were coded according to published guidelines that reconcile differences in the terminology used to describe components of exercise (ACSM, 2013;Norton, Norton, & Sadgrove, 2010 Length of exercise intervention: 1. up to 12 weeks; 2. more than 12 weeks.
Due to the likelihood that the measures of exercise intensity would vary across RCTs, it was decided that exercise intensity would be coded based on the exercise intensity criteria used within each study. While this may interfere with best practices, we assume it will allow an exploration of an exercise prescription variable that has been cited as having particularly strong benefits in comparison to exercise duration and frequency. Moderator and subgroup effects.
These predetermined exercise intervention variables (type, frequency, session duration, volume, length, intensity) and participants' dementia severity and type were coded to determine predictors of improved (or slowed progression) cognitive function.
This coding was determined based on how common these outcomes and variables were represented within and across the included studies.

Sensitivity analysis.
Sensitivity analyses were also considered and used to explore possible causes of methodological heterogeneity, such as including studies at high risk of bias due to extreme effect sizes and small sample sizes that could distort the mean effect size or variance. Forest plots were used for outlier analysis and assessment of publication bias. Please see Appendix D ('Characteristics of included studies').
All trials required a diagnosis of dementia for recruitment. All trials required participants to be 60 years or older. Eggermont et al. (2009a) and Eggermont et al. (2009b) required participants to be 70 years or older.
The period of time that the exercise programs were offered varied greatly, ranging from six weeks Eggermont et al., 2009b) weeks , 18 weeks , six months Hernandez et al., 2010;Venturelli et al., 2011), and 12 months .

Control groups.
The control groups for 14 of the studies received usual care with no additional interventions Barnes et al., 2015;Christofoletti et al., 2008;Coehlo et al., 2012;Hernandez et al., 2010;Hoffman et al., 2016;Kemoun et al., 2010;Kwak et al., 2008;Lamb et al., 2018;Miu, Szeto, & Mak, 2018;Prick et al., 2017;Thurm et al., 2011;Venturelli et al., 2011;Vreugdenhil et al., 2012). The control group for six 34 studies included social contact Eggermont et al., 2009a;Eggermont et al., 2009b;Lee & Kim, 2018;Toots et al., 2017;Van de Winckel et al., 2004). The control group in one trial received psychoeducation about healthy lifestyle (Holthoff et al., 2016). research settings to measure cognitive impairment (Tombaugh & McIntyre, 1992). It is commonly used in medicine and allied health fields to screen for dementia. It is also used to estimate the severity and progression of cognitive impairment and to follow the course of cognitive changes over time, which makes it effective in documenting an individual's response to treatment. Questions relate to a variety of domains, including attention, language, word recall, and orientation to time and place. The MMSE is composed of questions grouped into seven categories, each one designed to evaluate specific cognitive functions: time orientation, place orientation, three-word register, attention and calculation, immediate and delayed recall of the three words, language, and visuoconstructive praxis (Folstein, Folstein, & Mchugh, 1975). Scores range from 0 to 30 points and lower values represent a possible cognitive decline.

Alzheimer's Disease Assessment Scale Cognitive subscale (ADAS-Cog) (Global cognitive function).
Following the MMSE, ADAS-Cognitive Score was used in a total of seven trials Hoffman et al., 2016;Lamb et al., 2018;Miu et al., 2008;Thurm et al., 2011;Toots et al., 2017;Vreugdenhil et al., 2011), The ADAS-Cog is one of the most commonly used primary outcome measures in dementia trials. The ADAS-Cog was designed to improve assessment of subtle changes in symptoms. The ADAS-Cog can both be used as an overall measure of cognitive functioning, and as a direct assessment of different cognitive domains, including learning (word list), naming (objects), following commands, constructional praxis (figure copying), ideational praxis (mailing a letter), orientation (person, time, plan), recognition memory and remembering test instructions (Rosen, Mohs, & Davis, 1984). The ADAS-Cog is a 70-point scale, with a higher score indicating greater impairment.

Category Verbal Fluency (CVFT) (Language). CVFT was represented in seven
trials Christofoletti et al., 2008;Eggermont et al., 2009a;Eggermont et al. 2009b;Hoffman et al., 2016;Prick et al., 2017;Toots et al., 2017). This is a subtest from the Groninger Intelligence Test (GIT;Snijders & Verhage, 1983) that assesses the language domain of cognitive function. In this test the participant is asked to name as many animals and professions within one (separate) minute (Lezak, Howieson, & Loring, 2004). The outcome measure is the total number of animals and professions produced (range 0 -infinity). including planning, abstraction, logical sequencing, and monitoring of the executive processing (Sunderland, Hill, & Mello, 1989).   Eggermont et al., 2009b;Prick et al., 2017). The 8 Words Test of the Amsterdam Dementia Screeningtest (ADS; Lindeboom & Jonker, 1989) is used to measure episodic anterograde memory and learning domain of cognitive function. During the first part, eight unrelated words are read aloud to the subject five times. Immediately after each presentation, recall is tested, with the total number of words recalled after the five trials being used as the score (immediate recall, score range 0-40). In addition, recall is again tested after a delay of 10 minutes (delayed recall, score range 0-8). Next, a recognition test is followed in which the eight words are intermixed with eight "distractor" words (recognition score, score range 0 -16).

Stroop Color and Word Test (SCWT) (Executive functioning). The Stroop Test
was used in two trials Hoffman et al., 2016). The Stroop Color and Word Test is a neuropsychological test extensively used to assess the ability to inhibit cognitive interference that occurs when the processing of a specific stimulus feature impedes the simultaneous processing of a second stimulus attribute (well-known as the "Stroop Effect") (Golden, 1978). It appeals to the executive functioning domain of cognition.

Cambridge Cognitive Examination (CAMCOG) (Global cognitive function). The
CAMCOG was used in one study ; The CAMCOG is a brief neuropsychological battery designed to assess global cognitive function by evaluating the range of cognitive functions required for a diagnosis of dementia, and to detect mild degrees of cognitive impairment (Roth et al., 1986).

Rapid Evaluation of Cognitive Function (ERFC) (Global cognitive function)). The
ERFC was used in one study . The test consists of 12 subtests: spatial orientation, attention span, immediate and deferred memory, mental calculation, reasoning and judgment, comprehension, denomination, repetition, a written order, verbal fluency, apraxia, visual decoding and writing. The maximum score is 50, and a score lower than 46 indicates a significant probability of cognitive deficit (Gil, et al., 1986). In this way, this test measures global cognitive function taken as a whole.

Loewenstein Occupational Therapy Cognitive Assessment for Geriatric
Population (LOTCA-G) (Global cognitive function). The LOTCA-G battery was used in one study (Lee & Kim, 2018). The LOTCA-G is a tool to assess cognitive functions that was developed for rehabilitation in a hospital in Israel in 1974 (Katz, Elazar, & Itzkovich, 1996). The LOTCA-G has subsequently been used widely in many countries, including the USA. It is for adults and has been adjusted for the elderly population. The LOTCA-G is an assessment of global cognitive function, and assesses six cognitive areas: orientation, visual perception, special perception, praxis, visuomotor organization, and thinking operation.
The FIM-Cog was used in one study (Lee & Kim, 2018). The FIM refers to the assessment tool to measure daily living operations, which consists of either tests of self-care, five tests of mobility, and five tests of communication and social cognition (Granger et al., 1990). The score criteria consist of seven-point scales from one, which is a total dependent operation, to seven, which is independent operation without the help of others. The higher the score, the more the patient can perform daily living operations independently. From the full FIM score, the FIM-Cognitive Scale was used to assess global cognition. The intra-rater reliability for the cognition area is 0.83.
Frontal Assessment Battery (FAB) (Executive functioning). The FAB was used in one study . The FAB is an assessment of frontal cognitive functions (executive functions) for patients with neurodegenerative disorders. The battery consists of six subtests: (i) similarities (abstract reasoning); (ii) lexical fluency (mental flexibility); (iii) series motor (motor programming); (iv) conflicting instructions (sensitivity to interference); (v) go-no go (inhibitory control); (vi) conflicting instructions (sensitivity to interference); and (vi) prehension behavior (primitive reflex). It varies on a scale of 0-18 points, and higher scores represent better performance in frontal functions (Beato, Nitrini, Formigoni, & Caramelli, 2007).
Symbol Search-Subtest (Attention). Symbol Search was used in one study (Coelho et al., 2012). The Wechsler Adult Intelligence Scale-III Symbol Search-Subtest assesses focused attention (Wechsler, 2004).

Symbol Digit Modalities (SDMT) (Attention). Symbol Digit Modalities was used
in one study . This measure assesses the cognitive domain of mental speed and attention (Smith, A., 1982). Using the number and symbol key at the top of the test page, participants are asked to correctly decode several lines of symbols.
The total number of correct de-codings in 120 seconds is used as the outcome.
2. Six were complex, multimodal interventions in which exercise was combined with additional non-exercise treatments or training so that groups did not differ in exposure to exercise alone (Bayer et al., 2017;Bossers et al., 2016;Pitkala et al., 2013;Schwenk et al., 2010;Tanaka et al., 2017;Yoon et al., 2013).
For example, one study explored the effects of acupuncture on cognitive function (Gu et al., 2014), while several others involved cognitive training and exercises (Yaguez et al., 2011;Yokoyama et al., 2015).

Risk of bias in included studies.
See 'Cochrane risk of bias tool for randomized controlled trials' (Appendix C), 'Characteristics of included studies' (Appendix D) and 'Risk of bias summary: review author's judgments about each risk of bias item for each included trial' (Figure 2).
Eight trials were judged to be at low risk of bias for this domain, as sufficient information about the way the allocation sequence was generated was available Cheng et al., 2014;Eggermont et al., 2009a;Hoffman et al., 2016;Lamb et al., 2018;Prick et al., 2017;Toots et al., 2017;Thurm et al., 2011;Van de Winckel et al., 2004;Vreugdenhil et al., 2012).

Blinding of participants and personnel (performance bias).
With the exception of one trial, trials were at high risk of performance bias, as blinding of participants and personnel to the intervention was not possible, due to the nature of rehabilitation trials. In the trial by Venturelli et al. (2011, members of the research team did not know to which group each participant had been assigned and one on the research team was present during the walking exercise.

Blinding of outcome assessment (detection bias).
It was not clear whether and how outcome assessments had been blinded in multiple trials Cheng et al., 2014;Coelho et al., 2012;Hernandez et al., 2010;Holthoff et al., 2014;Kwak et al., 2008;Lamb et al., 2018;Lee & Kim, 2018;Thurm et al., 2011;Venturelli et al., 2011). Van de Winckel et al. (2004) was rated as being at high risk for detection bias for cognition outcomes as "the physiotherapist who was conducting both treatments evaluated the patients on cognition. However, the nurses who scored the patients were all blind to the group assignment." Remaining trials were deemed low risk for detection bias since outcome assessors were blinded Eggermont et al., 2009a;Eggermont et al., 2009b;Vreugdenhil et al., 2012).

Incomplete outcome data (attrition bias).
Attrition rates (drop-outs from the trials) varied from 0% to 38.8% in the included trials. The drop-out rates were higher in the experimental arms for Christofoletti et al.,  (2018), Kwak et al. (2008), Vreugdenhil et al. (2011) had 0% attrition in both experimental and control arms. Reasons for attrition were provided, including: death, illness, increased disability, disinterest, physician's disapproval, family withdrawal of consent, moving, and refusal to continue to participate.
In summary, the majority of the trials were found to be at low risk of attrition bias. Several trials had unclear risk A high risk of attrition bias was reported for five of the included studies for a variety of reasons that included: failure to report attrition rates for individual groups; a high attrition rate; or an imbalance of attrition between the groups, or failure to provide reasons for attrition, or both Kemoun et al., 2010;Miu, Szeto, & Mak, 2008;Toots et al., 2017); see Appendix D.
No trials used ITT principles of analysis to estimate missing data.  and Eggermont et al. (2009b)  participants, but only 61 were included in the ITT analysis. The reported completers' data from the included studies was used in the analyses. Thus, there was a potential risk of attrition bias in these studies.

Selective reporting (reporting bias).
With the exception of Cheng et al. (2014), all included trials were deemed to be at low risk of reporting bias.
Other potential sources of bias.
See Figure 2. 'Risk of bias summary: review author's judgments about each risk of bias item for each trial.'

Main analysis: Pooled exercise intervention effects on overall cognitive function.
See: 'Summary of findings for the main comparison: Exercise programs for dementia' (Table 1) and 'Forest plot: Exercise vs usual care: Cognition' (Figure 3).

Primary outcome -Cognition.
Each of the 21 studies in this meta-analysis included at least one cognitive outcome. Data from these 21 studies were included in the meta-analysis of the overall pooled exercise effect on cognition. Both pre-and post-intervention cognitive measures were required to be included in the meta-analyses of overall cognition (Objective 1), as well as meta-analyses by specific cognitive measure (Objective 2) and cognitive domain (Objective 3), as applicable to outcomes available in each study. If a study included multiple outcome measures of cognition, the outcome of primary interest ("primary outcome") identified in that study was used in the overall pooled mean of cognitive function across all studies. This was in accordance with the cognitive outcome criteria determined a priori and outlined in Chapter 2. Pre-and post-intervention measures following six weeks to 12 months of exercise intervention were included.
The author assumed that the effect sizes of individual studies would be comparable but not identical across studies due to expected heterogeneity across studies (different intervention types and outcomes of cognitive function), and therefore a random-effects model was used. The estimated SMD between exercise and control groups was 0.49 (95% CI [0.24 -0.75], P = 0.0002, 21 studies; Figure 3. 'Forest plot: and risk of bias (see Table 1. 'Summary of findings for the main comparison: Exercise programs for dementia').
Further exploration revealed very large effect sizes for two studies with very small sample sizes Venturelli et al., 2011). Asymmetrical funnel plots (Egger et al., 1997)  Cognition -19 studies), representing a large size treatment effect although slightly smaller than the treatment effect calculated from all 21 studies. This reduced the heterogeneity slightly (I 2 = 74%), though it remained substantial.
Heterogeneity was further investigated in subgroup analyses.

Main analysis: Pooled exercise effects by cognitive measure.
In accordance with Objective 2 of this meta-analytic review, exercise effects were then assessed within each measure of cognitive function that was represented in the included studies using MD. SMD was also reported to compare intervention effectiveness across different outcomes. Measures represented by at least two studies were analyzed.
Measures that were only used by a single study were not analyzed for MDs.

Mini-Mental State Examination (MMSE) (Global cognitive function). The data
from the MMSE showed that exercise was associated with improved scores compared to Venturelli et al., 2011Thurm et al., 2011

Main analysis: Pooled exercise intervention effects by cognitive domain.
In accordance with Objective 3 of this meta-analytic review, exercise effects within different domains of cognition were examined by pooling means of measures assessing the respective cognitive domains. This was also intended to target anticipated heterogeneity across studies.

Attention.
Three studies used measures of attention in assessing cognitive function (Coelho

Memory and Learning.
Six studies used measures within the memory and learning cognitive domain Christofoletti et al., 2008;Eggermont et al., 2009a;Lee & Kim, 2018;Prick et al., 2017). Effect of exercise on cognition for the memory and learning domain was not statistically significant; SMD = -0.28 (95% CI

Secondary analyses: Subgroup analyses.
In accordance with Objective 4 and Objective 5 of this meta-analytic review, prespecified exercise intervention variables and study design variables were examined as potential causes of heterogeneity. Potential reasons for high heterogeneity were explored by conducting meta-analyses that included only trials with: 1) aerobic-only exercise interventions; 2) exercise interventions using multiple, combined types of exercise (e.g., strength and aerobic exercises, balance and aerobic exercises);  10 exercise interventions for more than 24 weeks; 11) exercise programs less than 24 weeks; 12) participants with mild-to-moderate dementia severity; 13) participants diagnosed with AD only.
None of these analyses reduced the heterogeneity below 65% (moderate range).

Study design moderators.
Dementia severity. A majority of the trials included participants with "mild to moderate" dementia, which revealed significant random effects on cognitive performance; SMD = 0.36 (95% CI [0.12 -0.61], P = 0.004, I 2 = 76%; Figure 6.12. Cognition: Dementia severity -mild and moderate). It is important to note that despite multiple significant subgroup findings, the substantial heterogeneity (I 2 > 70%) remained unresolved.

Final summary.
A summary of findings from all analyses can be found in Table 2   This study conducted the most comprehensive systematic review and metaanalysis of RCTs exploring exercise intervention effects on cognitive functioning in adults >60 years of age with dementia to date. Importantly, it did not limit inclusion of studies by specific exercise type, publication date, or cognitive measure. The study also incorporated a multilevel meta-analysis method that included exploration of subgroups and moderator variables. The key finding from this study is that exercise interventions are Test (CDT) both demonstrated significant exercise treatment effects on cognition.
Objective 3. No evidence of exercise effects was found within the six specific cognitive domains that the review was intending to assess (attention, executive functioning, memory and learning, language, social cognition, visuospacial/perceptual-motor function). weeks. Subgroup analysis for aerobic-only type of exercise was not significant.
Objective 5. RCTs' specific type of dementia and severity of dementia at baseline were also variables coded for subgroup analyses. Subgroup analyses combining "mild-tomoderate" dementia severities revealed significant exercise effects on cognitive function (RCTs' pooled mean outcome measures of cognitive function). Similarly, subgroup analyses were performed using trials composed of Alzheimer's disease patients only (a majority of the trials), and this also revealed a significant exercise effect on pooled cognitive outcome measures. Given that Alzheimer's disease is the most common type of dementia, these results are promising. Given the exercise effects on cognitive function for mild-to-moderate dementia severity, this is also promising as individuals with less severe dementia may be more independent and capable of initiating, establishing, and maintaining a regular exercise program.

Overall completeness and applicability of evidence.
Clearly, additional research is needed that examines these important outcomes and provides the data needed for meta-analysis. Eight studies were based in the community/homes (Coehlo et  The participants within the trials were not homogeneous in terms of their diagnosis (e.g., AD, vascular dementia, mixed dementia, other) though most were qualified as having "moderate" severity dementia. As dementia is not a singular disease entity, and there is some evidence that exercise might affect the risk of these conditions in different ways (Rockwood 2007).
Also, the exercise programs were not homogeneous in terms of the type (e.g. aerobic, strength, balance, combined), duration (range: 8 weeks to 12 months), frequency (range: two times per week to daily) of activities, and intensity (low to high). Therefore, type, duration (less than 12 weeks versus longer than 12 weeks), frequency (less than three times per week versus more than three times per week), and intensity of the exercise programs were compared in further subgroup analyses. However, the examination of moderators depended on the variables available in the studies selected, which is an inherent limitation of meta-analysis.

This review was conducted as outlined in the Cochrane Handbook for Systematic
Reviews of Interventions (Higgins & Green, 2011), therefore, the introduction of bias during the review process was minimized. However, not all of the included trials reported data that could be used in the meta-analysis (e.g., data pertinent to risk of bias assessment, exercise intensity), and most authors did not respond to requests for this data.
This meant that the certain variables from these trials could not be included in the metaanalyses. This was unfortunate as the total number of trials that have examined the evidence of the benefit or lack of benefit of exercise programs in improving the symptoms of dementia is limited.
The limited number of studies included in this meta-analytic review resulted in even fewer variables in the subgroup analyses, as many of the included studies were missing variables of interest to this review (e.g., exercise intensity, exercise adherence, attrition). The Cochrane Handbook for Systematic Reviews of Interventions version 5.20 (2017) notes the importance of ensuring that there are adequate studies to justify subgroup analyses. The typical advice for undertaking simple regression analyses: that at least 10 observations (i.e. ten studies in the meta-analysis) should be available for each characteristic, and that even this will be too few when the covariates are unevenly distributed. The final number of studies included stands as a considerable weakness.
However, this could not be fully anticipated or controlled for by the author.
As studies were included only if exercise was the sole intervention, a large number of studies were excluded that used exercise as an adjunct component to another intervention (e.g., combined cognitive and exercise program) were excluded. It is 84 unfortunate such studies did not have exercise-only intervention arms, as this would significantly have increased the number of included studies, and therefore strengthened the generalizability of our findings and allowed more robust subgroup and moderator analyses.
Another potential source of study bias involves the small sample sizes in half of the included studies (10 studies had fewer than 30 participants). Furthermore, two studies with very small sample sizes (under 20 participants) and very large treatment effects could likely be contributing to the large heterogeneity of effect sizes found in the analyses. If such small studies had not found significant effect sizes, it is likely that they would not have been published due to such small sample sizes. This contributes to inherent publication bias, such that studies with small sample sizes are only available if they also found a larger effect size in comparison to the population effect size, which secures effect size heterogeneity.
The constant of heterogeneity was evident in the included studies. Heterogeneity was present across a variety of variables involving participant variables, exercise intervention variables, and cognitive outcome measures. However, this weakness of heterogeneity could not be fully anticipated or controlled for by the author.
As the author was the only coder, interrater reliability could not be assessed to ensure consistency and clarity at the title/abstract screening, full text screening, and data extraction stages. To ensure rating consistency, the author blindly re-rated and re-coded several studies, as well as comparing several studies' risk of bias ratings to other authors' ratings of the same studies in other published reviews. However, to help ensure that these 85 judgements are reproducible, it is desirable for more than one author to repeat parts of the process at various stages in a systematic review to enhance quality appraisal.
While this meta-analysis did not find any exercise effects within any of the Lastly, exercise influences cognitive function through multiple mechanisms.
Without direct measurements of these pathways, we cannot discern from the current meta-analyses which mechanisms underlie these findings from individuals diagnosed with dementia.

Agreements and disagreements with other reviews.
The primary result of this study aligns with the findings from a recent metaanalysis by Panza et al. (2018), which examined nineteen studies and found a favorable effect of exercise on cognitive function (SMD = 0.47, 95% CI [0.26 -0.68], P = 0.00).
The exercise effect in this study (SMD = 0.49, 95% CI [0.24 -0.75], P = 0.0002) is similar in magnitude to Panza et al.'s (2018) findings. However, Panza et al. (2018) included participants with and without AD (at risk of AD), while this study included only those with existing dementia diagnoses. In contrast to the findings of this study supporting exercise effects in combined-type (multimodal) but not aerobic-only exercise interventions, Panza et al. (2018) found aerobic exercise had the most favorable effect.
The findings from this study are of importance as they include new and additional relevant studies, multiple exercise intervention variables, a multilevel analysis examining outcomes by cognitive measure and domain, and are specific to those diagnosed with dementia. The findings of this meta-analysis are in contrast to a Cochrane Review by Forbes et al. (2015), which included nine trials examining exercise intervention effects on cognitive functioning and neuropsychiatric symptoms. Forbes et al. (2015) found no clear evidence of benefit from exercise (SMD = 0.43, 95% CI [-0.05 -0.92], P =0.08).

Implications for practice.
The findings of this meta-analysis and systematic review could have implications contributing to future policy, practice, and research. The results reveal the collected findings of the best evidence available on exercise for the treatment or management of dementia, a debilitating major chronic disease. They may help to eventually inform clinicians and policymakers of the exercise conditions for which exercise provides an improvement in cognitive function, and those for which there is no clear benefit or where evidence of benefit is lacking. By taking a step toward quantifying the benefits of exercise in treating one of the most common deadly and costly chronic diseases, healthcare providers and policymakers may be more willing or likely to take action on implementing more accessible, feasible, and cost-effective exercise interventions delivered on a population level. Barriers to exercise interventions and implementation in practice may also be discussed. Further elucidating the influence of these variables could help in tailoring exercise programs for the maximal patient benefit.

Moreover, dissemination and communication with the public can increase
awareness about the potential benefits of exercise for those with dementia. Family caregivers and providers may not be aware of these findings, or these findings may challenge existing beliefs about older adults with dementia and exercise (e.g., exercise is unsafe, exercise is not beneficial at this age). The promising evidence that exercise improves cognition may be added to exercise's synergistic impact on well-being. That is, exercise has been shown to prevent, delay or treat other chronic diseases, as well as increasing the likelihood of changing and engaging in other health behaviors (e.g., healthy diet, smoking cessation).

Implications for research and future directions.
The overview and results help to identify specific considerations that should be taken into account in future studies that contribute to a reliable basis for clinical application. Future studies should define, control for, and clearly report variables such as: OR "brain development" OR "cognitive performance" OR "cognitive function" OR "cognitive functioning" OR "information retrieval" OR "information processing" OR "perceptual skills" OR "intelligence quotient") Running OR cycling OR jogging OR walking OR bicycling OR bicycling OR treadmill OR treadmills OR ergometer OR ergometers OR "resistance training" OR "weight lifting" OR "weightlifting" OR "strength training" OR "workout" OR "workouts" OR dancing OR dance OR swim OR swimming OR yoga OR "tai chi" OR tai chi OR "t'ai chi" OR taijiquan OR qigong OR dancing OR "resistance training" OR "weight lifting" OR "strength training" OR sport OR sports AND (in title) intelligence OR "IQ" OR cognition OR cognitive OR brain OR memory OR attention OR attentiveness OR concentration OR concentrate OR learn OR learning OR cognitive OR attentiveness OR concentration OR concentrate OR learn OR learning OR neurocognition OR neurocognitive OR neuro-cognition OR neurocognitive OR "executive function" OR "executive functions" OR "executive functioning OR "problem solving" OR "brain development" OR "cognitive performance" OR "cognitive function" OR "cognitive functioning" OR "information retrieval" OR "information processing" OR "perceptual skills" OR "intelligence quotient"

AND
Clinical AND trial OR random* NOT cancer OR neoplasm* OR hypertensi* OR "high blood pressure" OR diabetes OR diabetic* OR HIV OR "acquired immunodeficiency syndrome" OR "cerebral palsy" OR parkinson's OR parkinson Running OR cycling OR jogging OR walking OR bicycling OR bicycling OR treadmill OR treadmills OR ergometer OR ergometers OR "resistance training" OR "weight lifting" OR "weightlifting" OR "strength training" OR "workout" OR "workouts" OR dancing OR dance OR swim OR swimming OR yoga OR "tai chi" OR tai chi OR "t'ai chi" OR taijiquan OR qigong OR dancing OR "resistance training" OR "weight lifting" OR "strength training" OR sport OR sports AND (in title) intelligence OR "IQ" OR cognition OR cognitive OR brain OR memory OR attention OR attentiveness OR concentration OR concentrate OR learn OR learning OR cognitive OR attentiveness OR concentration OR concentrate OR learn OR learning OR neurocognition OR neurocognitive OR neuro-cognition OR neurocognitive OR "executive function" OR "executive functions" OR "executive functioning OR "problem solving" OR "brain development" OR "cognitive performance" OR "cognitive function" OR "cognitive functioning" OR "information retrieval" OR "information processing" OR "perceptual skills" OR "intelligence quotient"

AND
Clinical AND trial OR random* NOT cancer OR neoplasm* OR hypertensi* OR "high blood pressure" OR diabetes OR diabetic* OR HIV OR "acquired immunodeficiency syndrome" OR "cerebral palsy" OR parkinson's OR parkinson

PsycARTICLES
Search Date: April 30, 2018 Search Dates: Date of inception -April 30, 2018 Returned Hits: 6 (in title) dementia OR Alzheimer OR alzheimers OR alzheimer's OR alzheimer AND (in title) exercise OR "physical activity" OR Running OR cycling OR jogging OR walking OR bicycling OR bicycling OR treadmill OR treadmills OR ergometer OR ergometers OR "resistance training" OR "weight lifting" OR "weightlifting" OR "strength training" OR "workout" OR "workouts" OR dancing OR dance OR swim OR swimming OR yoga OR "tai chi" OR tai chi OR "t'ai chi" OR taijiquan OR qigong OR dancing OR "resistance training" OR "weight lifting" OR "strength training" OR sport OR sports AND (in title) intelligence OR "IQ" OR cognition OR cognitive OR brain OR memory OR attention OR attentiveness OR concentration OR concentrate OR learn OR learning OR cognitive OR attentiveness OR concentration OR concentrate OR learn OR learning OR neurocognition OR neurocognitive OR "neuro-cognition" OR "neurocognitive" OR "executive function" OR "executive functions" OR "executive functioning" OR "problem solving" OR "brain development" OR "cognitive performance" OR "cognitive function" OR "cognitive functioning" OR "information retrieval" OR "information processing" OR "perceptual skills" OR "intelligence quotient"

AND
Clinical AND trial OR random* NOT cancer OR neoplasm* OR hypertensi* OR "high blood pressure" OR diabetes OR diabetic* OR HIV OR "acquired immunodeficiency syndrome" OR "cerebral palsy" OR parkinson's OR parkinson Running OR cycling OR jogging OR walking OR bicycling OR bicycling OR treadmill OR treadmills OR ergometer OR ergometers OR "resistance training" OR "weight lifting" OR "weightlifting" OR "strength training" OR "workout" OR "workouts" OR dancing OR dance OR swim OR swimming OR yoga OR "tai chi" OR tai chi OR "t'ai chi" OR taijiquan OR qigong OR dancing OR "resistance training" OR "weight lifting" OR "strength training" OR sport OR sports AND (in title, abstract, keyword) intelligence OR "IQ" OR cognition OR cognitive OR brain OR memory OR attention OR attentiveness OR concentration OR concentrate OR learn OR learning OR cognitive OR attentiveness OR concentration OR concentrate OR learn OR learning OR neurocognition OR neurocognitive OR "neuro-cognition" OR "neurocognitive" OR "executive function" OR "executive functions" OR "executive functioning" OR "problem solving" OR "brain development" OR "cognitive performance" OR "cognitive function" OR "cognitive functioning" OR "information retrieval" OR "information processing" OR "perceptual skills" OR "intelligence quotient"

AND
Clinical AND trial OR random* NOT cancer OR neoplasm* OR hypertensi* OR "high blood pressure" OR diabetes OR diabetic* OR HIV OR "acquired immunodeficiency syndrome" OR "cerebral palsy" OR parkinson's OR parkinson

Characteristics of included studies
(-) = unknown, AD = Alzheimer's disease, MD = mixed dementia, VaD = vascular dementia, UD = undefined dementia, MID = Multiple Infarct Dementia, NINCDS-ADRDA = national institute of neurological and communicative disorder and stroke & Alzheimer's diases and related disorders association, ICD-10 = international classification of diseases and related health problems, DSM=diagnostic and statistical manual of mental disorders.

Support for judgment
Random sequence generation (selection bias) Unclear risk Randomization process not described. Arcoverde emailed on April 30, 2019, no response.

Allocation concealment (selection bias)
Low risk Randomized with blind design by researcher who did not participate in initial assessments. Blinding (performance bias and detection bias) All outcomes

High risk
Not possible to blind participants and the personnel to the intervention allocated.
Blinding of outcome assessment (detection bias) All outcomes

Unclear risk
Blinding of outcome assessors not described.
Incomplete outcome data (attrition bias) All outcomes

Support for judgment
Random sequence generation (selection bias) Low risk "Residents recruited from 9 nursing homes were randomized by home" Allocation concealment (selection bias) High risk "Cluster design deemed necessary to avoid treatment contamination within homes" Blinding (performance bias and detection bias) All outcomes

High risk
Not possible to blind participants and the personnel to the intervention allocated ("open-label design was inevitable because activities could not be masked and it was not possible to prevent residences from talking to interviewers about the activities") Blinding of outcome assessment (detection bias) All outcomes

Unclear risk
Blinding of the outcome assessors not described.
Incomplete outcome data (attrition bias) All outcomes Unclear risk "Few attritions over time" Selective reporting (reporting bias) High risk Attrition not reported.
Other bias Unclear risk "Residents recruited from 9 nursing homes were randomized by home"

6-month RCT Participants
Country Inclusion criteria: "primary diagnosis of dementia" using ICD-10 criteria and confirmed by MMSE and Katz ADL score, medically fit for participation in intervention, resident of psychiatric institution. Exclusion criteria: cognitive impairment associated with other neuropsychiatric conditions or neurological diagnosis; antidepressant prescriptions with sedative or anticholinergic actions; impairment of cognition or balance related to drugs Interventions Experimental Group: physiotherapy kinesio-therapeutic exercises-strength, balance, memory, and recognition exercise using balls, elastic ribbons, and proprioceptive plates--provided by

Bias
Authors' judgement

Support for judgment
Random sequence generation (selection bias) Unclear risk Unclear process of randomization: (quote) "A sealed envelope with an identification number was assigned to each subject, each one filled with a slip giving the group. When a patient was registered and given a number, the appropriate envelope was opened" Allocation concealment (selection bias)

Low risk
Used sealed envelop, though did not specify if envelops were opaque Blinding (performance bias and detection bias) All outcomes

High risk
Not possible to blind participants and personnel to the intervention allocated: "As a common bias presented on most rehabilitation trials, it was not possible to 'blind' the subjects regarding the treatments" Blinding of outcome assessment (detection bias) All outcomes

Low risk
Blinded outcome assessors. Attrition: 0% Exercise program adherence: --Therapy for Controls: Standard care (kept to their same daily routine and did not participate in any regular or structured exercise programs) Outcomes

Support for judgment
Random sequence generation (selection bias) Low risk Randomized in blocks of 4-10 per participating center, using a computerized random-number generator. Allocation concealment (selection bias)

Low risk
Assessors blinded to group assignment throughout the study period completed the baseline assessments. Blinding (performance bias and detection bias) All outcomes

High risk
Not possible to blind participants and the personnel to the intervention allocated.
Blinding of outcome assessment (detection bias) All outcomes Low risk Raters performing the outcome measurements were blinded to group assignment, and patients and caregivers were advised not to disclose group assignment during the test sessions. Incomplete outcome data (attrition bias) All outcomes Low risk 4% total attrition. 76% of participants in intervention group attended more than 80% of exercise sessions, 78% exercised with intensity of more than 70% of maximal heart rate, 62% fulfilled both criteria ("high exercise subjects") Selective reporting (reporting bias)

Bias
Authors' judgement Support for judgment

Random sequence generation (selection bias)
Unclear risk Unclear process of randomization ("in the memory clinic of a regional hospital . . . consecutive patients with a diagnosis of dementia according to the DSM-IV-TR criteria were randomized into one of two groups") Allocation concealment (selection bias)

Unclear risk
Methods to conceal allocation not described Blinding (performance bias and detection bias) All outcomes High risk Not possible to blind participants and personnel to the intervention allocated Blinding of outcome assessment (detection bias) All outcomes Low risk Physiotherapist and occupational therapist who were blinded to allocation groups conducted the assessments at baseline and post-intervention.

Bias
Authors' judgement

Support for judgment
Random sequence generation (selection bias) Low risk Randomly assigned by blocked randomization (block size 20) to intervention or comparison group. Allocation concealment (selection bias)

Low risk
The allocation schedule was made by an independent researcher with a computer-generated block randomization using random Allocation Software (Version 1).

124
Blinding (performance bias and detection bias) All outcomes High risk Not possible to blind participants and the personnel to the intervention allocated.
Blinding of outcome assessment (detection bias) All outcomes Inclusion criteria: older adults with a record of dementia who were physically frail but cognitively and physically eligible for participating in the neuropsychological and physical examinations as well as in the physical movement training. Exclusion criteria: No indication of dementia (MMSE > 24, DSM-IV-TR criteria), salient behavioral problems or lack of minimally sufficient daily functioning, severe sensory impairments, absence of or severe impairments in written or spoken German and lack of minimal physical eligibility.

Interventions
Experimental Group: Type: Combined (physical exercise mainly conducted in a seated position but gradually increased in level of difficulty and complexity; the training combined strengthening, coordination, balance, flexibility, stamina) Frequency (sessions/week): 2 Duration (min/session): 45 min Volume (total min/week): 90 min Time period: 10 weeks Intensity: "moderate" Attrition: 21.2% Exercise program adherence: --("high" but not quantitatively specified) Control group: standard care Outcomes

Bias
Authors' judgement Support for judgment

Random sequence generation (selection bias)
Low risk "Group assignment randomly assigned by residency" Allocation concealment (selection bias) High risk "Randomized allocation to groups was not possible as participants were not able to visit the other building for training because of their physical frailty" Blinding (performance bias and detection bias) All outcomes

High risk
Not possible to blind participants and personnel to the intervention allocated Blinding of outcome assessment (detection bias) All outcomes Inclusion criteria: MMSE score≥10, a dementia diagnosis, age≥65 y, dependent on assistance in≥1 personal activity of daily living according to the Katz Index [29], ability to stand up from a chair with armrests with assistance from≤1 person, physician's approval, and ability to hear and understand spoken Swedish sufficiently to participate in assessments. All individuals included in the study gave informed oral consent to participation, which was confirmed by their next of kin.

Bias
Authors' judgement

Support for judgment
Random sequence generation (selection bias) Low risk Drawing lots "Clusters (N=36) of 3-8 participants each (that lived in same wing, unit, or floor) were formed. Randomization was stratified in all nursing homes except one, which had single cluster (aimed to have participants in both groups in each nursing home and reduce risk of factors associated with site to influence the outcome" Allocation concealment (selection bias)

Low risk
Participants were randomized after completion of enrollment process and baseline assessment to ensure concealed allocation. "Researchers not involved in study performed randomization drawing lots using sealed opaque envelopes." Blinding (performance bias and detection bias) All outcomes

High risk
Not possible to blind participants and the personnel to the intervention allocated Blinding of outcome assessment (detection bias) All outcomes

Bias
Authors' judgement Support for judgment

Random sequence generation (selection bias)
Unclear risk Method of random selection not described Allocation concealment (selection bias) Low risk Quote: "The head nurse of the ACU (not involved in the residents assessments) did the participants' randomization using StatsPlus for Macintosh" Blinding (performance bias and detection bias) All outcomes

Low risk
Quote: " . . . members of the research team did not know to which group each participant had been assigned . . . No one on the research team was present during the walking exercise" Blinding of outcome assessment (detection bias) All outcomes Unclear risk Quote: "Evaluation was done before and after the experiment period in a blind way"-not described well Incomplete outcome data (attrition bias) All outcomes Inclusion criteria: diagnosed with dementia using DSM-IV criteria; diagnosed with AD with NINCDS-ARDRA criteria; from outpatient memory disorder clinic; community dwelling with live-in care provider/caregiver who could visit daily. Exclusion criteria: physical condition that could prevent participation; evidence of neurodegenerative disorder (other than AD); already in exercise program more than once a week (resistance or aerobic training); started dementia medications in last 3 months.

Interventions
Experimental Group: 10 simple exercises and 30 minutes of brisk walking; home-based exercises--progressively became more challenging, and targeted strength and balance--and brisk walking under supervision of carer

Risk of bias
Bias