Physical Activity Interventions and Their Effects on Cognitive Function in People with Dementia: A Systematic Review and Meta-Analysis

Background: Physical activity (PA) has emerged as an alternative nonpharmacological approach to effectively address the effects of dementia. The primary aim was to identify and summarize PA interventions and their effects on cognitive function among persons with dementia (PwD). Methods: A systematic review was conducted with a meta-analysis using different electronic databases, such as PubMed, Embase, APA PsycNET, and the Web of Science. The identified and selected studies were randomized controlled trials (RCTs) that were written in English, published between 2000 and 2020, and implemented among PwD who received a PA intervention and whose cognitive function was measured at baseline and during a follow-up. Results: Twenty-two PA intervention studies met the eligibility criteria and showed a medium-size effect on the cognitive function of PwD, 0.4803 (95% CI = 0.1901–0.7704), with a high percentage of heterogeneity (I2 = 86%, p ≤ 0.0001). Moreover, this review complements other reviews by including eight studies that have not previously been considered. Overall, studies have methodological limitations. However, six studies implemented in the past five years have shown more robust methodological designs, including larger sample sizes and more comprehensive measurement tools. Conclusion: It is not yet possible to draw a conclusion on the ideal PA intervention for this population due to the high proportion of heterogeneity within the included studies. More emphasis is needed on the intensity of PA monitoring and adherence to such programs.


Introduction
Recent findings indicate that the population has been rapidly ageing during the last century due to improvements in health care, increase in life expectancy, and decrease in fertility rates [1]. As people age, body organs, tissues, and cells undergo change. Histological studies have shown that ageing affects the central nervous system (CNS) since it experiences neuroanatomical alterations, including an overall reduction in brain activity [2,3]. Therefore, changes and damage in the CNS are worrisome, due to its decisive role in controlling and coordinating essential functions of the body, including cognitive functions [4]. The physiological characteristics of dementia, an umbrella term for multiple neurodegenerative diseases [5], has been linked to the severe degeneration of brain cells and synapses in certain areas of the CNS, including the temporal, parietal and frontal cortices [6]. Damage in these areas manifests itself through memory and learning deficits [6]. In addition, dementia affects emotional regulation, social functioning, and activities of daily living [5]. According to the World Health Organization [7], there are 47 million people with dementia worldwide today, and it is expected that by 2030 this number will rise to 75 million and in 2050 to 135 million. Considering dementia's impact, researchers have concentrated efforts to minimize the burden associated with this disease by studying dementia risk factors and evidence-based dementia prevention and treatments [8].
The causes of dementia onset are not fully understood, but notably, the mechanism underlying dementia is associated with abnormal protein deposits that coexist with neurovasculature at different stages of the disease, which affect the functioning of the brain [9]. Depending on the type of dementia, different protein accumulations are observed. For instance, alpha-synuclein protein is linked to Lewy body dementia, whereas beta-amyloid and tau proteins are both related to Alzheimer's disease (AD), the most common form of dementia. Inadequate blood flow can lead to vascular dementia [9]. Other non-modifiable factors linked to dementia include age, sex, inflammation, and comorbidity, and genetic, environmental, and lifestyle factors [10]. Particularly in recent years, substantial epidemiological studies have provided evidence for lifestyle-related risk factors that trigger the development of dementia [11][12][13]. In light of this, the Lancet Commission presented a model describing nine modifiable risk factors (e.g., physical inactivity) that may contribute as much as 35% to the risk of dementia across the lifespan. Thus, by modifying these risk factors, one has a higher chance of preventing or delaying dementia progression [14].
In particular, PA during midlife and late life has been considered a cognitive reserveenhancing factor associated with a decreased risk of developing dementia [11,12,14]. This is mainly because regular PA improves the strength of cells and tissues to respond to oxidative stress, vascularization, and energy metabolism and also allows neurotropic effects through neurotrophic factor (BDNF) concentrations, which contribute to brain plasticity, memory improvement, neurogenesis, and synaptic plasticity [15]. These processes attenuate for the loss of brain tissue while the brain is ageing [14]. Thus, PA is linked with the concept of increased cognitive reserve, which indicates the brain's resilience. Persons who present this condition are more likely to cope with nervous system tissue damage without cognitive degeneration [14]. Moreover, the positive effects of PA on cognition appear to be influenced by preventing cardiovascular risk factors (e.g., obesity, hypertension, diabetes) which, at the same time, are linked with greater probability of dementia progression [16]. Additionally, neuroimaging methods add further evidence of the impact of PA on brain activity and cognitive function [16]. For instance, an enlarged level of connection was detected between the default mode network (DMN), which is a control structure widely known to be responsible for introspection and memory retrieval, after PA training [17]. Precisely, animal models of Alzheimer's disease (AD) illustrate that PA is an effective way to positively modify pathophysiological processes, including β-amyloid (Aβ) burden, tau phosphorylation, and neuronal loss [18].
In this way, PA plays a crucial role in the healthcare system. Including preventive and care strategies for dementia that promote resilience and healthy lifestyles, such as PA, may delay the onset and progression of dementia [14]. PA is understood "as any bodily movement produced by skeletal muscles that require energy expenditure above and beyond resting energy expenditure (one metabolic equivalent = 1 MET) and it can be undertaken in many different ways: walking, cycling, sports and active forms of recreation" [19]. Additionally, PA can be classified into different intensity levels: light (1.6-2.9 MET), moderate (3-5.9 MET) and vigorous (≥6 MET), each of which are based on the subjective intensity perception of an individual. Thus, this classification denotes, through MET values, the energy expenditure and/or the amount of oxygen consumed while sitting or performing a PA [20].
Although the positive effects of exercise on cognition in older adults have been researched, the influence of PA on cognitive function of PwD is still not well understood [16]. Scientific intervention studies have emerged to provide evidence for the efficacy of PA as a cognitive reserve-enhancing factor and to assess its potential in delaying cognitive decline in PwD. In two recent meta-analyses, [21,22] considering evidence up to 2018, one showed that 13 RCTs with 673 subjects diagnosed with AD presented statistically significant improvements in cognition after participating in PA interventions (SMD = 1.12 CI: 0.66~1.59) [21]. The second meta-analysis [22] involved 13 RCTs with 659 subjects with AD and reported that PA had a positive effect on cognitive function among persons with AD (p = 0.003). Overall, previous reviews have reported that PA might positively affect the cognition of PwD given its potential to delay cognitive impairment. However, these studies revealed inconclusive results associated with methodological issues and heterogeneity. Such conclusions are in line with other reviews published in recent years [23,24]. For instance, Forbes et al. [25] stated that no clear evidence was found regarding the effects of PA on cognitive activity (95% CI −0.05 to 0.92, p-value 0.08; 9 studies, 409 participants) due to considerable heterogeneity (I 2 value 80%) and deficient quality of the reported evidence.
Therefore, in order to obtain more conclusive results, multiple reviews [21][22][23][24][25] have emphasized that new trials should address methodological barriers by including larger sample sizes [21][22][23][24] and other strategies as follows: providing standardized intervention characteristics [21,24]; providing more information about randomization processes, blinding, attrition rates, and adverse events [25]; conducting different measurements throughout the intervention period [23]; implementing long-term follow-up measures [21][22][23]; using improved and more sensitive cognitive measures [23]; targeting the type of the disease [24]; targeting stage of the disease [22]; separately assessing subjects with Alzheimer's disease and vascular dementia [24]; including different types of PA [23]; and ensuring that the control group does not perform the same amount of PA as the experimental group [23].
Although the effects of PA on dementia patients' cognition have been widely studied over the last few years, it remains unclear whether these recommendations have been integrated into the latest trials and whether increasing methodological quality influences the homogeneity of the results obtained, particularly since the last existing meta-analyses [21,22] mostly included studies conducted before 2015. Therefore, we wanted to provide an update concerning the latest occurrences regarding the new RCTs implemented in the field.

Primary Objective
To identify the effects of PA interventions on cognitive function in individuals diagnosed with dementia compared to those in the control group.

Secondary Objective
To recognize if recent PA interventions address methodological barriers reported in previous reviews and provide clearer conclusions about the effects of PA on cognition in PwD.

Methodological Approach
To have clear guidance while conducting the systematic review, we followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) set of items to report systematic reviews and meta-analyses [26].

Criteria for Inclusion
Studies were considered eligible if they were RCTs in which participants were randomly assigned to a PA group or a control group. The exercise group required implementing a PA program, including strength, aerobic, and balance exercises, as well as interventions combining physical and cognitive exercises for improving the cognitive performance in PwD. In addition, there was no time restriction; interventions could cover any length and duration. In contrast, the control group consisted of usual care, social activities, or handicrafts. Moreover, participants had to be diagnosed utilizing valid criteria, including the Mini-Mental State Examination [MMSE] (cut-off scores for MCI ≤ 24, ≤21, and ≤19); the Montreal Cognitive Assessment [MoCA] (cut-off scores for MCI were ≤25, ≤24) [27]; the Diagnostic and Statistical Manual of Mental Disorders [28]; the National Institute of Neurological and Communicative Disorders and Stroke; and the Alzheimer's Disease and Related Disorders Association [29], or ICD-10 [30]. All forms of dementia diagnosis and severity were included. Trials measured cognitive function with a neuropsychological or cognitive test at baseline and follow-up. Finally, studies that were published in English between 2000 and 2020 were included. The primary outcome involved individuals with dementia and addressed their cognitive function.

Criteria for Exclusion
Studies excluded were pilot RCTs, systematic reviews, meta-analyses, study protocols, and conference publications. Studies were also excluded if the intervention was targeted at participants with mild cognitive impairment, PA training was implemented without assessing cognition, or multimodal interventions were conducted without a PA component.

Search Strategy
A search strategy was conducted on two different occasions (January and May 2020). Moreover, the search was performed for RCTs studying the efficacy of PA in four different databases: PubMed, Embase, APA PsycNET, and the Web of Science from the 1st of January 2000 until May 2020. To obtain the search results, we combined relevant English keywords such as physical activity, dementia, cognition, and RCTs (see Supplementary Materials-Additional File S1 for full electronic search). Furthermore, in May 2020, we performed an additional hand search screening of pertinent studies' bibliographies to identify articles that the initial search strategy did not recognize. Two independent reviewers (MC and AA) conducted this search, screened initial titles and abstracts, and retrieved the full text of potential papers. A third author was consulted when discrepancies emerged.

Study Selection
Initially, titles and abstracts were imported to EndNote; then, they were screened, and duplications or studies that were determined as irrelevant were omitted. Subsequently, full-text articles from the possible pertinent studies were screened in detail. At this point, studies that met the inclusion criteria were included. All data were independently scanned and selected by two reviewers. In the case of discrepancies, a third evaluator was consulted. This process for selecting studies is shown in the PRISMA flow diagram in Figure 1 [26].

Data Extraction
A data extraction sheet was designed to provide accurate data on PA programs among PwD. Information regarding participants, dementia severity at baseline according to the Mini-Mental State Examination (MMSE), the intervention group, the control group, lengthfrequency-duration, PA intensity, cognitive assessment, follow-up, adherence rate, and the impact on cognition was documented in a tabular form. Moreover, the means and standard deviations were extracted from global cognition measurements at baseline and at the end of the study. A t-test was used to determine statistical significance for global cognition. In some studies, this data was not available. Hence, corresponding authors were contacted, and those who did not respond were not considered for inclusion in the analyses.

Data Extraction
A data extraction sheet was designed to provide accurate data on PA prog among PwD. Information regarding participants, dementia severity at baseline acco to the Mini-Mental State Examination (MMSE), the intervention group, the control g length-frequency-duration, PA intensity, cognitive assessment, follow-up, adherence and the impact on cognition was documented in a tabular form. Moreover, the mean standard deviations were extracted from global cognition measurements at baselin at the end of the study. A t-test was used to determine statistical significance for g cognition. In some studies, this data was not available. Hence, corresponding au were contacted, and those who did not respond were not considered for inclusion analyses.

Synthesis of Results
A random-effect meta-analysis was carried out to evaluate global cognition outc in PwD due to heterogeneity among the studies. Furthermore, considering that st

Synthesis of Results
A random-effect meta-analysis was carried out to evaluate global cognition outcomes in PwD due to heterogeneity among the studies. Furthermore, considering that studies reported continuous outcomes, assessed at baseline and follow-up, we pooled means and standard deviations.
Moreover, a qualitative synthesis of the results was performed to understand what kind of PA components might be most effective in improving cognitive function among PwD. This summary is articulated based on the content characteristics and methodological aspects of PA interventions and their effects on the cognition of PwD.

Methodological Quality Assessment
Two independent reviewers (MC and AA) assessed the risk of bias of the included trials using the Effective Public Health Practice Project (EPHPP) Quality Assessment Tool for Quantitative Studies [31], considering sections A to F (A. selection bias; B. study design; 6 of 22 C. confounders; D. blinding; E. data collection method; and F. withdrawals and dropouts). According to the instrument dictionary, each of these components were rated using the codes "strong", "moderate" and "weak". An overall strong score was given when there were no weak ratings, a moderate overall score when there was one weak rating, and a weak overall score when there were two or more weak ratings.
Additionally, to provide a more detailed overview of the methodological aspects of the studies, we completed a systematic assessment based on previous review recommendations [21][22][23][24]. We included aspects such as the application of comprehensive cognitive measures [23], measurements throughout the intervention period [23], long-term followup [21][22][23], target dementia type [24], target dementia stage [24], and the provision of clear and available information on PA dose responses [21,24]. For this assessment, we counted and reported the number of recommendations fully incorporated into each study. An additional file shows more in detail previous reviews recommendations on methodological aspects (see Supplementary Materials-Additional File S2).

Study Selection
After conducting the electronic search in different databases using the established search terms, 5204 results were yielded. To this total amount, four articles from the hand search thought to be relevant were added. After screening titles and abstracts and removing duplicates, 4884 studies were excluded from further analysis. The remaining 324 studies were selected for full-text screening. Of those, 302 articles did not meet the inclusion criteria. Therefore, 22 studies were included in the present systematic review. Figure 1 illustrates the study selection process according to the PRISMA flow diagram [32].

PA Interventions
Regarding control groups, 43 [35] received relaxation and flexibility exercises, and 4.3% [41] had daily one-and-one conversations with a therapist. For the experimental groups exposed to PA training, the following characteristics were found regarding PA modality, frequencies and intensities.
PA interventions had a medium-size effect on the cognitive function of PwD of 0.4803 (95% CI = 0.1901-0.7704). Heterogeneity between studies was statistically significant (I 2 = 86%, p = < 0.0001) (see Figure 2). To assess publication bias between trials, a funnel plot was carried out (see Figure 3). Evidence was found to be skewed or asymmetric; thus, there was publication bias among the sixteen studies.

Quality Assessment
Each study's quality was assessed to avoid the risk of bias and provide consistent results. According to the Effective Public Health Practice Project (EPHPP) Quality Assessment Tool for Quantitative Studies (see Table 3), most of the studies presented an overall quality score between strong and moderate. However, in some cases, weak scores were associated with small sample sizes [39,41,44,45,47,49], lacking withdrawal and dropout reports [33,39,40]. In contrast, some studies showed stronger quality scores associated with the selection of bias since they included larger sample sizes (>100 participants) [35,36,38,42,43,46,48,51].  86%, p = < 0.0001) (see Figure 2). To assess publication bias between trials, a funnel plot was carried out (see Figure 3). Evidence was found to be skewed or asymmetric; thus, there was publication bias among the sixteen studies.

Quality Assessment
Each study's quality was assessed to avoid the risk of bias and provide consistent results. According to the Effective Public Health Practice Project (EPHPP) Quality Assessment Tool for Quantitative Studies (see Table 3), most of the studies presented an overall quality score between strong and moderate. However, in some cases, weak scores were associated with small sample sizes [39,41,44,45,47,49], lacking withdrawal and dropout reports [33,39,40]. In contrast, some studies showed stronger quality scores associated with the selection of bias since they included larger sample sizes (>100 participants) [35,36,38,42,43,46,48,51].

Discussion
This systematic review identified twenty-two RCTs aiming to test the effect of PA on the cognition of PwD. It provides methodologically sounder designs and new results than other studies conducted in recent years. Thus, it adds evidence to other reviews by including new RCTs [35,36,38,45,46,48] that have not previously been included. Overall, the meta-analysis found that PA interventions had a medium-size effect on the cognitive function of PwD. This indicates general positive effects of PA on cognition in PwD. However, the included trials presented a high percentage of heterogeneity (I 2 = 86%, p ≤ 0.0001) as they showed differences in the number of participants, intervention settings, cognitive measurement tools, follow-up periods, PA dose-responses, and reported outcomes. Therefore, these differences between the studies' methodologies limited the possibility of solid conclusions about the effects of PA on the cognition of PwD. These findings were consistent with a review by Forbes et al. [25], which showed considerable heterogeneity (I 2 value 80%) and thus inconclusive results.
Even though these results resemble those obtained by Forbes et al. [25], if we look at the newly added trials implemented in recent years, a slight difference is revealed. In particular, six recent trials [35,38,42,43,46,48] showed more powerful designs since they included larger sample sizes. In this way, this finding enables us to see small developments and progress in this particular field of research, including more solid methodological designs and higher statistical power in the most recent studies. Therefore, the results presented in these trials might lead to more precise conclusions about the effects of PA on the cognition of PwD.
Various features of PA interventions could play a crucial role in mediating the effects of PA on cognition, such as PA modalities, dose responses, and intensity. Based on the type of exercise and intensity, changes in the brain's structure have been obtained [58].
For instance, interventions implemented three types of PA modalities: (1) only PA training (cardiovascular or strengthening), (2) combined PA training (cardiovascular and strengthening), and (3) combined PA with cognitive training. According to Bossers et al. [36], combining aerobic PA with resistant training led to improvements in executive functions and memory functions. Thus, the study recommended combining both modalities to stimulate cognitive improvements in both. Öhman et al. [41] attributed improvements in executive functions to dual-tasking (e.g., talking while walking, singing while dancing) and other combined PAs performed at home (strength, balance, and endurance exercises). This study suggested that combined training may enhance the frontal lobe, which is the brain area in charge of executive functions. These results are consistent with one systematic review implemented by Lauenroth et al. [51], who claimed that multimodal PA interventions that consider cardiovascular training combined with resistance training and cognitive tasks resulted in better and more significant outcomes than individual PA training. Moreover, this type of intervention has contributed to improving frontal cognitive functions, global cognition, working memory, episodic memory, executive function, and processing speed [59].
Additionally, studies presented variations in their PA dose responses. Their session durations ranged between fifteen and one hundred and twenty minutes; their frequency per week fluctuated between two and seven times, and their total length took between less than three months and longer than twelve months. Particularly, studies in which longer periods of PA were undertaken were more likely to display positive effects. For example, Öhmann et al. [48] implemented a 12-month PA program, which led to positive effects on executive functions among community-dwelling PwD. Moreover, Hoffmann et al. [46] affirmed that PA seems to affect executive function (mental speed and attention) when implemented for at least six months. In the same way, Toot et al. [38] confirmed that for cognition effects, the interventions' duration seems to play a decisive role. Thus, a four-month program was not enough time to induce cognitive changes. Likewise, Kassermeijer et al. [35] did not show significant effects due to exergaming training. According to the authors, a possible explanation for these results was that people probably needed more time to master the challenges from this type of program, and this RCT implemented a short intervention period of 12 weeks. These results were aligned with one meta-analysis outcome [60] and with one study, which proved that six to twelve months of PA increased cognitive scores and affected brain structure [58].
Regarding PA intensities of trial interventions, the majority implemented moderate, followed by light and vigorous PA intensities. However, there is no consensus among studies as to which intensity level might be ideal. Karssemeijer et al. [35] stated no significant effects on executive functions, working memory or episodic memory after implementing a light intensity combined cognitive and PA intervention among communitydwelling persons with mild dementia. In contrast, The Dementia and PA trial [42], which had the largest sample size of the included studies, applied a moderate-to-high intensity PA program; however, these PA intensities also did not result in positive outcomes. Thus, Lamb et al. [42] specified that an exercise program of moderate-to-high intensity improved physical fitness but did not slow cognitive deterioration. Furthermore, participants who took part in the PA arm and had a high intervention attendance displayed worse cognitive decline than the control group. Therefore, according to these authors, there is a possibility that PA may have worsened cognitive impairment. In particular, these negative effects were associated with inflammation and inadequate oxygen supply to certain cortical areas. In this way, this study suggested that high-intensity aerobic and strength exercise should not be used as a method for addressing cognitive deterioration, and future research should examine other forms of PA among dementia patients. Likewise, Toots et al. [38] indicated that high-intensity training did not result in significant differences in global cognition or executive functions. These results are in line with one systematic review [60] that specified that PwD are fragile patients, and excessive and vigorous intensities of PA should be avoided to prevent other health complications. Moreover, monitoring a steady heart rate of 60% of the maximum heart rate might prevent excess complications and burden among patients. In addition, this range might be enough to activate neurobiological responses that benefit the brain functioning of PwD [60]. Further research is needed to clarify the role of intensity in mediating PA effects.
Another relevant aspect for effective PA interventions described in trials was PA engagement and adherence to programs due to high numbers of withdrawals in the trials. Thus, studies stated that bad adherence to their program was associated with a lack of motivation [37,49], low emphasis on PA in geriatric facilities, and a lack of knowledge regarding the benefits of PA [51]. Moreover, a high number of persons declined to participate in one study due to a lack of attractiveness of PA, particularly women [42]. Furthermore, one RCT stated that only one specific segment of institutionalized patients joined the study because they were already motivated to perform PA [43]. A current review showed that for healthy adults aged 80 years and older, it was necessary to initiate and adhere to PA to identify its health benefits, overcome physical-activity-associated fear, recognize and prioritize individual PA preferences, receive social support, and minimize environmental barriers [61]. However, considering that PwD present low functional activity and cognitive functioning, it is probable that variables mediating their PA engagement are different compared to those reported by healthy adults [62]. Recent literature lacks evidence on PA participation and adherence-related factors in PwD [63].
Thus, it can be observed that different factors, such as PA modalities, dose responses, intensities, and engagement and adherence, play an important role in facilitating effects on cognition in PwD. However, due to the variety of methodologies, contents and results reported in the included studies, the effects of PA on the cognition of PwD remain unclear. Additional evidence is needed, particularly concerning ideal PA modalities, dose-response intensity, and adherence.
Based on the findings from the most recent studies exploring the effects of PA on cognition of PwD, these might be some implications to consider for future research and policy. However, it is essential to consider them with caution, as the reported studies still present certain limitations:

•
Alternative forms of exercise need to be explored for PwD. For example, additional exercises designed to improve functional activity, a variable that has been proven to be influenced by PA among PwD, are needed [42]; • Exergaming combined with cognitive training is a method that promotes participants' initiation and adherence to PA through the innovative combination of technology and exercising [35]; • Engaging in long-term, individualized, home-based training may have some effect on the executive functions of PwD [48]; • Future programs should also examine the individual characteristics of participants (type and severity of dementia), as they may influence the effects of PA on cognition. It is also essential to examine who may benefit the most from PA [35].

Limitations
This review aimed to identify current studies and update the scientific evidence on the effects of PA on cognition; it included extensive eligibility criteria. For example, it included participants of all types and severity of dementia. These criteria, therefore, contributed to a high proportion of heterogeneity within the study. Likewise, through the funnel plot, an asymmetric plot was observed, which represented publication bias. Additionally, there was incomplete retrieval regarding effects across all the included studies due to missing data and a lack of responses from the authors. This may limit the quality of the evidence and, thus, should be considered when discussing the results. We interpreted our results carefully to avoid over-or under-estimating the scientific evidence of the methodologically weak RCTs.

Conclusions
The evidence for the benefits of PA for PwD remains unclear despite the fact that there is increased research activity within the studies identified in this review Furthermore, the selected studies contained stronger methodological aspects compared to reviews conducted in previous years. In addition, considering that certain prerequisites may affect PA programs, further research is needed. In particular, ideal PA modalities, duration, adherence to interventions, and exercise intensity monitoring should be considered.

Supplementary Materials:
The following are available online at https://www.mdpi.com/article/10 .3390/ijerph18168753/s1. Additional File S1: PubMed full electronic search; Additional File S2: Previous reviews recommendations on methodological aspects; Additional File S3: Cognitive domains and measurement tools.
Author Contributions: M.I.C. was a major contributor to conceptualizing the systematic review. Moreover, M.I.C. participated in screening and extracting data, assessing methodological quality, and drafting and writing the manuscript. A.A. collected, screened, and extracted data, and assessed the methodological quality of the included studies. N.L. revised the manuscript and provided feedback on the content and structure of the manuscript. J.R.T. was a major contributor in conceptualizing the review, revising and giving feedback on knowledgeable content, and giving the final approval of the manuscript. All authors have read and agreed to the published version of the manuscript. Data Availability Statement: The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.