A Cross Sectional Analysis of Physical Activity Engagement in Adults with Mild Cognitive Impairment

Background To address the paucity of literature regarding the relationship between mild cognitive impairment and physical activity engagement, this study aimed to understand the relative contribution of cognitive, demographic, physical and psychological variables related to physical activity engagement in individuals with mild cognitive impairment. Method This was a descriptive, cross – sectional study of secondary data from 62 participants with MCI (mean age 70.53, SD = 6.34), 53.2% female, median MoCA 23 (IQR: 20, 24)) from the NeuroExercise study, a 12 – month PA intervention on the outcome of cognitive function. The independent variable of interest was global cognitive function. Age, gender, years of education, number of medications, handgrip strength, depression, and quality of life were treated as covariates. The dependent variable was PA engagement in minutes per week, using the LAPAQ physical activity questionnaire and the Actigraph triaxial accelerometer device. Hierarchical regression analyses showed no signicant effect of cognitive function on physical activity engagement after controlling for the effects of covariates. Physical activity engagement was low relative to global physical activity guidelines ((M = 111.38, SD = 94.29) Actigraph ( t (51) = -2.95, p < .005) and the LAPAQ (M = 51.71, SD = 22.80), t (61) = -33.94, p = < .001)). A Bland- Altman measure of agreement demonstrated that objective and subjective measures of physical activity were not equivalent. for covariates, to assess the individual contributions of these variables and the contributions of these variables as domain constructs. A one-sample t test was used to assess differences in means of PA minutes per week compared to recommended physical activity guidelines. A Bland -Altman plot was used to analyse level of agreement between objective and subjective measures of PA engagement. Alpha level was set at .005 following Bonferroni adjustment for multiple comparisons.


Abstract Background
To address the paucity of literature regarding the relationship between mild cognitive impairment and physical activity engagement, this study aimed to understand the relative contribution of cognitive, demographic, physical and psychological variables related to physical activity engagement in individuals with mild cognitive impairment.

Method
This was a descriptive, cross -sectional study of secondary data from 62 participants with MCI (mean age 70.53, SD = 6.34), 53.2% female, median MoCA 23 (IQR: 20, 24)) from the NeuroExercise study, a 12 -month PA intervention on the outcome of cognitive function. The independent variable of interest was global cognitive function. Age, gender, years of education, number of medications, handgrip strength, depression, and quality of life were treated as covariates. The dependent variable was PA engagement in minutes per week, using the LAPAQ physical activity questionnaire and the Actigraph triaxial accelerometer device.

Results
Hierarchical regression analyses showed no signi cant effect of cognitive function on physical activity engagement after controlling for the effects of

Conclusions
This sample of adults with MCI were not su ciently physically active. Further, there was substantial variability between objective and subjective measures of physical activity engagement. Objective measurement of PA data may be more reliable for adults with mild cognitive impairment.

Background
Physical activity (PA) in mid -late life may help to support brain health in adults with mild cognitive impairment (MCI), a risk state for dementia ( According to global PA guidelines, adults aged 18 + should engage in at least 30 minutes of moderate intensity PA a day ve days per week, or 150 minutes per week. The most recent guidelines have increased the minimum amount of recommended PA to 150-300 minutes of moderate or 75-150 minutes of vigorous intensity PA per week (Bull et al., 2020), however as this study was conducted prior to the release of these updated guidelines, this report will focus on recommended PA as per the previous guidelines. Previous research suggests that adults with subjective cognitive impairment ( (Stuckenschneider et al., 2018) are positively correlated with PA engagement. However, Wettstein et al (2015) reported no correlation between cognitive function and PA engagement (Wettstein et al., 2015). Few studies have examined the demographic correlates of PA engagement in adults with MCI, with contradictory ndings. Age was negatively signi cant in one study (Vancampfort et al., 2018) and non -signi cant in another (Rovner, Casten, & Leiby, 2016). In those studies, being male was positively (Rovner et al., 2016) and non -signi cantly (Vancampfort et al., 2018) associated with PA engagement. Similarly, level of education was non -signi cant but level of literacy was positively associated in one study (Rovner et al., 2016), whereas Vancampfort et al (2018) found level of education to be negatively related to PA engagement in Chinese adults but positively related in Ghanaian adults. Both studies reported negative correlations with PA engagement and depression. In addition, some evidence regarding domains relating to physical health and function suggests indicators of physical health such as measures of chronic conditions, injury and weak grip strength may be negatively correlated with PA engagement (Vancampfort et al., 2018), and mobility and instrumental activities of daily living (IADL`s) may positively correlated (Rovner et al., 2016). This evidence is suggestive of associations between cognitive, demographic, physical and psychosocial variables on the outcome of PA engagement in adults with MCI, but it is contradictory and does not provide evidence of the relative contributions of the variables to the outcome of interest. . This highlights the need for studies to assess the reliability of self -reported PA data and its ability to accurately characterise PA engagement in a group of cognitively impaired adults.
This cross -sectional, descriptive study aimed to analyse secondary data from the NeuroExercise Study, a 12 -month PA intervention on the outcome of cognitive function ( See Appendix 1 for NeuroExercise Study protocol summary (Devenney et al., 2017)) to quantify the association between PA engagement and demographic, cognitive, physical and psychological domains in adults with MCI, and to make two key comparisons: between levels of PA in an MCI population relative to global PA guidelines, and between the reliability and convergent validity of subjective and objective measures of PA engagement in this sample. It is necessary to better understand these issues to facilitate the use of PA as a risk factor treatment target for addressing early cognitive decline.
It was hypothesised that global cognitive function would be positively associated with PA engagement, after controlling for the effects of demographic, physical and psychological variables. Secondary hypotheses were that PA engagement would be below the global recommended PA guidelines of 150 minutes per week, and that objective and subjective measures of PA engagement in this cohort would not be convergently valid. 2. Use a one-sample t -test to compare levels of PA in MCI sample to global Physical Activity Guidelines.
3. Assess the relationship and level of agreement between subjective (LAPAQ physical activity questionnaire) and objective ( Actigraph triaxial accelerometer) measures of PA engagement in individuals with MCI using a Bland -Altman scatterplot.

Design
This was a descriptive cross-sectional correlation study using secondary data. The independent variable of interest was level of global cognitive function.

Measures
Physical activity engagement Physical activity was measured using the self-report LASA Physical Activity Questionnaire (LAPAQ) (see Appendix 3.) and the Actigraph triaxial accelerometer device.
The LAPAQ is a researcher administered recall survey which records moderate to vigorous PA (MVPA) for the previous 14 days. Scoring is calculated by totalling minutes spent in each activity. Weekly totals were calculated to achieve minutes per seven days. The LAPAQ was designed for use in older populations and is valid and reliable in non -impaired older adults (Harris et al., 2009;Stel et al., 2004). In this sample Cronbach's alpha coe cient was .44.
The Actigraph GT3X (Actigraph, Pensacola, Fl, USA), is a medical -grade triaxial accelerometer worn on the hip for seven consecutive days. It is valid and reliable for measuring PA in non -impaired adults ( . Two participants failed to complete the CES-D, one was unable to complete the handgrip left measure, two were unable to complete the handgrip right measure and ten participants failed to supply Actigraph data. Table 1 provides an overview of all missing data.

Statistical Analysis
Baseline participant data was analysed using Spss V.25. Descriptive statistics were used to describe the distribution of the data. Kolmogorov -Smirnov values were inspected to assess normality of the distribution of scores ( Appendix 7, Table A1). Inspection of outliers showed some outliers which were not determined to be signi cant. The relevant boxplots for age, years of education, number of medications and the LAPAQ are presented in Appendices 8 -11 ( Figures A1 -A4). Scatterplot matrices were created to explore the linearity between each DV (LAPAQ (Appendix 12, Figure A5) and Actigraph (Appendix 13, Figure A6)) and the independent variables. Bar charts were used to inspect potential patterns in each IV across the categorical IVs (gender, recruitment source). Visual inspection of bar charts showed scores were similar when strati ed (Appendix 14, Figure A7 and Appendix 15, Figure A8).
Hierarchical linear regression analyses were used to explore associations between the independent and dependent variables of the LAPAQ and the Actigraph data, controlling for covariates, to assess the individual contributions of these variables and the contributions of these variables as domain constructs. A onesample t test was used to assess differences in means of PA minutes per week compared to recommended physical activity guidelines. A Bland -Altman plot was used to analyse level of agreement between objective and subjective measures of PA engagement. Alpha level was set at .005 following Bonferroni adjustment for multiple comparisons.

Descriptive Statistics
Variables that were not normally distributed were number of medications, MoCA and Actigraph scores. The sample was 53.2% female ( 33 women), mean age of 70.53 ( SD = 6.34). The mean minutes of PA engaged in per week were 51.72 ( SD = 22.82) (LAPAQ ) and 111 ( SD = 94.30) (Actigraph data). Means, medians, standard deviations and interquartile ranges are presented in Table 2, with normative data where available. Frequencies for all variables are presented in Appendices 16-23 (Tables A2 -A9).   Table 3.   Table 4. The mean minutes of PA per week were 51. 72 (+/-SD = 22.82) for the LAPAQ and 111 ( SD = 94.30). for the Actigraph. Frequency data shows that 29% of participants achieved 150 minutes or greater of PA engagement per week versus 0% of participants according to LAPAQ data. Frequencies for the LAPAQ and Actigraph data are presented in Table 5.   Table 6 presents descriptive statistics for participants who achieved recommended levels of PA per week versus those who did not as measured using Actigraph data. The Spearman rho correlation coe cient showed a medium positive correlation between variables ( r = .33, n = 52, p = .015) indicating a moderate level of agreement between the LAPAQ questionnaire and the Actigraph. A Bland -Altman scatterplot {Bland, 1986) to further investigate the level of agreement between measures of PA engagement (Fig. 1.) showed that the limits of agreement expressed using 95% con dence intervals were wide (240.

Discussion
This study found no associations between cognitive function and PA engagement as measured using either the Actigraph or LAPAQ. However, beta coe cients suggest some level of variance, although non -signi cant, in the effect of cognitive function on the outcome in both regressions after controlling for covariates, albeit higher in regression two where the outcome was objectively measured. This is also true of number of medications where the beta value was high in models two, three and four using Actigraph data as the outcome measure. Previous studies have reported a positive association between cognitive function, stage of MCI and PA engagement in MCI populations (Kobayashi et al., 2016;Stuckenschneider et al., 2018;Vancampfort et al., 2018). In comparison to previous studies however, the sample used here was small and lack of statistical power may account for the non -signi cant association in this case. Further research in larger samples of Irish adults with MCI is needed to overcome this limitation of this study. Therefore, with regard to the primary research hypothesis, based on this analysis we fail to reject the null hypothesis but suggest that further exploration is needed here.
Both subjective and objective measures of PA demonstrate that participants did not achieve 150 minutes of PA per week. Twenty nine percent of participants as recorded objectively and 0% of participants as measured subjectively engaged in 150 minutes. This is in contrast with previous MCI studies, two of which reported low/ no physical activity (< 150 minutes per week) in between 20-27 % of adults (Rovner et al., 2016;Vancampfort et al., 2018), and another which reported mean PA levels of between 164-270 minutes of PA per day (Stuckenschneider et al., 2018). Our ndings indicate that PA engagement in this sample of Irish adults with MCI is much lower than in previously sampled MCI cohorts in other countries. However, PA engagement in Irish adults in general has also been found to be low. Murtagh et al (Murtagh et al., 2015) report that 62% of cognitively healthy Irish adults did not engage in 150 minutes of PA per week and this is supported by a recent report stating 67% of Irish adults over 18 years were achieving < 150 minutes per week (Factsheets., 2018. ). The low level of PA engagement seen here may be re ective of normative population values among Irish adults. Studies of global levels of PA con rm that Irish adults are less physically active than their European (35% inactive), American (43% inactive), African (28% inactive) and Southeast Asian (17% inactive) counterparts (Hallal et al., 2012). This suggests a potential area for improvement in public health initiatives designed to address and promote brain health in ageing through increasing physical activity.
Finally, correlational analysis showed a moderate positive association between objective and subjective PA engagement measures. However, further assessment showed a high level of bias and lack of agreement between the measures, suggesting an overall lack of convergent validity There was a trend toward the difference between measures decreasing as the mean values get higher, suggesting that at lower levels of PA engagement there is less disagreement between measures. The level of variability would also suggest this is the case. Based on these ndings, we fail to accept the null hypothesis that measures will be convergently valid. These ndings may be due to the fact that the maximum value of the LAPAQ recorded was far below that recorded objectively. It may also suggest that participants in this sample under -estimated the level of PA that they engaged in, in support of Siebelings` nding that self -reported PA is prone to underestimation in cognitively impaired populations (Siebeling, 2012), or did not accurately recall past activity. Given that the internal reliability of the LAPAQ was shown to be low in this sample, and that previous studies of PA in Irish samples, in addition to previous reports of PA engagement in MCI cohorts, were concurrent with the ndings of levels of PA engagement as recorded via Actigraph (29 %), it is likely that objective measures of PA engagement are more accurate in this case.

Limitations
The sample size of this study was small, and the decision to adjust the Alpha value for multiple comparisons may have in ated the risk of type II error. Examination of effect sizes suggests the possible presence of an effect between the independent and dependent variables of interest which studies using larger samples may be better placed to detect.

Conclusions
It is recommended that future studies gather data from larger samples with the possible inclusion of comparison groups to enable more in-depth analysis that can further elaborate on the relative contribution of various factors on physical activity engagement in MCI. As MCI is often a pre-cursor to dementia, people with MCI are ideally placed to use physical activity as a way of delaying the progression of cognitive decline and/ or managing symptoms of cognitive decline. Therefore, health policy should address the low levels of PA in the population in general, with an added focus on adults with MCI who are at a greater risk of progressing to dementia and who may particularly bene t from the protective effects of regular PA engagement on brain health. Finally, objective measurement of PA engagement may be a more reliable method of measurement in adults with MCI, and research requiring PA measurement in this cohort should incorporate its use.  (2) and was carried out in accordance with institution guidelines. Ethics approval was granted on 18/05/17. Written informed consent was obtained from all participants.

Consent for publication
Not applicable.

Availability of data and materials
The dataset used and analysed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.
Funding L.C is a PhD scholar funded by the Health Research Board under SPHeRE/2013/1. The funders had no role in the study design, data collection and analysis, decision to publish or preparation of the manuscript.
Author contributions LC conducted data analysis and drafted the manuscript. BL and JMcHP oversaw and contributed comments to the writing of the manuscript.