Alzheimer’s Disease Biomarkers Interactively Influence Physical Activity,Mobility,and Cognition Associations in a Non-Demented Aging Population

Abstract

Background:

Alzheimer’s disease (AD) risk-reduction strategies (e.g., increasing physical activity, improving mobility) have garnered increasing attention in the literature. However, the effect of such modifiable factors on the preclinical trajectories of brain and cognitive health may be moderated by non-modifiable biomarkers associated with AD.

Objective:

In a longitudinal sample of non-demented older adults, we examine the independent predictors everyday physical activity (EPA) and mobility on executive function (EF) performance and change. Next, we test whether these predictions are modified by interactions between age and AD genetic risk.

Methods:

This accelerated longitudinal design included adults (n = 532, M age = 70.4, age range 53–95) from the Victoria Longitudinal Study. We tested the independent effects of EPA and Mobility (i.e., gait, balance), moderation by Apolipoprotein E (i.e., APOE ɛ4+, ɛ4–) and age (young-old, middle-old, old-old), and interactions between APOE and age on performance and 9-year change in an EF latent variable.

Results:

First, higher levels of both EPA and Mobility were associated with better EF performance and less decline. Second, the interaction between age and APOE showed that low EPA and older age was associated with poorer EF performance and steeper EF decline for APOE ɛ4 + carriers, and low mobility was associated with poorer EF performance and steeper EF decline for APOE ɛ4 + carriers than the non-risk carriers.

Conclusion:

In non-demented older adults, age moderated the effects of both EPA and Mobility on EF performance and change. However, this moderation occurs differentially across APOE4 status.

Keywords

Apolipoprotein E everyday physical activity executive function mobility Victoria Longitudinal Study

INTRODUCTION

As a result of accumulating evidence that Alzheimer’s disease (AD) processes develop over many years and that pharmacological treatments have been ineffective, attention has shifted toward identification of AD risk-reduction and prevention strategies [1]. Prevention strategies include reducing the risk of cardiovascular disease and increasing cognitive and physical activity [2, 3]. Notably, variables representing these risk-reduction targets may interact with other health, biological, and lifestyle factors to influence the level and slope of non-demented trajectories of cognitive change [4, 5]. For example, the cognitive benefits garnered by engaging in beneficial health and lifestyle factors (such as physical activity) may be restricted by increasing chronological age and genetic risk, two non-modifiable factors associated with an increased risk for AD. In addition, interactions between non-modifiable risk factors may differentially influence overlapping or related health and lifestyle factors (i.e., mobility and physical activity).

The multifactorial complexity of these dynamic networks of influence presents a challenge for researchers and clinicians. Recent developments have linked theoretical frameworks for assessing networks of influence [2] with modern research designs to advance the cause of identifying patterns, distinct contributions, timing of relationships, and interactions between multiple changing variables [6] representing both modifiable and non-modifiable risk factors for cognitive impairment and AD. Such research may potentially identify specific activities representing individualized risk or protection, subgroups with elevated risk for cognitive impairment, and time periods across older adulthood in which brain and cognitive processes are particularly vulnerable to modifiable factors. Such precision assessment could lead to tailored recommendations for risk reduction [1]. We systematically examine whether an interaction between age group and Apolipoprotein ɛ4 (APOE) influences the effects of two separate but possibly overlapping lifestyle and health factors (i.e., physical activity and mobility (gait)) on executive function (EF) performance and 9-year change in a large, well-characterized sample of non-demented older adults.

Age is the greatest non-modifiable risk factor for AD [7]. Unlike chronological age, biological aging does not unfold linearly, but involves a multiplicity of coordinated systems, mechanisms, transition states, and interacting processes [8 –10]. For example, vascular age-related changes, such as thickening of the arteries, co-occur with age-related changes in the immune system, such as atrophy of the thymus gland, a vital organ for T-cell development, which increases vulnerability to disease states [11]. Age-related changes in the cerebrovascular system have been shown to be related to changes in mobility [12], which itself is the product of multiple network interactions of the central nervous system [13] and physiological age-related changes such as sarcopenia [12].

The dynamic and interactive nature of multiple biological aging processes indicates potential malleability, perhaps differentially across older adulthood. Therefore, cognitive trajectories may be differentially modified by risk-reducing lifestyle activities at varying times across the broad band of late adulthood. Moreover, the extent and effectiveness of lifestyle modifications may be further moderated by genetic risk factors which could vary with advancing age [14].

Apolipoprotein E (APOE) is a lipoprotein involved in lipid metabolism and transportation. The APOE gene has three allelic variations, ɛ2, ɛ3, and ɛ4, yielding six possible genotypes: ɛ2ɛ2, ɛ2ɛ3, ɛ2ɛ4, ɛ3ɛ3, ɛ3ɛ4, and ɛ4ɛ4 [15]. APOE has been associated with multiple trajectories and clinical outcomes of aging, from normative cognitive decline to AD [16 –19]. Additionally, APOE influences health and lifestyle factors associated with cognitive aging, such as mobility and physical activity [20 –23]. Furthermore, when examining these variables within the more complex networks in which they operate, APOE differentiates mobility-cognition and physical activity-cognition trajectories over time spans of 1 to 3 years [24, 25].

Mobility refers to a cluster of functional health biomarkers, such as timed gait and balance. Deficits in mobility biomarkers have been associated with decline in EF and other cognitive domains in non-demented aging [26, 27] as well transitions to mild cognitive impairment (MCI) and AD [29 , 31–33]. Recent evidence reported that higher levels of physical functioning (as measured by a composite of indicators of mobility and muscle strength) were associated with better EF performance [28]. Mobility may be supported in cognitively normal aging by functional physical activity or fitness [29]. In fact, evidence suggests that maintaining minimum levels of physical activity over a period of three years attenuates declines in mobility for older adults [30]. Taken together, this evidence indicates that mobility is a modifiable risk factor associated with differential cognitive trajectories in aging.

Physical activity is defined as any skeletal muscle movement, which results in energy expenditure [31]. Recently, there has been growing interest in everyday physical activity (EPA), a modifiable lifestyle factor that encompasses daily leisure participation in a variety of activities available to older adults in voluntary low to moderate doses. Examples include leisure walking, tennis, and gardening. Some longitudinal research focusing only on EPA has found that higher participation levels are associated with beneficial cognitive effects— and these effects apply not only for non-demented older adults but also to MCI and AD groups [32 –34]. Specifically, for non-demented older adults, higher baseline EPA has been associated with better scores and less decline in multiple cognitive domains, such as reasoning, episodic memory, verbal fluency, and EF [34 –36].

EPA overlaps with and may support (or be restricted by) functional abilities [29] especially insofar as physical activities involve movement in forms and combinations of gait, balance, and postural transference. However, for present purposes, EPA can be differentiated conceptually from mobility in that the latter typically involves standardized assessment of movement-related fundamental abilities whereas the former typically involves movement as expressed in everyday life. To our knowledge, no studies have yet compared the potential beneficial cognitive effects of mobility and EPA markers in the same longitudinal study. Notably, the inclusion of both EPA and Mobility in the present study allows us to identify differential, distinct relationship patterns for these movement-related variables while statistically adjusting for influence of the other variable outside of covariate analysis.

EFs are among the most age-sensitive cognitive functions [4, 37] due to significant age-related neurodegeneration occurring in the prefrontal cortices [38]. EFs encompass higher-level cognitive processes required to make and execute plans, solve problems, set goals, shift between stimulus and response, and inhibit responses [39]. Notably, accelerated rates of EF decline have been observed up to 5 years prior to diagnosis in patients who develop AD [40] indicating that AD-related neuropathology may reach the point of functional limitation of EF prior to diagnosis. Considered together, accelerated rates of EF decline, long-term bidirectional interplay between EF and mobility [41], and the established relationship between EF and EPA [34] highlight the importance of longitudinal studies of EF in non-demented older adults.

Research goals

The overall purpose of this study was (1) to establish the extent to which EPA and mobility affected EF performance in terms of both level (at a statistical centering age) and slope (longitudinal change) and then (2) to examine whether an interaction between two key moderating variables (i.e., APOE risk and Age) further modified these observed prediction patterns. We assembled a 3-wave (up to 9 years) data set, covering a 40-year age span (55–95 years) of non-demented aging and used structural equation modeling to investigate three research goals (RG). RG1 was to use conditional growth models to explore how EPA and Mobility independently affect level and change in EF. RG2 was to use conditional growth models to explore whether APOE (risk, non-risk) or Age [i.e., young-old (YO), mid-old (MO), old-old (OO)] separately moderated the level and longitudinal EPA-EF and Mobility-EF relationships. RG3 was to examine the effects of the interaction between APOE and Age on both the level and longitudinal EPA-EF and Mobility-EF relationships.

MATERIALS AND METHODS

Participants

Participants were community dwelling older adult volunteers drawn from the Victoria Longitudinal Study (VLS). The VLS is a large-scale, long-term investigation of neurocognitive aging, impairment, and dementia as influenced by biological, medical, genetics, health, lifestyle, environmental, and other factors [42]. The VLS and all present data collection procedures were in full and certified compliance with prevailing human research ethics guidelines and boards. All participants provided informed written consent. Three main sequential cohorts (initially aged 55–95 years) are followed at about 4-year intervals. As a focus of this study is to examine change in EF as moderated by a genetic variant, participants were limited to a source subsample who had provided biofluid for genotyping between 2009 and 2011 (n = 700). This source subsample consisted of three sequential cohorts, with data collection occurring in the 2001–2015 period, Cohort 1 waves 6, 7, and 8, Cohort 2 waves 4 and 5, and Cohort 3 waves 1, 2, and 3, for a total individualized duration of up to 9 years [4]. The wave-to-wave retention rates by cohort ranged from 74% to 89%.

The following exclusionary criteria were applied at baseline to the source sample: (1) diagnosis of AD or other forms of impairment and dementia, (2) Mini-Mental State Exam score of <24 [43], (3) self-report of “severe” for a health condition cluster (high blood pressure, low blood pressure, diabetes, epilepsy, spinal or thyroid conditions, depression, head injury), (4) reported alcohol or drug dependence, (5) reported use of anti-psychotic medications, (6) self-reported “moderate” cases of neurological conditions (Parkinson’s or stroke), and (7) insufficient data on EF, Mobility or EPA. The resulting study sample comprised 534 non-demented participants (M age = 70.47; age range = 53.25–95.43; 66.5% female; see Table 1).

Table 1

Descriptive statistics for sample by APOE genotype

APOE	ɛ4 + (risk)	ɛ4 –(non-risk)
	W1	W1
n	132	382
Age	69.71 (8.27)	70.54 (8.71)
Range	54.63–87.36	53.24–95.25
Gender (% female)	64.4	65.9
Education	15.69 (3.09)	15.29 (2.93)
Range	8.0–24.0	6.0–23.0
EPA	16.27 (5.35)	15.94 (5.16)
Range	3.0–30.0	0.0–31.0
Mobility	0.09 (0.80)	0.034 (0.78)
Range	–2.62–1.58	–2.81–1.54
BMI	26.40 (3.58)	26.82 (4.17)
Range	18.58–37.51	14.95–43.33
PP	51.35 (9.81)	51.09 (9.41)
Range	34.00–82.25	32.13–91.50
MMSE	28.68 (1.27)	28.68 (1.25)
Range	24 –30	24–30

Results presented as Mean (Standard Deviation). W1, Wave 1; W2, Wave 2; W3, Wave 3; EPA, everyday physical activity; BMI, body mass index; PP, pulse pressure; MMSE, Mini-Mental State Exam. The genotypic distribution for APOE is in Hardy-Weinberg equilibrium, χ² = 0.84.

Everyday physical activity

The measure was the four-item physical activity subscale from the VLS-Activity Lifestyle Questionnaire (VLS-ALQ) [44, 45]. Each item indexed frequency of participation in an everyday physical activity over a period of two years. Frequency was reported according to the scale: 0 (never), 1 (less than once a year), 2 (about once a year), 3 (2 or 3 times a year), 4 (about once a month), 5 (2 or 3 times a month), 6 (about once a week), 7 (2 or 3 times a week), or 8 (daily). The four activities were common in the geographic location of the VLS: gardening indoors or outdoors; exercise activities such as jogging, swimming, bicycling, or walking; outdoor activities such as fishing, sailing, or backpacking; and recreational sports such as tennis, bowling, or golf. Responses were totaled, producing a continuous measure with scores ranging from 0–32. Higher scores indicate more participation in everyday leisure physical activities. Notably, the range (0–31) within this sample comparably encompassed the scope of possible EPA engagement in each of the age groupings used. Psychometric and other details are available [34 , 44].

Mobility

We used two standard measures of gait and balance, both of which have been described and examined previously in the VLS [46, 47] and elsewhere [48]. Gait was measured with a timed walk over a distance of 20 feet. Specifically, participants were instructed to stand behind a white line on the floor and walk in a straight line as quickly and safely as possible just past a tape on the floor at a distance of 10 feet, turn around and walk as quickly and safely as possible back to the starting position. Balance was measured with a timed turn task, which assessed the speed with which a person could make a complete circle from a standing position. Participants were instructed to stand directly behind a white line on the floor, with their toes lined up along the line and feet slightly apart, then make one complete turn in place, returning to the starting position with the toes lined up once again directly behind the white line. Time to perform each task was recorded in seconds. As scores were speed values the composite scores were reverse-coded for ease of interpretation. A composite mobility score was formed with unit-weighted z-scores of the two indicators. Therefore, a higher mobility score indicated better mobility performance. Time to perform each task was measured in seconds.

DNA extraction and genotyping

Saliva samples were collected according to standard procedures from Oragene-DNA Genotek. Samples were stored at room temperature in the Oragene disks until DNA extraction. DNA was manually extracted from the saliva sample mix using the manufacturer’s protocol and quantified using a NanoDrop ND-1000 Spectrophotometer (Wilmington, DE, see [4] for further information). Genetic analyses included genotype categorization based on the presence or absence of the risk allele. Dichotomous APOE genotype categories were: ɛ4+ (risk) consisted of ɛ4ɛ4 and ɛ3/ɛ4 allele combinations and ɛ4–(non-risk) consisted of ɛ2ɛ2, ɛ2/ɛ3, and ɛ3ɛ3 allele combinations. For all analyses including APOE, standard practice is to remove the genotype which combines the risk and protective alleles (ɛ2ɛ4; n = 20, final N = 514) [4]. The genotypic distribution for APOE was in Hardy-Weinberg equilibrium, χ² = 0.84.

Executive function

We assembled a robust EF latent variable comprised of four manifest EF indicators: Hayling sentence completion test [49], Stroop test [50], Brixton spatial anticipation test [49], and Color Trails Test part two [51]. All four measures have all been frequently used and validated in standard form with older adults in VLS and other studies [4 , 53].

Statistical analyses

Structural equation modeling (SEM) was conducted using Mplus 7 [54]. Structural equation modelling is a modern statistical approach that compares a structural model, comprised of a system of regression equations, to empirical data. This statistical approach allows analysis of latent variables and their relationships within the measurement model, offering the opportunity to analyze latent constructs concurrently and longitudinally while incorporating the measurement error of the variables in the model. SEM also simultaneously and efficiently models direct and indirect effects within the methodology of path analyses. Consistent with recommended procedures and other VLS research, we coded chronological age as a continuous variable and used as the metric of change for all analyses. We centered Age at 75 years, the approximate mean of the 40-year span of data, and a commonly observed inflection period in non-demented cognitive aging [45, 55] and roughly the mid-point of the present age distribution. We estimated missing EF data were by multiple imputations using Mplus 7 [54 , 57]. As per VLS protocol, 50 imputations were included.

Preliminary analyses

First, confirmatory factor analysis, longitudinal measurement invariance, and latent growth modeling were used to verify a previously modeled EF latent variable and growth model from a similar sample. We determined model fit with standard indices: (1) χ² for which a good fit would produce a non-significant test (p > 0.05), indicating the data were not significantly different than the model estimates, (2) comparative fit index for which≥0.95 was judged a good fit and between 0.90 and 0.94 was judged an adequate fit, (3) root mean square error of approximation, for which≤0.05 was judged good and between 0.06 and 0.08 was judged adequate, and (4) standardized root-mean-square residual for which good fit was judged by a value of≤0.08 [57, 58]. Longitudinal measurement invariance was tested by comparing models with free and constrained parameters using a chi-square difference test (Δχ²).

RG1: Independent effects of EPA and Mobility on the EF growth model

The best unconditional EF growth model from the preliminary analyses was used as the benchmark against which conditional growth models with the independent predictors of change (EPA, Mobility) were tested [57]. Path analyses were used to determine the effects of each predictor (separately) on level of EF performance at the centering age 75 and 9-year EF change.

RG2: Moderation of EPA-EF and Mobility-EF relationships by, separately, Age or APOE

A series of steps to test age moderation was followed. First, a model which tested the effect of EPA on level of EF performance and 9-year EF change was estimated, with all the parameter estimates constrained to be equal across three age groups of older adults (i.e., YO < 65, MO 65–75, and OO > 75). Second, the parameters were free to vary between age groups to examine moderation. Evidence of moderation was indicated by a significant difference test which compared the fully constrained to the unconstrained model [58]. These steps were repeated using Mobility as the predictor of EF level and 9-year change. The same series of steps was used to test APOE moderation across risk and non-risk groups using EPA and then Mobility as the predictor of EF level and 9-year change.

RG3: Interaction effects of APOE×Age on the EPA-EF and Mobility-EF relationships

We used the APOE risk/non-risk groups and applied the standard analytical moderation steps to examine the APOE-Age and APOE-Age interactions, on the EPA-EF and Mobility-EF relationships. Moderation was examined using three steps. First, interaction terms were created to identify six different APOE×Age groupings (i.e., APOE ɛ4–YO, APOE ɛ4–MO, APOE ɛ4–OO, APOE ɛ4 + YO, APOE ɛ4 + MO, and APOE ɛ4 + OO). Second, overall Age moderation was established using the standard moderation analysis procedure to identify differential moderation patterns among the three groups within APOE risk or non-risk status (i.e., the APOE ɛ4–YO, MO, and OO groups were compared to each other). Third, groups were compared across the APOE genetic status according to Age (i.e., the APOE ɛ4–YO group was compared to the APOE ɛ4 + YO group) to examine differences between APOE risk and non-risk carriers. These steps were repeated with Mobility as the predictor of EF level and 9-year change.

Covariates

Education, vascular health, measured by pulse pressure, a proxy measure of arterial stiffness, [4 , 59–61] and body mass index were included as covariates, as these factors have been associated with EF in non-demented aging. We performed the analyses with all covariates included and found that none of the covariates affected the main results. Therefore, the final models presented in this study did not include covariates.

RESULTS

Preliminary analyses: Latent growth modeling for EF

The analyses verified an EF latent variable and an EF growth model over a 40-year period. The latter included three results of particular relevance to this study. First, at the model centering age (75), older adults varied significantly in level of EF performance (b = 1.05, p < 0.001). Second, there was significant decline in EF performance (M = –0.016, p = 0.003). Third, there was significant individual variability in the rate of decline (b = 0.003, p < 0.001; see Table 2 for full EF growth model results).

Table 2

Confirmatory factor analyses and latent growth goodness of fit indexes for executive function models

Model	AIC	BIC	χ ²	df	p	RMSEA	CFI	SRMR	Δ χ ²	Δdf
CFA for One Factor Model
Configural Invariance	19493.1	19729.2	35.55	35	0.44	0.01 (0.00–0.03)	1	0.03
Metric Invariance	19042.9	19253.1	41.83	41	0.043	0.01 (0.00–0.03)	0.99	0.05	8.15	6
Scalar Invariance	19137.4	19313.3	152.37	49	<0.001	0.06 (0.05–0.07)	0.92	0.1	110.54*	8
Partial Scalar Invariance^a,b	19056.2	19249.2	63.17	45	0.038	0.03 (0.01–0.04)	0.99	0.06	21.34*	4
Model	AIC	BIC	–2LL	D	Δdf	p
Latent Growth Model
Fixed intercept	5188.45	5197.02	5184.44	–	–	–
Random intercept	3126.45	3139.32	3120.45	2064	1	<0.001
Random intercept, fixed slope	2848.41	2865.57	2840.41	280.04	1	<0.001
Random intercept, random slope^b	1793.95	1819.69	1781.95	1058.46	2	<0.001

AIC, Akaike information criterion; BIC, Bayesian information criterion; χ², chi-square test of model fit; df, degrees of freedom for model fit; RMSEA, Root Mean Square Error of Approximation; CFI, Comparative Fit Index; SRMR, Standardized Root Mean Square Residual; Δχ², change in chi-square; Δdf, change in degrees of freedom; CFA, Confirmatory Factor Analysis; –2LL, –2 Log likelihood; D, difference statistic (using –2LL). *p < 0.001. ^aBrixton and Color Trails free to vary. ^bBest fitting mode.

RG1: Independent effects of EPA and Mobility on the EF growth model

Using the models established in the preliminary analyses, we tested two growth models (one with EPA and one with Mobility) with time-invariant predictors of EF level and change. For EPA, baseline level revealed significant predictions for both EF performance at age 75 (b = 0.340, p < 0.001) and 9-year change (b = 0.016, p = 0.002; see Fig. 1). Specifically, lower baseline levels of EPA were associated with significantly worse EF performance (M = –0.347) at age 75 than were higher levels of EPA (M = –0.007). Moreover, lower baseline levels of EPA were associated with greater 9-year EF decline (M = –0.042) than were higher levels (M = –0.026).

Fig.1

Predicted growth curve model of Executive Function (EF) with continuous Everyday Physical Activity (EPA) as a predictor. The three categories of EPA are depicted for convenience. Age in years was the metric of change. The age variable was centered at 75 years.

For Mobility, baseline level was also a significant predictor of EF performance (b = 0.412, p < 0.001) and 9-year decline (b = 0.025, p < 0.001; see Fig. 2). Specifically, lower baseline levels of Mobility were associated with significantly worse EF performance (M = –0.201) than were higher levels of Mobility (M = 0.211). Moreover, lower baseline levels of Mobility were associated with greater 9-year EF decline (M = –0.039) than were higher levels (M = –0.014).

Fig.2

Predicted growth curve model of Executive Function (EF) with continuous Mobility as a predictor. The three categories of mobility are depicted for convenience. Age in years was the metric of change. The age variable was centered at 75 years.

RG2: Moderation of EPA-EF and Mobility-EF relationship, separately, by Age or APOE

We conducted four sets of moderation analyses to examine whether Age or APOE differentially moderated the previously observed EPA-EF and Mobility-EF relationships. Figures for RG2 are included in the supplementary material (Supplementary Figures 1–4).

Age moderation of the EPA-EF relationship

First, regarding Age, the fully constrained model produced a significantly worse fit than the unconstrained model (D = 423.82, Δdf = 14, p < 0.000), indicating Age moderated the EPA-EF relationship. Specifically, baseline level of EPA predicted both EF performance (b = 0.470 p < 0.001) and 9-year EF change (b = 0.017, p = 0.009) for the YO group only. Within the YO group, those with low baseline levels of EPA exhibited poorer EF performance (M = 0.524) and steeper 9-year decline (M = 0.01) than did YO individuals with high levels (M = 0.994, and M = 0.027, respectively). For the MO and OO age groups, level of EPA did not alter level of EF, nor the 9-year EF change.

Age moderation of the Mobility-EF relationship

A significant interaction also noted for Age within the Mobility-EF relationship, as evidenced by the fully constrained model producing a significantly worse fit than the unconstrained model (D = 337.34, Δdf = 14, p < 0.000). Specifically, for the MO and OO age groups, baseline mobility predicted both level of EF performance (b = 0.471 p < 0.001; b = 0.284, p = 0.001, respectively) and 9-year EF change (b = 0.031, p < 0.001; b = 0.023, p = 0.001, respectively). Within the MO group, adults with low baseline mobility exhibited poorer EF performance (M = –0.467) and steeper 9-year decline (M = –0.048) than did MO adults with high levels (M = 0.004, and M = –0.017, respectively; see Fig. 4b). Within the OO group, adults with low baseline mobility exhibited poorer EF performance (M = –0.381) and steeper 9-year decline (M = –0.072) than did OO adults with high levels (M = –0.097, and M = –0.049, respectively). For the YO age group, level of mobility did not alter level of EF or 9-year change.

Fig.3

Predicted growth curve of Executive Function (EF) using Everyday Physical Activity (EPA) as a predictor and moderated by age and APOE ɛ4–(i.e., ɛ2/ɛ2, ɛ2/ɛ3, ɛ3/ɛ3). The three categories of EPA are depicted for convenience. Age in actual years was the metric of change. The age variable was centered at age 75. Moderation analysis D = 441.38, Δdf = 35, p < 0.001.

Fig.4

Predicted growth curve model of Executive Function (EF) using Everyday Physical Activity (EPA) as a predictor and moderated by age and APOE ɛ4+ (i.e., ɛ3/ɛ4, ɛ4/ɛ4). The three categories of EPA were depicted for convenience. Age in actual years was the metric of change. The age variable was centered at age 75. Moderation analysis D = 441.38, Δdf = 35, p < 0.001.

APOE moderation of the EPA-EF relationship

Second, regarding APOE, a significant interaction was evidenced for APOE within the EPA-EF relationship, as the fully constrained model produced a significantly worse fit than the unconstrained model (D = 24.5, Δdf = 7, p < 0.000). Specifically, for the APOE non-risk carriers, baseline EPA predicted both level of EF performance (b = 0.407 p < 0.001) and 9-year EF change (b = 0.019, p < 0.001). This effect was not observed for the APOE risk carriers.

APOE moderation of the Mobility-EF relationship

A significant interaction was evidenced for APOE within the Mobility-EF relationship, as the fully constrained model produced a significantly worse fit than the unconstrained model (D = 22.74, Δdf = 7, p < 0.002). Differential interaction patterns were observed for both the APOE risk and non-risk groups. Specifically, for the APOE non-risk carriers, baseline mobility predicted both level of EF performance (b = 0.443, p < 0.001) and 9-year EF change (b = 0.029, p < 0.001). This effect was also seen for the APOE risk carriers, as baseline mobility also predicted both level of EF performance (b = 0.876, p < 0.001) and 9-year EF change (b = 0.051, p < 0.001).

RG3: Interaction effects of APOE×Age on the EPA-EF and Mobility-EF relationships

We conducted four sets of moderation analyses to examine whether an interaction between APOE×Age moderated the effect of EPA or Mobility on EF. In these analyses, an APOE×Age interaction would be observed if the EPA-EF or Mobility-EF relationships varied across APOE and age groups, thus producing a fully constrained model that was a significantly worse fit than the unconstrained model.

APOE×Age moderation of the EPA-EF relationship

For EPA, an APOE×Age interaction was evidenced by the fully constrained model producing a significantly worse fit than the unconstrained model (D = 441.38, Δdf = 35, p < 0.001). Differential interaction patterns were observed within the APOE non-risk group. Within the APOE non-risk YO and MO groups, level of EPA at baseline predicted both level of EF performance at age 75 (b = 0.517, p = 0.001; b = 0.310, p = 0.011) and 9-year EF change (b = 0.018, p = 0.001; b = 0.019, p = 0.027; see Fig. 3). Specifically, for the APOE non-risk YO and MO groups, low baseline EPA predicted poorer EF performance (M = 0.451; M = –0.286, respectively) and steeper 9-year decline (M = 0.008; M = –0.034, respectively) than did higher baseline EPA (M = 0.968, M = 0.024; M = 0.026; M = –0.015, respectively; see Fig. 3a and 3b). As can be seen by the similar trajectories depicted in Fig. 3c, level of EPA did not influence EF performance or 9-year change for the non-risk OO group. The interaction was not observed within the APOE risk group. Specifically, level of EPA did not influence EF performance or 9-year change for any adults (YO, MO, or OO) with the risk allele (ɛ4+; as can be seen by the similar trajectories depicted in each panel ofFig. 4).

Next, to examine further this interaction, we compared the respective age groups (i.e., YO APOE risk to YO APOE non-risk, MO APOE risk to MO APOE non-risk, OO APOE risk to OO APOE non-risk). Results indicated, at age 75, the OO APOE risk carriers with low levels of EPA had lower EF performance (M = –0.181) and had steeper EF declines (M = –0.073) than the OO APOE non-risk carriers with low levels of EPA (M = –0.26; M = –0.059, respectively). However, EF performance and rates of 9-year decline did not differ significantly between the YO APOE risk and non-risk groups with low levels of EPA, or the MO APOE risk and non-risk groups with low levels of EPA.

In sum, an interaction between APOE and Age was observed by differential patterns seen for the EPA-EF relationship across age and risk status groups. Whereas cognitive benefits of having high levels of EPA were seen for groups with no risk factors (i.e., APOE non-risk, younger ages), individuals with one or two risk factors (i.e., old age or/and genetic risk for AD) did not reap cognitive benefits of participating in EPA. Additionally, the group with low EPA and a combination of the risk factors (old age and AD genetic risk) exhibited the worst cognitive performance and steepest 9-yeardecline.

APOE×Age moderation of the Mobility-EF relationship

For Mobility, an APOE×Age interaction was evidenced by the fully constrained model, which produced a significantly worse fit than the unconstrained model (D = 356.28, Δdf = 35, p < 0.001). Differential interaction patterns were observed for the groups. First, within the APOE non-risk MO and OO groups, level of mobility at baseline predicted both level of EF performance (b = 0.371 p < 0.001; b = 0.217 p = 0.027, respectively) and 9-year EF change (b = 0.025, p = 0.001; b = 0.019 p = 0.011, respectively; see Fig. 5). Specifically, within the APOE non-risk MO and OO groups, adults with low baseline mobility exhibited poorer EF performance (M = –0.385; M = –0.262, respectively) and steeper 9-year decline (M = –0.042; M = –0.066, respectively) than did their peers with high levels of mobility (M = –0.014; M = –0.045; M = –0.017; M = –0.047, respectively; see Fig. 5b and 5c). As can be seen by the identical trajectories for all levels of mobility depicted in Fig. 5a, level of baseline mobility did not alter EF performance or 9-year EF change in the non-risk YO group. Second, for the APOE risk carriers, level of baseline mobility predicted EF performance for the YO and MO groups (b = 0.611, p = 0.029; b = 0.799, p < 0.001, respectively) and 9-year EF change for the MO group (b = 0.052, p < 0.001). Specifically, within the APOE risk YO and MO groups, adults with low baseline mobility exhibited poorer EF performance (M = 0.134; M = –0.684, respectively) than did their peers with high baseline mobility (M = 0.745; M = 0.115, respectively). Additionally, within the APOE risk MO group, those with low baseline mobility exhibited steeper 9-year decline (M = –0.066) than did their peers with high baseline mobility (M = –0.014). This pattern can be seen in Fig. 6a (by the differing levels of EF performance at age 75 for all levels of mobility) and in Fig. 6b (by the differing trajectories of 9-year change for the MO APOE risk carriers of all the levels of mobility). For the OO APOE risk carriers, level of mobility did not predict level of EF (b = 0.429, p = 0.101); however, a trend was noted for 9-year EF change (b = 0.031, p = 0.066), as can be seen by the differing 9-year change trajectories depicted in Fig. 6c.

Fig.5

Predicted growth curve model of Executive Function (EF) using Mobility as a predictor and moderated by age and APOE ɛ4–(i.e., ɛ2/ɛ2, ɛ2/ɛ3, ɛ3/ɛ3). The three categories of mobility were depicted for convenience. Age in actual years was used as the metric of change. The age variable was centered at age 75. Moderation analysis D = 356.28, Δdf = 35, p < 0.001.

Fig.6

Predicted growth curve model of Executive Function (EF) using Mobility as a predictor and moderated by age and APOE ɛ4+ (i.e., ɛ3/ɛ4, ɛ4/ɛ4). The three categories of mobility are depicted for convenience. Age in actual years was the metric of change. The age variable was centered at age 75. Moderation analysis D = 356.28, Δdf = 35, p < 0.001.

In addition, we compared the respective age groups (i.e., YO APOE risk to YO APOE non-risk, MO APOE risk to MO APOE non-risk, OO APOE risk to OO APOE non-risk). The YO, MO, and OO APOE risk carriers with lower levels of mobility experienced lower EF performance (M = 0.134; M = –0.684; and M = –0.66, respectively) and steeper EF decline (M = –0.005; M = –0.066; and M = –0.087, respectively) than their non-risk peers with lower levels of mobility (M = 1.0; M = –0.385; M = –0.262; M = 0.03; M = –0.042; M = –0.066).

In sum, an interaction between APOE and Age was observed by differential patterns seen for the Mobility-EF relationship across age and risk status groups. In contrast to the effects seen for EPA, cognitive benefits of high mobility were seen for individuals with one risk factor (i.e., either older age or AD genetic risk). Similar to the EPA results, the group with low mobility and both risk factors (i.e., the APOE risk OO group) had the lowest EF performance and steepest 9-year change. Notably, individuals with AD genetic risk, regardless of age, with low levels of mobility performed worse cognitively and had steeper 9-year cognitive decline than their non-risk peers.

DISCUSSION

The overall purpose of this study was to examine how two non-modifiable biomarkers for AD (age and APOE) interacted to influence the concurrent and longitudinal relationships between two modifiable factors (EPA, Mobility) and EF trajectories over a 40-year band of non-demented aging. We distributed this aim into three RGs.

RG1 (Independent effects of EPA and Mobility on the EF growth model) revealed two main results. First, for EPA, the results verified our previous research indicating that older adults with higher baseline levels of EPA had better initial EF performance and more gradual decline over the three waves [34]. This is consistent with emerging research indicating higher levels of EPA may attenuate cognitive decline and buffer against cognitive impairment and dementia [32 , 62–66], possibly through positive effects on brain volume, vascular health, and AD pathology [67 –69]. Moreover, habitual physical activity (defined as leisure and household behaviors, such as walking) has been found to be positively associated with a composite measure of EF in an AD population [70]. Thus, engagement in everyday physical activity may be a relatively accessible strategy for older adults of varying cognitive status to protect against cognitive declines and perhaps delay dementia.

Regarding Mobility, our results indicated that individuals with higher baseline levels of Mobility (as indicated by a composite measure of gait and balance) had better EF performance and less 9-year decline than did their peers with lower baseline levels of Mobility. Emerging research indicates that gait is not solely a motor task, but has neural correlates with EF performance [71]. Interestingly, the link between EF and gait speed has been demonstrated across both non-impaired and MCI populations [72, 73]. This highlights the possibility that mobility measures may be sensitive to changes in attention, revealing EF deficits before clinical detection [74] and may be contribute to a “motor-signature” for early cognitive impairment [75]. The parallel EPA-EF and Mobility-EF results of RG1 provided the groundwork for testing the independent moderating and interaction effects needed for RG2 and RG3.

Briefly, RG2 (Moderation of EPA-EF and Mobility-EF relationships by Age or APOE) revealed that both Age and APOE independently moderated the effect of both EPA and Mobility on EF trajectories. The results indicate that the cognitive benefits from EPA and mobility are not distributed evenly across all ages (in the 55–95-year range) or AD-related genetic risk status but appear selectively for older adults under certain conditions. For example, APOE non-risk carriers with higher EPA had better EF performance and more gradual EF decline than their non-risk peers with lower EPA. Additionally, APOE risk carriers with higher levels of mobility had better EF performance and less 9-year decline than their APOE risk peers with lower levels of mobility (for detailed interpretation, see the Supplementary Material). In the interest of precision in observational and intervention research, identifying mechanisms that operate under specific conditions will lead to more optimal strategies for delaying cognitive decline in non-demented aging. The critical identification of the independent moderating effects of Age and APOE leads us to RG3, in which we observed novel APOE×Age interactions on the EPA-EF and Mobility-EF relationships.

First, our results indicated an interaction between APOE and Age moderated the EPA-EF relationship. Specifically, YO and MO individuals who were also APOE non-risk carriers had high levels of EF performance and less EF decline if they also had high levels of EPA. The OO APOE non-risk carriers and APOE risk carriers of any age (YO, MO, OO) did not experience the same EF benefits of having high EPA. Notably, all four factors (i.e., APOE, Age, EPA, and EF) are involved and we know of no other studies that consider them simultaneously. Although cognitive benefits of physical activity are widely accepted, results of research on physical activity and APOE status have proven to be inconsistent. Various studies have indicated a positive association for only the risk (ɛ4) carriers [63 –65], whereas others have not found a significant association [76, 77]. Consider research by Podewils and colleagues (2005), who observed that engaging in higher levels of physical activity reduced the dementia risk only for APOE ɛ4 non-carriers. We found the APOE-physical activity relationship is further influenced by age-related processes and has a direct effect on cognition.

Regarding possible mechanisms, recent work has indicated that higher levels of physical activity are associated with lower levels of amyloid-β (Aβ), insulin, triglycerides, and higher levels of high-density lipoprotein [78]. Notably, lower plasma Aβ levels resulting from physical activity have been found for only the APOE ɛ4 non-carriers [78]. In addition, higher plasma Aβ has been associated with increased age [79], bilateral thinning of the prefrontal cortex, low cognitive performance, and cognitive decline [80]. Therefore, as seen in the current study, the younger ɛ4 non-carriers may exhibit the cognitive benefits of engaging in physical activity through reduced circulating levels of Aβ.

Alternatively, the absence of an effect of physical activity on OO APOE non-risk or APOE risk carriers could be considered with evidence that the double dose of risk factors (i.e., OO age, APOE risk) combined with low EPA produced the worst EF performance and the steepest decline. In this case, a combined effect of cerebrovascular dysfunction from age-related changes and APOE risk could nullify the cognitive benefits of high physical activity. More research examining cerebrovascular changes associated with APOE and physical activity is needed.

Second, our results indicated an interaction between APOE and Age moderated the Mobility-EF relationship. Results indicated while all groups (with exception of the YO APOE non-risk carriers) experienced cognitive effects of having differential levels of Mobility, the APOE risk carriers of all age groups with low levels of Mobility experienced lower EF performance and steeper EF decline than their non-risk peers with low levels of Mobility.

We know of one previous study [21] that has used a combination of Mobility, APOE status, and Age. Their results indicated that OO (mean age of 80 at baseline) APOE ɛ4 + carriers had more rapid 10-year motor decline than did the OO APOE non-risk carriers. We expanded on their results in two key ways. First, we added a measure of cognition known to be sensitive to both age-related neural changes and neurodegenerative processes. Second, we examined the two AD biomarkers both independently, and most importantly, interactively. Our results have many implications. First, the temporal sequence of change over time indicates a strong relationship between mobility and EF into late age. Second, mobility performance may be a sensitive early indicator of EF change in non-demented older adults. Third, for those with higher genetic risk for AD, mobility performance is a strong predictor of EF performance and 9-year decline. As the risk of AD increases dramatically for those over the age of 85, the OO (≥75 years of age at the first time point) APOE risk group with low mobility (and the lowest EF and steepest EF decline) could be a potential group to monitor for clinical symptoms of AD.

Slower gait speed has been linked with smaller prefrontal cortex volumes [81]. As the prefrontal cortex mediates EF processes, it is possible that level of mobility has an effect on level of EF as a function of prefrontal brain volume. Moreover, APOE ɛ4 has been associated with reduced prefrontal volumes [82]. Therefore, it is possible that atrophy associated with genetic risk exacerbates the relationship between mobility and brain volume. Recently, Aβ burden has been associated with a risk of mobility decline [41], as both older age and APOE ɛ4 confer higher risk [83, 84] Aβ accumulation could also explain the current results. Alternatively, the relationship between gait speed, physical activity, and regional atrophy could be mediated by cerebrovascular dysfunction such as increased white matter hyperintensities [85, 86]. Therefore as mentioned previously, cerebrovascular underpinnings may be at the root of the relationship seen between APOE, Age, Mobility, EPA, and EF. It is evident further research is needed in this area.

There are several limitations to this study. First, the participants of the VLS are initially selected to be relatively healthy and free of neurodegenerative disease, and may possess several risk-reducing factors, such as access to national health care, above average in years of education, and community-dwelling status. As a group, they may not be representative of the broadest population of older adults; however, they could reflect a growing proportion of older adults in western countries. Second, due to the design of the VLS, the third data point is not available for Cohort 2. Therefore, only participants from the first and third VLS cohort contributed three data points. A more complete design would have included three data points from all cohorts. However, we used multiple imputation (a known, effective, statistically sound method) [57] to accommodate this missing data. Third, we use a self-report measure of EPA and thus not all aspects of the construct domain are represented. Future research may consider including both observational (actigraphy) and self-report indicators in order to establish validity and create composite indicators. We note, however, that we included a physical measure (mobility, gait) and found complementary results. Fourth, as one of the exclusionary criteria was a MMSE score indicative of impairment (<24), it is possible carriers of the ɛ4 allele who are at higher risk of becoming impaired had already developed cognitive impairment and therefore were not included in the study. As mobility impairment could also be a phenotype of cognitive impairment, this exclusion could result in a sample of higher physically functioning APOE ɛ4 carriers who are protected from the risks associated with this gene.

Several strengths are associated with this study. First, we used modern statistical approaches to analyze three programmatic and integrated RGs. Second, we had a relatively large and well-characterized longitudinal sample (W1 n = 532), which spanned 40 years of aging. Third, as can be seen in Table 1, many characteristics of this community-dwelling sample were similar across genetic risk and non-risk carriers. For example, the risk and non-risk groups had similar mean ages, age ranges, gender representation, education, baseline Mobility levels, baseline EPA levels, and MMSE scores. Fourth, we included four standard, reliable neuropsychological manifest variables in a factorially validated EF latent variable. Fifth, we conducted parallel analyses (with EPA and Mobility) over a longitudinal timeframe that included a 40-year band of aging. Such an approach adds complexity to the written report, but we have carefully organized (into three sequential RGs) and streamlined the presentation for this article. We emphasize that novel, theoretically integrative, and potentially translatable results were obtained. For example, EPA was associated with cognition for only APOE non-risk carriers below the age of 75, whereas mobility was associated with cognition for APOE non-risk carriers over the age of 65, and APOE risk carriers of all ages. Such specificity of effects can lead to precision intervention for non-demented older adults. For example, younger APOE non-risk adults may be advised to increase their everyday physical activity across a wide variety of activities, whereas older APOE risk carriers may be advised to do activities that specifically target and improve mobility aptitude.

In conclusion, we examined the effects of interactions between non-modifiable AD biomarkers (i.e., APOE genetic risk and Age) on the predicted relationships of physical activity (EPA) and mobility on cognitive change (EF) in non-demented older adults. Our results indicated that (1) EPA and mobility predict EF performance and change, and (2) AD risk biomarkers interact to contribute to the observed EF variability. Notably, two major risk factors for AD, Age and APOE genotype, interacted to produce differential EF outcomes when considered with physical activity and mobility in this non-demented sample. Clinically, our results suggest precision interventions may be tailored to non-demented older adults based on both age and APOE genetic risk status. These results indicate that the influence of protective lifestyle factors may differ in relation to non-modifiable factors. This highlights the importance of longitudinal examination of multiple AD risk and protective factors on cognitive performance and change across a broad band of non-demented aging.

Footnotes

ACKNOWLEDGMENTS

This research was supported by grants from the National Institutes of Health/National Institute on Aging (R01 AG008235) to RAD, and the Canada Research Chairs program to RAD.

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Anstey

, Eramudugolla

, Hosking

, Lautenschlager

, Dixon

(2015) Bridging the translation gap: From dementia risk assessment to advice on risk reduction. J Prev Alzheimers Dis 2, 189–198.

Anstey

(2014) Optimizing cognitive development over the life course and preventing cognitive decline: Introducing the Cognitive Health Environment Life Course Model (CHELM). Int J Behav Dev 38, 1–10.

Beard

, Officer

, de Carvalho

, Sadana

, Pot

, Michel

J-P

, Lloyd-Sherlock

, Epping-Jordan

, Peeters

GMEE

, Mahanani

, Thiyagarajan

, Chatterji

(2016) The World report on ageing and health: A policy framework for healthy ageing. Lancet 387, 2145–2154.

McFall

, Wiebe

, Vergote

, Jhamandas

, Westaway

, Dixon

(2014) IDE (rs6583817) polymorphism and pulse pressure are independently and interactively associated with level and change in executive function in older adults. Psychol Aging 29, 418–430.

Fotuhi

, Hachinski

, Whitehouse

(2009) Changing perspectives regarding late-life dementia. Nat Rev Neurol 5, 649–658.

McFall

, Sapkota

, McDermot

, Dixon

(2016) Risk-reducing Apolipoprotein E and Clusterin genotypes ptotect against the consequences of poor vascular health on executive function performance and change in non-demented older adults. Neurobiol Aging 42, 91–100.

National Institute on Aging National Institutes of Health. Risk factors for Alzheimer‘s disease, https://www.nia.nih.gov/alzheimers/publication/preventing-alzheimers-disease/risk-factors-alzheimers-disease, September 2012, Accessed February 4, 2017.

Chung

, Cesari

, Anton

, Marzetti

, Giovannini

, Seo

, Carter

, Yu

, Leeuwenburgh

(2009) Molecular inflammation: Underpinnings of aging and age-related diseases. Ageing Res Rev 8, 18–30.

Kirkwood

(2005) Understanding the odd science of aging. Cell 120, 437–447.

10.

van Deursen

(2014) The role of senescent cells in ageing. Nature 509, 439–446.

11.

Clark

(2002), A means to an end: The biological basis of aging and death, Oxford University Press, Oxford, UK.

12.

Sorond

, Cruz-Almeida

, Clark

, Viswanathan

, Scherzer

, De Jager

, Csiszar

, Laurienti

, Hausdorff

, Chen

, Ferrucci

, Rosano

, Studenski

, Black

, Lipsitz

(2015) Aging, the central nervous system, and mobility in older adults: Neural mechanisms of mobility impairment. J Gerontol A Biol Sc 70, 1526–1532.

13.

Rosso

, Studenski

, Chen

, Aizenstein

, Alexander

, Bennett

, Black

, Camicioli

, Carlson

, Ferrucci

, Guralnik

, Hausdorff

, Kaye

, Launer

, Lipsitz

, Verghese

, Rosano

(2013) Aging, the central nervous system, and mobility. J Gerontol A Biol Sc 68, 1379–1386.

14.

Lindenberger

, Nagel

, Chicherio

, Li

S-C

, Heekeren

, Bäckman

(2008) Age-related decline in brain resources modulates genetic effects on cognitive functioning. Front Neurosci 2, 234–244.

15.

Lahiri

, Sambamurti

, Bennett

(2004) Apolipoprotein gene andits interaction with the environmentally driven risk factors:Molecular, genetic and epidemiological studies of Alzheimer’sdisease. Neurobiol Aging 25, 651–660.

16.

Brainerd

, Reyna

, Petersen

, Smith

, Taub

(2011) Is the apolipoprotein e genotype a biomarker for mild cognitive impairment? Findings from a nationally representative study. Neuropsychology 25, 679–689.

17.

Dixon

, DeCarlo

, MacDonald

, Vergote

, Jhamandas

, Westaway

(2014) APOE and COMT polymorphisms are complementary biomarkers of status, stability, and transitions in normal aging and early mild cognitive impairment. Front Neurosci 6, 1–11.

18.

Praetorius

, Thorvaldsson

, Hassing

, Johansson

(2013) Substantial effects of apolipoprotein E ɛ4 on memory decline in very old age: Longitudinal findings from a population-based sample. Neurobiol Aging 34, 2734–2739.

19.

Small

, Rosnick

, Fratiglioni

, Bäckman

(2004) Apolipoprotein E and cognitive performance: A meta-analysis. Psychol Aging 19, 592–600.

20.

Schiepers

, Harris

, Gow

, Pattie

, Brett

, Starr

, Deary

(2012) APOE E4 status predicts age-related cognitive decline in the ninth decade: Longitudinal follow-up of the Lothian Birth Cohort 1921. Mol Psychiatry 17, 315–324.

21.

Buchman

, Boyle

, Wilson

, Beck

, Kelly

, Bennett

(2009) Apolipoprotein E e4 allele is associated with more rapid motor decline in older persons. Alzheimer Dis Assoc Disord 23, 63–69.

22.

Podewils

, Guallar

, Kuller

, Fried

, Lopez

, Carlson

, Lyketsos

(2005) Physical activity, APOE genotype, and dementia risk: Findings from the Cardiovascular Health Cognition Study. Am J Epidemiol 161, 639–651.

23.

Sapkota

, Bäckman

, Dixon

(2017) Executive function performance and change in aging is predicted by Apolipoprotein E, intensified by Catechol-O-methyltransferase and Brain-derived neurotrophic factor, and moderated by age and lifestyle. Neurobiol Aging 52, 81–89.

24.

Schuit

, Feskens

, Launer

, Kromhout

(2001) Physical activity and cognitive decline, the role of the apolipoprotein e4 allele. Med Sci Sports Exerc 33, 772–777.

25.

Sakurai

, Montero-Odasso

(2017) Apolipoprotein E4 allele and gait performance in mild cognitive impairment: Results from the gait and brain study. J Gerontol A Biol Sci Med Sci doi: 10.1093/gerona/glx075

26.

Atkinson

, Rapp

, Williamson

, Lovato

, Absher

, Gass

, Henderson

, Johnson

, Kostis

, Sink

(2010) The relationship between cognitive function and physical performance in older women: Results from the women’s health initiative memory study. J Gerontol A Biol Sci Med Sci 65, 300–306.

27.

Gale

, Allerhand

, Sayer

, Cooper

, Deary

(2014) The dynamic relationship between cognitive function and walking speed: The English Longitudinal Study of Ageing. Age 36, 1–11.

28.

Desjardins-Crépeau

, Berryman

, Vu

TTM

, Villalpando

, Kergoat

M-J

, Li

, Bosquet

, Bherer

(2014) Physicalfunctioning is associated with processing speed and executivefunctions in community-dwelling older adults. J Gerontol BPsychol Sci Soc Sci 69, 837–844.

29.

Paterson

, Warburton

(2010) Physical activity and functional limitations in older adults: A systematic review related to Canada’s Physical Activity Guidelines. Int J Behav Nutr Phys Act 7, 38.

30.

Martinez-Gomez

, Bandinelli

, Del-Panta

, Patel

, Guralnik

, Ferrucci

(2017) Three-year changes in physical activity and decline in physical performance over 9 years of follow-up in older adults: The Invecchiare in Chianti Study. J Am Geriatr Soc 65, 1176–1182.

31.

World Health Organization, Physical activity, http://www.who.int/topics/physical_activity/en/, 2017, Accessed May 20, 2017.

32.

Buchman

, Boyle

, Yu

, Shah

, Wilson

, Bennett

(2012) Total daily physical activity and the risk of AD and cognitive decline in older adults. Neurology 78, 1323–1329.

33.

Grande

, Vanacore

, Maggiore

, Cucumo

, Ghiretti

, Galimberti

, Scarpini

, Mariani

, Clerici

(2014) Physical activity reduces the risk of dementia in mild cognitive impairment subjects: A cohort study. J Alzheimers Dis 39, 833–839.

34.

Thibeau

, McFall

, Wiebe

, Anstey

, Dixon

(2016) Geneticfactors moderate everyday physical activity effects on executivefunctions in aging: Evidence from the Victoria Longitudinal Study. Neuropsychology 30, 6–17.

35.

Lindwall

, Cimino

, Gibbons

, Mitchell

, Benitez

, Brown

, Kennison

, Shirk

, Atri

, Robitaille

, MacDonald

, Zelinski

, Willis

, Schaie

, Johansson

, Praetorius

, Dixon

, Mungas

, Hofer

, Piccinin

(2012) Dynamic associations of change in physical activity and change in cognitive function: Coordinated analyses of four longitudinal studies. J Aging Res 2012, 1–12.

36.

Wang

H-X

, Jin

, Hendrie

, Liang

, Yang

, Cheng

, Unverzagt

, Ma

, Hall

, Murrell

(2013) Late life leisure activities and risk of cognitive decline. J Gerontol A Biol Sci Med Sci 68, 205–213.

37.

de Frias

, Dixon

, Strauss

(2006) Structure of four executive functioning tests in healthy older adults. Neuropsychology 20, 206–214.

38.

Raz

, Rodrigue

(2006) Differential aging of the brain: atterns, cognitive correlates and modifiers. Neurosci Biobehav Rev 30, 730–748.

39.

Luszcz

(2011) Executive function and cognitive aging. In: Handbook of the Psychology of Aging, Schaie

, Willis

, eds. Elsevier Inc., pp. 59–72.

40.

Grober

, Hall

, Lipton

, Zonderman

, Resnick

, Kawas

(2008) Memory impairment, executive dysfunction, and intellectual decline in preclinical Alzheimer’s disease. J Int Neuropsychol Soc 14, 266–278.

41.

Tian

, An

, Resnick

, Studenski

(2016) The relative temporal sequence of decline in mobility and cognition among initially unimpaired older adults: Results from the Baltimore Longitudinal Study of Aging. Age Ageing. doi: 10.1093/ageing/afw185

42.

Dixon

, de Frias

(2004) The Victoria Longitudinal Study: From characterizing cognitive aging to illustrating changes in memory compensation. Aging Neuropsychol Cog 11, 346–376.

43.

Folstein

, Folstein

, McHugh

(1975) “Mini-mental state”: A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 12, 189–198.

44.

Hultsch

, Hertzog

, Small

, Dixon

(1999) Use it or lose it: Engaged lifestyle as a buffer of cognitive decline in aging? Psychol Aging 14, 245–263.

45.

Small

, Dixon

, McArdle

(2011) Tracking cognition-healthchanges from 55 to 95 years of age. J Gerontol B Psychol SciSoc Sci 66, 153–161.

46.

MacDonald

, Dixon

, Cohen

, Hazlitt

(2004) Biological age and 12-year cognitive change in older adults: Findings from the Victoria Longitudinal Study. Gerontology 50, 64–81.

47.

MacDonald

SWS

, Hundza

, Love

, DeCarlo

, Halliday

DWR

, Brewster

PWH

, Lukyn

, Camicioli

, Dixon

(2017) Concurrent indicators of gait velocity and variability are associated with 25-Year cognitive change: A retrospective longitudinal investigation. Front Neurosci 9, 17.

48.

Dumurgier

, Artaud

, Touraine

, Rouaud

, Tavernier

, Dufouil

, Singh-Manoux

, Tzourio

, Elbaz

(2017) Gait speed and decline in gait speed as predictors of incident dementia. J Gerontol A Biol Sci Med Sci 72, 655–661.

49.

Burgess

, Shallice

(1997) The Hayling and Brixton tests, Thames Valley Test Company, Thurston, Suffolk.

50.

Taylor

, Kornblum

, Lauber

, Minoshima

, Koeppe

(1997) Isolation of specific interference processing in the Stroop task: PET activation studies. Neuroimage 6, 81–92.

51.

D’Elia

, Satz

, Uchiyama

, White

(1996) Color Trails Test: Professional Manual, Psychological Assessment Resources, Odessa, FL.

52.

Bielak

, Mansueti

, Strauss

, Dixon

(2006) Performance on the Hayling and Brixton tests in older adults: Norms and correlates. Arch Clin Neuropsychol 21, 141–149.

53.

de Frias

, Dixon

, Strauss

(2009) Characterizing executive functioning in older special populations: From cognitively elite to cognitively impaired. Neuropsychology 23, 778–791.

54.

Muthén

, Muthén

(1998) Mplus User’s Guide (7thedn), Muthén & Muthén, Los Angeles, CA, USA.

55.

Dixon

, Small

, MacDonald

, McArdle

(2012) Yes, memorydeclines with aging— but when, how, and why? In Memoryand Aging: Current Issues and Future Directions, Naveh-Benjamin

, Ohta

, eds. Psychology Press, New York, pp. 325–347.

56.

Enders

(2011) Analyzing longitudinal data with missing values. Rehabil Psycholl 56, 267–288.

57.

Little

(2013) Longitudinal structural equation modeling, Guilford Press, New York, NY.

58.

Kline

(2011) Principles and practice of structural equation modeling, Guilford Press, New York, NY.

59.

Benetos

, Thomas

, Joly

, Blacher

, Pannier

, Labat

, Salvi

, Smulyan

, Safar

(2010) Pulse pressure amplification: A mechanical biomarker of cardiovascular risk. J Am Coll Cardiol 55, 1032–1037.

60.

McFall

, Wiebe

, Vergote

, Westaway

, Jhamandas

, Bäckman

, Dixon

(2015) ApoE and pulse pressure interactively influence level and change in the aging of episodic memory: Protective effects among ɛ2 carriers. Neuropsychology 29, 388–401.

61.

Raz

, Dahle

, Rodrigue

, Kennedy

, Land

(2011) Effects of age, genes, and pulse pressure on executive functions in healthy adults. Neurobiol Aging 32, 1124–1137.

62.

Blasko

, Jungwirth

, Kemmler

, Weissgram

, Tragl

, Fischer

(2014) Leisure time activities and cognitive functioning in middle European population-based study. Eur Geriatr Med 5, 200–207.

63.

Ferencz

, Laukka

, Welmer

A-K

, Kalpouzos

, Angleman

, Keller

, Graff

, Lövdén

, Bäckman

(2014) The benefits ofstaying active in old age: Physical activity counteracts thenegative influence of PICALM, BIN1, and CLU risk alleles onepisodic memory functioning. Psychol Aging 29, 440–449.

64.

Rovio

, Kåreholt

, Helkala

E-L

, Viitanen

, Winblad

, Tuomilehto

, Soininen

, Nissinen

, Kivipelto

(2005) Leisure-time physical activity at midlife and the risk of dementiaand Alzheimer’s disease. Lancet Neurol 4, 705–711.

65.

Woodard

, Sugarman

, A Nielson

, Carson Smith

, Seidenberg

, Durgerian

, Butts

, Hantke

, Lancaster

, Matthews

(2012) Lifestyle and genetic contributions to cognitive decline andhippocampal structure and function in healthy aging. CurrAlzheimer Res 9, 436–446.

66.

Blondell

, Hammersley-Mather

, Veerman

(2014) Does physical activity prevent cognitive decline and dementia?: A systematic review and meta-analysis of longitudinal studies. BMC Public Health 14, 510.

67.

Tamura

, Nemoto

, Kawaguchi

, Kato

, Arai

, Kakuma

, Mizukami

, Matsuda

, Soya

, Asada

(2015) Long-term mild-intensity exercise regimen preserves prefrontal cortical volume against aging. Int J Geriatr Psychiatry 30, 686–694.

68.

Gligoroska

, Manchevska

(2012) The effect of physical activity on cognition –physiological mechanisms. Mater Sociomed 24, 198–202.

69.

Kirk-Sanchez

, McGough

(2014) Physical exercise and cognitive performance in the elderly: Current perspectives. Clin Interv Aging 9, 51–62.

70.

Farina

, Tabet

, Rusted

(2014) Habitual physical activity (HPA) as a factor in sustained executive function in Alzheimer-type dementia: A cohort study. Arch Gerontol Geriatr 59, 91–97.

71.

Hsu

, Best

, Chiu

, Nagamatsu

, Voss

, Handy

, Bolandzadeh

, Liu-Ambrose

(2016) Structural neural correlates of impaired mobility and subsequent decline in executive functions: A 12-month prospective study. Exp Gerontol 80, 27–35.

72.

Ble

, Volpato

, Zuliani

, Guralnik

, Bandinelli

, Lauretani

, Bartali

, Maraldi

, Fellin

, Ferrucci

(2005) Executive function correlates with walking speed in older persons: The InCHIANTI study. J Am Geriatr Soc 53, 410–415.

73.

Shimada

, Makizako

, Tsutsumimoto

, Uemura

, Suzuki

(2015) Apolipoprotein E genotype and physical function among older people with mild cognitive impairment. Geriatr Gerontol Int 15, 422–427.

74.

Inzitari

, Newman

, Yaffe

, Boudreau

, De Rekeneire

, Shorr

, Harris

, Rosano

(2007) Gait speed predicts decline in attention and psychomotor speed in older adults: The health aging and body composition study. Neuroepidemiology 29, 156–162.

75.

Montero-Odasso

, Oteng-Amoako

, Speechley

, Gopaul

, Beauchet

, Annweiler

, Muir-Hunter

(2014) The motor signature of mild cognitive impairment: Results from the gait and brain study. J Gerontol A Biol Sci Med Sci 69, 1415–1421.

76.

Luck

, Then

, Luppa

, Schroeter

, Arelin

, Burkhardt

, Thiery

, Loffler

, Villringer

, Riedel-Heller

(2015) Association of the apolipoprotein E genotype with memory performance and executive functioning in cognitively intact elderly. Neuropsychology 29, 382–387.

77.

Paillard-Borg

, Fratiglioni

, Xu

, Winblad

, Wang

H-X

(2012) An active lifestyle postpones dementia onset by more than one year in very old adults. J Alzheimers Dis 31, 835–842.

78.

Brown

, Peiffer

, Taddei

, Lui

, Laws

, Gupta

, Taddei

, Ward

, Rodrigues

, Burnham

, Rainey-Smith

, Villemagne

, Bush

, Ellis

, Masters

, Ames

, Macaulay

, Szoeke

, Rowe

, Martins

(2012) Physical activity and amyloid-beta plasma and brain levels: Results from the Australian Imaging, Biomarkers and Lifestyle Study of Ageing. Mol Psychiatry 18, 875–881.

79.

Toledo

, Vanderstichele

, Figurski

, Aisen

, Petersen

, Weiner

, Jack

Jr , Jagust

, Decarli

, Toga

(2011) Factors affecting Aβ plasma levels and their utility as biomarkers in ADNI. Acta Neuropathol 122, 401–413.

80.

Llado-Saz

, Atienza

, Cantero

(2015) Increased levels of plasma amyloid-beta are related to cortical thinning and cognitive decline in cognitively normal elderly subjects. Neurobiol Aging 36, 2791–2797.

81.

Rosano

, Studenski

, Aizenstein

, Boudreau

, Longstreth

, Newman

(2012) Slower gait, slower information processing and smaller prefrontal area in older adults. Age Ageing 41, 58–64.

82.

Bender

, Raz

(2012) Age-related differences in episodic memory: A synergistic contribution of genetic and physiological vascular risk factors. Neuropsychology 26, 442–450.

83.

Liu

C-C

, Kanekiyo

, Xu

, Bu

(2013) Apolipoprotein E and Alzheimer disease: Risk, mechanisms, and therapy. Nat Rev Neurol 9, 106–118.

84.

Jack

Jr , Wiste

, Weigand

, Rocca

, Knopman

, Mielke

, Lowe

, Senjem

, Gunter

, Preboske

, Pankratz

, Vemuri

, Petersen

(2014) Age-specific population frequencies of cerebralβ-amyloidosis and neurodegeneration among people withnormal cognitive function aged 50-89 years: A cross-sectionalstudy. Lancet Neurol 13, 997–1005.

85.

Espeland

, Erickson

, Neiberg

, Jakicic

, Wadden

, Wing

, Desiderio

, Erus

, Hsieh

M-K

, Davatzikos

, Maschak-Carey

, Laurienti

, Demos-McDermott

, Bryan

(2016) Brain and white matter hyperintensity volumes after 10 years of random assignment to lifestyle intervention. Diabetes Care 39, 764–771.

86.

Fleischman

, Yang

, Arfanakis

, Arvanitakis

, Leurgans

, Turner

, Barnes

, Bennett

, Buchman

(2015) Physical activity, motor function, and white matter hyperintensity burden in healthy older adults. Neurology 84, 1294–1300.