PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
Neurobiol Aging. Author manuscript; available in PMC 2017 April 1.
Published in final edited form as:
PMCID: PMC4792130
NIHMSID: NIHMS756148

Impact of lifestyle dimensions on brain pathology and cognition

Abstract

Single lifestyle factors affect brain biomarkers and cognition. Here we addressed the covariance of various lifestyle elements and investigated their impact on positron emission tomography (PET)-based β-amyloid (Aβ), hippocampal volume and cognitive function in aged controls. Lower Aβ burden was associated with a lifestyle comprising high cognitive engagement and low vascular risk, particularly in apolipoprotein E (APOE) ε4 carriers. While cognitive function was related to high lifetime cognitive engagement and low vascular risk, Aβ load had no relation to current cognitive function. The covariance between high adult socioeconomic status, high education and low smoking prevalence predicted better cognitive function, and this was mediated by larger hippocampal volume. Our data show that lifestyle is a complex construct composed of associated variables, some of which reflect factors operating over the lifespan and others which may be developmental. These factors affect brain health via different pathways, which may reinforce one another. Our findings moreover support the importance of an intellectually enriched lifestyle accompanied by vascular health on both cognition and presumed cerebral mediators of cognitive function.

Keywords: aging, amyloid, APOE, hippocampus, PIB

1 Introduction

Up to half of all worldwide dementia cases may be attributable to modifiable lifestyle factors (Barnes and Yaffe, 2011). Many investigators have attempted to address the effects of lifestyle factors on both the risk of developing Alzheimer’s disease (AD) (Snowdon et al., 1996; Verghese et al., 2003) as well as the specific effects of lifestyle on the brain that may mediate AD risk (Landau et al., 2012; Reed et al., 2014; Valenzuela et al., 2008; Vemuri et al., 2012; Villeneuve et al., 2014). In these studies, factors such as cognitive engagement, physical activity, leisure activities, diet, and disease-related variables that may reflect health behaviors (such as cardiovascular risk) have been examined. “Lifestyle” is thus a very complex set of behaviors and exposures that are related to one another and to many other factors including genetics and socioeconomic status (Jagust and Mormino, 2011). Furthermore, many such lifestyle factors often serve as proxies for unmeasurable constructs such as brain reserve or resilience (Murray et al., 2011).

While the individual impact of various lifestyle factors has been repeatedly investigated, only a few studies have addressed their covariance and combined relationships to brain and cognition. The evaluation of multiple lifestyle variables - education, occupational and leisure cognitive/social activities combined into a composite lifestyle variable - has revealed an effect of lifestyle on cerebral small vessel disease features, neuronal density, cortical thickness and brain weight (Valenzuela et al., 2012), and, on global cognitive function (Vemuri et al., 2012; Vemuri et al., 2014). In statistical models the interrelation between lifestyle variables has been mainly addressed by treating other lifestyle factors as co-variates or moderator variables, by defining lifestyle indices, e.g. by summing binary scores for each lifestyle variable (Aleksandrova et al., 2014; Chiuve et al., 2008; Hamer and Stamatakis, 2008), or by using combinations of a priori grouping of variables and principal component analysis (Vemuri et al., 2012; Vemuri et al., 2014). All of these methods introduce assumptions about the relationships between variables that may not be valid.

The present study approached the problem of lifestyle as a coherent and broader construct of separate domains in order (i) to try to parse how various lifestyle elements co-vary with or differ from one another, and (ii) to investigate how those co-varying lifestyle variables differently affect brain biomarkers and cognition by evaluating multiple pathways at once.

2 Material and Methods

2.1 Selection of participants

The study included 152 cognitively normal older people representing a convenience sample, the Berkeley Aging Cohort Study (BACS). Subjects were recruited from the community by advertisements and word of mouth. Criteria for study inclusion were a Geriatric Depression Scale (GDS) score ≤ 10, a Mini Mental Status Examination (MMSE) score ≥ 25, normal cognitive functions (all cognitive scores within −1.5 standard deviation [SD] of age-, gender-and years of education-adjusted norms on two delayed recall memory tests: California Verbal Learning Test Long Delay Free Recall and Wechsler III Visual Recall Long Delay Free Recall), and an age between 60 and 90 years at the first visit. Each participant underwent a standardized neuropsychological test session, as well as MRI and [11C] Pittsburgh-compound-B (PIB) and [18F] Fludeoxyglucose (FDG)-PET scanning. None of the individuals reported current serious medical, neurologic, or psychiatric illnesses. Participants indicated whether they had a past or present medical history of arterial hypertension, hyperlipidemia and diabetes diagnosed by a healthcare professional and whether they had ever smoked cigarettes. Apolipoprotein E (APOE) genotyping was performed using DNA obtained from blood samples, and subjects were classified as heterozygote or homozygote APOEε4 (APOEε4+), as homozygote APOEε3 (APOEε3+) and as heterozygote or homozygote APOEε2 (APOEε2+) allele carriers. The available sample included all 118 individuals from previous publications (Landau et al., 2012; Wirth et al., 2014b; Wirth et al., 2014a) that focused on the investigation of the relationships between cognitive activity, physical activity and cortical amyloid-β (Aβ) retention. Written informed consent was obtained from each participant in accordance with the Institutional Review Boards of the University of California, Berkeley and Lawrence Berkeley National Laboratory (LBNL).

2.2 Cognitive activity, physical activity and socioeconomic status

Assessment of cognitive activity, physical activity and socioeconomic status (SES) has been reported in detail previously.

Lifetime cognitive activity was assessed using a validated 25-item questionnaire (Wilson et al., 2003) recording the frequency of common cognitively demanding activities (e.g. reading, writing letters) at various age epochs (6, 12, 18, 40 years retrospectively, and current age). Responses were provided on a five-point frequency scale (one = once a year or less, five = every day or almost every day). Three cognitive activity measures were created by calculating the mean of each age epoch for every participant: early life (average over the age epochs 6, 12, and 18), middle life (average over the age epoch 40), and current life (average over the current age epoch) cognitive activity. Based on 118 individuals who fully completed two or more cognitive activity measurements (mean [SD] time interval between two measurements 18 [6.3] months), calculated intraclass correlation coefficients (ICC, test-retest reliability) were 0.94 (95% CI: 0.92, 0.96) for early life cognitive activity, 0.82 (95% CI: 0.73, 0.87) for middle life cognitive activity and 0.79 (95% CI: 0.69, 0.85) for current life cognitive activity. Current physical activity was quantified using the modified Minnesota leisure-time activities questionnaire (Geffken et al., 2001; Taylor et al., 1978). The participants indicated the frequency they participated in physical and leisure activities during a typical, recent two-week period and during how many months per year. Frequency and duration information were multiplied using an activity-specific intensity code indicating calorie expenditure (Taylor et al., 1978) and summed to represent the intensity of physical activity (total kilocalories of energy expended) during the last year. Separately, the subjects assessed their walking miles or walking blocks (10 walking blocks equated with one walking mile) during a typical, recent one-week period, and their hours seated (including e.g. sleeping, eating and any other time sitting down) during a usual 24-hour period. Based on 115 individuals who fully completed two or more physical activity measurements (mean [SD] time interval between two measurements 18 [6.3] months), calculated ICCs were 0.80 (95% CI: 0.71, 0.86) for total kilocalories of energy expended during the last year, 0.82 (95% CI: 0.74, 0.87) for walking miles during a one-week period and 0.68 (95% CI: 0.54, 0.78) for hours seated during a 24hour period.

Socioeconomic status was estimated from the participants’ self-reported professional backgrounds, based on the 1990 occupation classification systems of the U.S. Bureau of the Census (Hauser and Warren, 1997).

2.3 Lifestyle dimensions

We applied a factor analysis for mixed (quantitative, qualitative) data (FAMD) using FactoMineR v1.27 (Husson et al., 2014) to identify uncorrelated clusters of variables that segregate into various lifestyle domains, to intentionally capture covariance patterns between distinct lifestyle variables. FAMD can be roughly considered as a composite of principal component analysis for quantitative variables and multiple correspondence analysis for qualitative data, balancing the influence of both continuous and categorical variables in the analysis. FAMD extracts components or dimensions, which represent clusters of variables that correlate highly with one another.

We included the following lifestyle variables in the analysis: quantitative cognitive lifestyle variables (early life cognitive activity, middle life cognitive activity, current life cognitive activity, socioeconomic status, education), quantitative measures of physical activity (total kilocalories of energy expended during the last year, walking miles during a one-week period, hours seated during a 24-hour period) and qualitative variables referring to the subjects’ vascular risk profile (arterial hypertension, hyperlipidemia, diabetes, smoking status).

For individuals with multiple assessments, variables closest to PET scanning were chosen.

We extracted four components with eigenvalues greater than one, which explained 54% of the variance in the data. To best illustrate results, components were named based on the individual variables that expressed the highest dimension loading score for the respective dimension: dimension 1 - high lifetime cognitive activity and low vascular risk profile (high CogAct/low VascRisk), dimension 2 - low current physical activity and high vascular risk profile (low PhysAct/high VascRisk), dimension 3 - high socioeconomic status /education and low smoking prevalence (high SES/high edu/low smoking), dimension 4 - high vascular risk profile (high VascRisk) (Table 1).

Table 1
Factor loadings of the quantitative lifestyle variables and category loads of the qualitative variables contributing to each lifestyle dimension.

Exploratory factor analysis by stepwise heuristic specification search, and subsequent confirmatory factor analysis, each conducted using the IBM SPSS Amos statistical software package (version 22.0), revealed a good model fit of this four-factor solution (BCC0 [Browne-Cudeck criterion] = 0, BIC0 [Bayes information criterion] = 0, χ2/degrees of freedom [df] = 0.6, Bentler-Bonett normed fit index [NFI] > 0.9, RMSEA [root mean square error of approximation] < 0.03). The largest sample size for which one would accept the hypothesis that this model is correct was 272 (Hoelter’s critical N, p = 0.05).

2.4 Cognitive factor scores

Cognitive factor scores were extracted from a baseline sample of 307 healthy individuals (mean [SD] age 74 [7] years, education 17 [2.1] years, MMSE 29 [1.3], number of males 98 [32 %]) with complete data on various cognitive tests, and with a GDS score ≤ 10, a MMSE score ≥ 25, normal cognitive functions (all cognitive scores within −1.5 SD of age-, gender-and years of education-adjusted norms), and an age between 60 and 90 years at the first visit. This approach was chosen to generate robust and reliable data on a sample as large as possible. The 307 subjects included all 152 participants who were included in the present study. The additional 155 subjects were BACS participants recruited identically but who did not undergo PET imaging.

Using that baseline sample, we performed a maximum likelihood estimation factor analysis with oblique rotation to extract factors representing various cognitive functions/domains from 19 different cognitive tests.

Cognitive tests were selected as appropriate for the factor analysis inclusion if 1) all scores/variables were normally distributed, 2) no excessive co-linearity existed between any two variables, 3) all variables were correlated with at least one other variable, 4) data loss was minimized. Finally, 14 cognitive tests were entered into the factor analysis, from which one (Letter Fluency (Spreen and Benton, 1977)) was subsequently removed as it did not load significantly into any factor.

The scree method (Cattell, 1966), parallel analysis (Humphreys and Montanelli, 1975) and Kaiser’s criterion (Kaiser, 1960) each suggested a three factor solution, which demonstrated a significant goodness-of-fit (χ2 = 67.9, df = 42, p < 0.01, RMSEA ≤ 0.05) and explained 56% of the total data variance.

Table 2 lists the cognitive tests that loaded onto each factor. Based on the pattern of test loading, factor 1 was interpreted as episodic memory, factor 2 as working memory and factor 3 as processing speed/executive function. Factor scores were calculated by multiplying the sum of each z-scored variable and its analysis-derived, variable-specific factor weight.

Table 2
Factor loading scores of the individual cognitive tests contributing to each cognitive factor.

2.5 Co-variates

GDS was related to lifestyle variables/dimensions (see Results). Depression might moreover influence self-assessment, and thus the participants’ reports on lifestyle variables that were based on self-assessment. Statistical models were thus adjusted for GDS, used as measure of depression (Yesavage et al., 1982), to assess robustness of the data and to control for potential bias or confounding effects.

2.6 Neuroimaging biomarkers

2.6.1 MRI and assessment of hippocampal volume

Structural MRI scans were acquired at LBNL on a 1.5 T Magnetom Avanto system (Siemens Medical Systems, Iselin, NJ, USA; 12 channel head-coil, triple mode). High-resolution T1-weighted magnetization-prepared rapid gradient echo axial scans were collected using the following parameters: repetition time = 2110 ms, echo time = 3.6 ms, flip angle = 15°, field of view = 256 × 256 mm, number of slices = 160 with a 50% gap, voxel size = 1 × 1 x1 mm3. For each individual the hippocampal volume was obtained from the native-space MRI scans using the automated subcortical FreeSurfer 5.1 parcellation (Fischl et al., 2002). All reconstructed data were visually checked for segmentation accuracy. Hippocampal segmentation was considered to have failed, (i) if parts of the cerebrospinal fluid (CSF), white matter, cortex or amygdala were identified as hippocampus or, (ii) if parts of the hippocampus were identified as CSF, white matter, cortex or amygdala (Mulder et al., 2014). Segmentation failures were corrected manually in all participants. For subsequent statistical analysis, the manually corrected hippocampal volume segmentation was summed across hemispheres and normalized to the FreeSurfer 5.1–derived total intracranial volume (ICV).

2.6.2 FDG- and PIB-PET

[11C] PIB and [18F] FDG PET scanning was performed with 15 mCi of [11C] PIB followed by 6–10 mCi of [18F] FDG (2 hours timespan between injections) at LBNL (Siemens ECAT EXACT HR PET or Siemens Biograph PET/CT scanner). Images were attenuation corrected using either a transmission scan or CT. PIB data were collected over a 90-minute measurement interval (four frames × 15 s, eight × 30 s, nine × 60 s, two × 180 s, eight × 300 s, three × 600 s), and FDG over a 30-minute measurement interval (six frames × 5 minutes, starting 30 minutes postinjection). After smoothing (4 mm Gaussian kernel with scatter correction) and preprocessing, FDG-images were intensity normalized to the mean glucose metabolism within the pons, manually edited from the subcortical FreeSurfer 5.1 brainstem segmentation (Fischl et al., 2002). PIB data processing used Logan graphical analysis with a cerebellar gray reference region to construct distribution volume ratio (DVR) images (Landau et al., 2012; Wirth et al., 2014b).

2.6.3 Assessment of PIB index and mean FDG

A global cortical PIB index (continuous variable) was calculated to estimate the subjects’ Aβ burden. The native-space PIB-PET images were co-registered to the native-space MRI scans and a mean DVR value was derived from FreeSurfer parcellated frontal, temporal, parietal, and anterior/posterior cingulate regions of interest (ROIs, for details see supplemental Table 1) using the automated Desikan-Killiany parcellation (Desikan et al., 2006).

FDG mean uptake was determined within a set of predefined and validated metaROIs (for details see (Landau et al., 2011)). ROI-specific values were averaged and for each subject, a composite ROI measure (mean FDG) was generated from the mean of the metaROIs (right and left inferior temporal and lateral parietal regions, bilateral posterior cingulate-precuneus region) relative to the mean of the pons (reference region).

2.7 Statistical analysis

Using the IBM SPSS Amos statistical software package (version 22.0) direct effects (i) of lifestyle dimensions on biomarkers, (ii) of lifestyle dimensions on cognitive functions, and (iii) of biomarkers on cognitive functions, as well as indirect (mediation) effects of lifestyle dimensions on cognitive factor scores with biomarkers as mediator variables were modeled. Initial models included at once (A) all four lifestyle dimensions (exogenous composites), all three biomarkers (intermediate endogenous observed variables: PIB index, mean FDG, hippocampal volume, and the residual covariances among the three biomarkers) and our three cognitive factor scores (downstream endogenous composites: episodic memory, working memory and processing speed/executive function, and the residual covariances among the three cognitive factor scores), or (B) all four lifestyle dimensions, all three biomarkers and their residual covariances as well as global cognition as a downstream endogenous latent variable (composed of episodic memory, working memory and processing speed/executive function). Both models were overidentified (for A df = 6, for B df = 21) and they did exhibit good measures of fit (χ2/df < 1, NFI > 0.9, RMSEA < 0.01). Reduced models were created by dropping several paths, resulting in further df increase (for A df = 29, for B df = 29) while maintaining same good model fit. Simplified models are displayed in Figure 1A & B and all final analyses were conducted on those reduced models. Multi-group analysis included the pairwise comparison of path coefficients (= regression weights) between APOEε3 carriers (Eε3Eε3) and APOEε4 carriers; critical ratios ≤ or ≥ 1.96 were deemed to be significant at p < 0.05. Unstandardized (B) and standardized estimates (β) of direct and indirect effects are reported, and indirect effects were evaluated for significance at p ≤ 0.05 using 95% bias-corrected confidence intervals (Monte Carlo method, 5000 bootstrap samples). Unless otherwise reported all direct and indirect effects remained significant after outlier exclusion (z-score < 2.68 or > 2.68). In cases where significant direct lifestyle dimension effects were found, follow-up linear regression models were computed to understand what lifestyle variable was driving the significant effect; in each model, the respective outcome variable was regressed simultaneously on all lifestyle variables. Linear regression models were conducted using SPSS, version 22.0 and statistical significance was defined as p ≤ 0.05. All statistics were performed using the standardized regression residuals of each variable controlling for age and gender, and results refer to those analyses. Unless otherwise reported, direct and indirect effects, as well as critical ratios derived from multi-group analysis, remained unchanged when the standardized regression residuals of each variable were additionally controlled for GDS.

Figure 1
Reduced final models used for path analysis and multi-group analysis

3 Results

3.1 Characteristics of participants

Descriptive statistics of the participants’ demographics, lifestyle variables and biomarkers are given in Table 3. Self-reports of physician-diagnosed arterial hypertension, hyperlipidemia and diabetes were present in respective 34%, 38% and 7% of all investigated cases, and a history of cigarette smoking was indicated by 30% of the participants. Forty-two participants (29%) had at least one APOEε4 allele, with two (1.4%) homozygous APOEε4 carriers; 13 (9%) of participants had at least one APOEε2 allele, and there were no homozygous APOEε2 carriers; 90 (63%) had two APOEε3 alleles. APOE carrier status was unknown in 9 subjects.

Table 3
Characteristics of participants.

Compared to APOEε3 subjects, APOEε4 carriers were significantly younger (t (101) = −2.2, p = 0.03), had a higher PIB index (t (46) = 3.6, p = 0.001), larger hippocampal volumes (t (118) = 2.4, p = 0.02) and higher processing speed/executive function scores (t (105) = 2.3, p = 0.02). While higher processing speed/executive function scores in APOEε4 carriers were explained by their younger age (t (105) = 1.7, p = 0.09), larger hippocampal volumes were not (t (118) = 2.1, p = 0.04) (age and gender adjusted independent samples t-test). Further demographics, lifestyle variables, lifestyle dimensions, mean FDG and the remaining cognitive factor scores did not differ between APOEε4 and APOEε3 carriers (independent samples t-test, χ2-test) (supplemental Table 2).

Age was positively correlated with lifestyle dimension low PhysAct/high VascRisk (r = 0.2, p < 0.01), and negatively correlated with hippocampal volume (r = −0.34, p < 0.01), mean FDG (r = −0.27, p < 0.01), episodic memory (r = −0.34, p < 0.01), processing speed/executive function (r = −0.34, p < 0.01) and global cognition (r = −0.29, p < 0.01).

GDS was positively correlated with lifestyle dimension low PhysAct/high VascRisk (r = 0.23, p < 0.01) and with hours seated during a 24-hour period (r = 0.18, p = 0.03); smokers had higher GDS (t (149) = 2.4, p = 0.02) compared to non-smokers.

Subjects who reported arterial hypertension were significantly older than participants without arterial hypertension (t (150) = 2.6, p = 0.01). Compared to men, women had significantly larger relative hippocampal volumes (t (138) = 2.6, p = 0.01), higher mean FDG values (t (144) = 5.1, p < 0.01) and episodic memory scores (t (114) = 3.1, p < 0.01), a higher cigarette smoking prevalence (χ2 (1) = 4.1, p = 0.04), but lower scores for lifestyle dimension high SES/high edu/low smoking (t (150) = −2.7, p < 0.01).

There were no further relationships between age, gender, GDS and the remaining lifestyle variables, lifestyle dimensions, biomarkers or cognitive factor scores (nonparametric correlations, independent samples t-test, χ2-test).

3.2 Direct effects

Statistical models revealed a significant inverse effect of lifestyle dimension high CogAct/low VascRisk on PIB index (B = −0.23, β = −0.23, p < 0.01; Figure 2A), which was most strongly related to a significant association between middle life cognitive activity and PIB (B = −0.25, β = −0.24, p < 0.04). A significant positive effect of lifestyle dimension high SES/high edu/low smoking on hippocampal volume (B = 0.2, β = 0.2, p < 0.02; Figure 2B) was also found; when hippocampal volume was regressed simultaneously on education, socioeconomic status and smoking, no single lifestyle variable explained the relationship between lifestyle dimension high SES/high edu/low smoking and hippocampal volume. Excluding socioeconomic status from the regression model, however, results in a significant association between education and hippocampal volume (B = 0.22, β = 0.21, p < 0.02).

Figure 2
Associations between lifestyle dimensions, biomarkers and cognitive domains

Lifestyle dimension high CogAct/low VascRisk was moreover related to better episodic memory (B = 0.22, β = 0.23, p < 0.01), resulting from a significant association between current life cognitive activity and episodic memory (B = 0.26, β = 0.26, p < 0.03), and to better global cognition (B = 0.16, β = 0.29, p < 0.04), resulting from a significant relationship between early life cognitive activity and global cognition (B = 0.26, β = 0.25, p = 0.02) (Figure 2C).

There were additional significant effects of hippocampal volume on episodic memory (B = 0.24, β = 0.25, p < 0.01), on processing speed/executive function (B = 0.21, β = 0.21, p < 0.02) and on global cognition (B = 0.19, β = 0.35, p < 0.01) (Figure 2D).

3.3 Mediation effects

There were significant indirect effects of lifestyle dimension high SES/high edu/low smoking on episodic memory (β = 0.05, 95% CI: 0.01, 0.11, p ≤ 0.05), on processing speed/executive function (β = 0.04, 95% CI: < 0.01, 0.1, p ≤ 0.05) and on global cognition (β = 0.07, 95% CI: < 0.01, 0.09, p ≤ 0.05), mediated by hippocampal volume (Figure 3A – C). After outlier exclusion only the indirect effect of high SES/high edu/low smoking on global cognition remained significant (β = 0.06, 95% CI: < 0.01, 0.08, p ≤ 0.05), while the indirect effects of high SES/high edu/low smoking on episodic memory (β = 0.04, 95% CI: < −0.01, 0.1, p ≥ 0.05) and on processing speed/executive function (β = 0.03, 95% CI: < −0.01, 0.09, p ≥ 0.05) did not.

Figure 3
Indirect effects of lifestyle dimension high SES/high edu/low smoking on cognitive domains

Table 4 provides an overview of all direct, indirect, and total effects derived from the mediation models.

Table 4
Standardized direct, indirect and total effects derived from mediation models.

3.4 Multi-group analysis

Multi-group analysis revealed a significant inverse association between high CogAct/low VascRisk and PIB index in APOEε4 carriers (B = −0.51, β = −0.37, p < 0.01), but not in APOEε3 carriers (B = −0.04, β = −0.06, p = 0.59). Difference on path coefficients (= critical ratio) was 2.3 (p < 0.05), which remained on a trend-level after outlier exclusion (critical ratio = 1.9). Detailed results of the multi-group analysis are presented in supplemental Table 3.

4 Discussion

In cognitively normal elderly, we identified dimensions of lifestyle characterized by common variance between (1) high lifetime cognitive activity and a low vascular risk profile, (2) low current physical exercise and high cardiovascular risk, (3) high socioeconomic status, high education and low smoking status, and (4) the simultaneous presence of various vascular risk factors. Beside their explanatory power for certain “lifestyles” and behaviors, these dimensions were particularly associated with brain pathology and cognitive functions. The composite of lifetime cognitive activity and a low vascular risk profile was related to lower cortical Aβ retention, and preserved episodic memory and global cognition. Additionally, the composite of high socioeconomic status, high education and low lifetime smoking status was associated with larger hippocampal volumes, which mediated indirect effects on episodic memory, processing speed and global cognition (Figure 4). The composite of high lifetime cognitive activity and a low vascular risk profile was related to lower Aβ burden in APOEε4 carriers.

Figure 4
Overview of significant associations between lifestyle dimensions, amyloid retention, hippocampal volume and cognitive functions

Our results suggest that lifestyle should be thought of as a complex construct composed of discrete elements, including cognitive and physical activity, education, socioeconomic status, and vascular risk factors. Importantly, in our study, not all lifestyle variables were related to one another, but certain lifestyle variables demonstrated meaningful covariance reflecting interrelated lifestyle and behavioral patterns. Of note is the strong common variance between cognitive activity and vascular health, and independently from this particular lifestyle, the strong covariance between physical inactivity and vascular risk.

Our data further support the growing literature reporting on an interaction between genetics and lifestyle on markers of brain health and disease. We replicate previous findings from our cohort that cortical Aβ load in elderly APOEε4 carriers can be moderated by lifetime cognitive activity (Wirth et al., 2014b). This supports the notion that individuals at higher genetic risk for AD may be especially sensitive to the biological effects of lifestyle conditions and behaviors.

Our data also corroborate previous findings reporting on the association between vascular risk factors and cortical Aβ accumulation (Reed et al., 2012; Reed et al., 2014; Rodrigue et al., 2013). We found Aβ retention was not independently associated with vascular risk variables, but was instead associated with a composite variable containing information associated with both leisure activity and vascular risk profile, especially in subjects at higher genetic risk for AD. Vascular health helps to maintain paravascular amyloid drainage/clearance mechanisms, and lifelong cognitive activity has been conjectured to result in lower neural activity-linked Aβ release (Hawkes et al., 2012; Iadecola, 2014; Jagust and Mormino, 2011; Snyder et al., 2014; Zlokovic, 2008). Those multiple mechanisms may be important in lowering amyloid burden, or mitigating its increase, especially in APOEε4 carriers. Vascular risk may therefore have played a role in prior findings stating relationships between cognitive activity and cortical Aβ load (Landau et al., 2012; Wirth et al., 2014b; Wirth et al., 2014a) or, alternatively, cognitive activity may have played a role in previous reports of the effects of vascular risk (Reed et al., 2012; Reed et al., 2014; Rodrigue et al., 2013).

The relationship between education, adult socioeconomic status, smoking status and hippocampal volume supports common evidence that early life environmental influences may affect brain structure development. Higher childhood socioeconomic status has previously been reported as associated with larger hippocampal volume and brain size in late life (Staff et al., 2012). Childhood socioeconomic status seems moreover to predict smoking habits in adulthood (Giesinger et al., 2014; Laitinen et al., 2013) and duration of education, which in turn is a proxy of adult socioeconomic status (Staff et al., 2012). Our results thus suggest that enduring developmental effects on brain structure may be explained by the association of early life experience with education, adult socioeconomic status and adulthood smoking prevalence.

Based on our results, we suggest that various lifestyle dimensions and brain biomarkers are related via two distinct pathways. The first pathway includes the association between lifelong and current life conditions and behaviors on amyloid burden and cognition, which we consider as a “dynamic lifestyle” pathway because it is presumably changing over the lifecourse. While the dynamic pathway affects episodic memory and global cognition, those associations are not explained by the pathway’s Aβ lowering impact but rather perhaps through a shared intermediate mechanism. Such mechanisms might include neural compensation (Elman et al., 2014) or neural efficiency (Jagust and Mormino, 2011). The second pathway includes the ongoing relationship between early life environmental influences and hippocampal volume, and this can be considered as a “static developmental” pathway. In our study, this latter “static” pathway was not related to cortical amyloid load, and its substantial association with various cognitive domains was mediated by sustained hippocampal volume. The “static development” pathway may confer protective benefits on brain structure, which keeps subjects in the normal category even in the presence of neurodegeneration/amyloid burden (Brayne et al., 2010; Erten-Lyons et al., 2009). These two pathways may reinforce one another, especially in APOEε4 carriers. As an example, larger hippocampal volumes in APOEε4 positives – compared to APOEε3 carriers - may result from subjects with better static endowments who have remained cognitively normal despite an Aβ load.

Our study has several limitations. Our cohort is characterized by particularly high levels of lifetime leisure activity, socioeconomic status and education, and participant recruitment by advertisement or word of mouth may have excluded subjects of lower socioeconomic status. In addition, our sample exhibits a lower variance in education level and a lower vascular risk prevalence compared to more population-based samples. The lack of associations between mean FDG and e.g. cognitive function, for instance, may result from missing meaningful variations in the predefined AD signature meta-ROIs (used to generate mean FDG) in that particularly healthy and well-educated cohort. The generalizability of some of the results presented has thus yet to be demonstrated. Furthermore, no assumption on causality can be drawn from the covariance pattern between various lifestyle variables, and the causality assumption of the detected associations between lifestyle, biomarkers and cognition cannot be proven by the present cross-sectional study design and must be confirmed by longitudinal surveys. Another study limitation is that certain lifestyle variable data - in particular, cognitive activity, physical exercise and smoking status - are self-reported, and thus might have been influenced by undetected bias. A final limitation is presented by the fact that our sample size may be too small to detect more subtle relationships between various lifestyle factors, brain biomarkers and cognition. For example, a recent meta-analysis of 16 cross-sectional studies including around 1,300 cognitively normal elderly revealed that amyloid burden has a modest, but significant effect on episodic memory (Hedden et al., 2013). However, a sample size nearly three times larger than the number of participants included in the present analysis would be necessary to detect those small effect sizes (Hedden et al., 2013). This may, at least in part, account for the absent association between Aβ load and cognition in our data set.

In conclusion, our data show that particular modifiable lifestyle factors share common variance that together result in various independent healthier or less healthy lifestyle patterns. These lifestyle patterns are related to brain health via a “dynamic lifestyle” pathway, including the association between lifelong cognitive activities and amyloid retention, in particular in APOEε4 carriers, and via a “static developmental” pathway, including the associations between early life experiences and hippocampal volume and cognition. Overall, our results emphasize the mutual importance of an intellectually enriched lifestyle accompanied by vascular health on both cognition and presumed neural mediators of cognition.

Highlights

  • Lifestyle is a complex construct of distinct covarying lifestyle elements
  • Lifestyle affects the brain via two pathways that reinforce one another
  • Cortical Aβ retention is related to APOEε4 carrier status
  • This relationship is modified by leisure activities and vascular risk factors
  • Hippocampal volume is related to early life environmental influences

Supplementary Material

Acknowledgments

This work was supported by a German Research Foundation (DFG) research fellowship (SCHR 1418/3-1) offered to SS and NIH grant AG034570.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

1. Aleksandrova K, Pischon T, Jenab M, Bueno-de-Mesquita H, Fedirko V, Norat T, Romaguera D, Knuppel S, Boutron-Ruault MC, Dossus L, Dartois L, Kaaks R, Li K, Tjonneland A, Overvad K, Quiros J, Buckland G, Sanchez M, Dorronsoro M, Chirlaque MD, Barricarte A, Khaw KT, Wareham NJ, Bradbury KE, Trichopoulou A, Lagiou P, Trichopoulos D, Palli D, Krogh V, Tumino R, Naccarati A, Panico S, Siersema PD, Peeters P, Ljuslinder I, Johansson I, Ericson U, Ohlsson B, Weiderpass E, Skeie G, Borch K, Rinaldi S, Romieu I, Kong J, Gunter MJ, Ward HA, Riboli E, Boeing H. Combined impact of healthy lifestyle factors on colorectal cancer: a large European cohort study. BMC Med. 2014;12(1):168. [PMC free article] [PubMed]
2. Barnes DE, Yaffe K. The projected effect of risk factor reduction on Alzheimer’s disease prevalence. Lancet Neurol. 2011;10(9):819–828. [PMC free article] [PubMed]
3. Brayne C, Ince PG, Keage HA, McKeith IG, Matthews FE, Polvikoski T, Sulkava R. Education, the brain and dementia: neuroprotection or compensation? Brain. 2010;133(Pt 8):2210–2216. [PubMed]
4. Cattell RB. The scree test for the number of factors. Multivariate Behavorial Research. 1966;1(2):245–276. [PubMed]
5. Chiuve SE, Rexrode KM, Spiegelman D, Logroscino G, Manson JE, Rimm EB. Primary prevention of stroke by healthy lifestyle. Circulation. 2008;118(9):947–954. [PMC free article] [PubMed]
6. Desikan RS, Segonne F, Fischl B, Quinn BT, Dickerson BC, Blacker D, Buckner RL, Dale AM, Maguire RP, Hyman BT, Albert MS, Killiany RJ. An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. Neuroimage. 2006;31(3):968–980. [PubMed]
7. Elman JA, Oh H, Madison CM, Baker SL, Vogel JW, Marks SM, Crowley S, O’Neil JP, Jagust WJ. Neural compensation in older people with brain amyloid-beta deposition. Nat Neurosci. 2014;17(10):1316–1318. [PMC free article] [PubMed]
8. Erten-Lyons D, Woltjer RL, Dodge H, Nixon R, Vorobik R, Calvert JF, Leahy M, Montine T, Kaye J. Factors associated with resistance to dementia despite high Alzheimer disease pathology. Neurology. 2009;72(4):354–360. [PMC free article] [PubMed]
9. Fischl B, Salat DH, Busa E, Albert M, Dieterich M, Haselgrove C, van der KA, Killiany R, Kennedy D, Klaveness S, Montillo A, Makris N, Rosen B, Dale AM. Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain. Neuron. 2002;33(3):341–355. [PubMed]
10. Geffken DF, Cushman M, Burke GL, Polak JF, Sakkinen PA, Tracy RP. Association between physical activity and markers of inflammation in a healthy elderly population. Am J Epidemiol. 2001;153(3):242–250. [PubMed]
11. Giesinger I, Goldblatt P, Howden-Chapman P, Marmot M, Kuh D, Brunner E. Association of socioeconomic position with smoking and mortality: the contribution of early life circumstances in the 1946 birth cohort. J Epidemiol Community Health. 2014;68(3):275–279. [PMC free article] [PubMed]
12. Hamer M, Stamatakis E. The accumulative effects of modifiable risk factors on inflammation and haemostasis. Brain Behav Immun. 2008;22(7):1041–1043. [PubMed]
13. Hauser RM, Warren JR. Socioeconomic indexes for occupations: a review, update, and critique. Soc Methodol. 1997;27:177–298.
14. Hawkes CA, Sullivan PM, Hands S, Weller RO, Nicoll JA, Carare RO. Disruption of arterial perivascular drainage of amyloid-beta from the brains of mice expressing the human APOE epsilon4 allele. PLoS One. 2012;7(7):e41636. [PMC free article] [PubMed]
15. Hedden T, Oh H, Younger AP, Patel TA. Meta-analysis of amyloid-cognition relations in cognitively normal older adults. Neurology. 2013;80(14):1341–1348. [PMC free article] [PubMed]
16. Humphreys LG, Montanelli RG. An investigation of the parallel analysis criterion for determining the number of common factors. Multivariate Behavioral Research. 1975;10:193–206.
17. FactoMine R. Multivariate Exploratory Data Analysis and Data Mining with R. 2014
18. Iadecola C. Hypertension and dementia. Hypertension. 2014;64(1):3–5. [PMC free article] [PubMed]
19. Jagust WJ, Mormino EC. Lifespan brain activity, beta-amyloid, and Alzheimer’s disease. Trends Cogn Sci. 2011;15(11):520–526. [PMC free article] [PubMed]
20. Kaiser H. The application of electronic computers to factor analysis. Educ Psychol Meas. 1960;20:141–151.
21. Laitinen TT, Pahkala K, Venn A, Woo JG, Oikonen M, Dwyer T, Mikkila V, Hutri-Kahonen N, Smith KJ, Gall SL, Morrison JA, Viikari JS, Raitakari OT, Magnussen CG, Juonala M. Childhood lifestyle and clinical determinants of adult ideal cardiovascular health: the Cardiovascular Risk in Young Finns Study, the Childhood Determinants of Adult Health Study, the Princeton Follow-Up Study. Int J Cardiol. 2013;169(2):126–132. [PMC free article] [PubMed]
22. Landau SM, Harvey D, Madison CM, Koeppe RA, Reiman EM, Foster NL, Weiner MW, Jagust WJ. Associations between cognitive, functional, and FDG-PET measures of decline in AD and MCI. Neurobiol Aging. 2011;32(7):1207–1218. [PMC free article] [PubMed]
23. Landau SM, Marks SM, Mormino EC, Rabinovici GD, Oh H, O’Neil JP, Wilson RS, Jagust WJ. Association of lifetime cognitive engagement and low beta-amyloid deposition. Arch Neurol. 2012;69(5):623–629. [PMC free article] [PubMed]
24. Mulder ER, de Jong RA, Knol DL, van Schijndel RA, Cover KS, Visser PJ, Barkhof F, Vrenken H. Hippocampal volume change measurement: quantitative assessment of the reproducibility of expert manual outlining and the automated methods FreeSurfer and FIRST. Neuroimage. 2014;92:169–181. [PubMed]
25. Murray AD, Staff RT, McNeil CJ, Salarirad S, Ahearn TS, Mustafa N, Whalley LJ. The balance between cognitive reserve and brain imaging biomarkers of cerebrovascular and Alzheimer’s diseases. Brain. 2011;134(Pt 12):3687–3696. [PubMed]
26. Reed BR, Marchant NL, Jagust WJ, DeCarli CC, Mack W, Chui HC. Coronary risk correlates with cerebral amyloid deposition. Neurobiol Aging. 2012;33(9):1979–1987. [PMC free article] [PubMed]
27. Reed BR, Villeneuve S, Mack W, DeCarli C, Chui HC, Jagust W. Associations between serum cholesterol levels and cerebral amyloidosis. JAMA Neurol. 2014;71(2):195–200. [PMC free article] [PubMed]
28. Rodrigue KM, Rieck JR, Kennedy KM, Devous MD, Diaz-Arrastia R, Park DC., Sr Risk factors for beta-amyloid deposition in healthy aging: vascular and genetic effects. JAMA Neurol. 2013;70(5):600–606. [PMC free article] [PubMed]
29. Serrano-Pozo A, Qian J, Monsell SE, Betensky RA, Hyman BT. APOEepsilon2 is associated with milder clinical and pathological Alzheimer disease. Ann Neurol. 2015;77(6):917–929. [PMC free article] [PubMed]
30. Snowdon DA, Kemper SJ, Mortimer JA, Greiner LH, Wekstein DR, Markesbery WR. Linguistic ability in early life and cognitive function Alzheimer’s disease in late life Findings from the Nun Study. JAMA. 1996;275(7):528–532. [PubMed]
31. Snyder HM, Corriveau RA, Craft S, Faber JE, Greenberg SM, Knopman D, Lamb BT, Montine TJ, Nedergaard M, Schaffer CB, Schneider JA, Wellington C, Wilcock DM, Zipfel GJ, Zlokovic B, Bain LJ, Bosetti F, Galis ZS, Koroshetz W, Carrillo MC. Vascular contributions to cognitive impairment and dementia including Alzheimer’s disease. Alzheimers Dement. 2014 [PMC free article] [PubMed]
32. Spreen O, Benton AL. Neurosensory Center Comprehensive Examination for Aphasia: Manual of instructions (NCCEA)(rev. ed.) Victoria, BC: University of Victoria; 1977.
33. Staff RT, Murray AD, Ahearn TS, Mustafa N, Fox HC, Whalley LJ. Childhood socioeconomic status and adult brain size: childhood socioeconomic status influences adult hippocampal size. Ann Neurol. 2012;71(5):653–660. [PubMed]
34. Taylor HL, Jacobs DR, Schucker B, Knudsen J, Leon AS, Debacker G., Jr A questionnaire for the assessment of leisure time physical activities. J Chronic Dis. 1978;31(12):741–755. [PubMed]
35. Valenzuela MJ, Matthews FE, Brayne C, Ince P, Halliday G, Kril JJ, Dalton MA, Richardson K, Forster G, Sachdev PS. Multiple biological pathways link cognitive lifestyle to protection from dementia. Biol Psychiatry. 2012;71(9):783–791. [PubMed]
36. Valenzuela MJ, Sachdev P, Wen W, Chen X, Brodaty H. Lifespan mental activity predicts diminished rate of hippocampal atrophy. PLoS One. 2008;3(7):e2598. [PMC free article] [PubMed]
37. Vemuri P, Lesnick TG, Przybelski SA, Knopman DS, Roberts RO, Lowe VJ, Kantarci K, Senjem ML, Gunter JL, Boeve BF, Petersen RC, Jack CR., Jr Effect of lifestyle activities on Alzheimer disease biomarkers and cognition. Ann Neurol. 2012;72(5):730–738. [PMC free article] [PubMed]
38. Vemuri P, Lesnick TG, Przybelski SA, Machulda M, Knopman DS, Mielke MM, Roberts RO, Geda YE, Rocca WA, Petersen RC, Jack CR., Jr Association of lifetime intellectual enrichment with cognitive decline in the older population. JAMA Neurol. 2014;71(8):1017–1024. [PMC free article] [PubMed]
39. Verghese J, Lipton RB, Katz MJ, Hall CB, Derby CA, Kuslansky G, Ambrose AF, Sliwinski M, Buschke H. Leisure activities and the risk of dementia in the elderly. N Engl J Med. 2003;348(25):2508–2516. [PubMed]
40. Villeneuve S, Reed BR, Madison CM, Wirth M, Marchant NL, Kriger S, Mack WJ, Sanossian N, DeCarli C, Chui HC, Weiner MW, Jagust WJ. Vascular risk and Abeta interact to reduce cortical thickness in AD vulnerable brain regions. Neurology. 2014;83(1):40–47. [PMC free article] [PubMed]
41. Wilson RS, Barnes L, Bennett D. Assessment of lifetime participation in cognitively stimulating activities. J Clin Exp Neuropsychol. 2003;25(5):634–642. [PubMed]
42. Wirth M, Haase CM, Villeneuve S, Vogel J, Jagust WJ. Neuroprotective pathways: lifestyle activity, brain pathology, and cognition in cognitively normal older adults. Neurobiol Aging. 2014a;35(8):1873–1882. [PMC free article] [PubMed]
43. Wirth M, Villeneuve S, La Joie R, Marks SM, Jagust WJ. Gene-environment interactions: lifetime cognitive activity, APOE genotype, and beta-amyloid burden. J Neurosci. 2014b;34(25):8612–8617. [PMC free article] [PubMed]
44. Zlokovic BV. The blood-brain barrier in health and chronic neurodegenerative disorders. Neuron. 2008;57(2):178–201. [PubMed]