Cardiorespiratory Fitness and Cognition: Longitudinal Associations in the FINGER Study

Abstract

Background:

Previous studies have found positive associations between cardiorespiratory fitness (CRF) and cognitive performance in older people but data are inconsistent and have methodological limitations.

Objective:

Our aim was to study the longitudinal associations of CRF with executive functions, processing speed and memory as well as with the overall cognitive function in older people at risk for cognitive impairment.

Methods:

Participants (n = 421), mean age 69.0, were a sub-sample of The Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability (FINGER). To be eligible, individuals were required to be 60–77 years old with a CAIDE (Cardiovascular Risk Factors, Aging and Dementia) Dementia Risk Score of at least 6 points and cognition at mean level or slightly lower than expected for age. CRF was assessed as peak oxygen uptake (VO_2peak, L/min) measured directly in a symptom-limited maximal exercise test on cycle ergometer at baseline and at 24 months. Cognitive performance was assessed using an extensive neuropsychological test battery (NTB) at baseline and at 24 months. NTB data were standardized to Z scores, and analyzed with the linear mixed model.

Results:

Over two years, VO_2peak was associated with NTB total score (β= 0.12, p = 0.01), executive functions (β= 0.16, p = 0.01), and processing speed (β= 0.25, p < 0.001), but not with memory (β= 0.11, p = 0.12).

Conclusion:

Over two years follow-up, CRF was associated with executive functions and processing speed, and was related also to the overall cognitive function.

Keywords

Aged cardiorespiratory fitness cognition executive function memory

INTRODUCTION

Late-life cognitive impairment and dementia due to neurodegenerative and vascular disorders are common challenges in our society. With a growing number of people suffering from dementia, research identifying its modifiable risk factors has become increasingly important [1]. One such risk factor is low cardiorespiratory fitness (CRF) which can be, however, improved by aerobic exercise. Previous studies in healthy older adults have reported association between high CRF and better overall cognitive function [2 –5] as well as memory [2 , 5–10], processing speed [5 , 11], and particularly, executive functions [2 , 13]. Furthermore, objectively measured CRF levels at midlife seem to associate with the development of dementia at later life [14].

While most of the evidence in older adults suggests that high CRF relates to better cognitive function, not all the studies, however, have found the association after controlling for age [15]. Previous studies are based mainly on cross-sectional data and only a few studies [2 , 10] have investigated the longitudinal associations between CRF and cognitive functions. Furthermore, these studies have methodological limitations restricting conclusions regarding cognitive change over time. For example, CRF and cognition has been assessed only at one time point [2, 8] or only one specific cognitive domain has been considered [10].

Our aim was to study the longitudinal associations between CRF, one of the predefined secondary outcomes of The Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability (FINGER) [16], and the primary and secondary cognitive outcomes during 2-year follow-up in at-risk older people. Specifically, we hypothesized that CRF longitudinally associates with executive functions, processing speed, and memory as well as with the overall cognitive function.

METHODS

Study design and participants

We used longitudinal data from the FINGER study, a 2-year population-based multidomain randomized controlled trial (RCT) (n = 1,260) executed in six centers in Finland (Helsinki, Vantaa, Kuopio, Oulu, Seinäjoki, and Turku). Participants were recruited from previous population-based non-interventional surveys [17, 18]. To be eligible for participation in the trial, individuals were required to be 60–77 years old and have a Cardiovascular Risk Factors, Aging and Dementia (CAIDE) Dementia Risk Score [15] of 6 points or higher (score based on age, sex, education, resting systolic blood pressure, body-mass index [BMI], total cholesterol, and physical activity; range 0–15 points). Cognitive screening was done with the Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) neuropsychological battery [19], and participants had to meet at least one of the following criteria: word list memory task (sum of three immediate 10-word list recalls) results of 19 words or fewer; word list delayed retention of 75% or less; or MMSE of 26 points or less out of 30 points. These criteria selected individuals with cognitive performance at the mean level or slightly lower than expected for age according to Finnish population norms [20]. Exclusion criteria have been previously described [16]. The protocol and the intervention of the FINGER study are presented in Supplementary Figure 1. The intervention group received four intervention components (nutrition, physical exercise, cognitive training, and management of metabolic and vascular risk factors) previously described in detail [16]. The control group received regular health advice. The FINGER study was approved by the coordinating ethics committee of the Hospital District of Helsinki and Uusimaa, and conducted according to Declaration of Helsinki. Participants gave written informed consent at screening and baseline visits.

A consortium diagram of the present sub-study is shown in Supplementary Figure 2. As predefined, CRF was intended to measure for all 443 participants [16] in 2 (Kuopio and Oulu) of the 6 participating centers. Baseline data on CRF were missing from 21 participants and they were excluded from the analyses. After excluding one more participant because of the technical error during the CRF measurement at 24 months, our sub-sample included 421 participants derived equally from the intervention and control groups.

Assessment of cardiorespiratory fitness

CRF was assessed as peak oxygen uptake (VO_2peak, L/min) measured directly by the breath-by-breath method using a paramagnetic oxygen analyzer (Vmax 29, SensorMedics Corporation, Yorba Linda, CA, USA) during a maximal symptom-limited exercise test on an electrically-braked cycle ergometer (Ergoline, Bitz, Germany) at baseline and at 24 months. VO_2peak is straightforward term for the highest value of VO₂ achieved on the particular incremental exercise test. The exercise test was considered maximal if the respiratory exchange ratio was ≥1.0 or if the test had to be terminated due to cardiovascular or pulmonary reasons, muscle fatigue, or overall fatigue. Participants were verbally encouraged to reach their maximum. Test started with a 6-min sitting period in the saddle followed by a 5-min warm-up at 0 watts (W). Thereafter, workload was increased 20W for men and 15W for women every 2 min. Self-reported ratings of perceived exertion were collected at 2-min intervals using the 20-item Borg Scale.

Assessment of cognition

A thorough cognitive assessment with standard neuropsychological tests (an extended version of the neuropsychological test battery [NTB]) [21] was done at baseline, at 12, and at 24 months by study psychologists. In the present study, however, we used cognition data only at the time points where CRF data were also available (i.e., baseline and 24 months). Outcomes were the change in cognitive performance measured with NTB total score, a composite score based on results from 14 tests (calculated as Z scores standardized to the baseline mean and standard deviation [SD], with higher scores indicating better performance) [16], and NTB domain Z scores for executive functions, processing speed, and memory. The executive functions domain included Category fluency test, Digit span test, Concept shifting test (condition C), Trail making test (shifting score: time in condition B – time in condition A), and a 40-stimulus version of the Stroop test (interference score: time in condition 3 – time in condition 2). The processing speed domain included Letter digit substitution test, Concept shifting test (condition A), and Stroop test (condition 2). The memory domain included Visual paired associates test-, immediate and delayed recall; Logical memory test immediate and delayed recall from the Wechsler Memory Scale-III; and word list learning and delayed recall from the CERAD test battery.

Other measurements

Symptoms of depression were assessed with the Zung self-rating depression scale (scores range from 25–100 with 25–49 as normal range) [22]. Height was measured in a standing position without shoes and systolic blood pressure (two measurements using a validated automatic device (Microlife WatchBP Office) in a sitting position, using the right arm, after 10 min of rest) by a trained study nurse. Total serum cholesterol and plasma glucose concentrations were determined enzymatically using commercial reagents from Abbott Laboratories on a clinical chemistry analyzer, Architect c8000 (Abbott Laboratories, Abbott Park, IL, USA). The detailed study protocol [16] and baseline characteristics [23] have been published previously.

Statistical analyses

Intervention and control groups were pooled and treated as one group in the analyses. To assess the longitudinal association between VO_2peak and cognitive functions, linear mixed model analyses with maximum likelihood estimation were conducted according to a two-level structure, i.e., repeated (baseline and 24 months) VO_2peak and cognition measures were clustered within subjects. Linear mixed model is suitable for longitudinal datasets containing correlated and unbalanced data.

We used Bayesian information criterion (BIC) as a measure of model adequacy so that a lower BIC indicates a better model with a better balance between complexity and good fit. We predeterminedly chose the model with a lowest BIC as our final model for a given outcome. That is, we did not force the more complex data structure to our model in case when it did not improve the model fit but instead brought on unnecessary complexity to the model.

We used a model adjusted for baseline age, sex, height, education, study group (intervention versus control), as well as time and depression as time-dependent covariates: $\begin{matrix} {OUTCOME}_{it} \\ = β_{0} + β_{1} age + β_{2} sex + β_{3} height \\ + β_{4} education + β_{5} study group + β_{6} time \\ + β_{6} depression + β_{7} {VO}_{2 peak} + ɛ_{it}, \end{matrix}$ where OUTCOME _it are observations for subject i at baseline and two-year observations; β₀ is the intercept; β₁, β₂, β₃, β₄, β₅, β₆, β₇, and β₈, are the regression coefficients for age, sex, height, education, study group, time, depression, and VO_2peak, respectively; and ɛ _it is the error for subject i at baseline and two-year observations. Additional adjustments included systolic blood pressure, total serum cholesterol and fasting plasma glucose concentration as time-dependent variables. As a measure of VO_2peak, we used absolute L/min value instead of more commonly used weight adjusted (ml/kg/min) value to avoid the possible bias from the weight change during lifestyle intervention. Height, although not a determinant of cognitive function per se, was included to remove variation (i.e., extra noise) in VO_2peak resulting from differences in body size.

Potential moderators of the association between cognition and VO_2peak were studied one after another by including the interaction term between VO_2peak and age, sex, education, symptoms of depression, systolic blood pressure, total serum cholesterol and fasting plasma glucose concentration in the adjusted model. To examine the extent to which covariates potentially mediated the association of VO_2peak with cognition, we next added the covariates, by turns, to the sex-, height-, and time-adjusted (basic) model and examined the magnitude of change in the β-coefficient for VO_2peak. The proportion of β-coefficient attenuation explained by each covariate was computed as follows; (β_{basic model} - β_{adjusted model})/ (β_{basic model})×100%.

The standard linear mixed model pool together within-subject (i.e., whether a change in one variable during the follow-up is associated with a change in another one) and between-subject (i.e., whether an overall level of one variable during the follow-up is associated with an overall level of another one) relationships in such a way that no separation can be made between the two aspects of longitudinal relationship. Because of this limitation, we carried out additional analyses to sort out the within- versus between-subject aspects of the relationship between VO_2peak and cognition by using a simple method described by van de Pol & Wright [24]. With that method, the subtracting the subject’s mean value from each observation value (i.e., within-subject centering) effectively eliminates any between-subject variation thus creating a new variable that expresses only the within-subject variation component. On the other hand, a second new variable expressing only the between-subject variation component is simply the mean of baseline and two-year observations (i.e., baseline and two-year observations for the same subject are both given the same value). Whenever the parameter estimates of these two effects seem to differ, it is then possible to compare them to see whether they are statistically different from each other.

Statistical analyses were performed using the IBM SPSS statistics for Windows, version 24.0 (IBM Corporation). The level of significance was less than 5% in all analyses.

RESULTS

Of the participants 53.9% were men, their mean age was 69.0, standard deviation (SD) 4.6 years (Table 1). Their VO_2peak decreased 5.5% during two years with no difference between study groups (study group*time, p = 0.5). We also calculated CRF*group*time interactions to test whether the effect of CRF on cognitive functions varies as a function of the study group. Results were non-significant (p for interaction 0.29–0.75) for all four cognitive outcomes. Symptoms of depression were at a normal range for a not depressed study population both at baseline and after two years in both study groups.

Table 1

Descriptive statistics of the study participants

	Participants, n	Baseline	Participants, n	2 years
Age, y	421	69.0 (4.6)	n/a	n/a
Sex, number of women (%)	421	195 (46.3)	n/a	n/a
Study group, intervention, n (%)	421	212 (50.4)	n/a	n/a
Education, years	421	9.6 (3.2)	n/a	n/a
Height, cm	415	167.1 (9.1)	n/a	n/a
Systolic blood pressure, mmHg	415	141.2 (17.0)	380	139.6 (16.0)
Total serum cholesterol, mmol/L	420	5.0 (1.0)	392	5.0 (1.0)
Fasting plasma glucose, mmol/L	420	6.2 (1.1)	393	6.3 (1.1)
Peak oxygen uptake, L/min	421	1.82 (0.5)	325	1.72 (0.5)
Peak oxygen uptake, ml/kg/min	415	23.4 (6.2)	322	22.4 (6.0)
Symptoms of depression, points	388	33.3 (7.3)	347	32.9 (7.6)
Cognition^*
NTB total score	421	–0.15 (0.57)	383	0.14 (0.66)
Executive functions	420	–0.13 (0.66)	382	–0.01 (0.70)
Processing speed	421	–0.11 (0.81)	382	0.09 (0.85)
Memory	421	–0.18 (0.65)	384	0.29 (0.80)

Data are mean (SD) or n (%). NTB = neuropsychological test battery. ^*Scores on the NTB total score, and on executive functions, processing speed, and memory are mean values of Z scores of the cognitive tests included in each cognitive outcome, with higher scores suggesting better performance.

CRF and cognitive performance over 2 years

Over two years, VO_2peak was associated with executive functions (β= 0.16, p = 0.01) and processing speed (β= 0.25, p < 0.001) but not with memory (β= 0.11, p = 0.12) (Table 2). VO_2peak was related also to the NTB total score (β= 0.12, p = 0.01). Results remained unchanged after adjusted for systolic blood pressure, total serum cholesterol, and fasting plasma glucose concentration one at a time or simultaneously. We also completed the supplementary analyses for the sub-group including only participants with complete VO_2peak data available (i.e., baseline and two years). Results remained unchanged with respect to processing speed (β= 0.28, p < 0.001) and memory (β= 0.06, p = 0.44) but altered slightly non-significant for NTB total score (β= 0.10, p = 0.07) and executive functions (β= 0.13, p = 0.07). The corresponding change in cognitive performance when VO_2peak increases 1.0 L/min is presented in Fig. 1.

Table 2

Association of cardiorespiratory fitness and covariates with cognition over two years

	NTB total score^a		Executive functions^b		Processing speed^b		Memory^a
	β	p	β	p	β	p	β	p
Peak oxygen uptake, L/min	0.12	0.01	0.16	0.01	0.25	<0.001	0.11	0.12
Sex^c	0.31	<0.001	0.12	0.19	0.52	<0.001	0.43	<0.001
Baseline age, y^c	–0.03	<0.001	–0.03	<0.001	–0.04	<0.001	–0.02	0.001
Education, y^c	0.06	<0.001	0.07	<0.001	0.07	<0.001	0.05	<0.001
Symptoms of depression	–0.01	0.05	–0.01	0.01	–0.01	0.002	–0.003	0.43
Study group (int. versus con.)^c	0.01	0.89	–0.02	0.77	–0.04	0.57	0.02	0.69

^aValues are derived from a time- and height-adjusted linear mixed model with intercept and time as random effects. ^bValues are derived from a time- and height-adjusted linear mixed model with only intercept as a random effect. ^cValues are from baseline. NTB, neuropsychological test battery. Age, sex, education, study group (intervention versus control), time and depression were included as covariates in all analyses. Additional adjustments included systolic blood pressure, total serum cholesterol and fasting plasma glucose concentration both one at a time and simultaneously.

Fig.1

Association between 1 L/min increase in VO_2peak and cognitive performance.

Systolic blood pressure moderated the association between VO_2peak and processing speed (β= 0.006, p for interaction = 0.03). The association was stronger in participants with higher systolic blood pressure (β= 0.31, p = 0.002) compared to participants with lower systolic blood pressure (β= 0.25, p = 0.02). Other covariates (age, sex, education, symptoms of depression, plasma glucose concentration, and total serum cholesterol) did not moderate the association between VO_2peak and cognition. Age was a strong mediator of the association between VO_2peak and cognition and explained attenuation of β-coefficient as follows: NTB total score by 37%, memory by 43%, executive functions by 37%, and processing speed by 31%. Education mediated the association between VO_2peak and cognition as follows: NTB total score by <1%, memory by 7%, executive functions by 4% and processing speed by 4%. Finally, total serum cholesterol mediated the association between VO_2peak and cognition as follows: NTB total score by 4.4%, memory by 2.9%, processing speed by <1% and executive functions by 5.1%. Symptoms of depression, systolic blood pressure and plasma glucose concentration did not mediate the association between VO_2peak and cognition.

Within- and between-subject effects between VO_2peak and cognition are presented in Table 3. The within-subject effects were not significant except for processing speed (p = 0.02). At the cognitive domain level, the between-subject effect was significant for executive functions (p = 0.008) and processing speed (p = 0.004), but not with memory (p = 0.18). The between-subject effect was significant also for NTB total score (p = 0.01). Within- versus between-subject effects were not statistically different from each other in any of the analysis. However, borderline significance was found for executive functions (p = 0.06) and NTB total score (p = 0.06).

Table 3

Within- and between-subject effects between cardiorespiratory fitness and cognitive function

	Within-subject effect		Between-subject effect		Difference
	β	95% CI	β	95% CI	β	95% CI
NTB total score	0.007	–0.13–0.15	0.21	0.05–0.36^*	0.20	–0.01–0.41
Memory	–0.07	–0.33–0.19	0.13	–0.06–0.32	0.20	–0.12–0.52
Executive functions	–0.01	–0.21–0.18	0.25	0.06–0.43^*	0.26	–0.007–0.53
Processing speed	0.24	0.04–0.44^*	0.33	0.10–0.56^*	0.09	–0.21–0.40

Difference denotes the difference between within- and between-subject effects [24]. ^*p<0.05. NTB, neuropsychological test battery.

DISCUSSION

The main finding of the present study was that over two years follow-up, high CRF was associated with better executive functions and processing speed as well as with better overall cognitive function in older people at risk for cognitive impairment. This longitudinal association between CRF and several cognitive outcomes highlights the importance of high CRF as one of the health factors in maintaining cognitive function in the late adulthood.

Systolic blood pressure moderated the association between VO_2peak and processing speed suggesting that the association may be more robust in individuals with higher systolic blood pressure. Age, sex, education, or symptoms of depression had no moderating effect on the association between VO_2peak and cognition. As expected, age was a strong mediator between VO_2peak and cognition and explained about one third of the relationship over two years. Furthermore, the positive association between CRF and cognitive function may be at least partly explained by the fact that persons with higher CRF are often also better educated. Total serum cholesterol mediated the association between VO_2peak and cognitive outcomes which suggests that also vascular risk factors contribute the fitness benefits for the brain health. Other covariates did not explain the association between VO_2peak and cognition.

The interpretation of the results may vary depending on the presence of within-subject and/or between-subject relationships between VO_2peak and cognition [24]. The within-subject relationship between VO_2peak and processing speed indicates that the change in VO_2peak was associated with the change in processing speed over time. The presence of within-subject relationship may reflect the potential for short-term changes suggesting that increase in CRF may contribute the corresponding improvement in processing speed in relatively short time period (two years in this case). The presence of between-subject relationship between VO_2peak and all cognitive outcomes except memory suggest that some latent, relatively fixed effects like heredity may underlie the association between CRF and cognition, albeit this cannot be confirmed in this data. It is also possible that the change in CRF may have an effect on cognition through structural brain changes which have evolved during such a long time period that a two-year interval has only negligible effect on those changes. While within-subject association reflects only a two-year period in the present study, between-subject association most probably reflects a longer time period, whereupon the direction of causality is more difficult to determine.

Our results agree with a previous longitudinal study [2] in older adults which reported longitudinal associations of baseline CRF with overall cognitive function, executive functions, and verbal memory over a 6-year period. However, cognitive assessment at baseline was restricted only to the overall cognitive function which limits conclusions regarding cognitive change over time [2]. Wendell et al. [8] studied the longitudinal associations between CRF and cognitive performance in a prospective observational study with a large sample (n = 1400) of subjects with a wide age range (19–94 years). Greater baseline CRF was associated with less memory decline across the lifespan which differs from our findings in that we did not find association between CRF and memory. Although a long follow-up (18 years) is a strength of that study, CRF measurement only at baseline may be considered as a limitation.

CRF is strongly and inversely associated with most cardiovascular risk factors [25] which in turn may play an important role in the etiology of Alzheimer’s disease [26]. This association between CRF and cardiovascular risk factors seems the most plausible mechanism to mediate the fitness benefits for the brain health. Interestingly, CRF can also modify the association between amyloid-β (Aβ) and cognitive function in late-middle-aged adults at risk for Alzheimer’s disease [27]. CRF was associated with immediate memory and verbal learning among persons with high Aβ burden but not in persons with minimal Aβ burden [27]. CRF is also related with changes in mechanisms such as cerebral blood flow, neurotrophic factors, neurotransmitter systems and neural architecture that have themselves been associated with cognitive performance [28].

Longitudinal study setting with notable sample size, comprehensive cognitive assessments and modern statistical approach appropriate for analyzing complex serial data are the strengths of this study. The change in the independent variable between the first and second visits was compared with the change in health outcome between visits. Results from this longitudinal design are used to infer that risk factor changes are related to changes in health outcomes. Linear mixed model handles erratic measurement intervals both within and across study participants, provides valid results despite the randomly missing data, and accounts for the correlation among recurrent measurements on the same participants [29]. We also used the objective, weight-independent measure of CRF (VO_2peak, L/min) to eliminate the possible effect of weight change during the two-year study period. Approximately one fifth of the participants did not attend CRF measurements at 24 months which is a limitation of this study. However, linear mixed model remains unaffected by randomly missing data as described above.

In summary, CRF was associated with executive functions and processing speed, and was related also to the overall cognitive function in participants who are representative of an important part of the general Finnish older population with several risk factors for dementia, but without pronounced cognitive impairment.

Footnotes

ACKNOWLEDGMENTS

This work was supported by Academy of Finland, La Carita Foundation, Alzheimer Association, Alzheimer’s Research and Prevention Foundation, Juho Vainio Foundation, Finnish Brain Foundation, Novo Nordisk Foundation, Finnish Social Insurance Institution, Ministry of Education and Culture, Salama bint Hamdan Al Nahyan Foundation, Axa Research Fund, EVO funding for University Hospitals of Kuopio, Oulu, and Turku and for Seinäjoki Central Hospital and Oulu City Hospital, Swedish Research Council, Swedish Research Council for Health, Working Life and Welfare, and af Jochnick Foundation. The blood pressure monitoring devices were provided by Microlife (Vilnius, Lithuania).

Authors’ disclosures available online ().

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

References

World Health Organization, Alzheimer’s Disease International (2012) Dementia: A public health priority. World Health Organization, Geneva. http://www.who.int/mental_health/publications/dementia_report_2012/en/, Accessed April 27, 2018.

Barnes

, Yaffe

, Satariano

, Tager

(2003) A longitudinal study of cardiorespiratory fitness and cognitive function in healthy older adults. J Am Geriatr Soc 51, 459–465.

Brown

, McMorris

, Longman

, Leigh

, Hill

, Friedenreich

, Poulin

(2010) Effects of cardiorespiratory fitness and cerebral blood flow on cognitive outcomes in older women. Neurobiol Aging 31, 2047–2057.

Netz

, Dwolatzky

, Zinker

, Argov

, Agmon

(2011) Aerobic fitness and multidomain cognitive function in advanced age. Int Psychogeriatr 23, 114–124.

Freudenberger

, Petrovic

, Sen

, Toglhofer

, Fixa

, Hofer

, Perl

, Zweiker

, Seshadri

, Schmidt

(2016) Fitness and cognition in the elderly: The Austrian Stroke Prevention Study. Neurology 86, 418–424.

Dougherty

, Schultz

, Boots

, Ellingson

, Meyer

, Van

, Stegner

, Edwards

, Oh

, Jean

, Korcarz

, Koscik

, Dowling

, Gallagher

, Carlsson

, Rowley

, Bendlin

, Sanjay

, Hermann

, Sager

, Stein

, Johnson

, Okonkwo

, Cook

(2017) Relationships between cardiorespiratory fitness, hippocampal volume, and episodic memory in a population at risk for Alzheimer’s disease. Brain Behav 7, e00625.

Hayes

, Forman

, Verfaellie

(2016) Cardiorespiratory fitness is associated with cognitive performance in older but not younger adults. J Gerontol B Psychol Sci Soc Sci 71, 474–482.

Wendell

, Gunstad

, Waldstein

, Wright

, Ferrucci

, Zonderman

(2014) Cardiorespiratory fitness and accelerated cognitive decline with aging. J Gerontol A Biol Sci Med Sci 69, 455–462.

Newson

, Kemps

(2006) Cardiorespiratory fitness as a predictor of successful cognitive ageing. J Clin Exp Neuropsychol 28, 949–967.

10.

Erickson

, Voss

, Prakash

, Basak

, Szabo

, Chaddock

, Kim

, Heo

, Alves

, White

, Wojcicki

, Mailey

, Vieira

, Martin

, Pence

, Woods

, McAuley

, Kramer

(2011) Exercise training increases size of hippocampus and improves memory. Proc Natl Acad Sci U S A 108, 3017–3022.

11.

Edwards

, Loprinzi

Combined associations of sedentary behavior and cardiorespiratory fitness on cognitive function among older adults. Int J Cardiol 229, 71–74.

12.

Weinstein

, Voss

, Prakash

, Chaddock

, Szabo

, White

, Wojcicki

, Mailey

, McAuley

, Kramer

, Erickson

(2012) The association between aerobic fitness and executive function is mediated by prefrontal cortex volume. Brain Behav Immun 26, 811–819.

13.

Gauthier

, Lefort

, Mekary

, Desjardins-Crépeau

, Skimminge

, Iversen

, Madjar

, Desjardins

, Lesage

, Garde

, Frouin

, Bherer

, Hoge

Hearts and minds: Linking vascular rigidity and aerobic fitness with cognitive aging. Neurobiol Aging 36, 304–314.

14.

DeFina

, Willis

, Radford

, Gao

, Leonard

, Haskell

, Weiner

, Berry

(2013) The association between midlife cardiorespiratory fitness levels and later-life dementia: A cohort study. Ann Intern Med 158, 162–168.

15.

Burns

, Mayo

, Anderson

, Smith

, Donnelly

(2008) Cardiorespiratory fitness in early-stage Alzheimer disease. Alzheimer Dis Assoc Disord 22, 39–46.

16.

Kivipelto

, Solomon

, Ahtiluoto

, Ngandu

, Lehtisalo

, Antikainen

, Backman

, Hanninen

, Jula

, Laatikainen

, Lindstrom

, Mangialasche

, Nissinen

, Paajanen

, Pajala

, Peltonen

, Rauramaa

, Stigsdotter-Neely

, Strandberg

, Tuomilehto

, Soininen

(2013) The Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability (FINGER): Study design and progress. Alzheimers Dement 9, 657–665.

17.

Saaristo

, Peltonen

, Keinanen-Kiukaanniemi

, Vanhala

, Saltevo

, Niskanen

, Oksa

, Korpi-Hyovalti

, Tuomilehto

, FIN-D2D Study Group (2007) National type 2 diabetes prevention programme in Finland: FIN-D2D. Int J Circumpolar Health 66, 101–112.

18.

Vartiainen

, Laatikainen

, Peltonen

, Juolevi

, Mannisto

, Sundvall

, Jousilahti

, Salomaa

, Valsta

, Puska

(2010) Thirty-five-year trends in cardiovascular risk factors in Finland. Int J Epidemiol 39, 504–518.

19.

Morris

, Heyman

, Mohs

, Hughes

, van Belle

, Fillenbaum

, Mellits

, Clark

(1989) The Consortium to establish a registry for Alzheimer’s disease (CERAD). Part I. Clinical and neuropsychological assessment of Alzheimer’s disease. Neurology 39, 1159–1165.

20.

Hanninen

, Pulliainen

, Sotaniemi

, Hokkanen

, Salo

, Hietanen

, Pirttila

, Poyhonen

, Juva

, Remes

, Erkinjuntti

(2010) Early detection of cognitive changes in memory diseases: New cut-off scores for the Finnish version of CERAD neuropsychological battery. Duodecim 126, 2013–2021.

21.

Harrison

, Minassian

, Jenkins

, Black

, Koller

, Grundman

(2007) A neuropsychological test battery for use in Alzheimer disease clinical trials. Arch Neurol 64, 1323–1329.

22.

Zung

(1965) A self-rating depression scale. Arch Gen Psychiatry 12, 63–70.

23.

Ngandu

, Lehtisalo

, Levalahti

, Laatikainen

, Lindstrom

, Peltonen

, Solomon

, Ahtiluoto

, Antikainen

, Hanninen

, Jula

, Mangialasche

, Paajanen

, Pajala

, Rauramaa

, Strandberg

, Tuomilehto

, Soininen

, Kivipelto

(2014) Recruitment and baseline characteristics of participants in the Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability (FINGER) – a randomized controlled lifestyle trial. Int J Environ Res Public Health 11, 9345–9360.

24.

van de Pol

, Wright

(2009) A simple method for distinguishing within- versus between-subject effects using mixed models. Anim Behav 77, 753–758.

25.

Erez

, Kivity

, Berkovitch

, Milwidsky

, Klempfner

, Segev

, Goldenberg

, Sidi

, Maor

(2015) The association between cardiorespiratory fitness and cardiovascular risk may be modulated by known cardiovascular risk factors. Am Heart J 169, 923.e1.

26.

de Bruijn

RFAG

, Ikram

(2014) Cardiovascular risk factors and future risk of Alzheimer’s disease. BMC Medicine 12, 130.

27.

Schultz

, Boots

, Almeida

, Oh

, Einerson

, Korcarz

, Edwards

, Koscik

, Dowling

, Gallagher

, Bendlin

, Christian

, Zetterberg

, Blennow

, Carlsson

, Asthana

, Hermann

, Sager

, Johnson

, Stein

, Okonkwo

(2015) Cardiorespiratory fitness attenuates the influence of amyloid on cognition. J Int Neuropsychol Soc 21, 841–850.

28.

Marmeleira

(2013) An examination of the mechanisms underlying the effects of physical activity on brain and cognition. Eur Rev Aging Phys Act 10, 83–94.

29.

Twisk

JWR

(2013) Applied Longitudinal Data Analysis for Epidemiology, 2nd edition, Cambridge University Press.