Cognitive and physical age-gaps in relation to mild cognitive impairment and behavioral phenotypes

doi:10.21203/rs.3.rs-2437986/v1

Download PDF

Article

Cognitive and physical age-gaps in relation to mild cognitive impairment and behavioral phenotypes

https://doi.org/10.21203/rs.3.rs-2437986/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this older preprint version

Read the latest preprint version →

Background: We can study how fast our biological aging clocks tick by calculating the difference (i.e., age-gaps) between machine learning estimations of biological age and chronological age. While this approach has been increasingly used to study various aspects of aging, few had applied this approach to study cognitive and physical age-gaps; not much is known about the behavioral and neurocognitive factors associated with these age-gaps. In the present study, we examined these age-gaps in relation to behavioral phenotypes and mild cognitive impairment (MCI) among community-dwelling older adults.

Methods: Participants (N=822, Age_mean=67.6) were partitioned into equally-sized training and testing samples. Cognitive and physical age-prediction models were fitted using nine cognitive and eight physical fitness test scores, respectively, within the training samples, and subsequently used to estimate cognitive and physical age-gaps for each subject in the testing sample. These age-gaps were then compared among those with and without MCI, and correlated with 17 behavioral phenotypes in the domains of lifestyle, well-being, and attitudes.

Results: Across 5,000 random train-test split iterations, we showed that older cognitive and physical age-gaps were significantly associated with MCI (versus cognitively normal) and worse outcomes across several well-being and attitude-related measures. Both age-gaps were also significantly correlated with each other.

Conclusions: These results suggest accelerated cognitive and physical aging were linked to worse well-being and more negative attitudes about the self and others, and reinforce the link between cognitive and physical aging. Importantly, we have also validated the use of cognitive age-gaps in the diagnosis of MCI.

Health sciences/Risk factors

Biological sciences/Neuroscience/Cognitive ageing

Biological sciences/Developmental biology/Ageing

Health sciences/Health care/Disease prevention/Lifestyle modification

cognitive aging

physical fitness

lifestyle

well-being

attitudes

mild cognitive impairment

The advent of machine learning has opened up novel approaches in the study of human aging. One such approach involves using biological/physical characteristics from a reference sample of healthy individuals to train an age-prediction model and then applying this model to predict the age of an unseen subject. The discrepancy between the subject’s predicted and actual age would tell us how fast their biological aging clock ticks; if the predicted age is larger than the actual age (i.e., older age-gaps), this would hint at an accelerated aging phenomenon. Conversely, younger age-gaps would suggest a slower and perhaps healthier pace of aging.

This approach has been increasingly used to study various aspects of biological aging, especially in that of brain aging ¹. Several studies have shown that cognitively impaired individuals, such as those with Alzheimer’s disease and mild cognitive impairment (MCI) tend to possess older brain age-gaps than their cognitively normal counterparts ². Furthermore, such brain age-gaps have also been validated against markers of cellular aging; shorter telomere lengths were found to be associated with older brain age-gaps ³. Relatedly, cognitive aging is another dimension of aging that can utilize/capitalize on such an ‘age-gap’ approach. We can similarly utilize machine learning models on cognitive test scores to estimate one’s cognitive age with reference to a healthy normative sample, and subsequently determine their cognitive age-gaps. Although cognitive aging is closely related to brain aging, only a few studies ^3,4 have attempted similar approaches to study cognitive age-gaps.

Apart from cognitive age-gaps, another ability-based age-gap metric can be estimated using physical fitness scores. There have been a handful of studies ^5–9 that attempted to estimate physical ages of individual participants using physical fitness test scores. Among them, only two ^5,8 went on to compute and study physical age-gaps in relation to other outcomes. Unlike the recent brain age studies that utilized machine learning paradigms involving independent training and testing datasets, these physical age studies did not partition their data into independent training and testing samples. In most of these studies, participants’ physical ages were usually estimated from age-prediction models constructed from their own data. This circularity in model-fitting and prediction is likely to produce overfitted models that generalize poorly to unseen subjects ¹⁰. In one study⁵, participants’ physical ages were estimated using formulae that were published more than two decades ago. Such formulae were not adequately validated beyond the dataset it was conceived from; it is unclear if these formulae can adequately generalize across populations and time.

In addition to addressing these methodological concerns, it is important to learn about the factors that modulate individual differences in these age-gaps, to inform future preventive interventions to slow down the aging process. In this regard, while many studies have identified disease-related factors, such as diabetes and stroke ¹¹, that were associated with older age-gaps, very few have looked at modifiable behavioral/lifestyle factors. One recent study investigated brain age-gaps in relation to smoking, alcohol consumption, social integration, and physical activity, and found that only frequent smoking and inadequate physical activity were associated with older brain age-gaps ¹². Another study found a weak but significant association between brain age-gaps and a composite lifestyle index; however, cognitive age-gaps were not significantly associated with this lifestyle index ⁴. Given the preliminary extent of these findings, more fine-grained and varied assessments of lifestyle factors need to be studied in relation to age-gaps. In particular, it will be useful to look into the well-being and psychological-related factors, since these are largely amendable to a vast range of psychosocial and psychoeducational interventions¹³.

To these ends, we used a machine learning paradigm to estimate cognitive and physical age-gaps in an older adult sample and investigated their associations with a diverse range of behavioral phenotypes—spanning across the domains of lifestyle, well-being, and attitudes. We hypothesized that older cognitive and physical age-gaps were associated with less healthy lifestyles, worse well-being, and more negative attitudes about themselves and others. As a secondary goal, we also sought to validate our cognitive age-gaps estimations, by showing that participants with MCI possessed significantly older age-gaps as compared to their normal aging counterparts.

Participants

Participants of the current study were recruited from the community as part of the Community Health and Intergenerational (CHI) study. The inclusion criteria and recruitment procedures for the CHI study have been described in detail elsewhere ¹⁴. Ethical approval for the CHI study was granted by the National University of Singapore Institutional Review Board. Written informed consent was obtained from participants before their participation.

The full sample consisted of 931 participants. Seven who met the criteria for dementia were excluded. A further 102 participants who had missing or invalid scores (e.g., did not follow instructions, were distracted, or not motivated) in any of the 17 cognitive and physical fitness tests were excluded. The remaining 822 participants were included in the analyses.

The included participants had a mean age of 67.6 years (SD = 5.7), and an average of 13.1 years of education (SD = 4.0). Among them, 172 met the criteria for MCI, as assessed by a consensus panel consisting of a team of psychiatrists and psychologists. Most participants (N = 786) were of Chinese ethnicity, the rest were of Indian (N = 22), Malay (N = 4), and other ethnicities (N = 8); two did not disclose their ethnicities. In terms of housing, 106 resided in 1–3 room public housing flats, 268 in 4–5 room public housing flats, 79 in executive/maisonnette apartments, 170 in private apartments and condominiums, 194 in landed apartments; five did not disclose their housing types. Housing type is commonly used as an approximate indicator of socio-economic status in the local context. Most participants (N = 685) reported living with other family members or relatives, 126 lived alone and 11 did not disclose their living arrangements.

Measures

A total of 34 behavioral measures (see Table 1) were used in the current study. Among them, nine cognitive test scores and eight physical fitness test scores were used for training the age-prediction models; five lifestyle, seven well-being and five attitude-related measures were used to study the behavioral phenotypes associated with the age-gaps. Locally adapted versions of these measures were used, where available and appropriate. The violin plots of all studied outcomes and the correlation heatmap of all measures are presented in the supplementary materials.

Table 1

Measures used in the current study
Cognitive test	Domains assessed	Physical fitness tests	Abilities assessed
WAIS III Digit span forward ¹⁵	Working memory span	Handgrip test ^{16 a}	Hand grip strength
WAIS III Digit span backward ¹⁵	Working memory manipulation	Modified functional reach test ^{17 b}	Stability
WAIS III Block design ¹⁵	visuospatial functions	30s chair stand ¹⁸	Leg strength and endurance
RAVLT Total learning ¹⁹	Immediate memory	Timed up and go ²⁰	Balance
RAVLT Delayed recall ¹⁹	Delayed memory	6m fast gait speed test ²¹	Walking speed
RAVLT Recognition ¹⁹	Recognition memory
Categorical fluency (animal)	Semantic fluency
Color Trails Test 1 ²²	Visual search and processing speed
Color Trails Test 2 ²²	Visual search, processing speed, divided attention and task switching
Behavioral phenotype	Scale/questionnaire
Lifestyle
Income adequacy	4-point likert scale response to “Are your present sources of income and allowances adequate to cover your monthly expenses”
Smoking frequency	5-point likert scale response to “Currently, how often do you smoke?”
Alcohol frequency	5-point likert scale response to “Currently, how often do you drink alcohol?”
Sleep quality	Global score from the Pittsburgh Sleep Quality Index ²³
MET	Estimated from the International Physical Activity Questionnaire ²⁴
Well-being
Perceived health	EuroQol Group’s 5-dimension questionnaire: visual analog scale ²⁵
Stress	5-point likert scale response to “Currently, how often do you feel stressed?”
Depression	Geriatric Depression Scale ²⁶
Anxiety	Geriatric Anxiety Inventory ²⁷
Social isolation	The Friendship Scale ²⁸
Quality of life	World Health Organization Quality of Life Assessment (AGE version) ²⁹
Life satisfaction	Satisfaction with Life Scale ³⁰
Attitudes
Loss	Attitude to Ageing Questionnaire: psychological loss subscore ³¹
Growth	Attitude to Ageing Questionnaire: psychological growth subscore ³¹
Physical decline	Attitude to Ageing Questionnaire: physical changes subscore ³¹
Gratitude	Gratitude Questionnaire ³²
Compassion	Compassion Scale³³

^aAssessed using a Jamar dynamometer separately on the dominant and non-dominant hands. ^bAssessed separately on the dominant and non-dominant sides, as well as the front. WAIS III = Wechsler Adult Intelligence Scale, Third Edition; RAVLT = locally adapted version of the Rey Auditory Verbal Learning Test; MET = metabolic equivalent of task.

Statistical analyses

Figure 1 shows the overview of the analytical framework. First, the data was partitioned into equally sized training and testing datasets, in a manner that the distributions of age, sex, and neurocognitive diagnoses (i.e., MCI or normal aging) were approximately balanced between the two partitions. Then, ridge regression (RR) models, as implemented in the R package ‘glmnet’³⁴, were fitted separately on the cognitive and physical fitness data to construct cognitive and physical age-prediction models within the training dataset. RR is intended for multiple regression scenarios in which the predictors are highly correlated with each other, such as in the case of our cognitive and physical fitness data. Unlike ordinary least squares regression, RR imposes an L²-norm penalty on its regression coefficients to minimize the residual sum of squares. In order to minimize overfitting, the optimal tuning parameter (λ) was determined via an internal 5-fold cross-validation (CV) procedure within the training dataset. The range of λ values tested was automatically determined and we selected the λ value that produced the smallest mean squared error in the 5-fold CV. Given that sex can significantly confound cognitive and physical fitness scores, sex was included as an unpenalized covariate in the RR models.

Then, these two age-prediction models were applied to the testing dataset to estimate the predicted cognitive and physical ages. In order to assess the age-prediction models’ accuracy in the testing dataset, the mean absolute error (MAE) and correlation between actual and predicted ages (r_{predicted−actual}) were calculated. Smaller MAE and larger r_{predicted−actual} suggest more accurate predictions.

Next, the age-gaps were calculated by subtracting the actual age from the predicted cognitive or physical age. Previous research in brain age predictions revealed a bias in the estimation of brain age-gaps ³⁵; the predicted ages of younger subjects tend to be overestimated while those of older subjects were underestimated. In other words, age-gaps were negatively associated with age. To correct this bias, we applied the age-bias correction procedure used in a recent brain age study ¹¹. Briefly, a linear regression model was fitted within the training dataset using the predicted and actual ages as the outcome and predictor, respectively. Subsequently, the predicted age in the testing dataset was corrected by subtracting the intercept and dividing by the slope obtained in the previously fitted regression model. Although this bias correction approach alters the magnitude of the age-gaps, it does not impact subsequent correlational analyses since the age-gaps essentially underwent a linear transformation.

Subsequently, two sets of analyses were carried out on the cognitive and physical age-gaps. First, the age-gaps were compared between MCI and normal aging. Given that the same 1-year age-gap would suggest a quicker pace of aging in a young subject than in a much older subject, we normalized each subject’s age-gap by dividing it by their age. Then, the age-gap difference between both diagnoses is calculated by subtracting the mean normalized age-gaps of the normal aging group from that of the MCI. In the second set of analyses, partial correlation analyses, controlling for age, were then carried out between the age-gaps and the lifestyle, well-being, and attitude-related measures.

It is expected that the RR prediction coefficients and the subsequent ‘downstream’ results would vary depending on the subset of participants assigned to the training and testing partitions. As such, we repeated these analyses 5000 times. In each iteration, participants were randomly assigned to the training and testing partitions and subsequently the mean age-gap differences and correlation coefficients (r) were averaged across these 5000 iterations. These statistics were tested against a null distribution generated via a permutation approach. That is, the same procedures were used to generate these age-gap differences and r values, however, participants’ labels were detached from their corresponding age-gaps and randomly shuffled across the testing dataset. The p values obtained from this permutation process were then corrected for false discovery rate (FDR). Statistical significance was set at FDR-corrected p < .05. All analyses were carried out in R 4.2.2. The R codes for the analyses and generating the figures are available at https://osf.io/szcpt/.

Data availability

The conditions of our ethics approval do not permit public archiving of the data supporting this study. Interested researchers seeking access to the anonymized data should contact the corresponding authors and complete a formal data sharing agreement. Access will be granted in accordance with ethical procedures governing the reuse of data.

Model characteristics

According to the prediction accuracy metrics generated from 5000 random train-test split iterations (see the top panel in Fig. 2), both age-prediction models performed reasonably well in the unseen testing dataset. The r_{predicted−actual} and uncorrected MAE obtained from both age-prediction models were comparable to those from brain age-prediction models trained and tested within independent samples in the UK Biobank dataset ¹¹. The cognitive age-prediction model performed slightly better than the physical age-prediction model, both in terms of MAE and r_{predicted−actual}. Both models revealed a strong age-prediction bias, as indicated by the strong negative correlations between the uncorrected age-gaps and age (see Fig. 2b). The bias-correction procedure that was applied subsequently, approximately doubled the predicted ages’ MAE (see Figs. 2c and d).

The distributions of obtained RR coefficients are illustrated in Fig. 2 (bottom panel). In the cognitive age prediction model, the block design test scores had the highest RR coefficients, suggesting that it has the strongest influence on predicting age amongst all other cognitive predictors. In the physical age prediction model, the timed up-and-go test was the strongest predictor of age.

Differences in age-gaps between MCI and normal aging

The permutation analyses revealed significant differences between diagnoses in the uncorrected normalized cognitive age-gaps (FDR-corrected p = .023), but not physical age-gaps (FDR-corrected p = .469). The bias-correction procedure magnified the group differences such that both groups differed significantly in the normalized cognitive (FDR-corrected p < .0002) and physical age-gaps (FDR-corrected p = .026). These results are illustrated in Fig. 3. These normalized age-gap differences can be used to estimate how much older an MCI subject’s age-gap is compared to his same-age unafflicted peer, by simply multiplying it by the subject’s age. For instance, the mean bias-corrected normalized cognitive age-gap difference is 0.18 according to Fig. 3b; we would expect a 65-year-old person with MCI to possess a bias-corrected cognitive age-gap that is 0.18*65 = 11.7 years older than his 65-year-old normal aging counterpart.

Age-gaps and behavioral phenotypes

Next, the age-gaps from both age prediction models were assessed for partial correlations with various lifestyle, well-being, and attitude-related outcomes, controlling for age, within the testing dataset. The distributions of correlation coefficients across the 5000 random train-test split iterations are shown in Fig. 3.

According to the obtained mean r values, older cognitive age gaps were significantly (FDR-corrected ps ≤ .048) associated with worse outcomes in terms of depression, social isolation, psychological loss, psychological growth, and gratitude. Next, older physical age-gaps were significantly associated (FDR-corrected ps ≤ .030) with worse outcomes in terms of perceived health, depression, anxiety, quality of life, and psychological growth. Interestingly, older physical age-gaps were also significantly associated with a lower frequency of alcohol consumption. Finally, cognitive and physical age gaps were significantly correlated (p < .0002).

The current study estimated subjects’ cognitive and physical age-gaps using a machine learning paradigm and studied these age-gaps in relation to a diverse range of behavioral phenotypes. The prediction metrics suggested that both age-prediction models generalized reasonably well to unseen subjects. The age-gap-associated behavioral profiles we have identified were mostly in line with our expectations; older cognitive and physical age-gaps were associated with poorer well-being and more negative attitudes, even if some of these associations do not reach statistical significance. Both types of age-gaps were also significantly correlated with each other. Additionally, our results also revealed significant physical and cognitive age-gap differences between participants with MCI and their normal aging counterparts. Overall, these findings identified comprehensive behavioral profiles in relation to accelerated cognitive and physical aging, and reinforced the link between both domains of aging.

Older age-gaps, especially physical age-gaps, were associated with poorer well-being, in terms of higher levels of depressive and anxiety symptoms, lower quality of life, and perceived health. While the cross-sectional nature of these findings prevents one from inferring the causal relationships, we speculate that the well-being-related consequences could occur as psychological reactions to the decline in physical and cognitive abilities ³⁶. Additionally, it is also plausible that the chronic stressors that worsen one’s well-being is linked to increased inflammation and a weakened immune system, thus putting one at risk for a wide range of health-related conditions and accelerating the age-related decline^37,38 in physical and cognitive functions ³⁹.

Older age-gaps were also observed to be associated with more negative attitudes about themselves and others— such as having a more pessimistic view of loss and growth, and less gratitude toward others. Similar explanations could account for such observations—these negative attitudes could arise as psychological reactions to the decline in physical and cognitive health. To some extent, these attitudes also resemble Beck’s cognitive triad— negative thoughts about the self, world, and future, and may predispose one to depression. Notwithstanding the health-related consequences of depression, negative appraisals alone can increase cortisol output and down-regulate immune system activity ⁴⁰. On a chronic basis, such immune system-related consequences will exacerbate brain aging ³⁹, and by extension cognitive aging. It is interesting to note that cognitive, compared to physical age-gaps, were more strongly related to negative attitudes, alluding to a possibility that cognitive aging may act on these attitudes in a different and perhaps more direct manner. In relation to this, one longitudinal study ⁴¹ showed that among older adults participants with high baseline levels of hopelessness, those who had better cognitive flexibility, as measured via a Stroop test, eventually felt less hopeless 10 months later. These results were interpreted to suggest that greater cognitive flexibility may enable one to be more receptive to contradictory evidence which could help to modify their pessimistic attitudes. Hence, when cognitive flexibility is impaired due to accelerated cognitive aging, it is likely that negative attitudes would become more resistant to change.

Compared to the well-being and attitudes-related outcomes, the lifestyle-related outcomes were not as strongly linked to both age-gaps. This could be due to the fact that these lifestyle-related outcomes were measured via self-reported assessments, which were largely subjected to retrospective recall bias. Such bias could affect the lifestyle-related measurements, more so than those of attitudes and well-being, due to the need to recall very specific details such as the frequency of behaviors, as well as onset and end times of events (e.g., sleep and physical activity). Counter-intuitively, less frequent consumption of alcohol was linked to older age-gaps. Apart from the issues associated with self-reported measurements, other possible explanations can account for this counter-intuitive result. First, increased alcohol consumption may reflect the participants’ higher socioeconomic status, especially since alcohol is relatively more expensive in the local context due to the liquor tax. Additionally, such alcohol consumption may also take place in the context of social interactions; thus alcohol consumption can be seen as an indirect measure of social interaction⁴². In relation to this possibility, one study reported older participants who drank alcohol 4 to 7 days a week, compared to those who do not consume alcohol in the past 3 months, felt significantly less lonely ⁴³. Consequently, frequent alcohol consumption would likely predict better cognitive and physical health, on account of the indirect effects of higher socioeconomic status and more frequent social interactions^44,45.

The significantly older cognitive age-gaps observed among participants with MCI are very much expected, since one of the criteria for the diagnosis requires the individual to score significantly lower, in one or multiple cognitive tests, than what would be expected of their age. Furthermore, bias-corrected physical age-gaps were also significantly older in MCI than in normal aging, though to a lesser extent. This partly reflects the small but significant association between cognitive and physical aging, as we have shown in our results. Longitudinal research ⁴⁶ has shown that physical fitness protects one from MCI and dementia. Physical fitness alludes to regular physical exercises, which increase the circulating levels of brain-derived neurotrophic factor, insulin-like growth factor 1, and vascular endothelial growth factor. These growth factors facilitate the repair and regeneration of the brain ⁴⁷, and thus are involved in the maintenance of cognitive health.

Overall, the current research presents some important implications. First, we identified several behavioral phenotypes that were associated with older physical and cognitive age-gaps. These findings will help to narrow down the search for modifiable behavioral and psychological factors that can be targeted in future interventions to slow cognitive and physical aging. Next, our results pertaining to the differences in cognitive age-gaps between MCI and normal aging validated the use of cognitive age-gaps in the neurocognitive diagnostic context. Essentially, cognitive age-gaps can be used as a single metric to characterize cognitive health meaningfully and objectively. Such age-gaps can be easily understood and interpreted even among laypersons. This is in stark contrast to the conventional and cumbersome use of multiple normed z-scores, which may appear somewhat abstract to laypersons. For instance, it is difficult to imagine what it means to have an impaired test score of 1.5 standard deviations below the mean. If we could quantify the cognitive impairment in terms of age-gaps, it would certainly help the diagnosed individual understand their situation more meaningfully.

The current study is subjected to some limitations. First, we excluded a significant proportion of the participants (~ 11%) due to incomplete data. There may be a possible selection bias, as the excluded sample (mean age = 69.7; mean years of education = 13.1) was significantly (ps ≤ .003) older and less educated than the included sample (mean age = 67.6; mean years of education = 11.5). Second, the cross-sectional nature of the study does not permit one to determine the causal relationships between age-gaps and behavioral phenotypes. Third, all lifestyle-related phenotypes included in the current study were measured using self-reports, which can be relatively less accurate among older participants due to the age-related decline in recall abilities, particularly in MCI ⁴⁸. Future studies may consider using wearables to track lifestyle-related outcomes, such as sleep and physical activity, more objectively.

Competing Interests

The authors declare no competing financial or non-financial interests

Author contributions

Junhong Yu: Writing - original draft, Formal analysis, Software, Visualization. Ted Kheng Siang Ng: Writing -review & editing. Rathi Mahendran: Writing - review & editing, Funding acquisition, Project administration, Supervision

Acknowledgments

The authors would like to thank Ee-heok Kua for his helpful comments on an earlier version of this manuscript. This work was supported by Research Donations from Kwan Im Thong Hood Cho Temple and Lee Kim Tah Holdings Pte Ltd, under the Mind Science Centre, Department of Psychological Medicine, National University of Singapore.

Cole, J. H. & Franke, K. Predicting age using neuroimaging: innovative brain ageing biomarkers. Trends Neurosci 40, 681–690 (2017).
Löwe, L. C., Gaser, C., Franke, K. & Initiative, for the A. D. N. The Effect of the APOE Genotype on Individual BrainAGE in Normal Aging, Mild Cognitive Impairment, and Alzheimer’s Disease. PLoS One 11, e0157514 (2016).
Yu, J. et al. Differences between multimodal brain-age and chronological-age are linked to telomere shortening. Neurobiol Aging 115, 60–69 (2022).
Anatürk, M. et al. Prediction of brain age and cognitive age: Quantifying brain and cognitive maintenance in aging. Hum Brain Mapp 42, 1626–1640 (2021).
Wang, Z. et al. Associations of various exercise types with health-related physical fitness: Focus on physical fitness age. J Phys Fit Sports Med 9, 75–82 (2020).
Nishijima, T. et al. A Comparison of Estimation Models of Physical Fitness Age for Elderly People using the Japan Fitness Test. International Journal of Sport and Health Science 4, 591–605 (2006).
Kimura, M., Mizuta, C., Yamada, Y., Okayama, Y. & Nakamura, E. Constructing an index of physical fitness age for Japanese elderly based on 7-year longitudinal data: Sex differences in estimated physical fitness age. Age (Omaha) 34, 203–214 (2012).
Toyoshima, K. et al. Difference between “Physical Fitness Age” Based on Physical Function and Chronological Age Is Associated with Obesity, Hyperglycemia, Depressive Symptoms, and Low Serum Albumin. The journal of nutrition, health & aging 2022 26:5 26, 501–509 (2022).
Nakagaichi, M., Anan, Y., Hikiji, Y. & Uratani, S. Developing an assessment based on physical fitness age to evaluate motor function in frail and healthy elderly women. Clin Interv Aging 13, 179 (2018).
Kriegeskorte, N., Simmons, W. K., Bellgowan, P. S. F. & Baker, C. I. Circular analysis in systems neuroscience: the dangers of double dipping. Nat Neurosci 12, 535–540 (2009).
Cole, J. H. Multimodality neuroimaging brain-age in UK biobank: relationship to biomedical, lifestyle, and cognitive factors. Neurobiol Aging 92, 34–42 (2020).
Bittner, N. et al. When your brain looks older than expected: combined lifestyle risk and BrainAGE. Brain Struct Funct 226, 621–645 (2021).
Bolier, L. et al. Positive psychology interventions: A meta-analysis of randomized controlled studies. BMC Public Health 13, 1–20 (2013).
Lee, R. Z. Y. et al. CHI study: protocol for an observational cohort study on ageing and mental health in community-dwelling older adults. BMJ Open 10, e035003 (2020).
Wechsler, D. WAIS-III: Administration and scoring manual: Wechsler adult intelligence scale. (Psychological Corporation, 1997).
Bechtol, C. O. Grip test: the use of a dynamometer with adjustable handle spacings. JBJS 36, 820–832 (1954).
Katz-Leurer, M., Fisher, I., Neeb, M., Schwartz, I. & Carmeli, E. Reliability and validity of the modified functional reach test at the sub-acute stage post-stroke. Disabil Rehabil 31, 243–248 (2009).
Rikli, R. E. & Jones, C. J. Development and validation of a functional fitness test for community-residing older adults. J Aging Phys Act 7, 129–161 (1999).
Rey, A. L’examen psychologique dans les cas d’encéphalopathie traumatique.(Les problems.). Arch Psychol (Geneve) (1941).
Mathias, S., Nayak, U. S. & Isaacs, B. Balance in elderly patients: the" get-up and go" test. Arch Phys Med Rehabil 67, 387–389 (1986).
Middleton, A., Fritz, S. L. & Lusardi, M. Walking speed: the functional vital sign. J Aging Phys Act 23, 314–322 (2015).
D’Elia, L. & Satz, P. Color trails test. (Psychological Assessment Resources, 1996).
Buysse, D. J., Reynolds, C. F., Monk, T. H., Berman, S. R. & Kupfer, D. J. The Pittsburgh sleep quality index: A new instrument for psychiatric practice and research. Psychiatry Res 28, 193–213 (1989).
Craig, C. L. et al. International physical activity questionnaire: 12-Country reliability and validity. Med Sci Sports Exerc 35, 1381–1395 (2003).
The EuroQol Group. EuroQol - a new facility for the measurement of health-related quality of life. Health Policy (New York) 16, 199–208 (1990).
Sheikh, J. I. & Yesavage, J. A. Geriatric Depression Scale (GDS): Recent evidence and development of a shorter version. Clin Gerontol 5, 165–173 (1986).
Pachana, N. A. et al. Development and validation of the Geriatric Anxiety Inventory. Int Psychogeriatr 19, 103–114 (2007).
Hawthorne, G. Measuring Social Isolation in Older Adults: Development and Initial Validation of the Friendship Scale. Soc Indic Res 77, 521–548 (2006).
Caballero, F. F. et al. Validation of an instrument to evaluate quality of life in the aging population: WHOQOL-AGE. Health Qual Life Outcomes 11, 1–12 (2013).
Diener, E., Emmons, R. A., Larsem, R. J. & Griffin, S. The Satisfaction With Life Scale. https://doi.org/10.1207/s15327752jpa4901_13 49, 71–75 (2010).
Laidlaw, K. et al. The attitudes to ageing questionnaire (AAQ): development and psychometric properties. Int J Geriatr Psychiatry 22, 367–379 (2007).
McCullough, M. E., Emmons, R. A. & Tsang, J. A. The grateful disposition: A conceptual and empirical topography. J Pers Soc Psychol 82, 112–127 (2002).
Martins, D., Nicholas, N. A., Shaheen, M., Jones, L. & Norris, K. The Development and Evaluation of a Compassion Scale. J Health Care Poor Underserved 24, 1235–1246 (2013).
Friedman, J., Hastie, T. & Tibshirani, R. Regularization Paths for Generalized Linear Models via Coordinate Descent. J Stat Softw 33, 1–22 (2010).
Liang, H., Zhang, F. & Niu, X. Investigating systematic bias in brain age estimation with application to post-traumatic stress disorders. Hum Brain Mapp 40, 3143–3152 (2019).
Bassuk, S. S., Berkman, L. F. & Wypij, D. Depressive symptomatology and incident cognitive decline in an elderly community sample. Arch Gen Psychiatry 55, 1073–1081 (1998).
Squassina, A., Pisanu, C. & Vanni, R. Mood Disorders, Accelerated Aging, and Inflammation: Is the Link Hidden in Telomeres? Cells 2019, Vol. 8, Page 52 8, 52 (2019).
Vitlic, A., Lord, J. M. & Phillips, A. C. Stress, ageing and their influence on functional, cellular and molecular aspects of the immune system. AGE 2014 36:3 36, 1169–1185 (2014).
di Benedetto, S., Müller, L., Wenger, E., Düzel, S. & Pawelec, G. Contribution of neuroinflammation and immunity to brain aging and the mitigating effects of physical and cognitive interventions. Neurosci Biobehav Rev 75, 114–128 (2017).
Denson, T. F., Spanovic, M. & Miller, N. Cognitive appraisals and emotions predict cortisol and immune responses: a meta-analysis of acute laboratory social stressors and emotion inductions. Psychol Bull 135, 823 (2009).
Yu, J. & Lee, T. M. C. Cognitive flexibility and changes in hopelessness across time: A moderation hypothesis. Applied Neuropsychology:Adult 24, (2017).
Alexander, F. & Duff, R. W. Social Interaction and Alcohol Use in Retirement Communities. Gerontologist 28, 632–636 (1988).
Canham, S. L., Mauro, P. M., Kaufmann, C. N. & Sixsmith, A. Association of Alcohol Use and Loneliness Frequency Among Middle-Aged and Older Adult Drinkers. http://dx.doi.org/10.1177/0898264315589579 28, 267–284 (2015).
Cacioppo, J. T. & Cacioppo, S. Social relationships and health: The toxic effects of perceived social isolation. Soc Personal Psychol Compass 8, 58–72 (2014).
Steptoe, A. & Zaninotto, P. Lower socioeconomic status and the acceleration of aging: An outcome-wide analysis. Proc Natl Acad Sci U S A 117, 14911–14917 (2020).
Sattler, C., Erickson, K. I., Toro, P. & Schröder, J. Physical Fitness as a Protective Factor for Cognitive Impairment in a Prospective Population-Based Study in Germany. Journal of Alzheimer’s Disease 26, 709–718 (2011).
El-Sayes, J., Harasym, D., Turco, C. v, Locke, M. B. & Nelson, A. J. Exercise-Induced Neuroplasticity: A Mechanistic Model and Prospects for Promoting Plasticity. The Neuroscientist 25, 65–85 (2018).
Knäuper, B. et al. How aging affects self-reports. European Journal of Ageing 2016 13:2 13, 185–193 (2016).

(Not answered)

supplmaterial.docx

Download PDF

Version 1

posted

You are reading this older preprint version

Read the latest preprint version →

Cognitive and physical age-gaps in relation to mild cognitive impairment and behavioral phenotypes

Status:

Version 1

Abstract

Figures

Main text

Methods

Participants

Measures

Statistical analyses

Data availability

Results

Model characteristics

Differences in age-gaps between MCI and normal aging

Age-gaps and behavioral phenotypes

Discussion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1