Framework for Terms Used in Research of Reserve and Resilience
We present the latest version of the Framework for Terms Used in Research of Reserve and Resilience.
We also include a summary table and project summaries which both describe the awarded pilot studies and how they incorporate the operational research definitions suggested by the Framework.
The study of factors such as genetics and life exposures that allow some humans and nonhumans to age more successfully than others has important implications for health policy and intervention. In this context, overarching concepts like reserve and resilience are often invoked for capturing differential susceptibility to brain aging and disease. However, design of studies and communication across investigators in this area has been hampered by a diversity of terminology. Several groups have published proposed nomenclature and operational definitions for concepts including cognitive reserve, brain reserve, brain maintenance, compensation, scaffolding, resistance and resilience. Across these papers there are often disparate definitions for the same term. In addition, most of these papers focused on human studies, so the definitions and nomenclature are not optimally suitable for nonhuman studies.
The Collaboratory on Research Definitions for Reserve and Resilience in Cognitive Aging and Dementia was established in 2019 to institute a three-year process of developing consensus definitions and research guidelines for cognitive reserve and related concepts. The present document is the result of and iterative process including three large annual Workshops, input from focused workgroups, extensive participation and consultation of over 45 selected expert investigators who utilize multiple research approaches and study both humans and nonhumans. In addition, the framework benefited from input provided by over 100 participants in the Collaboratory workshops. Here we present a framework that includes definitions for three concepts, cognitive reserve, brain maintenance and brain reserve, along with suggested operational definitions to help guide the design of research investigating these concepts.
Our aim is to present a reasonable, well-defined set of operational definitions to encourage, advance, and develop research on these concepts. Inherent in this goal is to encourage research from investigators who have different views or use different concepts.
It would be important to note whether the operational definition of a differently named concept coincides with one of those described here. Similarly, this framework provides a basis for describing how the operational definition of another concepts differs from those suggested here.
Our intention is not to limit the creativity or ingenuity of investigators, or to claim that the framework presents the only way to investigate these important concepts. We hope to encourage research that provides either evidence-based support for these concepts or that presents data that cannot be accommodated by the proposed operational definitions of these concepts. We also hope that referring to this framework will facilitate collaboration and comparison of findings across studies and species.
The Collaboratory also sponsored 12 studies that were intended to implement the suggested research guidelines. This disparate set of studies incorporates humans and nonhumans, as well as multiple approaches including epidemiologic, neuroimaging, and intervention. We include a table that summarizes the projects and how they incorporate the framework presented here. The full descriptions of the pilot projects are in the supplementary appendix. Our hope is that the presentation of how the study designs incorporated the suggested framework will provide useful real-world examples.
Our hope for this framework is that the use of a common vocabulary and operational definitions will facilitate even greater progress in understanding the factors that are associated with successful aging.
Cognitive reserve (CR) is a property of the brain that allows for cognitive performance that is better than expected given the degree of life-course related brain changes and brain injury or disease.
- Property of the brain refers to multiple potential mechanisms including molecular, cellular and network levels. The working hypothesis is that these mechanisms help cope with or compensate for brain changes and brain injury or disease.
- These mechanisms can be characterized via biological or cognitive-experimental approaches.
- Better than expected cognitive performance ideally refers to trajectories measured longitudinally.
CR can be influenced by multiple genetic and environmental factors, operating at various points or continuously across the lifespan.
2. Operational Definition: General considerations
Research aimed at further elucidating CR requires the inclusion of three components:
- measures of life course-related brain changes, insults or disease that theoretically impact cognitive outcomes,
- measures of associated change in cognition, and
- a variable that influences the relationship between 1 and 2.
Ideally, the aim is to demonstrate that any proposed CR measure (e.g., a sociocultural or functional brain measure) moderates the relationship between 1 and 2. For example, in an analysis where change in brain atrophy/pathology measures (1) predicts change in cognition (2), and includes education as a hypothesized CR proxy (3), there is a statistical interaction between brain measures and education in predicting change in cognition.
Even without evidence for moderation, it can also be sufficient to demonstrate that a hypothesized CR proxy or measure is associated with cognitive performance over and above (e.g., after adjusting for) the effects of brain change, pathology, or insult. For example, in a multiple regression analysis of change in cognition that includes brain atrophy/pathology measures and a hypothesized CR proxy, the proxy should account for variance in cognitive performance. In this analysis, the CR proxy simply adds predictive information (a protective factor), a weaker form of CR evidence than moderation.
All three components are needed when investigating cognitive reserve. For example:
Demonstrating that expression of higher connectivity within a specific resting BOLD network is associated with slower cognitive decline is not sufficient to conclude that expression of this network is reflecting cognitive reserve. This observation encompasses components 2 and 3. To make a claim about CR it must also include component 1, i.e., measures of age-related brain change, insult or disease that theoretically impact cognitive outcomes.
Similarly, a relationship between a particular genotype and rate of cognitive decline would not be sufficient to conclude that this genotype is associated with cognitive reserve. It would be important to demonstrate that the genotype’s relationship to reduced rates of cognitive decline is expressed through moderation of age-related brain change or reduction of the expected impact on cognitive performance of a given brain insult or disease.
3. Specification of the 3 components needed to elucidate CR
3.1 Measures of life course-related brain developmental changes, injury or disease that theoretically impact cognitive outcomes.
This could consist of measures of anatomic changes such as loss of brain volume or white matter tract integrity, onset and progression of disease pathology such as biomarkers of neurodegenerative disease.
These changes could be more extensively specified. Measures/mechanisms underlying aging that could impact cognitive outcomes include change in structure or function of synapses, oxidative damage/stress, impaired stress response signaling, Ca2+ dyshomeostasis, dysregulation, mitochondrial function, impaired waste disposal, inflammation, epigenetics, stem cell depletion, and altered neuronal activity/connectivity.
It is likely that unmeasured or unknown brain or pathologic changes contribute to inter-individual variance in the cognitive outcomes. Their eventual inclusion would increase the accuracy of the process of elucidating CR.
3.2 Measures of cognition
This term encompasses measures of cognition and day-to-day function that change with aging and disease. When possible, it would be useful to adopt cognitive tests that show changes with age or brain disease, and that can be used in humans and other animals. In this case, it is important to be mindful that pure operational similarity between human and nonhuman tasks is not sufficient; the tasks need to tap a similar underlying neurological system.
3.3 Cognitive Reserve Proxy/Mechanism: A hypothesized variable that influences the relationship between 1 and 2
As the definition of CR states, these mechanisms can be characterized via biological or cognitive experimental approaches.
Proxies for cognitive reserve in human studies have included features associated with both endowment and experience, including early age IQ, cognitively stimulating exposures across the age span, education, occupational exposures, leisure activity, social networks, or other exposures, hypothesized or to be discovered, that might impart cognitive reserve. Similar proxies such as behavioral training, physical exercise, environmental enrichment, social housing, or diet are applicable to nonhuman studies.
In addition, the nature of the cognitive reserve proxy or mechanism that influences the relationship between 1 and 2 can be explored. For example, investigators might explore whether differential expression of a specific functional network is associated the degree of sustained cognitive function in the face of age-related brain changes that impact cognition. More generally, mechanisms underlying cognitive reserve could be specified at the molecular, cellular or network levels.
4. Example of studies of CR
In studies of CR, longitudinal designs optimally address the three features underlying the concept of CR. However, rich information can be gained from cross sectional studies including discovering variables that appear to be critical for CR, establishing preliminary observations, providing insight into neurobiological mechanisms and developing research or conceptual approaches.
4.1 Longitudinal study incorporating measures of brain and cognitive change.
In a longitudinal study, one could explore whether some life exposure conceptually linked to CR moderates the relationship between change in brain status (e.g., volume, white matter tract integrity, white matter hyperintensity burden) and change in cognition. For example, one could establish a relationship between age-related changes in cortical thickness, brain volume, and white matter tract integrity with changes in cognition. The potential moderation by education of this relationship could then be explored. Such moderation would provide support for the idea that higher education is associated with cognitive reserve.
4.2 Neural implementation of cognitive reserve:
Although variables such as IQ, education, occupational attainment etc. can be associated with cognitive reserve as described in 4.1, more insight into the mechanisms underlying cognitive reserve might be obtained from studies that directly examine neural mechanisms.
In both human and non-human studies, imaging techniques including functional MRI (fMRI) spectroscopy, and EEG are uniquely suited for longitudinal measurements, providing in-depth assessments of brain structure, neural activity, and the chemistry in the aging brain. Extracellular vesicle biology in blood is advancing rapidly and may provide a translatable fluid biopsy for relevant brain changes in this context.
Thus, one goal might be to identify functional networks or circuits, whose differential expression moderates the relationship between age-related brain changes that impact cognitive outcomes and the associated change in cognition. For example, longitudinal studies of aging or neurodegenerative disease can investigate how the relationship between changes in structure/function and cognition/clinical status can be moderated by proposed reserve-related networks. It would be of interest to determine whether differential expression of this network is related to life exposures such as education or occupational experience. This would create a relationship between a proxy for CR and a brain potential mechanism underlying that proxy.
4.3 Intervention studies and natural experiments
Intervention studies can most directly test whether some exposure or mechanism underlies CR by examining whether the intervention moderates the effect of age-related brain changes on cognitive outcomes. These studies can help explore mechanisms underlying CR.
Similarly, controlled perturbations such as transcranial or direct current brain stimulation could model brain insult, stressor or disease. Alternately, they could be used to facilitate or perturb networks/circuits associated with CR.
Sometimes, environmental changes can be used as natural experiments. A natural experiment is a situation when some change occurs in the environment that are not under experimental control and approximates random assignment. An example of such a natural experiment is changes to compulsory schooling laws. Conversely, animal models that feature increased individual differences in cognitive aging, under conditions of tightly controlled life-course exposures, can test for inherent genetic and biological moderators or mediators of CR.
Brain maintenance refers to the relative absence over time of changes in neural resources or neuropatholgic changes as a determinant of preserved cognition in older age.
Brain maintenance can be influenced by multiple genetic and environmental factors, operating at various points across the lifespan.
2. Operational Definition
Brain maintenance is influenced by factors (genes, sex, early life influence or differential experiences) that slow or prevent both brain changes associated with aging and disease and the associated cognitive decline. The emphasis lies on change over time. Thus, brain maintenance may be operationalized as preservation of cognitive function associated with minimal changes in brain markers of aging or disease.
Research aimed at further elucidating BM requires the inclusion of three components.
- measures of age-related brain changes, injury or disease that theoretically impact cognitive outcomes,
- measures of associated change in cognition.
Demonstrating a link between less change in 1 and less change in 2 would be evidence of brain maintenance.
To investigate potential mechanisms of individual differences in BM one could examine:
- a hypothetical variable that influences 1.
This variable can encompass many of the same exposures potentially associated with CR. However, their impact on BM in this context would be specific to maintaining the structural and functional integrity of the brain.
4. Example of studies of BM
BM is optimally ascertained in longitudinal designs. A single time point measurement cannot definitively differentiate people who have maintained their brain from those who did not, but started at a higher baseline level. In both human and non-human studies this issue can be addressed to some degree by determining what level of brain status is expected for a particular age, or considering a given brain measure relative to the distribution seen in younger subjects. However, longitudinal designs are preferable to examine factors underlying interindividual differences in the change in neural resources that are in turn associated with differences in cognitive outcomes.
3.1 Longitudinal study of brain maintenance
A general approach to studying BM would be to examine longitudinally whether individual differences in the rate of age- or disease-related brain anomalies accumulated over time are related to individual differences in the rate of cognitive change.
3.2 Exposures related to brain maintenance
An extension of study 3.1 would be to assess potential proxies or mechanisms (e.g., genetic, lifestyle, neural) that are associated with these different trajectories of brain maintenance/change.
In summary, BM and CR are complementary concepts. BM accounts for individual differences in cognitive trajectories that are associated with differences in rate of brain change. In contrast, CR addresses individual differences in trajectories of cognitive change once BM is accounted for (i.e. given the same amount of BM).
Brain reserve has been used to reflect the neurobiological status of the brain (numbers of neurons, synapses, etc.). To the extent that BM is effective you have better BR at any point in time. BR does not involve active adaptation of functional cognitive processes in the presence of injury or disease as does CR.
2. Operational Definition
Research aimed at further elucidating BR requires the inclusion of 2 components:
- measures of brain features that theoretically are associated with cognition.
- associated measures of cognition.
3. Example of studies of BR
A goal of cross-sectional studies of BR could be to demonstrate that some brain features (e.g., regional volume, cortical surface area, patterns of cortical thickness, white matter microstructural properties) are associated with individual differences in cognitive performance.
Longitudinally, differences in BR could account for the observation that individuals starting at a different levels of cognition may show the same rate of age- or disease-related cognitive decline. This could reflect different initial levels (intercepts) due to variation in BR, but similar rates of change (slopes) due to similar depletion of BR.
BR has also been associated with individual differences in cognitive change given a certain amount of brain injury or disease, such as amyloid plaques and neurofibrillary tangles. This association relies on a threshold model, where there is a specific threshold of depletion of neurobiological capital that is required to induce disease-related changes. Those with higher BR can tolerate more injury before they show symptoms.
Cognitive Reserve CR
Cognitive Reserve CR
Cognitive Reserve CR
Factor Associated with CR
Cognitive Reserve CR
Brain Maintenance BM
Brain Maintenance BM
Brain Maintenance BM
Factor Promoting BM
Brain Maintenance BM
Lídia Vaqué-Alcázar, University of Barcelona
Decision tree testing cognition-MRI associations to define and differentiate CR and BM
CR and BM are complementary constructs, but a robust differentiation between them is challenging. Therefore, we designed a unified approach incorporating measures of brain magnetic resonance imaging (MRI) data and cognitive change, which will allow us to examine the neural mechanisms by identifying crucial functional circuits underlying CR and BM. Our experimental design is based on a decision tree where at each step a residual approach between episodic memory (EM) and multimodal MRI-based measures will provide a general metric rising to distinguish between BM: when there is a correspondence between EM stability and brain structure, and CR: when we detect a discrepancy between EM stability and multimodal MRI-based measures. This proposal will use already available longitudinal data of healthy participants aged 60 or over, including the 3 key components for each subject: (1) measures of age-related changes assessed by MRI acquisitions, (2) EM performance, and (3) typical CR proxies. A strength of this study is that here, in order to categorize the individuals, we will focus on a discrepancy (or not) between the age-related brain changes (component 1) and the associated changes in cognition (component 2). Further, we have planned to study how the ‘hypothetical variable’ (component 3) influences the relationship between 1 and 2 (i.e., CR) or influences 1 (i.e., BM). This could provide evidence whether the component 3 (i.e., ‘hypothetical variable’) proposed in both CR and BM operational definitions may be contributing at the same time to reduce rates of cognitive decline in the face of age-related brain changes (i.e., CR), and also may be associated with neuroprotective mechanisms, more related to BM (or BR). Overall, we think that the present approach will overcome some limitations in terms of clarifying the conceptual boundaries between CR and BM concepts.
Daniel Gray, University of Arizona
A test of the hypothesis that factors acting to protect synapse function are at the core of the biological basis of cognitive reserve
This study utilizes legacy brain tissue from a large experiment designed to study the molecular mechanisms of cognitive aptitude in F344 rats across the lifespan, including young, middle-aged, and old rats. All rats underwent (1) a series of structural and diffusion-weighted MRI scans, (2) a large battery of cognitive assessments that included tests of both medial temporal lobe and frontal cortical brain function that enabled the selection of those individuals that were average-, low- or high-performing for their age groups, and (3) histological brain sections were prepared of the hippocampus and will be labelled to identify two distinct proteins that are critical for synapse health and plasticity. These include the immediate-early gene neuronal pentraxin 2 (NPTX2) and markers for a specialized extracellular matrix structure called perineuronal nets (PNNs) that provide physical and biochemical support to synapses. These data will allow us to test via multiple regression models whether the expression of proteins associated with synapse health (3) moderates the relationship between structural changes in grey and white matter as assessed by MRI (1) and levels of cognitive function (2). We anticipate that some subset of the structural brain variables (total volume, ventricle size, lipofuscin) measured will be associated with synapse health and cognitive function. If our hypothesis is correct, we expect that the expression of our markers of synapse health will be highest in the high cognitive functioning rats across the lifespan (that is, in all 3 age groups). This might indicate that synapse health at all points of the lifespan helps define the trajectory that an individual’s cognitive function will take as age-associated changes in brain structure and function begin to accumulate. If we can establish this relationship, then we can design longitudinal studies to test our prediction that the slopes of cognitive decline will differ in animals that conform to our definition of these distinct cognitive statuses. This could be accomplished in future experiments by measuring cognitive status at regular intervals (in animals across the lifespan) to define the behavioral trajectories of these individuals. These animals can then be sacrificed at various points of the lifespan to assess synapse health in the context of the behavioral trajectories defined by the longitudinal cognitive testing.
Holly Hunsberger, Columbia University
Cognitive training to enhance cognitive reserve in aging mice
In this study cognitive reserve will be measured in the face of age-related changes. We can also include Alzheimer’s disease (AD) mice, which would be included as a brain insult/disease state. These studies will be longitudinal. The intervention here is, education or training. The question is: how does education enhance CR in the face of aging and disease? We will measure short, long, and working memory at 6-8 months of age as cognitive performance. We have the option to test these mice in short or working memory as a baseline measure at 2 months. We will measure the number of senescent cells within the hippocampus (and later whole brain) as a potential measure of brain change that influences memory. We know that senescent cells are increased with age. We will tag a contextual memory to examine differences in memory trace cell activation as a potential molecular mechanism of CR. We know that memory trace cell activation is decreased with age and AD pathology but anticipate that early-life training may help preserve it. The study can also test for potential brain maintenance. In this scenario mice will naturally age without training. Sex and mouse strain could impact cell senescence and memory trace activation, thus preserving memory. Other factors that may impact CR or BM will also be considered, including exercise in the training apparatus, handling of mice, novel environment of the training apparatus.
Anna Marseglia, Karolinska Institute
Defining a biological marker of cognitive reserve with deep learning from structural MRI
The overall aim of this study is to investigate whether a biological marker of cognitive reserve can modify the associations between measure(s) of brain pathology and changes in cognitive function. Age-related brain pathology measures include of Alzheimer’s disease (AD) neurodegeneration as assessed with a cortical thickness pattern and of vascular damage based on markers of SVD. Cognition is measured using a composite measure of global cognition (G-score), generated by averaging the Z-scores a battery of tests. Our biological measure of cognitive reserve will be based on differences between predicted brain age (PBA) and the person’s chronological age (CA) was developed given the same level of cognition, based on a deep learning model. Differences between PBA and CA are computed and categorized into 4 quartiles, with Q1 indicating younger-appearing brain (thus more reserve) to Q4 indicating older-appearing brain (less reserve). Then, within the Gothenburg H70 Birth Cohort Study-Birth cohort 1944 (including septuagenarians without dementia, cognitive impairment, or other neuro-psychiatric disorders, followed up after 5 yrs. from baseline [2014-2016]), we will investigate cross sectionally and longitudinally the potential modifying effect of the biological marker of cognitive reserve on the relationship between age-related brain pathology and changes in global cognitive function.
Rory Boyle, Trinity College Dublin
Improving the moderation and independent effect criteria of cognitive reserve
One criteria for candidate neuroimaging measure of CR is that it must be associated with cognition independently of brain structure (i.e. should display an ‘independent’ effect on cognition). The overall aim of this study is to assess imaging-based correlates of CR. We will first create a residual variable representing CR from the regression of global cognition on grey matter volume, hippocampal volume, mean cortical thickness, age, and sex. The residual from this regression reflects the degree to which an individual’s cognition is better or worse than expected given their brain structure, age, and sex. We will then develop a functional connectivity-based measure of CR (network-strength predicted CR) by applying connectome-based predictive modelling to predict the CR residual from functional connectivity data. We will test the validity of network-strength predicted CR by assessing whether it a) explains variation in longitudinal cognitive function above and beyond the effects of Alzheimer’s disease (AD) signature cortical thickness (brain change) and/or b) moderates the relationship between AD-signature cortical thickness and longitudinal cognitive function. If either of these criteria are satisfied, we will operationalize CR as network-strength predicted CR. If neither criteria is satisfied, we will operationalize CR as the CR residual. We will operationalize BM using the brain-predicted age difference, a robust machine learning residual, which compares an individual’s structural brain health, reflected by their voxel-wise grey matter density, to the state typically expected at that individual’s age. We have previously shown that larger brain-predicted age differences, reflecting worse BM, are associated with lower cognitive function. This previous study included the two required components for elucidating BM: 1) measure of age-related brain change = (cross-sectional) brain-predicted age difference; 2) measures of associated change in cognition = (cross-sectional) cognitive function. We will assess whether the operationalized measures of BM and CR, separately, are independently associated with longitudinal cognitive function after accounting for physiological and clinical health covariates. We hypothesize that CR, but not BM, will remain independently associated with longitudinal cognitive function.
Colin Groot, VU UMC Amsterdam
Reserve and Maintenance in AD: Effects on Individual Cognitive Trajectories (REMIND-ICT)
We will investigate the concepts of cognitive reserve, brain reserve and brain maintenance by evaluating individual cognitive trajectories, as assessed by Bayesian change point analyses. The first analysis will include longitudinal data from initially nondemented individuals with known AD pathology who then receive an AD diagnosis. The analysis looks at the initial level of performance, the time to the change point where cognitive decline begins, and the time to dementia. CR, as represented by the proxy education would be related to both the higher level of initial premorbid function and to the delay in inflection point. We hypothesize that brain reserve (as assessed with structural measures: gray matter atrophy and white matter hyperintensities) will be more strongly associated with higher initial performance (higher inflection points). However the time to the point of infection more driven to a greater degree by brain maintenance as measured by differential change in build-up of structural measures and AD pathology (CSF amyloid and tau).
The second analysis will focus on amyloid positive individuals diagnosed with AD. It will assess the relative effects of brain reserve and brain maintenance on cognitive trajectories after cognitive decline sets in. It will longitudinally assess the interplay of rates of structural degeneration and AD pathology build-up after a diagnosis of dementia. Cognitive reserve would be reflected by more rapid decline in individuals with higher education, since the pathology is more advanced at the time of diagnosis. However, we will also evaluate the rate of increase in AD pathology, since more rapid decline could be accounted for differential loss of brain maintenance which would result in the more rapid increase in AD pathology.
Gabriel Ziegler, University of Magdeburg
Exploring multivariate metrics to benchmark functional brain maintenance
In this project brain maintenance is based on a multivariate index which primarily measures youthlike appearance of distributed activity patterns in the FADE fMRI task in old age. For that purpose, a model of fMRI activity patterns (during performance of the same task) is learned in a healthy young reference cohort (n=100). The trained model accounts for multi-voxel correlations and heterogeneity within this group. This one-class classification (or novelty detection) approach then enables defining a similarity index (called FADE score) for new unseen subjects (which have performed the FADE task) which will be assessed in old age, e.g., in participants at risk of dementia (DELCODE DZNE cohort, n=543, 59-89 yo, characterized using biomarkers, WMH, longitudinal MRI and cognition). The proposed FADE score is cross-sectionally defined using individual differences at baseline rather than within-subject changes as required by the current definition of BM. However, it might play the role of a hypothetical variable predicting age- and/or pathology related brain changes, which are tied to cognitive decline on an individual level. More specifically, we use hippocampal (or AD-) network atrophy rates as measures of age-related brain changes, insults or disease that theoretically impact cognitive outcomes. The latter are assessed using clinical PACC score over follow-up measurements. First, we establish whether there is an association of hippocampal network and cognitive changes over follow-ups. Second, we test if FADE differences can predict these shared individual differences of brain-behavioral changes. The latter would support the role of FADE score as an important baseline predictor for BM (over time). Finally, the project aims to test whether the proposed FADE score (as a fMRI-based BM index) might play the role of a CR variable in participants facing ageing-related pathology. More specifically, the project implements the CR definitions by testing if effects of hippocampal network atrophy rates (as a longitudinal measure of aging-related brain pathology) on rates of cognitive decline (over follow-ups) are moderated by the FADE score.
Christian Habeck, Columbia University
Functional activation patterns to explain cognitive performance beyond brain structure and age
We propose to operationalize Cognitive Reserve is as an fMRI activation pattern that (1) accounts for cognitive performance and (2) is uncorrelated with brain structure and age. This is accord with the suggested operational definition for CR in that it includes: brain changes associated with cognition: multimodal characterization of brain structure changes, using regional cortical volume, regional cortical thickness, tract integrity, 2: change task performance, and 3 a putative measure of cognitive reserve: an fMRI activation, or functional-connectivity, pattern. In order to derive that CR pattern, we will Orthogonalize fMRI data and cognitive outcome with respect to brain structure and age. We will then use PCA-regression (Scaled Subprofile Modeling; SSM) to derive a residual activation pattern that accounts for residual cognitive performance; the pattern score will now be orthogonal to brain structure and age by design. We can then test the associations of pattern score with traditional CR proxies like education and verbal intelligence using standard Pearson correlation. We will also forward apply the derived patterns(s) to held-out data and test whether the resulting pattern score accounts for cognitive performance beyond age and brain structure. For genuine replication in de-novo data sets, we will use fMRI activation data from several other tasks collected in our laboratory, a verbal Working-Memory task, a set-switching task.
Melis Anaturk, UCL
Cross-sectional and longitudinal modelling of brain and cognitive age to study cognitive reserve and maintenance
Our study aims to use machine learning to estimate brain age based on MRI (i.e., T1-weighted, diffusion-weighted and FLAIR images) and cognitive age, based on neuropsychological measures. This will enable us to estimate the brain age gap (BAG) and cognitive age gap (CAG), which quantify how much an individual’s apparent brain/cognitive age deviates from healthy aging patterns. Our study then examines the heritability of cross-sectional estimates of BAG and CAG and their genomic correlations, as well as how traditional markers of cognitive reserve (i.e., premorbid IQ, education) and a composite lifestyle measure relate to changes in these metrics over time. The proposed project will address these research objectives by conducting a secondary analysis of three datasets: the UK Biobank study, Lothian Birth Cohort 1936 and Insight 46.
The cross-sectional analyses in our study premise that individual differences in BAG at a given time point reflect, at least partially, interindividual variation in age-related brain changes. However, a cross-sectional approach cannot truly disentangle whether the observed differences in BAG are due to pre-existing differences in brain structure or differences in the rate of brain ageing between individuals. Therefore, our study will also take longitudinal data into account, by estimating BAG at all available time points. This will allow us to statistically model changes in BAG and CAG, over time. In this way, stability in BAG over time could be indicative of brain maintenance while stable CAG trajectories combined with concurrent changes in BAG could indicate cognitive reserve. Overall, our study integrates the components of brain maintenance as (1) BAG potentially captures age-related brain changes, (2) will be investigated alongside cognitive trajectories and (3) examined in relation to lifestyle factors. Additionally, our study operationalizes cognitive reserve as CAG, after adjusting for BAG. In this way, we take into account age-related changes in brain structure (component 1) and in cognitive function (component 2). We also regress premorbid IQ, education and a lifestyle marker onto CAG, after adjusting for BAG (component 3).
Tom Foster, University of Florida
Molecular markers to operationally define cognitive reserve
Utilizing a public functional genomics data repository, Gene Expression Omnibus, we identified cross-sectional transcriptional data from the CA1 region of male rats. Using the water maze, aged animals were characterized as cognitively impaired (AI) or unimpaired (AU; i.e., component 2 from the consensus document). From 5 initial studies, we identified genes representing possible age-related brain changes (i.e., component 1 from the consensus document), defined as genes that were significantly (p<0.05) different (increasing or decreasing) between all aged animals and young. This set of potential aging genes was then tested against other data sets that examined gene expression in the whole hippocampus of rats and mice to evaluate reliability of these age-related genes. To determine possible mechanisms for cognitive reserve (i.e., component 3 from the consensus document), genes were identified that were significantly (p<0.025) different between AU and young, and these same genes were not different (p>0.1) from young and AI rats. In addition, we defined brain maintenance genes as those that were not different (p>0.1) between young and AU rats, and these same genes were significantly (p<0.025) different between AI and young. Finally, we examined the relationship between expression of activity dependent immediate early genes (IEG) in region CA1 and the medial prefrontal cortex (mPFC) of animals characterized on two different tasks, water maze and attentional set-shift. Shifts in network engagement, exhibited as efficiency (less activation of immediate early genes, which are associated with preserved cognition) and compensation (greater activation of immediate early genes which are associated with preserved cognition), were evaluated between young, AU, and AI rats. To ascertain shifts in regional recruitment as compensatory or debilitative, IEG activation in the CA1 and mPFC were correlated between the three groups.
Sara James, UCL
Disentangling the role of CR proxies in a longitudinal age-homogenous study incorporating multiple measures of brain health and cognitive change
Exposures associated with CR may in some cases not only moderate the expression of age-related brain changes, insults or disease (cognitive reserve), but could also contribute to the development of age-related brain changes, insults or disease (brain maintenance) Using a longitudinal age-homogenous population-based study we aim to investigate: 1: Which typical CR proxy exposures are directly associated with the development and change of a range of measures indexing age-related brain changes, insults or disease; and to what extent the associated brain health measures mediate the relationship between the CR proxy exposures and cognition and cognitive decline. 2. Which, and to what extent, a range of typical CR proxy exposures moderate the subtle cognitive expression (cognition and cognitive decline) of a range of measures indexing age-related brain changes, insults or disease. We will explore the unique contributions of a range of CR proxy exposures given that they could have differential effects. The CR proxy exposures of interest have been chosen based on prior evidence that they predict later-life cognition and include early social circumstances; childhood IQ; own educational attainment; occupational attainment; and crystallized ability. We will explore the effects from a range of pathological markers, characterizing overall brain health. Pathology markers include Aβ status (indicative of AD pathology); white matter hyperintensity volume (indicative of cerebral small vessel disease); hippocampal and brain volume, and cortical thickness (indicative of atrophy).
Eero Vuoksimaa, University of Helsinki
Aging and Memory – Origins of heterogeneity in cognitive trajectories study
We will use longitudinal cognitive and brain imaging data from the Vietnam Era Twin Study of Aging to investigate if brain maintenance and/or cognitive reserve explain heterogeneity in episodic memory trajectories from late middle age to early old age. A composite score of three episodic memory tests from three time points (at mean ages of 56y, 62y, 68y) will be used to indicate cognitive change (outcome / dependent variable). Predictors (independent variables) of cognitive change are changes in relative cortical surface area and thickness in regions implicated in Alzheimer’s disease (brain maintenance model). We will test whether young adult – at a mean age of 20y – general cognitive ability and lifetime years of education impact brain maintenance, cognitive reserve or both. For BM we will test if these exposures are associated with less brain change, resulting in better preserved cognition. For CR, we will test if these proxy measures of cognitive reserve moderate the effects of cortical surface area / thickness change on cognitive trajectories: cognitive reserve interaction model with three components to test if those with higher cognitive ability can tolerate brain changes better than those with lower cognitive ability.