The values indicate the loadings of the 5 factors to cognitive reserve. e1, e2, e3, e4, and e5 indicate the measurement error for each cognitive reserve factor. Structural equation modeling fit statistics: χ25 = 41.919; P < .001; comparative fit index = 0.981; standardized root mean squared residual = 0.026; root mean squared error of approximation = 0.068.
Incidence rates were adjusted for age, sex, smoking, alcohol consumption, physical activity, body mass index, heart disease, hypertension, cerebrovascular disease, diabetes, and apolipoprotein E ɛ4. AD indicates Alzheimer disease.
aP < .05 compared with lowest CR score tertile.
bP < .01 compared with lowest CR score tertile.
eMethods 1. Covariate details.
eMethods 2. SEM computation details.
eTable 1. Pearson correlation coefficient matrix of observed contributors to cognitive reserve indicator included in the calculation of a latent variable.
eTable 2. Baseline and neuropathological characteristics of the 611 autopsied participants by dementia.
eTable 3. Association of brain pathologies with middle and highest cognitive reserve at baseline using multinomial logistic regression.
eTable 4. Hazard ratios (HRs) with 95% CIs of dementia and Alzheimer disease–related dementia in relation to cognitive reserve among participants with autopsies.
eTable 5. Incidence rates (per 1000 person-years) of dementia and Alzheimer disease–related dementia by different levels of cognitive reserve and brain pathologies.
eTable 6. Hazard ratios (HRs) with 95% CIs of the association of levels of cognitive reserve with dementia and Alzheimer disease–related dementia with multiple imputation for missing data.
eTable 7. Hazard ratios (HRs) and 95% CIs of the association of cognitive reserve with dementia and Alzheimer disease–related dementia after excluding MCI.
eTable 8. Hazard ratios (HRs) and 95% CIs of the association of cognitive reserve with dementia and Alzheimer disease–related dementia among all participants: results from competing risk models.
eTable 9. Hazard ratios (HRs) and 95% CIs of the association of cognitive reserve with dementia and Alzheimer disease–related dementia among all participants.
eTable 10. Hazard ratios (HRs) and 95% CIs of the association of cognitive reserve with dementia and Alzheimer disease–related dementia among all participants.
eTable 11. Hazard ratios (HRs) and 95% CIs of the associations of individual cognitive reserve indicators with dementia and Alzheimer disease–related dementia among all participants.
eFigure 1. Participant flow chart.
eFigure 2. Kaplan-Meier survival curves of dementia by cognitive reserve.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Xu H, Yang R, Qi X, et al. Association of Lifespan Cognitive Reserve Indicator With Dementia Risk in the Presence of Brain Pathologies. JAMA Neurol. 2019;76(10):1184–1191. doi:10.1001/jamaneurol.2019.2455
Is high lifespan cognitive reserve (CR) indicator associated with a reduction in dementia risk, and how strong is this association in the presence of high brain pathologies?
In this cohort study including 1602 dementia-free older adults, high lifespan CR was associated with a decreased risk of dementia. This association was present in people with high Alzheimer disease and vascular pathologies.
Accumulative educational and mentally stimulating activities enhancing CR throughout life might be a feasible strategy to prevent dementia, even for people with high brain pathologies.
Evidence on the association of lifespan cognitive reserve (CR) with dementia is limited, and the strength of this association in the presence of brain pathologies is unknown.
To examine the association of lifespan CR with dementia risk, taking brain pathologies into account.
Design, Setting, and Participants
This study used data from 2022 participants in the Rush Memory and Aging Project, an ongoing community-based cohort study with annual follow-up from 1997 to 2018 (mean follow-up, 6 years; maximum follow-up, 20 years). After excluding 420 individuals who had prevalent dementia, missing data on CR, or dropped out, 1602 dementia-free adults were identified at baseline and evaluated to detect incident dementia. During follow-up, 611 died and underwent autopsies. Data were analyzed from May to September 2018.
Information on CR factors (education; early-life, midlife, and late-life cognitive activities; and social activities in late life) was obtained at baseline. Based on these factors, lifespan CR scores were captured using a latent variable from a structural equation model and was divided into tertiles (lowest, middle, and highest).
Main Outcomes and Measures
Dementia was diagnosed following international criteria. Neuropathologic evaluations for Alzheimer disease and other brain pathologies were performed in autopsied participants. The association of lifespan CR with dementia or brain pathologies was estimated using Cox regression models or logistic regression.
Of the 1602 included participants, 1216 (75.9%) were women, and the mean (SD) age was 79.6 (7.5) years. During follow-up, 386 participants developed dementia (24.1%), including 357 participants with Alzheimer disease–related dementia (22.3%). The multiadjusted hazards ratios (HRs) of dementia were 0.77 (95% CI, 0.59-0.99) for participants in the middle CR score tertile and 0.61 (95% CI, 0.47-0.81) for those in the highest CR score tertile compared with those in the lowest CR score tertile. In autopsied participants, CR was not associated with most brain pathologies, and the association of CR with dementia remained significant after additional adjustment for brain pathologies (HR, 0.60; 95% CI, 0.42-0.86). The highest CR score tertile was associated with a reduction in dementia risk, even among participants with high Alzheimer disease pathology (HR, 0.57; 95% CI, 0.37-0.87) and any gross infarcts (HR, 0.34; 95% CI, 0.18-0.62).
Conclusions and Relevance
High lifespan CR is associated with a reduction in dementia risk, even in the presence of high brain pathologies. Our findings highlight the importance of lifespan CR accumulation in dementia prevention.
The cognitive reserve (CR) hypothesis has been proposed as a compensatory mechanism to cope with age-related brain damage and to account for interindividual variability in the ability to maintain cognitive function in the presence of brain pathologies.1 Education, occupation attainment, and social and cognitive activities have been considered as proxy measures of CR.2-4 However, emerging evidence has suggested that CR is an active construct that develops from continued life experiences.5 One reserve-enhancing factor during a certain period alone could not fully explain the accumulation of cognitive activities over the life course.6 So far, evidence on whether and to what extent lifespan CR accumulation may reduce dementia risk is still limited.6
According to CR theory, Stern et al7 have suggested that 3 components are required for CR-related research: a measure of CR, clinical or cognitive performance outcomes, and the status of the brain (reflecting brain pathologies). However, as in vivo measures of neuronal pathology are not widely available, few studies presenting the association of the proxy of CR with dementia have taken brain pathologies into account. Several studies have shown that CR might be directly associated with neuropathology and resist the accumulation of brain pathologies.8-10 However, other studies have indicated that CR might bypass classic brain pathologies11 and represent other pathways, such as enhancing brain network efficiency to compensate for dementia pathology.12,13 Therefore, the role of brain pathologies in the association of CR with cognitive outcomes remains unclear.
We previously reported that more frequent cognitive activities from early to late life and social activities in late life were associated with slower cognitive decline.14-16 In the present study, we aim to verify the hypothesis that high lifespan CR accumulation is associated with a reduction in clinical dementia risk and to estimate the strength of this association in the presence of brain pathologies using data from a long-term community-based cohort study in which people donated their brains for autopsy.
The Rush Memory and Aging Project17 is an ongoing prospective cohort study that investigates risk factors for common chronic neurodegenerative conditions in older adults. At the time of enrollment and thereafter, all participants underwent a comprehensive clinical assessment, including medical history, neurological examination, and detailed cognitive function testing.17
Beginning in 1997 through 2018, a total of 2022 participants were enrolled. The participants were annually followed-up with, for a maximum of 20 years. In this study, among 2022 participants, a total of 420 were excluded, including 112 with prevalent dementia, 101 with missing data on CR-enhancing factors at baseline, 31 not eligible for their first follow-up because of mental disorder, 136 who dropped out before the first follow-up evaluation, and 40 who died. Thus, 1602 participants were available for the current study. During the follow-up period, 747 participants died, of whom 611 (81.8%) underwent autopsy (eFigure 1 in the Supplement).
The study was approved by the Institutional Review Board of Rush University Medical Center and was performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments. Written informed consent was obtained from all participants, and Uniform Anatomic Gift Act documentation was obtained for all participants who underwent autopsy. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.
Data on stimulating mental and social activities over the life course and social network collected at baseline were considered to construct an indicator of CR. These factors included education; early-life, midlife, and late-life cognitive activities; social activity in late life; and social network in late life.
For education, years of education was calculated based on the number of years of regular school reported at baseline.18 Among 1602 participants, the mean (range) years of education was 14.76 (0-29) years.
For early-life, midlife, and late-life cognitive activities, participants completed a 37-item cognitive activity questionnaire at baseline.14,16 The activities included reading books, visiting a library, and writing letters during childhood (aged 6 to 12 years), young adulthood (aged approximately 18 years), middle age (aged approximately 40 years), and late life (at the study enrollment). Frequencies of participation in each activity at different periods of life were rated from 1 (once a year or less) to 5 (every day or about every day).14 In the questionnaire, there were 7 late-life (ie, at baseline) activities and 21 early-life activities, including 11 regarding childhood, 10 regarding young adulthood, and 9 regarding midlife cognitive activity. Item scores were averaged to yield separate early-life, midlife, and late-life activity measures.19
For social activity in late life, frequency of activity was assessed using a 6-item scale, including (1) going to restaurants, sporting events, or teletract (off-track betting); (2) going on day or overnight trips; (3) doing unpaid community or volunteer work; (4) visiting relatives’ or friends’ houses; (5) participating in groups, such as at a senior center; and (6) attending church or religions services. Participants were asked to rate how often they participated in each activity based on a 5-point scale from 1 (once a year or less) to 5 (every day or about every day). Item scores were summed and averaged to obtain a composite measure of social activity.20,21
For social network in late life, participants were asked about the number of children they have and meet monthly. They were also asked about the number of relatives (besides spouse and children) and other close friends to whom they feel close and with whom they felt at ease and could talk to about private matters and could call on for help as well as how many of these people they see monthly. Social network size was the number of these individuals (children, family, and friends) seen at least once per month.22
Clinical diagnoses of dementia and Alzheimer disease (AD)–related dementia were based on criteria of the joint working group of the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer Disease and Related Disorders Association.23 The diagnosis of mild cognitive impairment (MCI) referred to persons with cognitive impairment diagnosed by the neuropsychologist but without a clinical diagnosis of dementia by the examining clinician.24,25
Postmortem neuropathologic evaluations were conducted on the 611 autopsied brains (median [interquartile range] postmortem interval, 6.90 [3.53-10.27] hours). Global AD pathology burden, β-amyloid plaques, and tangles were quantified26 and also categorized as being at low or high levels. Chronic infarcts, including gross infarcts and microinfarcts14; cerebral vascular disease pathology, including atherosclerosis, arteriolosclerosis, and cerebral amyloid angiopathy27; Lewy bodies14; and typical hippocampal sclerosis28 were assessed as present or absent.
A broad range of potential confounders and proxies for confounders were considered, including age, sex, smoking, alcohol consumption, physical activity, body mass index (BMI, calculated as weight in kilograms divided by height in meters squared), Mini-Mental State Examination (MMSE) score, heart disease, hypertension, diabetes, and apolipoprotein E (APOE) ε4 status. They are described in detail in eMethods 1 in the Supplement.
The characteristics of the study population by dementia or survival status were compared using t tests and nonparametric tests (χ2 and Wilcoxon rank sum tests). To construct the lifespan CR accumulation based on stimulating mental and social activities during life course, structural equation modeling (SEM) was performed to derive a best-fitting measurement model using full information maximum likelihood estimation (eMethods 2 in the Supplement). The lifespan CR score was operationalized as both a continuous and a categorical variable (based on the tertile distribution; reference, lowest tertile).
Cox regression models were used to estimate the hazard ratios (HRs) and 95% CIs for dementia incidence associated with the CR score. Follow-up time was calculated as the time from study entry to dementia diagnosis, death, or the last examination. The basic models were adjusted for age and sex. The multiadjusted models were further adjusted for smoking, alcohol consumption, physical activity, BMI, MMSE score, heart disease, hypertension, cerebrovascular disease, diabetes, APOE ɛ4, death, and brain pathologies. To test a dose-dependent association of CR with dementia risk, the categorized CR variable (1 for lowest tertile, 2 for middle tertile, and 3 for highest tertile) as a continuous variable was used in Cox regression models. Multiadjusted multinomial logistic regression was used to estimate odds ratios (ORs) with 95% CIs of the association of CR with the burden of AD and other brain pathologies. Incidence rates and 95% CIs of dementia per 1000 person-years were calculated for CR categories by different levels of brain pathologies, adjusting for the confounders.
In sensitivity analysis, multiple imputation by chained equation was used to impute data for 166 participants with missing values. Further, we excluded 420 individuals with MCI at baseline. Considering death as a competing risk for dementia, a competing risks model (the Fine-Gray model) was used to estimate HR for dementia associated with CR score. The level of statistical significance was set at a P value less than .05, and all P values were 2-tailed. All analyses were performed with Stata/SE version 15.0 (StataCorp).
Of the 1602 participants, 1216 (75.9%) were women, and the mean (SD) age was 79.6 (7.5) years. During follow-up (mean follow-up, 6 years; maximum follow-up, 20 years; total follow-up time, 9721.7 person-years), 386 participants developed dementia (24.1%), including 357 participants with AD-related dementia (22.3%). Table 1 presents the baseline characteristics of the study population by incident dementia.
The final CR indicator was generated based on the best-fitting SEM with the 5 CR-enhancing factors (education; early-life, midlife, and late-life cognitive activities; and social activities in late life), as presented in Figure 1. The model fit was satisfactory, and the correlation coefficients among the 5 factors varied from 0.17 to 0.64 (eTable 1 in the Supplement). A unique value for the CR score was obtained for each participant, and the score ranged from −8.00 to 5.81.
In both basic and multiadjusted Cox regression models, the CR score (as a continuous variable) was associated with a reduction in risk of dementia. After adjusting for age, sex, smoking, alcohol consumption, physical activity, BMI, MMSE score, heart disease, hypertension, cerebrovascular disease, diabetes, APOE ɛ4, and death, compared with those in the lowest CR score tertile, the HRs of dementia were 0.77 (95% CI, 0.59-0.99) for those in the middle CR score tertile and 0.61 (95% CI, 0.47-0.81) for those in the highest CR score tertile, and the HRs of AD-related dementia were 0.77 (95% CI, 0.59-1.00) for those in the middle CR score tertile and 0.61 (95% CI, 0.46-0.81) for those in the highest CR score tertile. The association of the CR score with dementia and AD-related dementia risk was dose dependent (Table 2).
Kaplan-Meier survival analysis showed that median (interquartile range) dementia onset time was 12.30 (8.78-15.82) years in participants in the lowest CR score tertile, 14.98 (11.14-18.82) years in participants in the middle CR score tertile, and more than 20 years in participants with highest CR. Participants in the highest CR score tertile had later dementia onset by more than 7 years compared with those in the lowest CR score tertile (eFigure 2 in the Supplement).
Of the 747 people who died, 611 (81.8%; 440 [72.0%] women; mean [SD] age, 83.0 [5.8] years) underwent autopsy, of whom 241 were diagnosed as having incident dementia. Baseline and neuropathological characteristics of the autopsied participants by dementia status are shown in eTable 2 in the Supplement.
In the post mortem data analysis, multiadjusted multinomial logistic regression analyses showed that compared with the lowest CR score tertile, the middle and highest CR score tertiles at baseline were not associated with the burden of AD pathology and other brain pathologies, except for gross infarcts (OR, 0.49; 95% CI, 0.31-0.78) (eTable 3 in the Supplement). There was no statistically significant association of MCI with global AD pathology (OR, 0.78; 95% CI, 0.37-1.67; P = .53) or gross infarcts (OR, 1.04; 95% CI, 0.38-2.82; P = .94) among those in the highest CR score tertile.
The multiadjusted Cox regression models showed that the highest CR score tertile was significantly associated with a reduction in risk of dementia (HR, 0.60; 95% CI, 0.42-0.86) and AD-related dementia (HR, 0.60; 95% CI, 0.41-0.87) compared with the lowest CR score tertile after additional adjustment for global AD pathology and other brain pathologies. The associations of CR score tertile with risk of dementia and AD-related dementia were dose dependent (eTable 4 in the Supplement).
Compared with those with high brain pathologies but in the lowest CR score tertile, the incident rates of dementia were about 38% to 55% lower in people both in the highest CR score tertile and with high brain pathologies (including global AD pathology, gross infarcts, and microscopic infarcts) (Figure 2; eTable 5 in the Supplement). In stratified analysis by level of brain pathology, the association of high CR score tertile with a reduction in dementia risk remained significant in participants with high AD pathology (HR, 0.57; 95% CI, 0.37-0.87) and any gross infarcts (HR, 0.34; 95% CI, 0.18-0.62) (Table 3).
The results were not altered much compared with those from initial analyses when we repeated the following analyses by (1) multiple imputation for missing values (eTable 6 in the Supplement), (2) excluding 420 individuals with MCI at baseline (eTable 7 in the Supplement), (3) using competing risks models in all participants (eTable 8 in the Supplement), (4) removing education from the SEM (eTable 9 in the Supplement), (5) removing late-life cognitive and social activities from the SEM (eTable 10 in the Supplement), and (6) assessing the association of individual factors included in CR with dementia risk (eTable 11 in the Supplement). Finally, we found that sex did not significantly affect the association of CR with dementia risk in all participants (HR, 1.05; 95% CI, 0.78-1.43; P = .74) or in autopsied participants (HR, 1.01; 95% CI 0.68-1.49; P = .97).
In this community-based prospective study of dementia-free older adults, we found that (1) high lifespan CR indicator accumulated through education, early-life cognitive activities, midlife cognitive activities, late-life cognitive activities, and social activities in late life was associated with a reduction in risk of dementia in a dose-dependent manner; (2) CR was not associated with most brain pathologies, and the association of CR with dementia remained significant after additional adjustment for brain pathologies; and (3) high CR could be associated with a reduction in dementia risk even in the presence of high AD burden and vascular pathologies. Neuropathological and neuroimaging studies have suggested that many people may tolerate considerable AD-related neuropathology without expressing the clinical syndrome.1 Indeed, about 25% of cognitively healthy older adults have increased levels of β-amyloid plaques in the brain.4 The concept of CR refers to the capacity to be resilient to age-related brain changes and the disease-related pathology in the brain without developing clinical dementia29 through enhanced brain network efficiency, capacity, or flexibility.13 Although a number of CR-related factors, including higher education attainment,2,30 complex occupation status,29 and rich cognitive and social activities,3,31,32 have been individually associated with a reduction in dementia risk, the association of each individual component with dementia could also be because of many alternative paths instead of a direct relation to the hypothesized CR. For example, lower education that is associated with dementia risk may also contribute to the deleterious effects of low socioeconomic status or cardiovascular disorders.33
In recent years, the use of CR indices has been suggested to evaluate the CR based on cumulative reserve factors,34,35 and the specific weight of each proxy indicator has been controversial. In the present study, to extract the CR score, we used SEM based on lifespan (ie, through early life, midlife, and late life) cognitive-enhancing activities and social activities in late life, and the weight of each CR factor was generated from SEM according to its contribution to the score, which was not equally weighted. We found that lifespan CR indicators in the middle and highest tertiles were associated with an approximately 23% to 39% reduction in risk of dementia. Furthermore, the association of CR with dementia was dose dependent, suggesting that accumulative educational and mentally stimulating activities throughout life are of great significance, given that there is currently no effective treatment for dementia.
So far, few studies presenting the association of the proxy of CR with dementia have taken brain pathologies into account. A 1999 study8 found that lower education was associated with the occurrence of cerebral infarcts. However, many other studies have failed to find a direct association of CR factors (such as education,11 cognitive activity,16 or cognitive lifestyle score36) with common dementia neuropathology. In the present study, we found that high lifespan CR indicator was not associated with most brain pathologies, except for gross infarcts, and baseline MCI status did not modify the association of brain pathology with CR. Further, high CR indicator was associated with a reduction in the risk of dementia independently of AD, vascular, and other brain pathologies. In addition, high lifespan CR indicator may be associated with a reduction in the risk of dementia even in the presence of high AD and vascular pathologies. These results were consistent with other studies11 and the CR theory7 that CR could reduce dementia risk and compensate for or cope with dementia pathology through other pathways rather than avoiding pathology directly.
This study has high rates of clinical evaluation and autopsy, which might minimize selective bias. Furthermore, the use of latent factors could capture the comprehensive effect of multiple CR factors across the lifespan. Nonetheless, some limitations need to be pointed out. First, the generalizability of the findings is limited because the study participants were volunteers. Second, as the brains were obtained at the end of the study, the causal inference of the neuropathologic basis in the association of CR with dementia must be further explored carefully. Third, CR-related factors were assessed by retrospective self-report, which could be subject to measurement error. However, use of a SEM-based latent variable approach allowed for the correction of unreliability in these factors. Fourth, as brain changes might occur nearly 15 years before clinical diagnosis of dementia, reverse causality between dementia and exposure studied at baseline could have occurred. We excluded those with incident dementia during the first follow-up, and the observed association remained significant. Fifth, nonresponse bias might have occurred owing to missing data. However, we repeated the analysis by multiple imputation for missing values, and the main results were not altered much.
This study provides evidence that high lifespan CR indicator, encompassing education, early-life, midlife, and late-life cognitive activities, and social activities in late life, is associated with a reduction in dementia risk, even in people with high AD and vascular pathologies. Our findings suggest that accumulative educational and mentally stimulating activities enhancing CR throughout life might be a feasible strategy to prevent dementia, even in people with high AD or vascular pathologies. Further large population-based longitudinal studies are warranted to establish the strategies of engagement in CR-related activities for the prevention of dementia.
Accepted for Publication: June 6, 2019.
Corresponding Author: Xiuying Qi, PhD (firstname.lastname@example.org), and Weili Xu, MD, PhD (email@example.com), Department of Epidemiology and Biostatistics, School of Public Health, Tianjin Medical University, 22 Qixiangtai Rd, Heping Qu, Tianjin 300070, China.
Published Online: July 14, 2019. doi:10.1001/jamaneurol.2019.2455
Author Contributions: Ms H. Xu and Dr W. Xu had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Mss H. Xu and Yang contributed as co–first authors. Drs Bennett and W. Xu contributed equally as co–senior authors.
Study concept and design: W. Xu.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: H. Xu, Yang.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: H. Xu, Yang.
Obtained funding: Bennett, W. Xu.
Administrative, technical, or material support: H. Xu, Yang, Qi, Song, Bennett, W. Xu.
Study supervision: Qi, Bennett, W. Xu.
Conflict of Interest Disclosures: None reported.
Funding/Support: Dr Bennett received grants R01AG17917 and UH2NS100599 from the National Institutes of Health. Dr W. Xu received grant 2017-00981 from the Swedish Research Council, grant 81771519 from the National Natural Science Foundation of China, Demensfonden, grants from the Konung Gustaf V:s och Drottning Victorias Frimurare Foundation, and grants from the Alzheimerfonden. This project is part of CoSTREAM (http://www.costream.eu) and received funding from the European Union’s Horizon 2020 Research and Innovation Programme under grant agreement No. 667375.
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Meeting Presentation: This paper was presented at the Alzheimer's Association International Conference; July 14, 2019; Los Angeles, California.
Create a personal account or sign in to: