Association of Lifelong Cognitive Reserve with Dementia and Mild Cognitive Impairment among Older Adults with Limited Formal Education: A Population-Based Cohort Study

Cognitive reserve (CR) refers to the individual adaptability of cognitive function in coping with brain aging or pathology [1]. The lifelong CR capacity can be derived from intellectual factors experienced over the life course, such as early-life educational attainment, adulthood marital status and occupational complexity, and late-life physical activity and social engagement [2]. Recently, a few population-based studies of urban populations from high-income countries, where people usually have high education, have consistently shown that higher CR capacity derived from lifelong cognitive enriching factors is related to lower incidence of dementia [3‒5] and mild cognitive impairment (MCI) [6]. However, population-based studies have rarely examined the associations of lifelong CR with dementia and MCI among rural residents with very limited education in low- or middle-income countries. Given that no or very limited formal education is associated with dementia [7] and that urban and rural residents have distinct lifestyle and CR profiles [8], it is important to investigate whether CR capacity may also benefit cognitive function even among rural older adults with no or very limited formal education. In addition, high educational attainment not only serves as a major contributor to CR [9] but also might affect other CR proxies over the lifespan (e.g., adulthood work complexity and late-life social activity) [10]. Thus, clarifying the potential role of lifelong composite CR capacity in late-life cognitive function among rural older adults with very limited education is highly relevant.

Furthermore, previous studies have suggested that individual CR proxies (e.g., education and social activity) appear to be differentially associated with subtypes of dementia and MCI [11‒13]. However, it is unclear whether lifelong composite CR capacity is preferably associated with certain subtypes of dementia and MCI. This is important because preferable associations of CR with specific types of dementia or MCI might shed light on the potential neuropathological mechanisms of CR, and thus inform the development of intervention strategies. In addition, given that the potential associations of CR proxies with the risk of dementia and MCI might vary by demographic features and genetic susceptibility (i.e., APOE genotypes) [14‒16]. It is relevant to clarify whether lifelong CR capacity could interact with certain demographic factors and APOE genotypes to affect the risk of dementia and MCI.

Therefore, in this population-based cohort study, we aimed to investigate the longitudinal associations of lifelong CR capacity with incidence of dementia and MCI, especially with regard to subtypes of dementia and MCI, among rural older adults with no or very limited formal education. We hypothesized that high lifelong CR capacity was associated with reduced risks of late-life dementia and MCI even among rural older adults with no or very limited formal school education and that the associations might vary by different subtypes of dementia and MCI.

This was a population-based cohort study. Data were derived from the Shandong Yanggu Study of Aging and Dementia (SYS-AD), which targeted people aged 60 years and above in the rural area of Yanlou Town, Yanggu County, Western Shandong Province, China [17]. In August 2014-September 2015, a total of 3,277 participants took part in the baseline examinations; of them, 278 participants were excluded due to prevalent dementia (n = 206) or insufficient information for the diagnosis of dementia (n = 72). The follow-up examination was performed in March-September 2018, as a part of the Multimodal Interventions to Delay Dementia and Disability in Rural China (MIND-China, registration no.: ChiCTR1800017758) project [18]. Out of the 2,999 dementia-free participants identified at baseline, 872 were excluded due to missing measures of CR (n = 167), death (n = 229), loss to follow-up (n = 528), or insufficient information for the diagnosis of dementia at follow-up (n = 14). Thus, a total of 2,127 participants (70.9% of all eligible participants) who were free of dementia at baseline completed the follow-up examination, which consisted of the sample for analyzing the association of lifelong CR with risk of dementia (analytical sample 1). In addition, 1,635 participants who were free of MCI at baseline consisted of the sample for analyzing the association of lifelong CR with risk of MCI (analytical sample 2). Online supplementary Figure 1 (for all online suppl. material, see https://doi.org/10.1159/000532131) shows the flowchart of study participants.

At baseline, we collected data through face-to-face interviews, clinical examinations, and laboratory tests. Data included social demographics (e.g., age, sex, education, occupation, and marital status), lifestyles (e.g., smoking, alcohol drinking, and physical and social activity), cardiometabolic health conditions (e.g., hypertension, diabetes, hyperlipidemia, coronary artery disease, and stroke), neuropsychological tests, and social support. We categorized early-life educational levels as no formal school education, elementary school (1–5 years), and middle school or above (≥6 years). We dichotomized adulthood occupation as farming versus non-farming, and marital status as married versus single, divorced, or widowed. Smoking and alcohol consumption were categorized into ever versus never smoking or drinking alcohol. We defined the frequency of late-life physical activity as at least weekly versus less than weekly. The late-life social activity was assessed by inquiring the participants about the question: “how often do you participate in the social activities?”. The options for answer included “never,” “occasionally,” “often,” and “frequently participate.” Based on participants’ answers, we dichotomized late-life social activity as never or occasional (combining options “never” and “occasionally”) versus frequent (combining options “often” and “frequently”). The late-life social support (e.g., living alone or with family members and support from family members, neighbors, or colleagues) was assessed using the Social Support Rating Scale that was validated among Chinese adult population and then was dichotomized into low (below the mean score) versus high social support, as previous reported [19].

Hypertension was defined as systolic pressure ≥140 mm Hg or diastolic pressure ≥90 mm Hg, current use of antihypertensive medication or self-reported history of hypertension [20]. Diabetes was defined according to self-reported history of diabetes diagnosed by a physician, fasting serum glucose ≥7.0 mmol/L, or current use of hypoglycemic medication [20]. Hyperlipidemia was defined as total cholesterol level ≥6.2 mmol/L, triglyceride ≥2.3 mmol/L, low-density lipoprotein cholesterol ≥4.1 mmol/L, high-density lipoprotein cholesterol <1.0 mmol/L, or having a self-reported history of hyperlipidemia [17]. We ascertained stroke and coronary heart disease according to self-reported history of the disease diagnosed by physicians during the annual health check-ups [17].

At follow-up, we collected data on neuropsychological status, clinical conditions, and genotyping [18, 21]. After an overnight fast, the trained staff collected venous blood samples and extracted and quantified genomic deoxyribonucleic acid [21]. The APOE genotype was determined using the TaqMan single-nucleotide polymorphism method and conducted by an operator who was blinded to all clinical data, as previously reported [22]. The distribution of APOE genotypes conformed to the Hardy-Weinberg equilibrium (p > 0.05).

At both baseline and follow-up examinations, dementia was diagnosed following the criteria outlined in the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, in which a 3-step diagnostic procedure was followed [17]. Briefly, in step 1, the trained medical staff performed the first face-to-face interview, clinical examination, and laboratory tests following a structured questionnaire. In step 2, the neurologists specialized in dementia care (L.C., T.H., S.T., X.H., and Q.Z.) reviewed all the records from step 1 to screen participants who were suspected to have dementia or who had insufficient information for the judgement of dementia. Finally, the neurologists conducted the second in-person or telephone interviews with participants, informants, or both who were selected in step 2, and the diagnosis was made based on all assessments.

Alzheimer’s disease (AD) was diagnosed according to the National Institute on Aging-Alzheimer’s Association criteria for probable Alzheimer’s dementia [23]. The diagnosis of vascular dementia (VaD) was made according to the criteria of the National Institute of Neurological Disorders and Stroke and Association Internationale pour la Recherche et l’Enseignement en Neurosciences for probable VaD [24], which was based on the self-reported, clinical, neuroimaging evidence of stroke, as well as the presence of a clear temporal relation between stroke and the onset of dementia. Dementia cases who could not be classified as either AD or VaD were considered to have other types of dementia.

Among participants who were free of dementia both at baseline and follow-up, MCI was clinically defined by the neurologists via reviewing all records from the interviews, clinical examination, and comprehensive assessments of sub-cognitive domains, following the Petersen’s criteria [21, 25]. At follow-up, participants with MCI were further categorized as having amnestic MCI (aMCI) if the memory domain was impaired or non-amnestic MCI (non-aMCI) if there was no impairment in the memory domain [21].

We compared the baseline characteristics of study participants between the two age groups (60–74 vs. ≥75 years), using the general linear regression model for continuous variables and the χ² test for categorical variables. We used the structural equation model to generate the lifelong composite CR score by taking into account early-life educational attainment, adulthood occupation and marital status, and late-life physical activity, social activity, and social support, among all participants with all individual CR proxies in the SYS-AD study (n = 3,060) (as shown in Fig. 1). We used the Cox proportional-hazards models to examine the association of lifelong composite CR score with incident dementia, AD, and VaD, in which the CR score was analyzed as both a continuous variable and a categorical variable (i.e., tertiles). Likewise, among participants who were free of MCI at baseline, we used the Cox models to examine the association of lifelong composite CR score with incident MCI, aMCI, and non-aMCI. Finally, we examined the statistical interaction of CR score with sex, age groups, and APOE genotypes on risks of dementia and MCI. If a statistical interaction was detected (p for interaction<0.05), we further performed the stratified analysis to assess the direction and extent of the interaction. We reported the results from statistical models that were adjusted for age (in years), ever smoking, ever alcohol consumption, hypertension, hyperlipidemia, diabetes, coronary heart disease, and stroke, and if applicable, for sex and APOE genotypes. We used Stata Statistical Software: Release16, for Windows (StataCorp LLC, College Station, TX, USA) for all the statistical analyses.

At baseline, the mean age of the 2,127 participants was 70.1 (standard deviation, 4.8) years, 59.4% were women, and 81.5% had no formal school education or only attended elementary school. Compared to participants aged 60–74 years, those aged 75 years and older were more likely to be farmers, and single, divorced, or widowed, to participate in physical activity, and to have a lower lifelong CR score, and less likely to have education, take part in social activity, get social support, drink alcohol, have diabetes or stroke, or carry the APOE ε4 allele (p < 0.05, Table 1). The two age groups did not differ significantly in the proportions of ever smoking, hypertension, hyperlipidemia, or coronary heart disease (p > 0.05).

During the total number of 8,330.6 person-years of follow-up (mean follow-up time per person: 3.7 years; standard deviation = 0.4), dementia was diagnosed in 101 persons, including 74 with AD, 26 with VaD, and 1 with other type of dementia. Each 1-point increment in the lifelong CR score at baseline was associated with an approximately 20% reduced risk of incident dementia and AD (p < 0.001), but not with incident VaD (Table 2). Compared to low tertile of CR score, the high tertile of CR score was associated with a reduced risk of incident dementia and AD (p for linear trend <0.001), but not with incident VaD (Table 2).

There was a statistical interaction between lifelong CR score and age groups on incident AD (p for interaction = 0.011). Stratified analysis by age groups suggested that each 1-point increment in composite CR score was significantly associated with a reduced risk of incident AD in people aged 60–74 years (multivariable-adjusted HR = 0.71; 95% CI: 0.59–0.86) but not in those aged 75 years and older (0.90; 0.74–1.10). Similarly, when categorizing the lifelong CR score into tertiles, compared to low CR tertile, the medium and high tertiles of CR score were related to a lower risk of AD only in people aged 60–74 years but not in those aged 75 years and older (Fig. 2). There was no statistical interaction of lifelong composite CR score with sex or APOE genotypes on incident dementia, AD, or VaD (all p for interactions >0.05).

Among participants who were free of MCI at baseline (n = 1,635), 331 were defined to have MCI at the follow-up, including 270 with aMCI and 61 with non-aMCI. Each 1-point increment in lifelong composite CR score was related to an approximately 15% lower risk of incident MCI and aMCI (p < 0.001) but not related to incident non-aMCI (Table 3). When lifelong CR score was treated as a categorical variable, compared to low CR tertile, the high tertile of CR score was related to a lower risk of incident aMCI (p for linear trend <0.05). Lifelong CR score was not associated with incident non-aMCI (Table 3). There was no statistical interaction of lifelong composite CR score with age, sex, or APOE genotypes on incidence of MCI, aMCI, or non-aMCI.

In this population-based cohort study among Chinese rural-dwelling older adults with no or very limited formal education, we found that high lifelong CR capacity was associated with reduced risks of late-life dementia and MCI, especially AD and aMCI, but not VaD or non-aMCI. In addition, the association between high lifelong CR capacity and the reduced risk of AD was evident only in people aged 60–74 years. These results underline the importance of high lifelong CR capacity in maintaining cognitive health into late life and in delaying cognitive dysfunction even among people with no or very limited educational attainment.

Most of the previous studies have investigated individual CR proxies (e.g., education, leisure activity, and social engagement) in middle age or late life in association with cognitive outcomes [26‒28]. By contrast, our study used the composited CR capacity derived from lifelong cognitively stimulating factors, which could represent the CR capacity accumulated over the life course. Furthermore, the association of high CR capacity with low risk of dementia has been well established among urban or highly educated populations [3‒5]. Our study extends these findings to the rural-dwelling older adults who have no or very limited formal school education. The clinicopathological data from the Brazilian Aging Brain Study showed that even a few years of formal education could contribute to CR, and thus, benefit late-life cognitive function [29]. This study supports the view that lifelong composite CR capacity was a relevant determinant of late-life cognitive phenotypes even among older adults with no or very limited education.

Our study further revealed that high lifelong CR capacity was associated with a low risk of Alzheimer type of dementia but not with VaD. Neuropathological and neuroimaging studies have suggested that high CR capacity could mitigate the detrimental effect of AD-related neurodegenerative pathology on cognitive performance and delay the onset of the dementia syndrome [30‒33], which is in line with our finding. On the other hand, data from neuroimaging-based studies appeared to show no evidence that higher CR capacity could modify the association of cerebrovascular injury (e.g., brain infarct and white matter hyperintensities) with poor cognitive function [34, 35]. Similarly, a clinic-based study among patients with subcortical vascular cognitive impairment suggested that high educational attainment might compensate for the detrimental effects of cortical thinning, but not cerebrovascular lesions (i.e., white matter hyperintensities and lacunes), on cognitive function [36]. Taken together, findings from the literature and our own study support the view that increased CR capacity could compensate mainly for the adverse cognitive consequences owing to neurodegenerative pathologies rather than cerebrovascular injury.

Previously, the Rush Memory and Aging Project has linked higher lifespan CR capacity with a lower risk of incident MCI [6]. Our study extends the previous findings by showing that higher CR capacity is associated with a lower risk of aMCI but not non-aMCI. This has important implications for understanding the neuropathological mechanisms of cognitive benefits of CR because the underlying neuropathology and evolution of aMCI and non-aMCI in older adults differ. Neuropathological studies have suggested that neurodegenerative lesions are the underlying pathology of aMCI [37], whereas cerebral macro- and microvascular lesions play a pivotal role in non-aMCI [38, 39]. Given that higher CR capacity could compensate mainly for the cognitive consequence of neurodegenerative lesions, but not cerebrovascular lesions [30, 34], it is plausible that higher CR capacity could be preferably associated with a reduced risk of aMCI rather than risk of non-aMCI.

Notably, we found that the association of high lifelong CR capacity with the reduced risk of AD existed mainly in people aged 60–64 years but not in those aged 75 years and older. This could be partly explained by the view that cognitive benefits of high CR capacity might be gradually diminished with the development and progression of brain pathology (e.g., amyloid and tau proteins) as people age; when the load of age-related brain pathology reaches the advanced stage or threshold for manifesting symptoms of cognitive impairment, the high CR capacity might no longer be able to tolerate or compensate for the accumulated brain pathology and delay cognitive decline and dementia [31, 40]. This could largely explain our finding that the association of higher CR capacity with the lower risk of AD is less pronounced in older than in younger old people. These findings may have implications for determining the target populations when designing the multimodal intervention programs that include CR components to delay cognitive decline and dementia.

Our population-based cohort study targeted the rural-dwelling older adults with no or very limited formal education in China, a sociodemographic group that has been substantially underrepresented in dementia research. In addition, the composite CR capacity was quantified from multiple cognitive-enhancing indicators experienced over the lifespan, and MCI was defined by integrating clinical assessments with standard neurocognitive tests. However, our study also had limitations. First, we had no biomarkers of brain pathology, which limited the potential to investigate the underlying neuropathological mechanisms of lifelong CR capacity in maintaining cognitive function. Furthermore, given the relatively short follow-up period, the long preclinical phase of dementia, and the nature of observational study, caution is needed when interpreting the observed associations between lifelong CR capacity and cognitive outcomes as causal relationship. Third, although a broad range of covariates were taken into account in our analysis, confounding bias effect might still play a part due to imperfect assessments of some covariates (e.g., self-reported lifestyle factors and health history). Finally, our single-center study targeted older adults with very limited education from only one rural area in western Shandong province, which should be kept in mind when generalizing our findings to other populations.

In conclusion, this population-based cohort study indicates that among rural-dwelling older adults with no or very limited formal education, high lifelong CR capacity is associated with low risks of late-life dementia and MCI, especially AD and aMCI, but not VaD and non-aMCI. In addition, the beneficial effects of lifelong CR capacity on reduced AD risk appeared to be evident mainly in young-old people (age 60–74 years). These findings could inform the development of intervention strategies to maintain late-life cognitive health and delay the onset of dementia among rural populations who have no or very limited formal education.

We would like to thank all study participants in the Shandong Yanggu Study of Aging and Dementia and the project of Multimodal Interventions to Delay Dementia and Disability in Rural China and all staff in Yanlou Town Hospital for their collaboration in the organization of field surveys, as well as our research group at Shandong Provincial Hospital for their collaboration in data collection and management.

The protocols of both SYS-AD and MIND-China projects were reviewed and approved by the Ethics Committee on Human Experimentation at Shandong Provincial Hospital in Jinan, Shandong (Approval Number: 2018-014, date: 2018-01-12). The MIND-China study was registered in the Chinese Clinical Trial Registry (Registration Number.: ChiCTR1800017758). Patient consents were not required as this study was based on publicly available data.

All authors declare no financial or other conflicts of interest.

This work was supported by the Science and Technology Program for Public Wellbeing of Shandong Province, China (Grant No. 2013kjhm180405), the National Key Research & Development Program of China (Grant No. 2017YFC1310100), the National Nature Science Foundation of China (Grant No. 82171175 and 82011530139), the Academic Promotion Program of Shandong First Medical University (Grant No. 2019QL020), and the Taishan Scholar Program of Shandong Province, China. Li Y received a scholarship from the China Scholarship Council (Grant No. 201906220042). Qiu C received grants from the Swedish Research Council (Grant No. 2017‐00740, 2017‐05819, and 2,020‐01574), the Swedish Foundation for International Cooperation in Research and Higher Education (Grant No. CH2019‐8320) for the Joint China‐Sweden Mobility program, and Karolinska Institutet, Stockholm, Sweden. The funding agency had no role in the study design, data collection, data analyses, manuscript drafting, revisions, or in the decision to submit the work for publication.

Y. Du and C. Qiu supervised the study and contributed to manuscript drafting. Y. Li, Y. Du, and C. Qiu planed the study. Y. Li, Y. Ren, L Cong, T Hou, L Song, M Wang, X Wang, X Han, S Tang, Q Zhang, and Y. Wang contributed to data collection and assessment. Y. Li, Y. Ren, and S. Dekhtyar contributed to data analysis. All authors contributed to critical revisions of the manuscript and approved the final version of the manuscript.

Data on which this study is based are derived from the population-based SYS-AD and MIND-China projects. Data are not publicly available due to ethical reasons. Further inquiries can be directed to the corresponding author.