Education as Proxy for Cognitive Reserve in a Large Elderly Memory Clinic: ‘Window of Benefit’

Abstract

Background:

The role of cognitive reserve (CR) to explain individual differences in cognitive functioning is unclear in memory clinic patients.

Objective:

To examine the cross-sectional effect of CR on cognition in relation to levels of neurodegeneration in a large elderly single-center memory clinic population.

Methods:

We included patients with subjective cognitive impairment (SCI, n = 481), mild cognitive impairment (MCI, n = 628) or Alzheimer’s disease (AD, n = 1,099). Education was used as proxy for CR and visually rated medial temporal lobe atrophy (MTA) on CT was used as parameter of neurodegeneration. Relations between CR, cognition, and MTA were analyzed with multiple linear regression adjusted for age, sex, and cerebral atrophy. In addition, we examined if education affects the relation between MTA and cognition using an interaction variable.

Results:

Education was significantly related to all measures of cognition including subtests with an explained variance of education as a determinant of cognition of 11%. More highly educated patients had more advanced levels of MTA at the same level of cognition. All these results were stronger or only present in demented compared to non-demented patients but appeared no longer significant in those with lowest overall cognition. The interaction effect was significant indicating that with more advanced MTA, less cognitive decline was shown in higher educated patients.

Conclusion:

Education is a very strong determinant of cognition in an elderly memory clinic population. The positive effect of education was stronger in demented than in non-demented patients but disappeared in those with the lowest cognitive scores indicating a “window of CR benefit”.

Keywords

Cognitive reserve dementia education medial temporal lobe atrophy memory clinic

INTRODUCTION

The concept of cognitive reserve (CR) is extensively investigated recently and is coined as a possible protective factor in cognitive decline [1 –3]. Theconcept is a complex theoretical construct to describe individual differences in susceptibility to cognitive, functional or clinical decline due to aging or brain disease [2]. The CR framework postulates that lifetime experiences in combination or interaction with genetic factors enable cognitive processes to cope better with brain disease or aging [3]. Although not interchangeable, level of education is widely used as a proxy for CR [3 –5]. Population-based studies demonstrated an association between higher level of education and lower risk of dementia [6 –8]. Groot et al. [4] showed that not only CR but also brain reserve have independent effects on cognitive function, with stronger effects in pre-dementia than in dementia patients in a sample of Alzheimer’s disease (AD) biomarker-positive participants. Another study using similar patient settings also demonstrated the beneficial effect of CR in the memory clinic [9]. Earlier, some small studies using position emission tomography showed effects of education on the impact of AD biomarkers, such as amyloid burden, hypometabolism, and tau pathology, on cognitive function [10 –12]. Generalizing these findings to memory clinics with patients at higher ages, for example in those over 80 years is problematic since no data are available in elderly cohorts. Furthermore, most studies have a relatively low number of included patients or selected study populations such as tertiary referral centers.

The present study examines the effect of education as proxy of CR in a large cohort of elderly in the spectrum from subjective cognitive decline to those with lowest cognitive scores in more severe dementia, analyzing how the relation between MTA and cognition is modulated by education.

METHODS

Subjects

Patients included were referred with cognitive complaints to the memory clinic at Tergooi Hospital, a general hospital in Blaricum, The Netherlands, between 2009 and 2018. Each patient received a standardized work-up including clinical examination by both neurologist and geriatrician. Based on the standardized diagnostic protocol, patients were diagnosed with mild cognitive impairment (MCI), AD, another type of dementia, psychiatric or other diseases causing cognitive disorders, or if all test were normal, complaints were considered as subjective (SCI). In this study, all consecutive patients with SCI (n = 481), MCI (n = 628), and AD (n = 1099) were included; patients diagnosed with other conditions were not taken along in our analyses (n = 478).

Standard protocol approvals, registrations, and patient consents

The study protocol was approved by the local Medical Ethics Committee.

Clinical diagnostic procedures

All patients underwent an extensive standardized diagnostic evaluation as described elsewhere [13], including complete medical and neurological examination by a neurologist and geriatrician, informant-based history, vital function assessment, laboratory testing, electrocardiogram, neuropsychological examination including the Cambridge Cognitive Examination (CAMCOG) and Visual Association Test, Geriatric Depression Scale, assessment of the Instrumental Activities of Daily Living scale and a CT-scan of the brain. Clinical diagnosis of MCI or AD was established by consensus in a multidisciplinary team according to the current criteria [14, 15]. A diagnosis of SCI was established in the absence of objective cognitive and psychiatric deficits.

Cognitive reserve

As a proxy for CR we used education defined as the level of education by the Verhage system. The Verhage system is a standardized index ranging from 1 to 7 used in the Netherlands, wherein ‘1’ means that primary school was not completed and ‘7’ means that an academic degree was obtained [16].

Cognition

Cognitive screening was performed using the CAMCOG part of the CAMDEX [17]. The CAMCOG total score includes a memory subscore containing orientation and memory, and a CAMCOG non-memory score containing language, attention, praxis, maths, abstraction, and perception. The Mini-Mental State Examination (MMSE) is obtained as part of the CAMCOG [18]. Cognition was divided into four cognitive domains (language, memory, attention, and visuospatial ability) and two subtests (MMSE, Clock). The variable visuospatial ability was computed with the two CAMCOG sub-scores: perception and praxis constructive.

In this study, we explored the CAMCOG total score, MMSE total score, and looked into the cognitive domains language, memory, attention, and visuospatial ability.

Computed tomography protocol

All subjects underwent a 64-detector row CT scan of the brain using a Siemens Somatom definition AS 64 slice scanner according to a CT brain protocol for the memory clinic (260 mAs, 120 kV, 64 * 0.6 mm collimation, pitch of 0.55, WC/WW = 40/80, CARE kV = on, dose optimization slider on non-contrast). Oblique coronal, sagittal, and transversal reconstructions were made with bone window 1.5 mm slices, axial slices of 5.0 mm, and oblique coronal slices of 3.0 mm.

All CT-scans were reviewed by a radiologist and all scans were visually assessed for atrophy and vascular scores in a consensus meeting by a neurologist and geriatrician.

The level of atrophy of the medial temporal lobe was assessed using the visual MTA-score by Scheltens (range 0–4; 0 = no medial temporal atrophy, 4 = maximal medial temporal atrophy) [19]. Presence of global cortical atrophy (GCA) was determined using the Pasquier scale (range 0–3, 0 = normal volume, 3 = knife blade atrophy and severe ventricular enlargement [20].

Statistical analyses

We used SPSS version 24.0. Baseline characteristics comparing SCI, MCI, and AD were analyzed with one-way ANOVA or with Chi-square tests when appropriate. A one-way ANOVA with a post hoc Tukey’s test was used to compare the variable means between SCI, MCI, and AD.

First, we examined the relationship between education as independent variable and measures of global cognition and subtests as dependent variables using multiple linear regression analysis with adjustment for age, sex, MTA, and GCA for combined groups SCI, MCI, and AD and for groups dementia (AD) versus non-dementia (SCI and MCI) separately. Explained variances (R²) were calculated to examine the magnitude of education as a determinant of cognition. Model 1 describes the relation between age, sex, GCA, MTA and cognition, whereas the R² change attributable to education is calculated in model 2 by adding education as a predictor of cognition.

Second, we test the hypothesis that for the same level of cognition, more highly educated patients have more advanced levels of MTA. This was performed by multiple linear regression, examining the relation between the independent variable education, and MTA as a dependent variable adjusted for age, sex, and global atrophy. To hold the level of cognition as a constant factor, we additionally adjusted for the total CAMCOG score.

In a separate stratification analysis, we calculated these regression coefficients (thus not adjusted for the CAMCOG total score in this analysis) for different strata of CAMCOG scores (CAMCOG total score into five groups: 0–50, 51–60, 61–70, 71–80, 81–90, 91–110).

Finally, to examine if education affects the relationship between MTA and cognition, we added an interaction variable between education and MTA in a General Linear Model analysis.

RESULTS

Sample characteristics

Demographic and clinical characteristics of the total population are described in Table 1. The total population had a mean age of 79 years, MCI patients were older than SCI patients (p < 0.01), while patients with AD were older than SCI and MCI patients (p < 0.01). Looking at levels of neurodegeneration, MTA and GCA scores were higher in MCI compared to SCI patients (p < 0.01), and higher in AD compared to SCI and MCI patients (p < 0.01). Comparing levels of education, SCI patients had a higher education compared to MCI and AD patients (p < 0.05) and MCI patients had higher education compared to AD patients (p < 0.05). In the total population, there were n = 562 patients with low education (Verhage score 1–3), n = 1,094 with middle level education (Verhage score 4 and 5), and n = 526 with high education (Verhage score 6 and 7).

Table 1

Demographic and clinical characteristics of the total sample according to diagnosis

	Total (n = 2208)	SCI (n = 481)	MCI (n = 628)	AD (n = 1099)
Demographics
Sex, n (%) female	1341 (61%)	253 (53%)	384 (61%)	705 (64%)
Age	78.5 (8.6)	71.9 (10.0)	79.0 (7.2)^a	81.1 (7.1)^b
Education, median (range)	5 (1–7)	5 (1–7)	5 (1–7)	4 (1–7)^c
Neurodegeneration
MTA	1.7 (1.0)	0.8 (0.8)	1.5 (0.9)^a	2.2 (0.8)^b
GCA	1.4 (.7)	1.0 (0.7)	1.3 (0.7)^a	1.7 (0.8)^b
Cognition
MMSE	21.9 (5.5)	27.2 (2.34)	23.8 (3.5)^a	18.4 (5.0)^b
Clock	2.1 (.9)	2.7 (0.56)	2.4 (0.8)^a	1.7 (0.9)^b
CAMCOG total	73.5 (16.5)	89.8 (7.2)	79.2 (9.5)^a	63.1 (15.2)^b
Memory	15.2 (5.6)	21.1 (2.7)	16.9 (3.8)^a	11.7 (4.9)^b
Attention	4.8 (2.2)	6.2 (1.2)	5.5 (1.7)^a	3.9 (2.2)^b
Language	24.3 (3.8)	26.9 (2.1)	25.3 (2.5)^a	22.6 (4.1)^b
Visuospatial ability	9.9 (3.1)	12.3 (2.0)	10.6 (2.5)^a	8.5 (3.0)^b

Data presented as mean (SD) unless otherwise specified. Group comparisons were performed with independent sample t-test, or Chi square. SCI, subjective cognitive impairment; MCI, mild cognitive impairment; AD, Alzheimer’s disease; MTA, medial temporal lobe atrophy; GCA, global cortical atrophy; MMSE, Mini-Mental State Examination. ^aMCI > SCI p≤0.01; ^bAD > SCI and MCI p≤0.01; ^cAD < SCI and MCI p≤0.01.

Table 2

Effect of education on cognition according to diagnosis

Domain	SCI (N = 481)	MCI (N = 628)	AD (N = 1099)	p
	Non-Dementia		Dementia
MMSE	0.63^a	0.68^a	1.09	0.003^b
Clock	0.12^a	0.08^a	0.15^a	0.074
CAMCOG total	2.69^a	2.63^a	3.95^a	0.001^a
Memory	0.60^a	0.58^a	0.96^a	0.002^b
Attention	0.22^a	0.23^a	0.38^a	0.021^b
Language	0.67^a	0.67^a	0.86^a	0.086
Visuospatial ability	0.46^a	0.40^a	0.61^a	0.096

Multiple linear regression analysis with education as the independent variable and cognition as the dependent variable, adjusted for age, sex, GCA, and MTA. Values depicted are regression coefficients (B). SCI, subjective cognitive impairment; MCI, mild cognitive impairment; AD, Alzheimer’s disease; MMSE, Mini-Mental State Examination. p-value: Comparison of the differences of regression coefficients between non-dementia and dementia stages. ^ap≤0.01; ^bp≤0.05.

Effect of education on cognition in different stages of cognitive decline

First, we analyzed the effect of education on different cognitive domains and tests in the total cohort. There was a positive statistically significant effect of education on cognition for combined groups of SCI, MCI, and AD for all tests including the different domains (education in relation to MMSE: regression coefficient (B) = 1.05, p < 0.001; clock drawing B = 0.14, p < 0.001; CAMCOG total score B = 3.83, p < 0.001; memory B = 0.97, p < 0.001; attention B = 0.36, p < 0.001; language B = 0.85, p < 0.001; visuospatial ability B = 0.58, p < 0.001). Regression coefficients for diagnosis subtypes are shown in Table 2. To examine the difference between patients according to their disease stage, we used the General Linear Modal to determine the differences between non-dementia and dementia. The patients with dementia had a significant higher regression coefficient than the non-dementia patients on most of the cognitive measurements, indicating that the effect of education is stronger in the dementia stage. This is further illustrated in Figs. 1 and 2. In addition, we examined the contribution of education as a predictor of cognition including global cognition (CAMCOG and MMSE) as well as respective cognitive subtests, using the R² change from model 1 to model 2. These results show that education is a strong determinant for all the cognitive tests (p < 0.001), with a high level of 11% explained variance for the total CAMCOG score (MMSE 8%, clock 5%, memory 6%, attention 6%, language 11%, visual ability 8%).

Fig.1

Effect size of education on cognition according to disease stage. Effect sizes are regression coefficients, adjusted for age, sex, MTA, and GCA; error bars indicate the standard error. *Significant effect at p < 0.05.

Fig.2

CAMCOG total effect size. Effect sizes are regression coefficients, adjusted for ae, sex, MTA, and GCA; error bars indicate the standard error. *Significant effect at p < 0.05.

To summarize, education is related to cognition, while regression coefficients tested statistically significantly higher in AD than SCI/MCI.

Effect CR on MTA

Second, multiple linear regression was used to test the hypothesis that for the same level of cognition, more highly educated patients have more advanced levels of MTA. The analysis revealed a significant positive effect of education on MTA at the same level of cognition for combined groups of SCI, MCI and AD (B = 0.07, p < 0.001), adjusted for age, sex, and global cortical atrophy. This indicates that highly educated patients have more advanced levels of MTA at similar cognition. Looking at the different subdiagnoses, this effect was only significant in AD patients (B = 0.06, p < 0.001) but not in SCI (B = 0.04, p = 0.14) and MCI (B = 0.01, p = 0.60). This further adds to differences according to disease stage (demented versus non-demented patients) examined in the previous paragraph with a different statistical analysis.

Next, we examined the effect of education on MTA calculating these regression coefficients for different strata of CAMCOG scores (0–50, 51–60, 61–70, 71–80, 81–90, 91–110) testing the hypothesis that after a certain decrease of cognitive function the protective value fades out. In the CAMCOG strata under 90 and above 50, education was significantly related to MTA, but this relation was not significant above 90 and below 50 (Table 3). These results suggest that education does not operate as a reserve at the both ends of the spectrum of the CAMCOG scores. For the low side of this spectrum, these data indeed indicate that at a certain level of cognitive performance, i.e., a CAMCOG total score of 50, education loses its compensatory mechanism (comparable MMSE score of 12). Analysis of the regression coefficients of the covariates of education in relation to MTA (age, sex, and GCA) in the different strata shows a consistent pattern of non-significance of sex and high significance of GCA (all p < 0.001). In addition, the regression coefficients of age in the 6 respective strata from the highest (CAMCOG 91–110) to the lowest (CAMCOG < 50) were: 0.03±0.01, p < 0.001; 0.04±0.01, p < 0.001; 0.03±0.01, p < 0.001; 0.03±0.01, p < 0.001; 0.02±0.01, p < 0.01; 0.01±0.01, p = 0.54. The stratum of those with CAMCOG 91–110 included 248 SCI, 47 MCI, and 16 AD patients, with a mean MMSE score of 28.2 and mean MTA of 0.8. Those with CAMCOG 50 or lower included 0 SCI, 9 MCI, and 193 AD patients with a mean MMSE of 11.6 and mean MTA of 2.5. The dynamic range of MTA scores (lowest mean of right and left score 0, highest 4) in the different CAMCOG strata was 0 to 4 in the strata 0–50, 51–60, 61–70, 71–80, 81–90, and 0 to 3.5 in the stratum 91–110. Thus, there were no patients with an MTA score of 4 in those with a CAMCOG score from 91 to 110. An MTA score of 4 is a very infrequent finding in our study with patient numbers of 9, 7, 8, 7, and 4 in the respective strata of 0–50, 51–60, 61–70, 71–80, and 81–90.

Table 3

Effect of education on medial temporal lobe atrophy at different CAMCOG total levels

CAMCOG total	B	SE	p	N
score
0–50	0.04	0.04	0.33	202
51–60	0.09	0.04	0.011	238
61–70	0.08	0.03	0.003	355
71–80	0.11	0.02	0.000	522
81–90	0.06	0.03	0.017	831
91–110	–0.02	0.04	0.58	311

Multiple linear regression analysis with education as the independent variable and MTA (medial temporal lobe atrophy) as the dependent variable, adjusted for age, sex, GCA (global cortical atrophy), stratified according to CAMCOG levels. Values depicted are regression coefficients (B).

Interaction effect between education and MTA

In the final analysis, a statistically significant interaction effect was demonstrated between MTA and education (p < 0.01) in relation to cognition. This shows that the relationship between MTA and cognition is modified by education (Table 4). At a higher level of education, the regression line was less steep, indicating that the relationship between MTA and cognition is stronger in lower educated patients and less strong in higher educated patients (Fig. 3).

Table 4

Multiple linear regression examining the relation of MTA and education on cognitive performance

	Model 1		Model 2
Model terms	B (SE)	p	B (SE)	p
Constant	86.67 (2.99)	0.00	94.84 (2.98)	0.00
MTA	–6.38 (0.32)	0.00	–6.42 (0.32)	0.00
Education	3.85 (0.19)	0.00	3.75 (0.19)	0.00
Age	–0.22 (0.04)	0.00	–0.24 (0.04)	0.00
Sex	–3.13 (0.19)	0.00	–3.11 (0.59)	0.00
Education x MTA	NA	NA	0.71 (0.19)	0.00

Model 1 includes separate terms for MTA (medial temporal lobe atrophy) and education to assess the independent contribution on cognition. Model 2 added an interaction term to determine if the relation between MTA and cognition is modified by education. Both models are adjusted for age and sex. The B-value is the degree the predictor affects the CAMCOG-total score if all the other predictors are held constant and the SE-value indicates the standard error in the regression analysis.

Fig.3

Interaction effect between cognition and MTA according to education level. Graphic showing the relation between MTA (medial temporal lobe atrophy) and cognition (CAMCOG total score) according to different education levels (Verhage score, 1 lowest level, 7 highest level. At a higher level of education, the regression line is less steep, indicating that the relationship between MTA and cognition is stronger in lower educated patients and less strong in higher educated patients.

DISCUSSION

In the present study we investigated the association of education as proxy for CR in relation to cognitive function and cerebral atrophy in a large elderly memory clinic population. Education was strongly related to cognition and the effect was larger in dementia patients compared to non-dementia patients. Furthermore, for the same level of cognition, highly educated patients had more advanced MTA compared to low educated individuals. The effect was only found in AD and not in SCI and MCI patients and remarkably the effect was no longer statistically significant in patients with very low levels of cognition and also absent in those with highest cognitive function. In addition, education modified the relation between MTA and cognition, showing less cognitive decline in higher educated patients with more MTA.

The main finding of education as strong determinant of cognitive function is consistent with most previous cross-sectional and longitudinal studies [3]. Generalization to clinical practice, however, is problematic since few studies specifically address this issue in memory clinic patients [4, 5]. There is one recent memory clinic study showing a comparable positive effect of CR on cognitive symptoms in AD, but these were selected AD biomarker-positive patients and the cohort is younger than our patients (mean age 66.2 versus 78.5 years) [4]. Our study therefore extends the concept of CR to unselected memory clinic patients at a much higher age. In addition, we show that there is a large contribution of CR to overall cognition in this setting in terms of a high explained variance. The confirmation of the CR hypothesis in our cohort is further based on strong statistical findings such as more atrophy of the medial temporal lobe in those with high CR for same level of cognition and the finding of interaction, also in line with earlier findings [9 , 22].

One of the most interesting findings in our study is that the CR effect was very strong in the dementia stage (AD) but not significant in the predementia stages (SCI and MCI). In addition, stratified analysis showed that the effect of education was absent in those with highest cognitive levels, is strongly significant in the large middle category of cognitive levels and that the effect disappears in those with lowest cognitive scores. This suggests that CR depends truly on the disease state and that there is a “window of CR benefit”. The graphic feature (Fig. 4a) below shows our hypothesis that the protective value of CR can be visualized as an inverted U-shaped curve.

Fig.4

Hypothesis of the protective value of CR according to presence of pathology. Image of the protective value of CR as an inverted U-shaped curve, representing that individuals with more CR can cope with more AD pathology to a certain extent but that this value is lost as disease progresses. a) general protective effect; b) protective effect in memory clinic at older age (blue line) and in amyloid positive patients (orange line).

Thus, individuals with more CR can cope with more AD pathology to a certain extent and this value is lost as disease progresses. This could explain different protective effects in different disease stadia, such as the protective effect Groot et al. found in SCI amyloid-positive patients (pathology already quite present) and no effect in dementia (already going down in the curve) [4]. At first glance, Groot et al. found opposite results compared to ours, namely a stronger effect of CR in the pre-dementia stage and no effect in dementia. However, including only amyloid positive patients, we hypothesize a comparable shape of curve of the protective effect in their study, but only more shifted to the right. In other words, we hypothesize that pathology load in amyloid selected patients is higher than our memory clinic at older age, what puts the protective effect of CR at a different level (Fig. 4b). Mungas et al. also suggested that the beneficial effects of CR are less effective as levels of neurodegeneration increase [5].

Interestingly, our results very well fit the proposed hypothetical compensation model by Gregory et al. [23]. They suggest that CR effectiveness increases and starts in reaction to increasing pathology and brain deterioration, then reaches a plateau phase and finally decreases in response to relentless disease effects [23]. A recent longitudinal clinical pathologic study investigated the very other end of the spectrum in non-diseased state since individuals had no dementia at baseline and the outcome was incident dementia [24]. The authors state that they could not confirm the CR hypothesis since education was not associated with slower rate of cognitive decline or later onset of cognitive decline, but the question is whether the CR hypothesis can be excluded by this type of study. Timing of observation is a crucial element in longitudinal trajectory studies with long follow-up necessary to be sensitive for cognitive decline in relation to CR [3]. If these data would still be compatible with our hypothetical model of a CR activity window, it is conceivable that pathology load was too low to activate the CR compensation mechanism in this study. Although in a different study population, the strongest confirmation of our findings was reported very recently by van Loenhoud et al. demonstrating that the protective effect of CR is dependent on disease stage. They show an attenuated clinical progression in predementia stages of AD, but accelerated cognitive decline after dementia onset in those with greater CR [25]. In addition to CR, another concept of reserve describing the capacity to preserve cognitive function is brain reserve, typically operationalized by intracranial volume in human neuroimaging studies. These data are not available in our study but an interrelation between CR and brain reserve was shown for example by Groot et al. (significant correlation of r = 0.17, p < 0.01) [4].

Strengths of our study consist of the large and unselected memory clinic sample (n = 2208) compared to other studies [4 , 26]. Furthermore, our patients represent an age group at risk of developing dementia and is much older than most other research populations (mean age of our population 79 years compared to 57–74 years [4 , 26]). This gives new information about the age-related spectrum across which CR serves as a protective factor providing external validity for generalization and practice in other general memory clinics. Some limitations of the current study have to be addressed. The main limitation is the cross-sectional design of our study, which does not allow interference about the direction of causality. Furthermore, we did not have availability of an amyloid biomarker in our cohort, what might have resulted in inclusion of other dementia pathologies. We examined explicit cognitive reserve and not brain reserve, which refers to the structural characteristics of the brain at a given point in time [3, 4]. Earlier studies found that cognitive reserve and brain reserve are uncorrelated what legitimates our focus solitary on CR [3]. In the literature, different proxies are used for CR; in line with most studies in dementia, we used education because of the robust association [3]. Using the Verhage score for education, this accounts mainly for education early in life not taking along lifetime experiences or lifestyle, although individuals of wealthier families are more likely to obtain higher levels of education and thereby, socioeconomic characters can influence CR [27]. MTA was assessed by visual rating which has proven to be reliable and replicable but volumetric MTA measurements might improve accuracy. CT-imaging was used instead of MRI which might have influenced our results. However, visual assessment on CT may lead to MTA underestimation as compared to MRI and this could possible lead to underestimation of relation between education, cognition, and MTA. If this effect plays a role, results in our study are not affected and would only have been more pronounced. Further, the absence of an effect of education on MTA at the upper and lower limits of the CAMCOG distribution could theoretically be explained by lack of dynamic range of the MTA scores. However, the MTA range is from the lowest score (0) to the highest score (4) in all strata, except for 0 to 3.5 in the highest CAMCOG stratum of 91–110. MTA scores of 4 are very infrequent in all strata (see results) and it is thus unlikely that lack of MTA range influenced our results. In the stratum of those with CAMCOG lower than 50, regression coefficients of GCA and sex were similar as other strata but age was not significant. This could not be explained by lack of range in MTA (0 to 4) or by lack of range in age (60–96 years) but may be related to the fact that this group represents a rather homogeneous group of severe AD patients.

The mechanisms by which CR operates still remain poorly understood. One hypothesis postulates better functional networks to perform same tasks in high CR individuals. Research focusing on network analysis such as MEG investigations could be of special interest in future, along with longitudinal studies helping us to understand the processes and impact of CR on disease progression over time.

In conclusion, our results demonstrate that CR plays a major role as determinant of cognitive function in elderly memory clinic patients. We suggest a “window of CR benefit” to compensate cognitive decline: some level of neuropathology needs to be present but in the more severe AD disease state, CR cannot compensate anymore. Our study may give impetus to future intervention studies examining the possible protective effects of cognitive stimulation and environmental enrichment to enhance neuronal plasticity at higher age.

DISCLOSURE STATEMENT

Authors’ disclosures available online (https://www.j-alz.com/manuscript-disclosures/19-1332r2).

References

Stern

(2012) Cognitive reserve in ageing and Alzheimer’s disease. Lancet Neurol 11, 1006–1012.

Pettigrew

, Soldan

(2019) Defining cognitive reserve and implications for cognitive aging. Curr Neurol Neurosci Rep 19, 1.

, Yang

, Qi

, Dintica

, Song

, Bennett

, Xu

(2019) Association of lifespan cognitive reserve indicator with dementia risk in the presence of brain pathologies. JAMA Neurol 76, 1184–1191.

Groot

, van

Loenhoud A

, Barkhof

, van

Berckel BNM

, Koene

, Teunissen

, Scheltens

, van

der Flier WM

, Ossenkoppele

(2018) Differential effects of cognitive reserve and brain reserve on cognition in Alzheimer disease. Neurology 90, 149–156.

Mungas

, Gavett

, Fletcher

, Farias

, DeCarli

, Reed

(2018) Education amplifies brain atrophy effect on cognitive decline: Implications for cognitive reserve. Neurobiol Aging 68, 142–150.

Stern

, Gurland

, Tatemichi

, Tang

, Wilder

, Mayeux

(1994) Influence of education and occupation on de incidence of Alzheimer’s disease. JAMA 271, 1004–1010.

Caamano-Isoma

, Corral

, Montes-Martinez

, Takkouche

(2006) Education and dementia: A meta-analytic study. Neuroepidemiology 26, 226–232.

Meng

, D’Arcy

(2012) Education and dementia in the context of the cognitive reserve hypothesis: A systematic review with meta-analyses and qualitative analyses. PLoS One 7, e38268.

Perneczky

, Wagenpfeil

, Lunetta

, Cupples

, Green

, Decarli

, Farrer

, Kurz

A; MIRAGE Study Group

(2009) Education attenuates the effect of medial temporal lobe atrophy on cognitive function in Alzheimer’s disease: The MIRAGE study. J Alzheimers Dis 17, 855–861.

10.

Hoenig

, Bischof

, Hammes

, Faber

, Fliessbach

, Eimeren

, Drzezga (2017) Tau pathology and cognitive reserve in Alzheimer’s disease. Neurobiol Aging 57, 1–7.

11.

Ewers

, Insel

, Stern

, Weiner

MW; The Alzheimer’s Disease Neuroimaging Initiative (ADNI)

(2013) Cognitive reserve associated with FDG-PET in preclinical Alzheimer disease. Neurology 80, 1194–1201.

12.

Morbelli

, Perneczky

, Drzezga

, Frisoni

, Caroli

, van

Berckel BN

, Ossenkoppele

, Guedj

, Didic

, Brugnolo

(2013) Metabolic networks underlying cognitive reserve in prodromal Alzheimer disease: A European Alzheimer disease consortium project. J Nucl Med 54, 894–902.

13.

Claus

, Staekenborg

, Roorda

, Stevens

, Herderschee

, van

Maarschalkerweerd W

, Schuurmans

, Tielkes

, Koster

, Bavinck

, Scheltens

(2016) Low prevalence of mixed dementia in a cohort of 2,000 elderly patients in a memory clinic setting. J Alzheimers Dis 50, 797–806.

14.

Petersen

, Smith

, Waring

, Ivnik

, Tangalos

, Kokmen

(1999) Mild cognitive impairment: Clinical characterization and outcome. Arch Neurol 56, 303–308.

15.

McKhann

, Knopman , D

Chertkow

, H , Hyman

, Jack

Jr , Kawas

, Klunk

, Koroshetz

, Manly

, Mayeux

, Mohs

, Morris

, Rossor

, Scheltens

, Carrillo

, Thies

, Weintraub

, Phelps

(2011) The diagnosis of dementia due to Alzheimer’s disease: Recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement 7, 263–269.

16.

Verhage

(1964) Intelligence and age in a Dutch sample. Hum Development 8, 238–245.

17.

Derix

, Hofstede

, Teunisse

, Hijdra

, Walstra

, Weinstein

, van

Gool WA

(1991) CAMDEX-N: The Dutch version of the Cambridge Examination for Mental Disorders of the Elderly with automatic data processing. Tijdschr Gerontol Geriatr 22, 143–150.

18.

Folstein

, Folstein

, McHugh

(1975) “Mini-mental State”. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 12, 189–198.

19.

Scheltens

, Leys

, Barkhof

, Huglo

, Weinstein

, Vermersch

, Kuiper

, Steinling

, Wolters

, Valk

(1992) Atrophy of medial temporal lobes on MRI in “probable” Alzheimer’s disease and normal ageing: Diagnostic value and neuropsychological correlates. J Neurol Neurosurg Psychiatry 55, 967–972.

20.

Pasqueir

, Leys

, Weerts

, Mounier-Vehier

, Barkof

, Scheltens

(1997) Inter- and intraobserver reproducibility of cerebral atrophy assessment on MRI scans with hemispheric infarcts. Eur Neurol 36, 268–272.

21.

Liu

, Julkunen

, Paajanen

, Westman

, Wahlund

, Aitken

, Sobow

, Mecocci

, Tsolaki

, Vellas

, Muehlboeck

, Spenger

, Lovestone

, Simmons

, Soininen

(2012) Education increases reserve against Alzheimer’s disease- evidence from structural MRI analysis. Neuroradiology 54, 929–938.

22.

Persson

, Eldholm

, Barca , M , Cavallin

, Ferreira

, Knapskog

, Selbæk

, Brækhus

, Saltvedt

, Westman

, Engedal

(2017) MRI-assessed atrophy subtypes in Alzheimer’s disease and the cognitive reserve hypothesis. PLoS One 12, e0186595.

23.

Gregory

, Long

, Kloppel

, Razi

, Scheller

, Minkova

, Papoutsi

, Mills

, Durr

, Leavitt

, Roos

RAC

, Stout

, Scahill

, Langbehn

, Tabrizi

, Rees

(2017) Operationalizing compensation over time in neurodegenerative disease. Brain 140, 1158–1165.

24.

Wilson

, Yu

, Lamar

, Schneider

, Boyle

, Bennet

(2019) Education and cognitive reserve in old age. Neurology 92, e1041–e1050.

25.

van

Loenhoud AC

, van

der Flier WM

, Wink

, Dicks

, Groot

, Twisk

, Barkhof

, Scheltens

, Ossenkoppele

(2019) Cognitive reserve and clinical progression in Alzheimer disease: A paradoxical relationship. Neurology 93, e334–e346.

26.

Soldan

, Pettigrew

, Lu

, Wang

, Selnes

, Albert

, Brown

, Ratnanather

, Younes

, Miller

(2015) BIOCARD Research Team Relationship of medial temporal lobe atrophy, APOE genotype, and cognitive reserve in preclinical Alzheimer’s disease. Hum Brain Mapp 36, 2826–2842.

27.

Israel

, Beaulieu

, Hartless

(2001) The influence of family and community social capital on educational achievement. Rural Sociol 66, 43–68.