Cognitive reserve predicts baseline tau burden in the U.S. POINTER trial imaging cohort

Abstract

INTRODUCTION

Higher cognitive reserve (CR) is associated with reduced dementia risk. We hypothesized that higher CR is associated with less baseline Alzheimer's disease (AD) pathology in the U.S. Study to Protect Brain Health Through Lifestyle Intervention to Reduce Risk (U.S. POINTER) cohort.

METHODS

A subsample of participants underwent amyloid beta and tau positron emission tomography imaging. Regression analysis was used to model the association between educational attainment (EA) as a CR proxy measure, amyloid positivity, and entorhinal cortex (ERC) and meta‐temporal region of interest (meta‐ROI) tau standardized uptake value ratio (SUVR).

RESULTS

In 911 participants with complete imaging data, higher CR was significantly associated with lower ERC tau SUVR. CR was not associated with amyloid status or meta‐ROI tau.

DISCUSSION

In the U.S. POINTER cohort, higher EA predicted lower tau burden in the ERC. Meta‐ROI tau levels may be too low in this cognitively unimpaired sample to reveal associations. CR may have a role in promoting biological resistance to AD pathology.

Highlights

• Higher cognitive reserve is associated with a lower dementia risk and better cognitive performance after a diagnosis of Alzheimer's disease.

• The U.S. Study to Protect Brain Health Through Lifestyle Intervention to Reduce Risk trial imaging cohort was cognitively unimpaired at baseline and a subsample received amyloid and tau positron emission tomography scans.

• Higher educational attainment, a proxy measure of cognitive reserve, was significantly related to tau levels in the entorhinal cortex region of interest.

1. BACKGROUND

Alzheimer's disease (AD) is a progressive neurodegenerative disorder that limits both the duration and quality of life for affected individuals. AD is characterized by the pathologic deposition of extra‐neuronal amyloid beta (Aβ) plaques and intracellular aggregation of misfolded tau protein. AD pathological change is by far the most common underlying cause of dementia, representing an estimated 60% to 80% of cases, although many persons with AD also have evidence of other pathological processes caused by cerebrovascular disease and Lewy body disease, among others, that may contribute to their cognitive symptoms. ¹

Because of the high prevalence of AD and its significant impact on individuals’ and families’ physical, emotional, and economic health, intense efforts have been directed toward identifying modifiable risk factors for AD and effective preventive interventions.

A construct that has garnered attention as a significant modifier of AD dementia risk is cognitive reserve (CR). This construct arose from observations of a protective effect of education attainment (EA) on age of onset and cognitive test performance in the early stages of AD. ² , ³ , ⁴ , ⁵ , ⁶ , ⁷ A National Institute of Health working group convened to clarify conceptual and operational definitions of CR defined it as a property of the brain that manifests as better than expected cognitive performance for a given level of brain injury or pathological change. ⁸ Kremen et al. ⁹ proposed a framework in which CR is viewed as an original reservoir of cognitive ability, “cognitive maintenance” is the longitudinal preservation of this original store of cognitive ability, and “cognitive resilience” is an absence of cognitive decline in the presence of aging or neurodegeneration. Cognitive maintenance and cognitive resilience are presumed to be underpinned by anatomical and cellular/molecular features of the brain. ¹⁰

CR is best captured by tests of cognitive ability obtained in early adulthood, such as high school achievement tests or military service entrance exams. Composite scores that include selected neuropsychological tests obtained in older adults have been used to estimate CR. ¹¹ , ¹² When these measures are not available, the variable most often used to reflect CR is EA. ¹³ Children's and adolescents’ measured cognitive ability is highly predictive of ultimate EA, ¹⁴ , ¹⁵ , ¹⁶ and of a dementia diagnosis risk in later life independent of other demographic and health‐related risk factors. ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² , ²³ Although other demographic variables, such as occupational complexity and socioeconomic status, have been shown to contribute to CR, EA has historically been the best socio‐demographic predictor of later life health outcomes, including AD. ¹³

Studies aimed at understanding the role of CR in cognitive resilience in the presence of AD pathologic changes have shown consistent support for a mitigating effect of higher CR on cognitive symptoms in the early stages of AD, with a more rapid rate of decline as pathology advances. ¹¹ , ²⁴ , ²⁵ , ²⁶ , ²⁷ , ²⁸ , ²⁹ Results are mixed for studies that examined the role of CR in delaying the onset and/or progression of AD pathologic changes that lead to cognitive decline, with some studies reporting evidence of an association between CR and baseline level and/or rate of change in AD biomarkers ³⁰ , ³¹ , ³² , ³³ and other studies finding no relationship between CR and AD biomarker levels or longitudinal change. ¹¹ , ³⁰ , ³¹ , ³² , ³³ , ³⁴ , ³⁵ , ³⁶

To clarify the role of CR in modifying levels of AD pathology prior to AD diagnosis, we tested the cross‐sectional association between amyloid positivity and cerebral tau levels in participants enrolled in The U.S. Study to Protect Brain Health Through Lifestyle Intervention to Reduce Risk (U.S. POINTER) trial who also enrolled in an ancillary imaging study that collected magnetic resonance imaging (MRI) and amyloid and tau positron emission tomography (PET) scans. ³⁷ , ³⁸

2. METHODS

2.1. Participants

Details of the study design and methods of the U.S. POINTER trial and an analysis of baseline characteristics of the imaging subsample have been reported. ³⁸ , ³⁹ The parent trial design called for randomization of 2000 participants recruited from the communities surrounding five sites: University of California Davis, Wake Forest University School of Medicine, Rush University and Advocate Aurora Health (combined Chicago, Illinois site), Baylor College of Medicine, Kelsey Research Foundation and Houston Methodist (combined Houston, Texas site) and Brown University/Merriam Hospital and Butler Hospital (combined New England/Rhode Island site). To be eligible for the parent trial, participants had to: 1) be between 60 and 79 years old inclusive; 2) be sedentary (not a regular exerciser, determined using the POINTER Physical Activity Questionnaire); 3) receive a low Mediterranean‐DASH Intervention for Neurodegenerative Delay (MIND) Diet score (determined using the MIND Diet Screener) ⁴⁰ , ⁴¹ ; 4) be presumed to be cognitively unimpaired based on a demographically adjusted score ≥ 32 on the Telephone Interview of Cognitive Status–Modified (TICSm) ⁴² and a global score ≤ 0.5, a Sum of Boxes ≤ 1.0 on the Clinical Dementia Rating Scale (CDR); ⁴³ 5) have ≥ 2 risk factors for cognitive decline, including: suboptimal cardiovascular health (treated or untreated): systolic blood pressure ≥ 125 mmHg or low‐density lipoprotein cholesterol ≥ 115 mg/dL or glycated hemoglobin (HbA1c) ≥ 6.0%; a first‐degree family history (mother, father, sister, brother) of memory impairment; race/ethnicity self‐identification as Black, Native American, or Hispanic/Latinx, Asian, other race/ethnicity; the 70–79 year age group; male sex. Recruitment methods were designed to ensure a racially and ethnically diverse study population. Of 2011 individuals randomized to the parent trial, 1052 participants agreed to participate in an imaging ancillary study that included baseline and longitudinal structural MRI, and amyloid and tau PET scans. Participants who completed both a baseline amyloid and tau PET scan were eligible for inclusion in the present analysis.

2.2. Variables

2.2.1. Demographic variables and CR

Age, sex, racial/ethnic group self‐identification, and years of education were collected on a baseline participant questionnaire. Participants were included if they self‐identified as Black, Asian, White of European ancestry, or Hispanic/Latino. There were insufficient numbers of representatives of other racial/ethnic groups to allow valid adjustment of regression models for this variable, which has a known association with AD risk.

RESEARCH IN CONTEXT

1. Systematic review: The authors used web‐based tools and reference lists of published papers to identify articles related to cognitive reserve (CR) and its relation to preserved cognitive performance Alzheimer's disease (AD) pathology.

2. Interpretation: This study found a significant relationship between higher educational attainment, a proxy for CR, and lower baseline tau levels in the entorhinal cortex in cognitively unimpaired participants recruited to a multimodal lifestyle intervention to reduce dementia risk (U.S. Study to Protect Brain Health Through Lifestyle Intervention to Reduce Risk [U.S. POINTER]). CR did not predict amyloid positivity or tau levels in the meta‐temporal region of interest. These results provide evidence that CR may directly influence AD pathological spread.

3. Future directions: Planned analysis of longitudinal positron emission tomography scans in the initially unimpaired U.S. POINTER cohort will further elucidate the role of CR in progression of AD pathology.

Operationalizing CR in older populations is a challenge because current cognitive measures may be confounded with ongoing neurodegenerative processes. Some POINTER participants were monolingual Spanish speakers with cognitive testing performed in Spanish. Although the trial endpoint consisted of a global cognitive composite score obtained from a test battery that included tests of memory and executive function and processing speed, ³⁷ the battery did not include subtests commonly used to predict pre‐morbid cognitive function. In the absence of a validated measure of pre‐morbid functioning in this diverse sample, we used EA as an acceptable measure of CR.

In the U.S. POINTER trial, EA was captured as a nine‐level categorical variable that reflected milestones beginning with no formal education to doctoral degree. No participant had obtained less than a high school education and for purposes of the analysis, the nine categories were collapsed to three levels: high school with or without some college experience, college graduate, and advanced degree.

2.2.2. Health behaviors and vascular risk factors

Health behaviors that have been linked to dementia risk were collected during three baseline screening visits. A dietary intake pattern associated with favorable cognitive outcomes was assessed with the MIND diet score. ⁴¹ Systolic and diastolic blood pressure were obtained with automated devices. Recorded values reflected the average of the last two of three readings. Body mass index (BMI) calculated from weight and height measurements. Physical fitness was assessed by means of a timed, 400 m walk test carried out at the baseline visit. The metric chosen to reflect overall physical fitness in the analysis was gait speed expressed as meters per second to complete the total number of laps. A reported diagnosis of type 2 diabetes, HbA1c, and BMI served as measures of metabolic status. Treatment with insulin at baseline was exclusionary.

2.2.3. Apolipoprotein E genotype and family history of dementia

Whole blood samples were collected and shipped ambient overnight to the Alzheimer's Therapeutic Research Institute Biomarker Core and Biorepository where whole blood was centrifuged to separate the blood into its components. DNA was extracted from the buffy coat using the Qiagen QIAcube‐HT instrument and manufacturer's instructions. QIAamp 96 DNA QIAcube‐HT kits were used per manufacturer's instructions to isolate and purify the DNA. Buffy coat samples were thawed on wet ice and briefly vortexed to prepare for DNA extraction on the instrument. Apolipoprotein E (APOE) genotype was determined by real‐time analysis of single nucleotide polymorphism (SNP) using TaqMan technology (Life Technologies). The real‐time analysis was performed using validated TaqMan assays on a CFX‐96 real‐time system (Bio‐Rad) per manufacturer's instructions. Analysis was performed using the Bio‐Rad CFX Manager software. Participant self‐reports of dementia in close family members, including mother, father, siblings, or any close family member, were obtained during the screening interview.

2.2.4. Amyloid and tau PET imaging

U.S. POINTER Ancillary Neuroimaging Study protocol included structural MRI, cerebral amyloid accumulation measured from 18F‐florbetaben (FBB) PET scans (acquired at 90–110 minutes post‐injection) and cerebral tau measured from 18F MK‐6240 PET scans (90–110 minutes post‐scan uptake). The FBB and MK‐6240 PET scans were co‐registered to structural MRIs, and anatomically defined regions were used to calculate total and region‐specific amyloid and tau standardized uptake value ratios (SUVRs), as described previously. ³⁸ , ⁴⁴ FBB and MK‐6240 images were pre‐processed according to a protocol based on the Alzheimer's Disease Neuroimaging Initiative to account for scanner differences, including smoothing to common 6 mm³ resolution. ⁴⁵ Amyloid positivity was defined as a global summary region SUVR of > 1.08 (intensity normalized by the whole cerebellum reference region), which represents 2 standard deviation above the mean global summary region SUVR of a group of young controls. ⁴⁶ Cross‐sectionally, amyloid levels have little association with cognitive symptoms. Thus, the binary classification of amyloid positivity is of greater value in identifying persons on the AD pathology trajectory than a continuous SUVR measure.

Tau pathology spreads from the transentorhinal region to the hippocampus and then to the ventral and lateral temporal lobe and other neocortical areas as cognitive symptoms progress. ⁴⁷ , ⁴⁸ Because the U.S. POINTER study population was cognitively unimpaired at baseline, we limited the analysis of tau burden to the entorhinal cortex (ERC) and a meta‐ROI which represents a weighted average of tau burden across several temporal regions of interest (ROIs): amygdala, ERC, parahippocampal gyrus, fusiform gyrus, inferior temporal gyrus, and middle temporal gyrus. This meta‐ROI has been found to differentiate controls from those with AD and captures/summarizes clinically meaningful tau burden across the clinical spectrum. ²⁰ , ⁴⁹ There is evidence that amyloid pathology influences tau accumulation in the ERC as well as beyond the ERC in the temporal lobes. ⁵⁰ Although it is a small region, signal in the ERC, even when divided into subregions, has been linked to meaningful differences in cognitive performance and amyloid positivity. ⁴⁹ , ⁵¹ , ⁵² Both tau ROIs were intensity normalized by inferior cerebellar gray matter.

2.2.5. Statistical analysis

Descriptive statistics were calculated for the total group and for subgroups defined by baseline amyloid status. We first evaluated the predictors of amyloid positivity using logistic regression analysis. Linear regression analysis was then used to evaluate the association between the two dependent variables of interest, ERC tau and meta‐ROI tau, and the independent variable, CR‐EA, adjusting for demographic, behavioral, genetic, and vascular covariates. We used 5 mm increments of systolic blood pressure in our models, and both systolic and diastolic pressure have similar relationships to cognitive outcomes, with systolic pressure showing a more pronounced effect. ⁵³ Our modeling strategy was to enter all candidate variables simultaneously as main effects. We then tested interactions between CR‐EA and variables that were statistically significant (p ≤ 0.05) when entered as main effects to identify possible differential influence of CR‐EA at different levels of the candidate covariates. We did not adjust for cognitive measures collected at baseline, because our hypothesis focused on the direct association between CR and AD PET biomarkers, not on the role of CR in altering symptomatic responses to AD biomarker levels. All analyses were conducted using STATA 17.

3. RESULTS

3.1. Sample characteristics

A total of 911 U.S. POINTER study participants completed both an amyloid and tau PET scan and provided information on their self‐identified race/ethnicity and education attainment. Two hundred sixty‐six of these participants (28.5%) were amyloid positive. The characteristics of study participants, stratified by amyloid positivity, are reported in Table 1. All three education levels were well represented, with nearly equal proportions of participants at each level. Although most participants were White/European (73%), there was substantial numeric representation of Black (n = 129), Hispanic/Latino (n = 70), and Asian (n = 28) participants. Sixty‐one participants selected their race or ethnicity as “other.” Of these, 43 (70.5%) identified their race as White plus another race or ethnicity. Approximately 20% of participants had a global CDR score of 0.5. A global CDR score of 0.5 was not exclusionary for study participation, because all participants had to pass an initial cognitive screening test and a global CDR of 0.5 requires further evaluation to determine whether the individual meets definitions for cognitive impairment. Participants’ cognitive performance was monitored during the trial and adjudicated as to whether they met criteria for mild cognitive impairment or early dementia. Eventual diagnostic classifications were not available in the baseline data set, and as has been done in other reports of baseline U.S. POINTER sample characteristics, we included them in the analysis.

TABLE 1.

Characteristics of the study sample by amyloid status. ^*

	Amyloid positive (n = 266)	Amyloid negative (n = 645)	Total sample (n = 911)	p
Demographic variables
Age (years)	69.67 ± 5.09	67.88 ± 5.22	68.40 ± 5.24	<0.001
Sex (n, % female)	166 (62.42)	392 (60.78)	911 (61.25)	0.646
Self‐identified racial and ethnic group (n, % within race‐ethnicity)				0.091
Black	29 (10.20)	100 (15.55)	129 (14.19)
Asian	6 (2.26)	22 (3.42)	28 (3.08)
White/European	192 (72.18)	429 (66.72)	621 (68.32)
Hispanic/Latino	16 (6.02)	54 (8.40)	70 (7.70)
Other †	23 (8.65)	38 (5.91)	61 (6.71)
Education: highest degree (n, % within degree attained)				0.796
High school	88 (33.08)	199 (30.85)	287 (31.50)
College graduate	89 (33.46)	226 (35.04)	315 (34.58)
Advanced degree	89 (33.46)	220 (34.11)	309 (33.92)
U.S. POINTER site (n, % within site)				0.015
1	30 (11.28)	86 (11.28)	116 (12.73)
2	65 (24.44)	140 (21.71)	205 (22.50)
3	71 (26.69)	115 (17.83)	186 (20.42)
4	63 (23.68)	190 (29.46)	253 (27.77)
5	37 (13.91)	114 (17.67)	151 (16.58)
Behavioral, metabolic, and genetic risk factors
Memory loss in mother (n, % yes)	156 (58.65)	339 (52.56)	495 (54.34)	0.093
Memory loss in family member (n, % yes)	207 (77.82)	481 (74.57%)	688 (75.52%)	0.300
APOE genotype (n,% of ε4 alleles)				<0.001
0	139 (52.45%)	495 (76.98%)	634 (69.82%)
1	108 (40.75%)	139 (21.62%)	247 (27.20%)
2	18 (6.79%)	9 (1.40%)	27 (2.97%)
Baseline MIND score	7.09 ± 1.38	6.99 ± 1.47	7.02 ± 1.45	0.349
Baseline gait speed (meters/second)	1.33 ± 0.20	1.35 ± 0.22	1.34 ± 0.21	0.125
Baseline systolic blood pressure (mmHg)	131.2 ± 16.48	131.75 ± 15.77	131.60 ± 15.97	0.660
Diagnosed type 2 diabetes	45 (16.92%)	96 (14.88%)	141 (15.48%)	0.440
BMI (Kg/m2)	30.45 ± 5.81	30.21 ± 5.22	30.28 ± 5.40	0.551
Imaging results
Meta temporal tau SUVR	1.12 ± 0.22	1.04 ± 0.11	1.06 ± 0.16	<0.001
Entorhinal cortex tau SUVR	1.25 ± 0.41	1.05 ± 0.20	1.11 ± 0.29	<0.001
Cognitive scores
MMSE	28.72 ± 1.29	28.82 ± 1.35	28.79 ± 1.31	0.284
Cognitive composite ‡ (z‐score)	−0.13 ± 0.99	−0.02 ± 1.01	−0.05 ± 1.01	0.118
Global CDR score (n, % with CDR = 0.5)	116 (18.0)	62 (23.3)	178 (19.5)	0.065

Abbreviations: APOE, apolipoprotein E; BMI, body mass index; CDR, Clinical Dementia Rating; MIND, Mediterranean‐DASH Intervention for Neurodegenerative Delay; MMSE, Mini‐Mental State Examination; SUVR, standardized uptake value ratio; U.S. POINTER, U.S. Study to Protect Brain Health Through Lifestyle Intervention to Reduce Risk.

^*Tabled values are mean (standard deviation) for continuous variables or n (%) for discrete variables. P value is for comparison of amyloid‐positive versus amyloid‐negative participants by analysis of variance for continuous variables or chi squared for discrete variables.

^† Forty‐one (67.21%) of the “other” racial/ethnic group participants selected “White/European” plus another racial/ethnic group.

^‡ Global composite score derived from tests of the modified Neuropsychological Test Battery, which include the Free and Cued Selective Reminding Test (FCSRT), Story Recall, Visual Paired Associates, Number Span, Word Fluency, Trail‐Making Test Parts A and B (TMT‐A and TMT‐B), and Digit Symbol Substitution Test (DSST). ⁵⁴

^§Clinical Dementia Rating Scale Global Score: 0 = unimpaired, 0.5 = isolated memory impairment, function not affected, consistent with mild cognitive impairment.

3.2. Association between amyloid positivity and education attainment

Table 2 presents the predictors of amyloid positivity in this sample, adjusted for all other study variables. Higher age (odds ratio [OR] = 1.10; 95% confidence interval [CI] = 1.06, 1.14; p ≤ .001), White/European ethnicity (OR = 1.71; 95% CI = 1.07, 3.13; p = 0.035), mixed/other race/ethnicity (OR = 2.37; 95% CI = 1.12, 5.00, p = 0.024), and having one or two APOE ε4 alleles (OR _{1 ε4} = 2.92; 95% CI = 2.09, 4.09; p ≤ .001; OR _{2 ε4} = 10.65; 95% CI = 4.37, 25.93; p < 0.001) conferred a higher risk of amyloid positivity. Neither EA nor the behavioral or metabolic variables were significantly associated with being amyloid positive, although the OR for advanced degree versus high school graduation approached significance and suggested a protective effect (OR_AdvDeg = 0.70, 95% CI = 0.47, 1.04, p = 0.074).

TABLE 2.

Predictors of amyloid positivity. ^*

Dependent variable: amyloid status
Demographic variables
Age (years)	1.10 (1.06, 1.14)	<0.001
Sex (male = reference)	1.21 (0.84, 1.73)	0.303
Education attainment (highest degree)
High school (ref)	−
College graduate	0.80 (0.54, 1.18)	0.260
Advanced degree	0.70 (0.47, 1.04)	0.074
Self‐identified racial and ethnic group
Black (reference)	−
Asian	1.34 (0.45, 3.95)	0.597
White/European	1.71 (1.01, 3.13)	0.047
Hispanic/Latino	1.28 (0.59, 2.78)	0.526
Mixed/Other	2.37 (1.12, 5.00)	0.024
Behavioral, metabolic, and genetic risk factors
Memory loss in mother (no = reference)	1.41 (0.94, 2.12)	0.097
Memory loss in family member (no = reference)	1.17 (0.71, 1.91)	0.535
APOE genotype (number of ε4 alleles)
0 (reference)	−
1	2.92 (2.09, 4.09)	<0.001
2	10.65 (4.37, 25.93)	< 0.001
Baseline MIND score (per 1 unit increment)	1.02 (0.92, 1.14)	0.671
Gait speed in walking test (per 1 unit increment)	1.34 (0.50, 3.62)	0.566
Baseline SBP (mmHg, per 5 unit mmHg)	0.99 (0.94, 1.04)	0.606
Diagnosed diabetes (no = reference)	1.17 (0.76, 1.81)	0.463
BMI (Kg/m2, per 1 unit increment)	1.03 (0.99, 1.06)	0.139

Abbreviations: APOE, apolipoprotein E; BMI, body mass index; MIND, Mediterranean‐DASH Intervention for Neurodegenerative Delay; SBP, systolic blood pressure; U.S. POINTER, U.S. Study to Protect Brain Health Through Lifestyle Intervention to Reduce Risk.

^*Model adjusted for U.S. POINTER site.

3.3. Association between tau SUVR and education attainment

The associations between meta‐ROI and ERC tau SUVR and CR‐EA are reported in Table 3. In a regression model that predicted meta‐ROI SUVR adjusted for amyloid status and all covariates, White/European race/ethnicity (beta ± standard error [SE]_{White vs. Black} = 0.032 ± 0.016, p = 0.046) and positive amyloid status (beta ± SE = 0.070 ± 0.012, p < 0.001) predicted high meta‐ROI SUVR. A diagnosis of type 2 diabetes was associated with lower meta‐ROI tau SUVR (beta ± SE = –0.032 ± .014, p = 0.029). Variables associated with higher ERC tau SUVR included older age (beta ± SE = 0.010 ± .002, p < 0.001), amyloid positivity (beta ± SE = 0.167 ± 0.021, p < 0.001), and APOE genotype (beta ± SE_{1 ε4 vs. 0 ε4} = 0.052 ± 0.021, p = 0.014; beta ± SE_{2 ε4 vs. 0 ε4} = 0.156 ± .055, p = 0.005). Lower ERC tau SUVR was significantly associated with CR‐EA (beta ± SE_{Coll vs. HS} = –0.046 ± .023, p = 0.042; beta ± SE_{AdvDeg vs. HS} = –0.065 ± .023, p = 0.005) and BMI (beta ± SE = –0.004 ± .002, p = 0.046). An interaction term for EA by amyloid status added to the full model was significant in predicting ERC tau SUVR (Table 3), and therefore we evaluated the role of CR‐EA in the sample stratified by amyloid status (Table 4). We excluded from these models the variables that were not associated with either meta‐ROI or ERC tau measures. As shown in Table 4, among amyloid‐positive participants, CR‐EA did not predict meta‐ROI tau SUVR, but higher CR‐EA was significantly associated with lower ERC tau SUVR (beta ± SE _{Coll vs. HS} = –0.143 ± 0.060, p = 0.019; beta ± SE _{AdvDeg vs. HS} = –0.168 ± 0.060, p = 0.006). The relationship between higher CR‐EA and lower predicted ERC tau SUVR is shown in Figure 1. Interaction terms for CR‐EA and age, race/ethnicity, sex, and number of APOE ε4 alleles were not significant in predicting meta‐ROI or ERC tau.

TABLE 3.

Predictors of tau SUVR in all participants. ^*

	Meta temporal tau SUVR		Entorhinal cortex tau SUVR
	Beta (SE beta)	p	Beta (SE beta)	p
Full model
Age (years)	0.002 (0.001)	0.116	0.010 (0.002)	<0.001
Sex (male = reference)	0.021 (0.011)	0.065	0.018 (0.021)	0.388
Education attainment (highest degree)
High school (ref)	−		−
College graduate	−0.017 (0.013)	0.183	−0.046 (0.023)	0.042
Advanced degree	−0.019 (0.133)	0.133	−0.065 (0.023)	0.005
Self‐identified racial and ethnic group
Black (ref)	−		−
Asian	0.019 (0.033)	0.561	0.105 (0.059)	0.074
White/European	0.032 (0.016)	0.046	0.042 (0.029)	0.144
Hispanic/Latino	0.010 (0.023)	0.654	0.011 (0.042)	0.785
Other	0.031 (0.024)	0.209	0.053 (0.044)	0.224
Amyloid status (reference = negative)	0.070 (0.012)	<0.001	0.167 (0.021)	<0.001
Family history of dementia (n, % Yes)	−0.0003 (0.016)	0.984	−0.001 (0.028)	0.965
Memory loss in mother (n, % yes)	0.002 (0.013)	0.897	0.011 (0.024)	0.631
APOE genotype (number of ε4 alleles)
0 (ref)	−		−
1	0.013 (0.012)	0.260	0.052 (0.021)	0.014
2	0.057 (0.031)	0.062	0.156 (0.055)	0.005
Baseline MIND score	0.0004 (0.004)	0.919	−0.005 (0.006)	0.467
Gait speed in walking test	−0.020 (0.032)	0.530	−0.025 (0.058)	0.670
Baseline SBP (5 mmHg increments)	0.001 (0.002)	0.403	0.001 (0.003)	0.859
Diagnosed type 2 diabetes (0 = no)	−0.032 (0.014)	0.029	−0.031 (0.026)	0.232
BMI	−0.001 (0.001)	0.282	−0.004 (0.002)	0.046
Amyloid‐EA interaction (reference levels omitted)
Amyloid+ by college graduate	−0.049 (0.028)	0.078	−0.134 (0.049)	0.007
Amyloid+ by advanced degree	−0.039 (0.028)	0.154	−0.134 (0.049)	0.006

Abbreviations: APOE, apolipoprotein E; BMI, body mass index; EA, education attainment; MIND, Mediterranean‐DASH Intervention for Neurodegenerative Delay; SBP, systolic blood pressure; SUVR, standardized uptake value ratio; U.S. POINTER, U.S. Study to Protect Brain Health Through Lifestyle Intervention to Reduce Risk.

^*Tabled values are unstandardized beta coefficients (standard errors) from linear regression equations predicting meta temporal tau SUVR and entorhinal cortex tau SUVR. Models adjusted for U.S. POINTER site. Interactions of EA with race/ethnicity were not significant and are not reported.

TABLE 4.

Relationship between educational attainment and tau SUVR in participants stratified by amyloid status. ^*

	Amyloid positive (n = 266)				Amyloid negative participants (n = 645)
	MetaROI tau SUVR		Entorhinal cortex tau SUVR		MetaROI tau SUVR		Entorhinal cortex tau SUVR
	Beta (SE)	p	Beta (SE)	p	Beta (SE)	p	Beta (SE)	p
Age (years)	0.002 (0.003)	0.505	0.010 (0.005)	0.053	0.002 (0.001)	0.015	0.009 (0.002)	<0.001
Sex (reference = male)	0.027 (0.029)	0.352	0.032 (0.053)	0.539	0.019 (0.009)	0.034	0.007 (0.016)	0.647
Education attainment (highest degree)
High school (reference)	−		−		−
College graduate	−0.048 (0.034)	0.155	−0.143 (0.060)	0.019	0.000 (0.011)	0.995	−0.003 (0.019)	0.861
Advanced degree	−0.048 (0.033)	0.153	−0.168 (0.060)	0.006	−0.006 (0.011)	0.562	−0.023 (0.019)	0.245
Self‐identified racial and ethnic group †
Black (ref)	−		−		−		−
Asian	0.030 (0.102)	0.768	0.178 (0.183)	0.333	0.006 (0.011)	0.809	−0.003 (0.047)	0.108
White/European	0.095 (0.047)	0.044	0.144 (0.084)	0.089	0.007 (0.013)	0.585	0.004 (0.023)	0.878
Hispanic/Latino	0.022 (0.070)	0.749	0.012 (0.004)	0.924	−0.004 (0.019)	0.849	.000 (0.034)	0.989
Other	0.092 (0.064)	0.151	0.137 (115)	0.236	−0.012 (0.021)	0.581	−0.004 (0.038)	0.918
Family history
APOE genotype (number of ε4 alleles)
0 (ref)	−		−		−		−
1	0.034 (0.029)	0.234	0.130 (0.052)	0.013	−0.004 (0.011)	0.698	0.009 (0.019)	0.615
2	0.120 (0.057)	0.037	0.278 (0.103)	0.007	−0.047 (0.037)	0.204	0.026 (0.066)	0.699
BMI (Kg/m2)	−0.005 (0.002)	0.038	−0.012 (0.004)	0.006	.001 (0.001)	0.101	0.000 (0.002)	0.756
Diagnosed diabetes	−0.025 (0.038)	0.516	0.042 (0.088)	0.547	−0.021 (0.012)	0.087	−0.035 (0.022)	0.112

Note: Bold font is used to highly statistically significant coefficients.

Abbreviations: APOE, apolipoprotein E; BMI, body mass index; ROI, region of interest; SUVR, standardized uptake value ratio; U.S. POINTER, U.S. Study to Protect Brain Health Through Lifestyle Intervention to Reduce Risk.

^*Tabled values are beta coefficients and standard errors obtained from results of multivariable linear regression equations predicting meta temporal and entorhinal cortex tau SUVR. Consistently non‐significant covariates in full models in Table 3 omitted. Models adjusted for U.S. POINTER site.

^† Interactions of education attainment with race/ethnicity and number of APOE ε4 alleles were tested and none were significant.

FIGURE 1.

Association between predicted entorhinal cortex (ERC) tau SUVR and education attainment in amyloid‐positive participants in the U.S. Study to Protect Brain Health Through Lifestyle Intervention to Reduce Risk cohort. SE, standard error; SUVR, standardized uptake value ratio.

Higher BMI was associated with both lower meta‐ROI tau and ERC tau levels in amyloid‐positive participants, and APOE ε4 allele positivity was associated with higher meta‐ROI and ERC tau SUVR. Higher age predicted higher meta‐ROI and ERC tau SUVR. Female sex was significantly associated with higher meta‐ROI SUVR, but not ERC tau SUVR. We evaluated whether education attainment interacted with APOE ε4 allele number or race/ethnicity in predicting EC tau SUVR in amyloid‐positive individuals, but none of the interaction terms were significant.

Although the ascertainment of education attainment in the U.S. POINTER cohort included considerable detail regarding number of years completed, higher levels of education were expressed as categories (e.g., some college, associate degree, etc.). To test whether the relationship between CR‐EA would remain significant if we modeled the variable as years of education, we assigned numeric values to the categories of post high school education to reflect typical completion times for each degree. As shown in Table S1 in supporting information, there was an inverse association between years of education and ERC tau SUVR in amyloid‐positive participants (beta ± SE = –0.022 ± .036, p = 0.032). We also confirmed that adjustment for a CDR global score consistent with a subjective memory complaint did not affect the results reported in Table 4 (see Table S2 in supporting information).

4. DISCUSSION

In this diverse sample of cognitively normal individuals between 65 and 79 years of age, elevated amyloid status was not associated with education attainment. However, education attainment was significantly associated with tau SUVR levels in a brain region most likely to be involved in tau deposition in pre‐symptomatic individuals. This relationship was restricted to amyloid‐positive participants only, suggesting that CR is most closely associated with reduced tau burden among those on the AD pathway. The only variables associated with both meta‐ROI and ERC tau SUVR in amyloid‐negative participants were age and study site, supporting both in vivo and post mortem studies that suggest that increases in tau deposition are primarily dependent on the presence of amyloid. There was no interaction between racial/ethnic group identification and education in either the meta‐ROI or the ERC tau SUVR levels.

The findings of our study support a possible role of CR not only in the preservation of cognitive performance in the early stages of AD, but in resistance to the onset or rate of deposition of at least one of the proteinopathies responsible for AD symptoms. A slower rate of tau deposition could explain the consistent finding in population‐based cohort studies of a relationship between CR and age of onset of AD symptoms, along with preservation of cognitive function for a time in persons with cognitive symptoms and documented amyloid pathology.

We did not find a cross‐sectional relationship between CR‐EA and amyloid deposition. This finding is consistent with those reported by Almeida et al. using cerebrospinal fluid (CSF) Aβ₄₂ as the amyloid marker. ³¹ A lack of association between CR‐EA and baseline amyloid levels can be inferred from descriptive tables reported in three longitudinal studies of amyloid and tau accumulation. ⁵⁵ , ⁵⁶ , ⁵⁷ Most longitudinal cohort studies report no evidence that amyloid deposition rates in normal and cognitively impaired individuals are significantly altered by baseline CR. ¹¹ , ³⁴ , ³⁶ , ⁵⁵ , ⁵⁶ .

The majority of published studies that examine the relationship between CR and tau deposition focus on slope differences in tau and cognition trajectories related to CR, and do not report cross‐sectional baseline associations between CR and tau level ¹¹ , ²⁷ , ²⁸ , ⁵⁸ , ⁵⁹ However, two studies have reported cross‐sectional baseline relationships between tau PET or CSF levels. Almeida et al. examined the moderating effect of CR‐EA on the relationship between CSF AD biomarkers and age in cognitively normal and cognitively impaired groups and found that CR‐EA attenuated age‐related adverse differences in total tau (t‐tau), phosphorylated tau (p‐tau), and the ratio of these biomarkers with Aβ₄₂ in all groups—that is, age‐adjusted biomarker levels were more favorable in high CR compared to low CR groups. Shimada et al., ³⁰ in a study of amyloid and tau PET levels in cognitively impaired and cognitively normal samples found that lower CR‐EA was independently associated with higher tau SUVR in all study groups.

Studies that examined the effect of CR on longitudinal rate of tau increase have reported mixed results. Jack et al. ³⁵ reported a marginally significant (p = 0.09) increase in rate of tau accumulation associated with a 4 year decrease in CR‐EA in a cognitively impaired sample, but not a cognitively normal sample from the Mayo Clinic cohort. Cai et al. found an accelerated rate of tau tangle formation in higher CR individuals with cognitive impairment, but no relationship between CR and rate of tau increase in cognitively normal individuals. ³³ Baseline SUVR levels were not reported. Pettigrew et al studied the role of CR in AD biomarker changes in an initially cognitively normal cohort and found no relationship between CR and either baseline level or longitudinal change in AD‐specific CSF biomarkers including Aβ_1‐42, t‐tau, and p‐tau₁₈₁. ³⁶ We included self‐identified race/ethnicity classification in our analysis because of documented differences in the prevalence of AD and other dementias across the major self‐identified racial/ethnic groups in the US population, as well as differences in prevalence of AD biomarker elevations and risk genes, including APOE. In the United States, non‐Hispanic Black and Hispanic older adults are more likely than White older adults to have AD or other dementias (Alzheimer's Association Facts and Figures 2024). However, the role of known AD biomarkers associated with this increased risk is unclear. Xiong et al. and other authors have documented differences in baseline and longitudinal Aβ and tau levels measured with a variety of modalities in Black and non‐Hispanic White cohorts consisting of cognitively unimpaired or mixed impaired and unimpaired individuals. ⁶⁰ , ⁶¹ , ⁶² , ⁶³ Molina‐Henry et al. reported that non‐Hispanic White participants had the highest likelihood and Hispanic Blacks the lowest likelihood of being eligible for continued screening based on a plasma Aβ42/40 test in the large sample of cognitively normal individuals screened to participate in a clinical trial of an anti‐amyloid agent. ⁶⁴ Differences in APOE and other AD risk gene prevalence, along with differences in the risk of developing AD with an APOE ε4 allele across population ancestry groups have been documented. ⁶⁴ , ⁶⁵ , ⁶⁶ , ⁶⁷

In the U.S. POINTER imaging sample, persons from the racial/ethnic categories White/European and “other” were at increased risk for amyloid accumulation compared to Black/African‐Americans after adjustment for covariates including APOE genotype. The majority of the participants classified as “other” self‐identified as White/European plus some other race, and it is possible that their predominant genetic ancestry was White/European. Achieving a clear understanding of the relationship among genetic, social, and environmental factors and AD risk biomarkers could be accelerated if population ancestry, AD polygenic risk scores, and socioeconomic metrics were consistently used in prediction models rather than the artificial construct of race/ethnicity.

Although amyloid‐positive participants classified as White/European had higher meta‐ROI tau SUVR, their ERC tau SUVR was not significantly elevated compared to the reference group. The relationship between baseline amyloid accumulation and tau aggregation in groups defined by self‐identified race/ethnicity in the U.S. POINTER imaging sample requires further elucidation, because tau levels are expected to be low in this cognitively normal population, and sample sizes in some racial/ethnic subgroups may not be large enough to detect small differences.

The relationship we found between lower BMI and tau SUVR in amyloid‐positive participants is consistent with previous studies that have reported that, although higher mid‐life BMI is associated with increased the risk of later dementia onset, ⁶⁸ , ⁶⁹ lower BMI is associated with near‐term dementia risk and with AD biomarker levels. ⁷⁰ , ⁷¹ , ⁷² , ⁷³ , ⁷⁴ , ⁷⁵ . An association between type 2 diabetes and lower meta‐ROI SUVR seen in the entire sample did not persist when the sample was stratified by amyloid status. Type 2 diabetes has been found to be associated with vascular pathology, but not AD pathology, although interest in establishing an association between insulin resistance and amyloid accumulation persists. ⁷⁶ Overall, the consistent epidemiological association reported between mid‐life cardiovascular risk factors, including elevated BMI, hypertension, and type 2 diabetes, and development of cognitive impairment ⁷⁷ , ⁷⁸ , ⁷⁹ , ⁸⁰ was not reflected in the cross‐sectional association between measured cardiovascular disease risk factors and AD imaging biomarkers in the U.S. POINTER sample. It is possible that these associations are not discernible in cross‐sectional analyses involving cognitively unimpaired individuals.

There were no consistent relationships between female sex and either amyloid positivity or tau level in the U.S. POINTER imaging sample. Females have a higher prevalence of AD, in part due to their greater longevity, and there is evidence of male and female differences in the relationship between amyloid and tau accumulation in samples with a range of cognitive impairment, from no impairment to dementia. ⁸¹ , ⁸² Planned longitudinal scans may clarify the relationship between sex and amyloid and tau accumulation in this sample.

We limited our analysis to tau SUVR in selected brain regions that were more likely to be informative in a cognitively normal population. The lack of association between CR and meta‐ROI in our sample may be due to a combination of factors, including low levels of tau in this non‐demented group, smaller variability of tau in the meta‐Roi which restricts detection of associations, and the reduction in the role of CR once tau spread has extended beyond the ERC.

Limitations of the study include the potential for selection bias as the U.S. POINTER sample has a higher dementia risk factor burden than the general population, which may constrain the range of study variables and alter associations that would be obtained in a more heterogenous population group. A more direct measure of CR than education attainment, such as early adolescent or adult test scores, could provide a more robust estimate of the effect of CR on tau levels. Repeated imaging of the U.S. POINTER sample after the 2 year intervention phase will provide further clarity on the role of CR‐EA in tau accumulation.

CONFLICT OF INTEREST STATEMENT

S.M. Landau is on the DSMB for KeifeRX and the NIH IPAT study, has consulted for Banner Health, and has received speaking honoraria from Eisai, J&J, and IMPACT‐AD. No other conflicts are reported by the authors. Author disclosures are available in the Supporting Information.

CONSENT STATEMENT

Written informed consent was obtained from participants under protocols approved by the central institutional review board (IRB) at Wake Forest University School of Medicine, with concurrence by local site IRBs.

Supporting information

Supporting Information

Supporting Information

Supporting Information

Pavlik VN, Weber CJ, Masdeu JC, et al. Cognitive reserve predicts baseline tau burden in the U.S. POINTER trial imaging cohort. Alzheimer's Dement. 2025;21:e70892. 10.1002/alz.70892