Evaluating the association of apolipoprotein E genotype and cognitive resilience in SuperAgers

Abstract

INTRODUCTION

“SuperAgers” are oldest‐old adults (ages 80+) whose memory performance more closely resembles middle‐aged adults. The present study examined apolipoprotein E (APOE) allele frequency in non‐Hispanic Black (NHB) and non‐Hispanic White (NHW) SuperAgers compared to controls and Alzheimer's disease dementia cases.

METHODS

In 18,080 participants from eight cohorts, harmonized clinical diagnostics and memory, executive function, and language domain scores were used to identify SuperAgers, cases, and controls across age‐defined bins.

RESULTS

NHW SuperAgers had significantly lower frequency of APOE‐ε4 alleles and higher frequency of APOE‐ε2 alleles compared to all cases and controls, including oldest‐old controls. Similar patterns were found in a small yet substantial sample of NHB SuperAgers; however, not all comparisons with controls reached significance.

DISCUSSION

We demonstrated strong evidence that APOE allele frequency relates to SuperAger status. Further research is needed with a larger sample of NHB SuperAgers to determine if mechanisms conferring cognitive resilience differ across race groups.

Highlights

• Apolipoprotein E (APOE) allele frequency differs between SuperAgers and cases

• APOE allele frequency differs between non‐Hispanic White SuperAgers and controls

• The relationship of APOE and non‐Hispanic Black SuperAger status is unclear

1. BACKGROUND

“SuperAgers” is a term used to describe oldest‐old (ages 80+) adults with episodic memory performance most closely resembling adults in their 50s to mid‐60s. ¹ , ² It is unclear whether SuperAgers’ high memory scores are due to resistance to age‐related pathologic processes or high brain reserve. ³ , ⁴ , ⁵ Further research is needed to elucidate factors conferring optimal memory performance in SuperAgers. Moreover, research is needed to explore optimal memory performance in non‐Hispanic Black (NHB) SuperAgers, as this group is largely understudied. ⁶

Apolipoprotein E (APOE)‐ε4 is the strongest genetic risk factor for late‐onset AD. ⁷ The Northwestern SuperAging project reported lower APOE‐ε4 allele frequency in SuperAgers (N = 10‐12) compared to non‐demented older adults. ² , ⁸ In contrast, most studies report no differences in APOE‐ε4 allele frequency between SuperAgers and oldest‐old adults with typical memory performance, both groups having lower APOE‐ε4 allele frequency compared to AD dementia cases. ⁴ , ⁵ , ⁹ , ¹⁰ , ¹¹ , ¹² Notably, these studies have small SuperAger samples (N = 25‐64) ⁴ , ⁵ , ⁹ , ¹⁰ , ¹¹ , ¹² oftentimes drawn from the same cohort, thus limiting their generalizability and reliability. Further, these studies exclusively include non‐Hispanic White (NHW) participants. ⁴ , ⁵ , ⁹ , ¹⁰ , ¹¹ , ¹² To our knowledge, only one study has been published with a NHB SuperAger sample (N = 61) and did not find a significant difference in APOE‐ε4 allele frequency between SuperAgers and same‐age controls. ⁶ Even fewer studies have explored the relationship of APOE‐ε2, the protective APOE allele, and SuperAger status, ¹¹ , ¹² , ¹³ likely due to the low minor allele frequency of APOE‐ε2. Studies of APOE‐ε2 allele frequency and superior memory in the oldest‐old have not found a significant relationship ¹¹ , ¹² , ¹³ ; however, questions of statistical power, generalizability, and reliability remain.

The present study aims to explore APOE‐ε4 and ‐ε2 allele frequency in SuperAgers compared to AD dementia cases and controls in a large, harmonized multicohort dataset from the Alzheimer's Disease Sequencing Project Phenotype Harmonization Consortium (ADSP‐PHC). Using harmonized clinical diagnoses and cognitive domain scores (e.g., memory, executive function, language), we classified NHW and NHB middle‐aged, old, and oldest‐old adults as cases, controls, or SuperAgers, and compared APOE‐ε4 and ‐ε2 allele frequency of SuperAgers to cases and controls by age bin. Although prior literature suggests that there is not a relationship between optimal memory in oldest‐old age and APOE genotype, this is likely due to a limitation of sample size. The ADSP‐PHC has enabled us to complete, to our knowledge, the largest and most racially diverse study to date of APOE allele frequency and SuperAger status. We hypothesize that NHW and NHB SuperAgers will possess a lower frequency of APOE‐ε4 alleles and a higher frequency of APOE‐ε2 alleles compared to both AD dementia cases and controls.

2. METHODS

2.1. Study population

The ADSP‐PHC was assembled in 2021 to provide large‐scale harmonization of ADSP cohorts, spanning markers of cognition, neuroimaging, fluid biomarkers, and neuropathology. Cohorts that are part of ADSP‐PHC and were included in the present study are: Adult Changes in Thought (ACT), ¹⁴ Alzheimer's Disease Neuroimaging Initiative (ADNI), ¹⁵ Biomarkers of Cognitive Decline Among Normal Individuals (BIOCARD), ¹⁶ National Alzheimer's Coordinating Centers (NACC), ¹⁷ National Institute on Aging Alzheimer's Disease Family Based Study (NIA‐AD FBS), ¹⁸ Religious Orders Study/Rush Memory and Aging Project/Minority Aging Research Study (ROS/MAP/MARS), ¹⁹ , ²⁰ , ²¹ Knight Alzheimer's Disease Research Center (Knight ADRC), ²² and Wisconsin Registry for Alzheimer's Prevention (WRAP). ²³

RESEARCH‐IN‐CONTEXT

1. Systematic review: The authors searched PubMed for studies related to SuperAgers and apolipoprotein E (APOE). Our review revealed that previous studies examining APOE‐ε4 allele frequency in SuperAgers have small SuperAger samples (N = 10‐64), only one study has characterized APOE‐ε4 allele frequency in non‐Hispanic Black (NHB) SuperAgers, and few studies have examined APOE‐ε2 allele frequency in SuperAgers.

2. Interpretation: We extend this work by examining the relationship of SuperAger status and APOE allele frequency in a large, harmonized multicohort dataset, including 18,080 total participants and 1,623 non‐Hispanic White (NHW) and NHB SuperAgers (NHW SuperAgers: N = 1,412; NHB SuperAgers: N = 211).

3. Future directions: We found that APOE allele frequency significantly differed between SuperAgers and Alzheimer's disease dementia cases, while differences in APOE allele frequency between SuperAgers and controls were only significant in NHW comparisons. Further research is needed to determine if mechanisms conferring resilience in SuperAgers differ across race groups.

Written informed consent was obtained from all participants in each cohort, and research was carried out with protocols approved by each site's institutional review board. These secondary analyses were approved by the Vanderbilt University Medical Center institutional review board.

2.2. Cognitive domain scores

Each cohort used different neuropsychological assessment tools to measure cognition which may not be on the same scale. Using pre‐established procedures, ²⁴ we derived co‐calibrated and harmonized cognitive domain scores for four domains (i.e., memory, executive function, language, and visuospatial) using confirmatory factor analysis. ²⁵ In summary, experts carefully reviewed the neuropsychological assessments administered by each cohort and assigned item scores to a cognitive domain: memory, executive function, language, or visuospatial (see https://vmacdata.org/adsp‐phc/data/data‐dictionary/cognition to explore the neuropsychological item scores assigned to each cognitive domain). Rigorous quality control procedures ensured consistent data coding across cohorts. Overlapping items with identical scoring and administration across cohorts were identified as “anchor items” and were used to anchor cognitive domain scores across cohorts. Then, confirmatory factor analyses were used to model memory, executive function, language, and visuospatial domains. Full details on cognitive domain harmonization and co‐calibration can be found in original papers published by Mukherjee and colleagues. ²⁴ , ²⁵

Consistent with previous definitions of SuperAgers, only memory, executive function, and language domains were used in present analyses. ¹ , ² To ensure inclusion of high‐quality harmonization, cognitive domain scores with a standard error of measurement > 0.6 (estimated during the co‐calibration and composite generation procedure) were excluded.

For the purpose of this study, regression‐based cognitive domain scores were created for participant classification. Age, years of education, sex, and cohort were regressed on cognitive domain score at each available timepoint. Residuals were used to capture cognitive performance not explained by demographic variables or cohort differences. Separate models were used for NHW and NHB participants.

2.3. Participant classification

Participants were grouped by race (self or study reported), clinical status, and age (Figure 1A). Age bins were decided based on the age‐related criteria of SuperAgers (ages 80+) and the middle‐aged adults whose memory performance is compared to oldest‐old adults to determine SuperAger status (ages 50‐64). ²⁶ Therefore, the age bins included middle‐aged (ages 50‐64), old (ages 65‐79), and oldest‐old (ages 80+) adults. Participant classification was determined separately for NHW and NHB adults. Of note, the demographic characteristics of the middle‐aged adult participants used to determine SuperAger status (NHW: N = 3,123, age = 59 ± 4, education = 16 ± 2, sex (female%) = 68%; NHB: N = 575, age = 60 ± 4, education = 15 ± 3, sex (female%) = 74%) were similar to the demographic characteristics of the larger oldest‐old adult sample from which our SuperAgers and other oldest‐old study participants were drawn (NHW: N = 13,197, age = 83 ± 4, education = 16 ± 3, sex (female%) = 58%; NHB: N = 1,797, age = 83 ± 3, education = 14 ± 3, sex (female %) = 77%).

FIGURE 1.

Flow diagram for participant classification of SuperAgers, cases, and controls. (A) Flowchart depicting inclusion and exclusion criteria for identifying SuperAgers, AD dementia cases, controls. (B) Flowchart depicting selection order of SuperAgers, cases, and controls. Age range of participants indicated by line segment with arrows on each end. Age of participant classification is indicated by position of shorter, labeled line segments. Closed circles at the end of line segments indicate inclusion of age, such that age range is less‐than‐or‐equal‐to or greater‐than‐or‐equal‐to the age with which the circle aligns, while open circles indicate exclusion of age, such that age range is less‐than or greater‐than the age with which the circle aligns. Sequence of selection is indicated by line height, higher lines indicating earlier selection. AD, Alzheimer's disease; ADSP‐PHC, Alzheimer's Disease Sequencing Project – Phenotype Harmonization Consortium; CN, cognitively normal; EXF, executive functioning; LAN, language; MEM, memory.

Given that longitudinal data were available for most participants, and some participants’ ages spanned multiple age bins, participant classification was decided using a predetermined schema (Figure 1B). Firstly, participants with a clinical diagnosis of AD dementia at least once were considered cases. Across all cohorts, a diagnosis of AD dementia followed standard published criteria (National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer's Disease and Related Disorders Association [NINCDS‐ADRDA]). ²⁷ , ²⁸ , ²⁹ Given that the present study focused on dementia cases likely due to AD, participants with comorbid neurological disorders (e.g., stroke, traumatic brain injury with significant loss of consciousness) or primary diagnoses of non‐AD dementia (e.g., dementia with Lewy bodies, vascular dementia) at any timepoint were excluded. Age bin was determined based on age at first AD dementia diagnosis.

Next, SuperAgers were identified using previously published criteria. ¹ , ² Across all cohorts, SuperAgers were defined as oldest‐old adults with (1) memory domain scores at or above the mean of middle‐aged adults (ages 50‐64), (2) executive function and language residual scores no more than 1 standard deviation below their same‐aged peers (ages 80+) at the same visit as their superior memory performance, and (3) whose diagnosis remained “cognitively normal” for the duration of study participation.

Regarding classification as a control, given the importance of oldest‐old (ages 80+) participants to our central analyses, we sequentially identified oldest‐old, middle‐aged, and then old controls. Criteria for controls included (1) memory, executive function, and language residual scores no lower than 1 standard deviation below their same‐aged peers at a single visit, and (2) who did not meet standard published criteria for mild cognitive impairment or AD dementia and thus were given a diagnosis of “cognitively normal,” for the duration of study participation.

2.4. APOE genotyping

APOE haplotypes were determined from the single nucleotide variants rs7412 and rs429358 for ACT, BIOCARD, NACC, NIA‐AD FBS, Knight ADRC, and WRAP and from pyrosequencing of APOE codons 112 and 158 for ADNI and ROS/MAP/MARS. ³⁰ , ³¹

2.5. Statistical analyses

Statistical analyses were performed using R Statistical Software (v4.2.3.). ³² Logistic regression models examined differences in APOE‐ε4 and ‐ε2 allele frequency of SuperAgers compared to cases and controls at all age bins. APOE allele positivity was determined by allele presence (0 = no allele present, 1 = one or more alleles present). Models covaried for sex and years of education due to their known modifying effects on the relationship of APOE and late‐life cognition. ³³ , ³⁴ Sensitivity analyses included analyses removing individuals with APOE‐ε2/ε4 genotype and adding APOE genotyping method as a covariate. Additionally, to determine whether differences in sample size explained differences in the significance of comparisons across NHW and NHB samples, we included sensitivity analyses randomly down‐sampling the NHW sample by participant classification and cohort. Correction for multiple comparisons was applied using Benjamini‐Hochberg false discovery rate (FDR) procedure. ³⁵ Results tables include odds ratios (OR), confidence intervals (CI), and FDR‐corrected p‐values.

3. RESULTS

3.1. Participant characteristics

In total, 18,080 participants were included in the present analyses with a total of 78,549 datapoints (Table 1). Participants completed an average of 4 ± 4 visits over 5 ± 5 years. The number of follow‐up visits and length of follow‐up varied by cohort due to differences in study design.

TABLE 1.

Participant characteristics by cohort.

Participant demographics	All	ACT	ADNI	BIOCARD	Knight ADRC	NACC	NIA‐AD FBS	ROS, MAP, MARS	WRAP
No. of participants	18080	2188	825	51	735	11851	645	1462	323
No. of observations	78549	13016	3568	455	3314	42004	956	13931	1305
Visits, mean (SD)	4 (4)	6 (3)	4 (3)	9 (3)	5 (4)	4 (3)	1 (1)	10 (6)	4 (2)
Follow‐up time, mean (SD), y	5 (5)	10 (6)	3 (3)	13 (4)	4 (5)	3 (4)	2 (3)	9 (6)	9 (5)
Baseline age, mean (SD), y	72 (10)	73 (6)	74 (7)	53 (9)	74 (8)	72 (10)	74 (12)	77 (8)	53 (6)
NHW race, no. (%)	15698 (87)	2116 (97)	753 (91)	51 (100)	635 (86)	10149 (86)	605 (94)	1086 (74)	303 (94)
Education, mean (SD), y	15 (3)	15 (3)	16 (3)	17 (2)	15 (3)	16 (3)	14 (3)	16 (4)	16 (3)
Female sex, no. (%)	11213 (62)	1295 (59)	401 (49)	30 (59)	453 (62)	7319 (62)	406 (63)	1091 (75)	218 (67)
APOE genotype
ε2/ε2, %	0.4	0.6	0.4	0	0.3	0.4	0.6	0.5	0.3
ε2/ε3, %	8.8	12.5	8.6	13.7	8.7	7.9	5.1	12.1	7.1
ε2/ε4, %	2.6	2.3	2.4	0	2.7	2.6	3.3	2.1	3.4
ε3/ε3, %	47.0	58.8	44.8	62.7	41.9	44.2	37.7	57.9	52.6
ε3/ε4, %	33.5	24.0	33.1	19.6	38.8	35.6	43.1	24.4	32.2
ε4/ε4, %	7.8	1.7	10.7	3.9	7.6	9.4	10.2	2.9	4.3
SuperAgers, no. (%)	1623 (9)	275 (13)	59 (7)	0 (0)	53 (7)	935 (8)	16 (2)	285 (19)	0 (0)
Controls, no. (%)	7628 (42)	1207 (55)	335 (41)	49 (96)	290 (39)	4605 (39)	309 (48)	510 (35)	323 (100)
Cases, no. (%)	8829 (49)	706 (32)	431 (52)	2 (4)	392 (53)	6311 (53)	320 (50)	667 (46)	0 (0)

Abbreviations: ACT, Adult Changes in Thought; ADNI, Alzheimer's Disease Neuroimaging Initiative; APOE, apolipoprotein E; BIOCARD, Biomarkers of Cognitive Decline Among Normal Individuals; Knight ADRC, Knight Alzheimer's Disease Research Center at Washington University; MAP, Memory and Aging Project; MARS, Minority Aging Research Study; NACC, National Alzheimer's Coordinating Centers; NHW, non‐Hispanic White; NIA‐AD FBS, National Institute on Aging Alzheimer's Disease Family Based Study; ROS, Religious Orders Study; WRAP, Wisconsin Registry for Alzheimer's Prevention.

Average baseline age varied by cohort (Age_{[all cohorts]} = 72 ± 10); the youngest cohorts on average, BIOCARD and WRAP, primarily recruited cognitively normal participants. Generally, cohorts were highly educated (Years of Education_{[all cohorts]} = 15 ± 3), mostly female (62%), and mostly NHW (87%). The frequency of each APOE genotype differed by cohort, with higher frequency of ε3/ε4 and ε4/ε4 genotypes in cohorts with a higher frequency of AD dementia cases.

3.2. Participant classification

SuperAgers made up 9% of all participants (N = 1,623). The two youngest cohorts, BIOCARD and WRAP, did not contribute any SuperAgers. Cognitively unimpaired controls comprised 42% of all participants (N = 7,628), and cases made up 49% of all participants (N = 8,829).

Table 2 displays participant characteristics of NHW and NHB SuperAgers, controls, and cases in age‐defined bins. On average, NHW SuperAgers (N = 1,412) were somewhat older, had more years of education, and included more males than NHB SuperAgers (N = 211). There was not a difference in the proportion of participants in each participant classification (SuperAger, control, case) across racialized groups ($\chi$ ² _[2] = 3.79, p = 0.15). Comparing NHW and NHB participants across age‐defined bins, all NHW bins had greater average years of education and a higher proportion of males than NHB bins.

TABLE 2.

Characteristics of SuperAgers, cases, and controls by race and age bin.

A	SuperAgers	Middle‐Aged controls	Old controls	Oldest‐old controls	Middle‐Aged cases	Old cases	Oldest‐old cases
No. of participants	1412	1622	3202	1213	1101	3528	3258
No. of observations	5640	6698	11320	3505	2534	8833	7255
Visits, mean (SD)	4 (3)	4 (3)	4 (3)	3 (2)	2 (2)	3 (2)	2 (2)
Follow‐up time, mean (SD), y	4 (4)	5 (5)	4 (3)	3 (3)	1 (2)	2 (2)	1 (2)
Baseline age, mean (SD), y	83 (3)	58 (4)	71 (4)	84 (4)	59 (5)	74 (4)	87 (5)
Education, mean (SD), y	16 (3)	16 (2)	16 (3)	15 (3)	15 (3)	15 (3)	15 (3)
Female sex, No. (%)	953 (67)	1119 (69)	1964 (61)	661 (54)	618 (56)	1905 (54)	1989 (61)
APOE‐ε2 frequency, No. (%)	246 (17)	216 (13)	467 (15)	179 (15)	50 (5)	184 (5)	321 (10)
APOE‐ε4 frequency, No. (%)	276 (20)	647 (40)	981 (31)	270 (22)	639 (58)	2495 (71)	1425 (44)

B	SuperAgers	Middle‐Aged controls	Old controls	Oldest‐old controls	Middle‐Aged cases	Old cases	Oldest‐old cases
No. of participants	211	296	752	145	85	413	444
No. of observations	864	1017	3017	426	164	1009	1052
Visits, mean (SD)	4 (3)	3 (3)	4 (3)	3 (2)	2 (1)	2 (2)	2 (2)
Follow‐up time, mean (SD), y	4 (4)	3 (4)	4 (3)	2 (3)	1 (1)	2 (2)	2 (2)
Baseline age, mean (SD), y	82 (3)	60 (4)	70 (4)	83 (3)	60 (4)	74 (4)	86 (4)
Education, mean (SD), y	15 (3)	15 (3)	15 (3)	13 (3)	14 (3)	14 (4)	13 (4)
Female sex, No. (%)	177 (84)	222 (75)	564 (75)	115 (79)	58 (68)	283 (69)	331 (75)
APOE‐ε2 frequency, No. (%)	56 (27)	68 (23)	155 (21)	27 (19)	8 (9)	36 (9)	53 (12)
APOE‐ε4 frequency, No. (%)	54 (26)	125 (42)	261 (35)	39 (27)	65 (76)	307 (74)	216 (49)

(A) Non‐Hispanic White participants; (B) Non‐Hispanic Black participants.

Abbreviations: APOE apolipoprotein E;

3.3. APOE allele frequency

In NHW comparisons (Table 3A), SuperAgers had a significantly higher frequency of APOE‐ε2 alleles (Figure 2A; middle‐aged controls: OR = 1.38, 95%CI[1.13, 1.68], P_FDR  = 0.002; old controls: OR = 1.24, 95%CI[1.05, 1.47], P_FDR  = 0.015; oldest‐old controls: OR = 1.28, 95%CI[1.03, 1.59], P_FDR  = 0.035; middle‐aged cases: OR = 4.55, 95%CI[3.30, 6.27], P_FDR  < 0.001; old cases: OR = 4.02, 95%CI[3.26, 4.96], P_FDR  < 0.001; oldest‐old cases: OR = 2.03, 95%CI[1.68, 2.44], P_FDR  < 0.001) and a significantly lower frequency of APOE‐ε4 alleles (Figure 2B; middle‐aged controls: OR = 0.37, 95%CI[0.31, 0.43], P_FDR  < 0.001; old controls: OR = 0.55, 95%CI[0.47, 0.64], P_FDR  < 0.001; oldest‐old controls: OR = 0.81, 95%CI[0.67, 0.99], P_FDR  = 0.044; middle‐aged cases: OR = 0.18, 95%CI[0.15, 0.21], P_FDR  < 0.001; old cases: OR = 0.09, 95%CI[0.08, 0.11], P_FDR  < 0.001; oldest‐old cases: OR = 0.32, 95%CI[0.27,0.37], P_FDR  < 0.001) compared to all cases and controls. In contrast, NHB SuperAgers (Table 3B) had a significantly higher frequency of APOE‐ε2 alleles only compared to cases (Figure 2C; middle‐aged cases: OR = 3.59, 95%CI[1.60, 8.06], P_FDR  = 0.005; old cases: OR = 4.56, 95%CI[2.75, 7.55], P_FDR  < 0.001; oldest‐old cases: OR = 2.73, 95%CI[1.75, 4.25], P_FDR  < 0.001), and a significantly lower frequency of APOE‐ε4 alleles compared to all cases and controls except oldest‐old controls (Figure 2D; middle‐aged controls: OR = 0.48, 95%CI[0.33, 0.71], P_FDR  = 0.001; old controls: OR = 0.66, 95%CI[0.47, 0.94], P_FDR  = 0.035; middle‐aged cases: OR = 0.12, 95%CI[0.07, 0.22], P_FDR  < 0.001; old cases: OR = 0.12, 95%CI[0.08, 0.18], P_FDR  < 0.001; oldest‐old cases: OR = 0.39, 95%CI[0.27, 0.57], P_FDR  < 0.001).

TABLE 3.

Logistic regression model results comparing APOE‐ε2 and ‐ε4 allele frequency among SuperAgers, cases, and controls.

	APOE‐ε2		APOE‐ε4
A	OR (CI)	PFDR	OR (CI)	PFDR
SuperAgers vs. Middle‐Aged controls	1.38 (1.13, 1.68)	0.002	0.37 (0.31, 0.43)	<0.001
SuperAgers vs. Old controls	1.24 (1.05, 1.47)	0.015	0.55 (0.47, 0.64)	<0.001
SuperAgers vs. Oldest‐Old controls	1.28 (1.03, 1.59)	0.035	0.81 (0.67, 0.99)	0.044
SuperAgers vs. Middle‐Aged cases	4.55 (3.30, 6.27)	<0.001	0.18 (0.15, 0.21)	<0.001
SuperAgers vs. Old cases	4.02 (3.26, 4.96)	<0.001	0.09 (0.08, 0.11)	<0.001
SuperAgers vs. Oldest‐Old cases	2.03 (1.68, 2.44)	<0.001	0.32 (0.27, 0.37)	<0.001

	APOE‐ε2		APOE‐ε4
B	OR (CI)	PFDR	OR (CI)	PFDR
SuperAgers vs. Middle‐Aged controls	1.19 (0.79, 1.80)	0.423	0.48 (0.33, 0.71)	0.001
SuperAgers vs. Old controls	1.41 (0.99, 2.01)	0.079	0.66 (0.47, 0.94)	0.035
SuperAgers vs. Oldest‐Old controls	1.63 (0.93, 2.85)	0.115	1.18 (0.70, 1.98)	0.564
SuperAgers vs. Middle‐Aged cases	3.59 (1.60, 8.06)	0.005	0.12 (0.07, 0.22)	<0.001
SuperAgers vs. Old cases	4.56 (2.75, 7.55)	<0.001	0.12 (0.08, 0.18)	<0.001
SuperAgers vs. Oldest‐Old cases	2.73 (1.75, 4.25)	<0.001	0.39 (0.27, 0.57)	<0.001

(A) Non‐Hispanic White; (B) Non‐Hispanic Black.

Abbreviations: APOE apolipoprotein E; CI, confidence interval (95%); FDR, false discovery rate; OR, odds ratio; P_FDR , FDR‐corrected p‐value.

FIGURE 2.

APOE allele frequency in NHW and NHB SuperAgers compared to cases and controls. Bar charts depicting APOE‐ε2 and APOE‐ε4 allele frequency for SuperAgers, cases, and controls across age‐defined subgroups. Participant classification is indicated by the X‐axis, while percent frequency is indicated by the Y‐axis. (A) APOE‐ε2 allele frequency in NHW participants. (B) APOE‐ε4 allele frequency in NHW participants. (C) APOE‐ε2 allele frequency in NHB participants. (D) APOE‐ε4 allele frequency in NHB participants. Asterisks denote significant differences in allele frequency compared to SuperAgers determined by logistic regression models covarying for sex and years of education (* $\le$ 0.05, ** $\le$ 0.01, *** $\le$ 0.001). APOE, apolipoprotein E; NHW, non‐Hispanic White; NHB, non‐Hispanic Black.

3.4. Sensitivity Analyses

Analyses were repeated removing individuals with an APOE‐ε2/ε4 genotype. Given their low frequency, very few participants were removed (Supplementary Table S1; NHW SuperAgers: N = 1,393; NHB SuperAgers: N = 200) and findings were preserved (Supplementary Table S2).

Findings were also preserved when APOE genotyping method was included as a covariate (Supplementary Table S3).

To determine the impact of sample size on the significance of NHB comparisons, we randomly down‐sampled the NHW sample by participant classification and cohort (Supplementary Table S4). In the smaller sample, differences in APOE‐ε2 and ‐ε4 allele frequency between NHW SuperAgers and controls were no longer significant, except for the comparison between NHW SuperAgers and middle‐aged controls (Supplementary Table S5).

4. DISCUSSION

Mechanisms conferring cognitive resilience in oldest‐old age are yet unknown. This is the largest published sample of SuperAgers and the largest study to date of the relationship of APOE‐ε4 and ‐ε2 allele frequency and SuperAger status. Across eight national aging cohorts, we identified 1,623 NHW and NHB SuperAgers with APOE genotyping using longitudinal harmonized cognitive and clinical data. As expected, we found that SuperAgers had a higher frequency of APOE‐ε2 alleles and a lower frequency of APOE‐ε4 alleles compared to individuals with AD dementia. Unlike most previous studies of APOE genotype in NHW SuperAgers, we found significant differences in APOE genotype compared to controls of all ages, including oldest‐old controls. Specifically, the odds of having 1 ε4 allele was about 20% lower in NHW SuperAgers compared to same‐age controls, and the odds of having 1 ε2 allele was about 30% higher in NHW SuperAgers compared to same‐age controls. This finding held when ε2/ε4 carriers were removed from analyses and when APOE genotyping method was added as a covariate. Although NHB SuperAgers also had a higher frequency of APOE‐ε2 alleles and a lower frequency of APOE‐ε4 alleles compared to oldest‐old controls, comparisons did not rise to the level of significance. Sensitivity analyses down‐sampling NHW SuperAgers suggest that sample size likely impacted the significance of comparisons with NHB SuperAgers.

APOE‐ε4 is the strongest genetic risk factor for late‐onset AD, ⁷ and has been shown to be related to increased entorhinal and hippocampal atrophy ³⁶ and amnestic cognitive impairment. ³⁷ , ³⁸ Our data are supportive of these findings, such that AD dementia cases had a higher frequency of APOE‐ε4 alleles compared to controls and SuperAgers. ³⁹ Unlike most studies of genetic resilience in SuperAgers, we found that NHW SuperAgers had a lower frequency of APOE‐ε4 alleles compared to oldest‐old adults with typical memory performance. This was found in two previous studies published from the same cohort (N = 10‐12), ² , ⁸ but was not replicated in subsequent studies with larger samples of SuperAgers. ⁴ , ⁵ , ⁹ , ¹⁰ , ¹¹ , ¹² This finding is relatively unexpected. Across racialized groups, the effect of APOE‐ε4 is most impressive prior to age 70. ⁴⁰ , ⁴¹ Additionally, APOE‐ε4 carriership is related to increased mortality. ⁴² , ⁴³ In line with these studies, we found lower APOE‐ε4 allele frequency in oldest‐old compared to middle‐aged and old controls. Despite NHW oldest‐old controls being older than NHW SuperAgers on average, we found that SuperAgers had a significantly lower frequency of APOE‐ε4 alleles compared to oldest‐old controls, indicating that APOE‐ε4 allele carriership influences memory even in adults who live past age 80. Moreover, dementia incidence increases exponentially with age, ⁴⁴ suggesting that the oldest‐old control group also represents exceptional aging. We found that NHW SuperAgers had a significantly lower frequency of APOE‐ε4 alleles compared to same‐age peers, thus underscoring the utility of the SuperAgers phenotype for uncovering factors conferring resilience in cognitive aging.

The protective APOE‐ε2 allele is related to lower likelihood of late‐onset AD dementia ¹³ , ⁴⁵ and better cognitive performance in older adults even in the presence of AD neuropathology. ⁴⁶ Unlike APOE‐ε4, APOE‐ε2 carriership was previously shown to affect cognition after age 80; more precisely, although oldest‐old APOE‐ε2 carriers were as likely as APOE‐ε4 carriers to meet neuropathologic criteria for AD, they were less likely to be diagnosed with dementia. ⁴⁵ Nonetheless, no previous studies have found a relationship of APOE‐ε2 allele frequency and SuperAger status. ¹¹ , ¹² , ¹³ The present study is the first to find that NHW SuperAgers had a significantly higher frequency of APOE‐ε2 alleles compared to controls. Our results suggest that APOE‐ε2 not only reduces the likelihood of dementia in oldest age but increases the likelihood one will possess optimal memory in oldest age. Future studies are needed to determine whether SuperAgers have similar levels of AD neuropathology compared to AD dementia cases as was previously found in oldest‐old APOE‐ε2 carriers without dementia. ⁴⁵

Genetic factors underlying the superior memory performance of NHB SuperAgers, a critically under‐diagnosed and understudied group, are relatively unknown. ⁶ Previous research suggests that, while NHB individuals have a higher frequency of APOE‐ε4 alleles compared to NHW individuals, APOE‐ε4 carriership is associated with attenuated risk for late‐onset AD, yet similar mortality risk, in NHB individuals compared to NHW individuals. ⁴¹ , ⁴⁷ , ⁴⁸ , ⁴⁹ The effect is likely related to global population ancestry; more specifically, APOE‐ε4 carriership is associated with a greater risk of AD in NHB older adults with decreased global African ancestry or increased global European ancestry. ⁴¹ Similar to APOE‐ε4, NHB individuals have a higher frequency of APOE‐ε2 alleles compared to NHW individuals. ⁴⁷ Unlike APOE‐ε4, researchers did not detect differences in the protective effect of APOE‐ε2 alleles related to global population ancestry. ⁴¹ In fact, a recent study from the MARS cohort found that NHB older adults (ages 65+) with more APOE‐ε2 alleles had slower cognitive decline over a 10‐year study period. ⁵⁰ Additionally, APOE‐ε2 carriership was related to better survival in a sample of NHB and NHW individuals from NACC with and without AD neuropathology. ⁵¹ In the present study, NHB SuperAgers had a significantly lower frequency of APOE‐ε4 alleles compared to cases and younger controls, and a significantly higher frequency of APOE‐ε2 alleles compared only to cases. Importantly, the present study included a substantially smaller sample of NHB SuperAgers compared to NHW SuperAgers, although the sample was still far larger than has been previously reported. Sensitivity analyses suggest that sample size likely impacted the significance of comparisons with NHB SuperAgers. It is also possible that the lack of significant difference in APOE allele carriership between NHB SuperAgers and oldest‐old controls is partially explained by survivorship bias; recent research indicates that NHB adults have a higher probability of survival from ages 70 and 80 to 100 compared to NHW adults. ⁵² Thus, NHB SuperAgers and cognitively normal oldest‐old adults may share environmental and genetic factors, including APOE‐ε2 allele carriership, that support survival and reduced mortality risk in older age, resulting in a lack of significant difference in APOE genotype between these two, very similar, exceptional aging groups. ⁶ Still, more research and targeted recruitment of high‐performing NHB oldest‐old adults is necessary to determine the role of APOE genotype in their sustained optimal memory performance.

The present study has several strengths, including being the largest of its kind to explore the relationship of APOE genotype and optimal memory performance in both NHW and NHB SuperAgers. This is the largest study to date to identify differences in APOE‐ε4 allele frequency based on SuperAger status, and the first study of SuperAgers to find a relationship between APOE‐ε2 allele frequency and SuperAger status. Our study also has several limitations. Across cohorts, participants were highly educated, limiting generalizability to lower‐educated or socioeconomically diverse populations. Additionally, studies have identified other genetic factors that may confer greater AD risk in NHB older adults compared to APOE, including ABCA7. ⁵³ , ⁵⁴ Future studies will need to consider other genetic factors that may also be relevant to exceptional aging in NHB older adults. Moreover, differences in the effect of genetic factors on AD risk have been associated with genetic ancestry, which was only available for a subset of participants in our sample. ⁴¹ Subsequent research leveraging advanced genetic analyses to consider admixture may further our understanding of genetic profiles underlying AD risk.

The present study reveals important information about genetic factors associated with SuperAger status. While significant findings were restricted to NHW comparisons, study results importantly direct our attention to the need for targeted recruitment of high‐performing NHB oldest‐old adults to more definitively determine the role of genetics in sustained optimal memory performance in oldest‐old age across racialized groups. There is an ongoing initiative that intends to answer complex questions about brain aging, resilience, and resistance in a well‐characterized SuperAging cohort through longitudinal, multimodal data collection. ⁵⁵ , ⁵⁶ SuperAging research will contribute tremendously to our understanding of factors conferring cognitive resilience in oldest‐old age, such as preserved regional brain volume, differences in lifestyle factors, and resistance to neuropathology.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest. Author disclosures are available in the supporting information.

CONSENT STATEMENT

All participants provided informed consent in their respective cohort studies.

Supporting information

Supporting Information

Supporting Information

Durant A, Mukherjee S, Lee ML, et al. Evaluating the association of apolipoprotein E genotype and cognitive resilience in SuperAgers. Alzheimer's Dement. 2026;22:e71024. 10.1002/alz.71024