Genetic Risk, Midlife Life's Simple 7, and Incident Dementia in the Atherosclerosis Risk in Communities Study

Abstract

Background and Objectives

Higher scores in Life's Simple 7 (LS7), a metric for cardiovascular and brain health, have been associated with lower risk of dementia. It is uncertain whether this association holds among those with high genetic risk of dementia. Our objective is to evaluate the extent that LS7 may offset dementia risk across the range of genetic risk.

Methods

Participants in the Atherosclerosis Risk in Communities (ARIC) Study were followed from 1987–1989 to 2019. We derived midlife LS7 scores and generated genetic risk scores (GRS) using genome-wide summary statistics of Alzheimer disease, which have been used to study the genetic risk for dementia. Incident dementia was ascertained based on the criteria of the National Institute on Aging–Alzheimer's Association workgroups and Diagnostic and Statistical Manual of Mental Disorders. The associations of the GRS and LS7 with incident dementia were evaluated using Cox regression.

Results

This study included 8,823 European American (EA) and 2,738 African American (AA) participants (mean age at baseline 54 years). We observed 1,603 cases of dementia among EA participants and 631 among AA participants (median follow-up 26.2 years). Higher GRS were associated with higher risk of dementia (EA, hazard ratio [HR] per SD 1.44, 95% CI 1.37, 1.51; AA, HR 1.26, 95% CI 1.16, 1.36). Among EA participants, higher LS7 scores were consistently associated with lower risk of dementia across quintiles of GRS, including the highest quintile (HR per point 0.91, 95% CI 0.87, 0.96). Among AA participants, the associations between LS7 and incident dementia within stratum of GRS had the same direction as among EA participants, although wide CIs and smaller sample sizes limited reliable inferences.

Discussion

Across strata of GRS, higher midlife LS7 scores were associated with lower risk of dementia. Larger sample sizes from diverse populations are needed to obtain more reliable estimates of the effects of modifiable health factors on dementia risk within genetic risk strata in each ancestry group.

Dementia results in progressive loss of cognitive abilities that debilitates daily function and costs 2.55 million disability-adjusted life-years in the United States annually.¹ Late-onset dementia is a complex syndrome determined by many health factors with genetics playing a contributing role. Individuals of European ancestry who are carriers of the APOE ε4 allele are estimated to have 2 to 5 times higher lifetime risk of dementia.² Because no effective treatments are available for dementia,³ modifiable health factors have been gaining increasing attention as potential interventions for the prevention and management of dementia.⁴ The associations between modifiable health factors and incident dementia among those with high genetic risk have been inconsistent.^5,6 Determining whether modifiable health factors have consistent effects across levels of genetic risk might help shape personalized prevention strategies.

The Life's Simple 7 (LS7) score consists of 7 modifiable health factors with higher scores associated with lower risk for dementia or cognitive decline.^7-11 Initially proposed by the American Heart Association (AHA) for maintaining cardiovascular health, this metric has also been recommended for brain health.^12,13 LS7 includes 4 behavioral factors (smoking, body mass index, physical activity, and diet) and 3 biological factors (total cholesterol, blood pressure, and fasting blood glucose). The total score ranges from 0 to 14 with each factor contributing 0 to 2 points.

Late-onset Alzheimer disease (AD) accounts for 60%–80% of dementia cases.¹⁴ Estimates of the genetic risk for AD can inform the genetic risk for dementia. Whereas the APOE ε4 allele confers substantial genetic risk for AD, risk variants in regions outside of APOE have been identified.^15,16 Using results of large-scale genome-wide association studies (GWAS) of AD to generate genetic risk scores (GRS) can enhance the estimation of genetic risk over using the APOE locus alone and has been applied in the studies of the genetic risk for dementia.^6,17,18 We generated GRS using results of large-scale GWAS of AD and derived LS7 scores at midlife among African American (AA) and European American (EA) participants in the Atherosclerosis Risk in Communities (ARIC) study. To assess the extent that LS7 could offset dementia risk across the range of genetic risk, we evaluated the association of LS7 with incident dementia within strata of GRS and APOE genotype.

Methods

Study Population

The ARIC study is a prospective cohort study of 15,972 adults (11,478 self-reported White participants, 4,266 self-reported Black participants, and 48 other races) from 4 communities (Washington County, Maryland; Forsyth County, North Carolina; northwestern suburbs of Minneapolis, Minnesota; and Jackson, Mississippi). The Jackson site enrolled only self-reported Black participants by design. Including the enrollment visit (1987–1989, visit 1), 7 in-person visits have been completed. On the race terms for ARIC participants, we use White and Black based on the self-identified race provided by the participants at study visits. When referring to genetic background, we use ancestral geographical regions (EA and AA) as is commonly done in genetic studies.¹⁹

Among ARIC participants, 11,915 had genetic data that passed quality control (9,044 EA and 2,871 AA), data for LS7 scores, and covariates. The ancestry of each participant was confirmed based on genetic principal component (PC) analysis within each self-reported racial group. Details of genotyping, quality control, and imputation are reported in the eMethods (links.lww.com/WNL/B975). After excluding participants with missing data for LS7 scores (n = 341) and education levels (n = 13), 11,561 participants (2,738 AA and 8,823 EA) were included (eFigure 1, links.lww.com/WNL/B973).

Standard Protocol, Approvals, Registrations, and Patient Consents

This study was approved by the institutional review board of each ARIC site: University of North Carolina at Chapel Hill; Wake Forest University, Winston-Salem, North Carolina; Johns Hopkins University, Baltimore, Maryland; University of Minnesota, Minneapolis; and University of Mississippi Medical Center, Jackson. All participants provided written informed consent at each visit and proxies provided consent for participants who were determined to lack capacity.

Genetic Risk Score Generation, Genotyping, and Imputation

The GWAS of AD used for GRS generation were conducted among AA and EA ancestry separately. Ancestry-specific genetic association studies take advantage of the more homogeneous genetic patterns within 1 ancestry, such as allele frequency and linkage disequilibrium, to identify risk variants, and do not assume that genetic risk differs inherently by ancestry.²⁰ We generated GRS among AA and EA participants separately using ancestry-matched summary statistics of 2 large-scale GWAS of AD (EA: 21,982 cases, 41,944 controls; AA: 2,784 cases, 5,222 controls).^15,16

The GRS for each participant is the weighted sum of the number of risk alleles of each selected single nucleotide polymorphism (SNP) using the effect size of the risk allele as the weight. These scores represent relative measures of genetic risk within an ancestry group. The scores were generated using PRSice-2 using near independent SNPs below a certain p value threshold in the GWAS summary statistics.²¹ The steps were the same among AA and EA participants. First, independent SNPs were selected based on linkage disequilibrium (r² < 0.1) calculated from imputed genetic dosage in the ARIC study guided by the strength of association (p value) in the summary statistics of the GWAS of AD. Next, p value thresholds (from 0.1 to 5 × 10⁻⁸ by an increment of 5 × 10⁻⁷) were used to select the independent SNPs for calculating a set of GRS for each p value threshold. Finally, PRSice-2 selected the p value threshold resulting in the GRS with the strongest association with incident dementia in ARIC. The selected p value thresholds in AA and EA participants were 5.5E-07 and 3.6E-06, respectively (eTable 1, links.lww.com/WNL/B974). These thresholds are in the range of the p value thresholds used in a recent risk prediction model of AD.²² eFigure 2 (links.lww.com/WNL/B973) displays the histograms of the GRS of the 2 ancestry groups. To assess whether the performance of the GRS selected by PRSice-2 might be sensitive to a particular p value threshold, we compared the unadjusted hazard ratio (HR) for dementia of the selected GRS with those generated using the p value thresholds that were 1 interval lower and higher in the ARIC study. Among AA participants, these unadjusted HRs were 99.4% and 98.6% of the HRs of the selected GRS. Among EA participants, these HRs were 99.9% of the HRs of the selected GRS, suggesting that the selected GRS were not sensitive to small changes in the p value threshold (eTable 2, links.lww.com/WNL/B974). The ARIC genotype data were imputed dosage of SNPs with minor allele frequency ≥1% and imputation quality ≥0.8. Details of genotyping, imputation, and generation of genetic PCs are reported in the eMethods (links.lww.com/WNL/B975).

Generation of LS7 Scores

LS7 scores were derived according to the AHA definitions (eTable 3, links.lww.com/WNL/B974).¹² Details have been reported previously.⁹ All data for deriving the LS7 scores were collected at midlife during the enrollment visit (1987–1989, visit 1) using standardized protocols. Diet was assessed by a modified 66-item Harvard food frequency questionnaire. Physical activity was assessed by the Baecke questionnaire.²³ Smoking status was self-reported. Body mass index was calculated using height and weight measured at the study visit. Measures of systolic blood pressure (SBP) and diastolic blood pressure were the averages of the second and third readings. Glucose and total cholesterol levels were measured from peripheral blood. The use of antihypertensive, cholesterol-lowering, or glucose-lowering medications within the past 2 weeks were based on self-report or inspection of medication bottles during the study visit.

All-Cause Dementia Ascertainment

Participants were followed from enrollment (1987–1989, visit 1) up to the end of visit 7 (December 31, 2019). The methods for dementia ascertainment in the ARIC study have been reported previously.²⁴ Briefly, dementia status was determined by an expert committee, which included physicians and neuropsychologists, based on the criteria from the National Institute on Aging–Alzheimer's Association (NIA-AA) workgroups and the DSM-5. The data included detailed cognitive and functional assessments (eTable 4, links.lww.com/WNL/B974) collected at ARIC visits (visit 5, 2011–2013; visit 6, 2016–2018, and visit 7, 2018–2019) and cognitive tests at visits 2 (1990–1992) and 4 (1996–1998). The Clinical Dementia Rating interview and the Functional Assessment Questionnaire were used in in-person and telephone interviews of participants and informants who could not attend the clinical visit. Participants who did not attend visit 5 were administered the modified Telephone Interview for Cognitive Status (TICS).^25,26 TICS scores were education-adjusted.²⁴ After visit 5, the 6-Item Screener was offered annually to all participants, and the Alzheimer's Dementia 8-Item Informant Questionnaire was administered to informants by phone.²⁷ Dementia status was also ascertained using ICD-9 dementia codes at hospitalization discharge and on death certificates obtained by cohort surveillance.

Dementia date was first set as the earliest of the hospitalization date with an ICD-9 code for dementia, death date if a dementia code was listed on the death certificate, date of telephone communication with the participant or proxy with indication of dementia, or date of the first visit when dementia was indicated. Dementia onset date ascertained from informant interviews, hospitalization, and death certificate was subtracted by 6 months to account for the expected lag in the reporting of the event. Participants who were classified as not having dementia were censored at the last study contact date when there was no indication of dementia or the date of death obtained by cohort surveillance.

Ascertainment of Other Variables

Race and education levels (< high school, high school graduate or equivalent or vocational school, and at least some college, graduate, or professional school) were self-reported. Prevalent diabetes mellitus was defined as having a fasting glucose level ≥126 mg/dL, nonfasting glucose level ≥200 mg/dL, self-reported diabetes medication use, or self-reported physician diagnosis of diabetes. Prevalent hypertension was defined as SBP ≥140 mm Hg, diastolic blood pressure ≥90 mm Hg, or self-reported use of antihypertension medications. Prevalent cardiovascular disease (CVD) was defined as prevalent stroke based on self-report and coronary heart disease and heart failure based on chart review by expert committees.^28,29 Genotyping of the 2 SNPs (rs429358, rs7412) that define the 6 APOE genotypes was performed using the TaqMan system (Thermo Fisher Scientific).

Association Analysis

Given that GRS was a relative measure of risk within each ancestry group, all analyses were stratified by ancestry. The GRS and the LS7 scores were analyzed as continuous variables and categorical variables. The GRS were categorized by quintiles among EA participants and tertiles among AA participants due to a smaller sample size. As descriptive statistics, we calculated the crude incidence of dementia per unit of the LS7 scores by strata of the GRS in both ancestry groups. The LS7 scores were categorized into low (0–6), intermediate (7–8), and high (9–14). These cut points were selected such that there were at least 100 incident dementia cases in each category in both ancestry groups and were also used by the multi-ethnic Reasons for Geographic and Racial Differences in Stroke (REGARDS) study for the analysis of LS7 and incident cognitive impairment.¹¹

Baseline characteristics by ancestry group were compared across strata of GRS using the χ² test for categorical variables, 1-way analysis of variance for nonskewed continuous variables, and Kruskal-Wallis test for skewed continuous variables. To investigate the effect of LS7 for dementia across levels of genetic risk, we used cause-specific Cox regression models to assess the association of the GRS and LS7 scores with incident dementia.³⁰ We assessed the proportional hazards assumption by testing for the time-dependent term of the predictor of interest. p Value for trend was obtained by using the rank of a categorical variable as continuous variable. Our primary analysis has 3 components (eFigure 3, links.lww.com/WNL/B973). First, we evaluated the main effect of the GRS and LS7 on incident dementia by ancestry. Second, to address the main study question of the effects of LS7 across genetic risk strata, we evaluated the association of the LS7 scores with incident dementia with each GRS strata in each ancestry group. Third, given that the APOE ε4 allele has been estimated to account for the majority of the known genetic risk of late-onset AD among EA individuals,^31,32 we evaluated the association of LS7 with incident dementia by APOE genotype among EA participants. The genotypes were categorized into 4 groups: ε2 carriers without ε4 (ε2/2, ε2/3), ε3 homozygotes (ε3/3), ε4 heterozygotes (ε2/4, ε3/4), and ε4 homozygotes (ε4/4). This analysis excluded 357 EA participants without ε4 genotype data, resulting in a sample size of 8,466, and was not conducted among AA participants. Compared with EA individuals, AA individuals have more complex haplotype structure involving variable length short repeats in the APOE region that may modify the association between the APOE alleles and dementia. Data required for generating these haplotype structures were not available in the ARIC study. In these 3 primary analyses, the covariates included age, sex, study center, and education levels. In models including the GRS as independent variable, the covariates also included 10 genetic PCs to control for potential confounding due to subpopulation stratification.

We conducted 2 secondary analyses. First, to evaluate the relevance of the genetic signals outside of the APOE region for incident dementia, we evaluated the association between GRS and incident dementia in addition to controlling for the APOE ε4 genotype. Second, for all analyses involving LS7 as a predictor, we conducted sensitivity analyses by adding prevalent CVD as a covariate, given that LS7 was initially proposed as a metric for maintaining cardiovascular health,¹² and CVD has been associated with incident dementia.³³ This sensitivity analysis excluded 270 participants with missing values for prevalent CVD, resulting in a sample size of 2,682 among AA participants and 8,609 among EA participants. Descriptive statistics and Cox regression analyses were conducted using R version 4.0.

Data Availability

ARIC data from visit 1 to visit 5 are available through the Biologic Specimen and Data Repository Information Coordinating Center (BioLINCC). Data that are not yet available through BioLINCC are available upon request through the ARIC Coordinating Center at the University of North Carolina.

The AD GWAS summary statistics from EA ancestry are available for download.³⁴ The AD GWAS summary statistics from AA ancestry are available at the National Institute on Aging Genetics of Alzheimer's Disease Data Storage Site (NIAGADS).

Results

At baseline, the mean age was 54.3 years and 53.5 years among EA and AA participants, respectively. The proportion of men was 46.8% among EA and 36.9% among AA participants. The mean LS7 score was 8.3 among EA and 6.6 among AA participants. The proportion of APOE ε4 carriers was 27.9% among EA and 40.4% among AA participants (Table 1). Within each ancestry group, the baseline characteristics were largely similar across GRS strata, except that the participants with higher GRS had higher total cholesterol levels and a higher proportion of APOE ε4 carriers (eTables 5 and 6, links.lww.com/WNL/B974).

Table 1.

Baseline Characteristics

Over a median follow-up of 26.2 years, 1,603 incident dementia events were observed in EA (18.2%) and 631 in AA (23.0%) participants. After adjusting for age, sex, study center, education levels, and the first 10 genetic PCs, each SD higher in GRS was associated with higher risk of incident dementia (EA, HR 1.44, 95% CI 1.37, 1.51; AA, HR 1.26, 95% CI 1.16, 1.36; Table 2). When analyzed by quintile of the GRS, compared with quintile 1, quintile 5 had 1.5- to 2.7-fold higher risk for incident dementia (EA, HR 2.70, 95% CI 2.31, 3.15; AA, HR 1.55, 95% CI 1.27, 1.88). With the addition of the APOE ε4 genotype as a covariate, the GRS remained associated with higher risk for dementia, although not significant among AA participants (EA, HR 1.19, 95% CI 1.09, 1.30; AA, HR 1.03, 95% CI 0.89, 1.19).

Table 2.

Association Between Genetic Risk Scores and Incident Dementia

Higher LS7 scores were associated with lower risk of dementia in both ancestry groups (EA, HR 0.90 per point, 95% CI 0.88, 0.91; AA, HR 0.95 per point, 95% CI 0.91, 0.99; Table 3). When the behavior and biological components were analyzed separately, the associations were similar to the overall LS7 score among EA participants but more variable among AA participants. When analyzed as a 3-category variable, higher LS7 showed a graded association with lower risk of dementia in both ancestry groups. Among EA participants, compared with the low category, the intermediate and high categories were associated with 30% and 43% lower risk for dementia, respectively (intermediate, HR 0.70, 95% CI 0.62, 0.81; high, HR 0.57, 95% CI 0.50, 0.64). Among AA participants, the intermediate and high categories were associated with 6% and 17% lower risk for dementia, respectively (intermediate, HR 0.94, 95% CI 0.79, 1.12; high, HR 0.83, 95% CI 0.66, 1.04). After controlling for prevalent CVD, the association of LS7 with incident dementia remained similar (eTable 7, links.lww.com/WNL/B974). The associations of GRS and LS7 with incident dementia were independent (p for interaction in both ancestry groups >0.2).

Table 3.

Association Between Life's Simple 7 and Incident Dementia by Ancestry

Within quintiles of the GRS, higher LS7 scores were consistently associated with lower risk of dementia among EA participants. For example, in quintile 5 of the GRS, each point higher in LS7 scores was associated with 8% lower hazard for dementia (HR 0.92 per point, 95% CI 0.88 to 0.96; Figure 1). The HR estimates in other quintiles of the GRS were similar. Among AA participants, these associations were largely in the same direction as among EA but with wide CIs, likely due to smaller sample size (Figure 1). These associations remained similar after controlling for prevalent CVD (eTable 8, links.lww.com/WNL/B974). eTables 9 and 10 (links.lww.com/WNL/B974) report the sample size and crude incidence of dementia per each unit of LS7 score within each stratum of the GRS.

Figure 1. Association Between Life's Simple 7 Scores and Incident Dementia Overall and by Stratum of the Genetic Risk Scores.

Hazard ratios were estimated per point increment of Life's Simple 7. Covariates: age, sex, study center, and education levels. Among African American (AA) participants, the genetic risk scores were categorized into tertiles due to the smaller sample size, and the smaller number of single nucleotide polymorphisms (n = 9; see Methods) selected for the genetic risk score led to a slightly rugged tertile distribution. EA = European American.

When LS7 scores were categorized into 3 levels, the associations of this LS7 categorical variable within the strata of the GRS were largely consistent with their associations as continuous variable. Among EA participants, the highest LS7 category was significantly associated with lower risk of dementia across all quintiles of the GRS (Figure 2, eTable 11, links.lww.com/WNL/B974). Among the smaller sample of AA participants, these associations had wide CIs, suggesting less certainty (Figure 2, eTable 12). These associations remained similar in both ancestry groups after controlling for prevalent CVD (eTables 11 and 12).

Figure 2. Association Between 3 Categories of Life's Simple 7 Scores and Incident Dementia by Stratum of Genetic Risk Scores.

Covariates: age, sex, study center, and education levels. AA = African American; EA = European American.

Among EA participants, within the APOE genotype groups, higher LS7 scores were largely associated with lower risk of dementia, although among the smaller sample of ε4 homozygotes, the results had wide CIs and were nonsignificant (Table 4, eTable 13, links.lww.com/WNL/B974).

Table 4.

Association Between Life's Simple 7 and Incident Dementia Among European American Participants by APOE Genotype (n = 8,466)

Discussion

In this large community-based cohort of AA and EA participants, across the GRS strata, higher LS7 scores were consistently associated with lower risk of dementia among EA participants. These associations among AA participants were largely in the same direction as those among EA participants but had wide CIs, likely due to smaller sample size.

Higher LS7 scores have shown protective associations for dementia among cohorts of European ancestry and for cognitive decline among multiethnic cohorts.^8,9,11,35,36 The ARIC study has previously reported the protective association of LS7 scores for cognitive decline, which is useful for monitoring the development of dementia and its progression.⁹ The diagnosis of dementia combines cognitive and functional criteria and is commonly used to assess health burdens. Inconsistent results on the associations between the behavior component of LS7 and incident dementia have been reported. Our result on the protective association between the behavioral component of LS7 and incident dementia among EA participants after a median follow-up of 26 years agrees with the findings from the Whitehall study, which reported a similar protective association after 25 years of follow-up.⁷ This protective association was not observed among participants in the UK Biobank after a median follow-up of 9 years.³⁷ Because dementia can take decades to develop,³⁸ the early preclinical phase of dementia could have effects on behaviors. Taken together, the findings from these 3 studies suggest the behavioral component might have long-term effects when these behaviors are practiced in midlife, or the one-time measure of these behaviors at midlife might reflect long-term practice.

Inconsistent results have been published on the associations of favorable lifestyle on dementia risk across genetic risk strata. In the Rotterdam Study with a median follow-up of 14 years, a favorable lifestyle (absence of smoking, depression, and diabetes, regular physical activity, avoidance of social isolation, and a healthy diet) had a protective association with incident dementia only among noncarriers of the APOE ε4 allele and not among ε4 carriers (ε2/4, ε3/4, or ε4/4).⁵ In the UK Biobank cohort with a median follow-up of 8 years, a favorable lifestyle (no current smoking, regular physical activity, healthy diet, and moderate alcohol consumption) was associated with lower risk of dementia across genetic risk strata defined by quintile of GRS.⁶ In the Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability (FINGER), the protective effects of multidomain intervention (diet, exercise, cognitive training, and vascular risk management) on cognitive outcomes did not differ by APOE ε4 carrier status after 2 years of follow-up.³⁹ The UK Biobank study did not analyze by APOE genotypes.⁶ The Rotterdam Study and FINGER combined all APOE ε4 carriers and did not analyze APOE ε4 homozygotes separately. Among the APOE ε4 genotypes, ε4 homozygotes are known to have the highest risk for dementia^2,40: ∼5 times higher risk among ε4 homozygotes and ∼2 times higher risk among ε4 heterozygotes compared with noncarriers of the ε4 allele.² We evaluated the association of LS7 scores within strata of APOE genotypes among EA participants. Our results suggest that favorable lifestyle and vascular factors were protective against dementia, including for ε4 heterozygotes. Among ε4 homozygotes, the association of LS7 had a wide CI, suggesting that more research is needed on the effect of favorable lifestyle and vascular factors for dementia in this high-risk group.

Our study has several strengths. First, the ARIC cohort is a well-characterized community-based cohort. In addition to EA participants, we included AA participants, among whom studies on the interplay of genetic risk and health factors and their effect on cognitive outcomes are very limited.³⁶ Second, our follow-up time spanned ∼30 years (median of 26 years). Because dementia takes decades to develop,³⁸ the prospective associations from this study are not likely to be affected by preclinical effects of dementia. Third, the incident dementia ascertainment was rigorous. Some limitations warrant mention. First, variables for calculating LS7 scores were measured at a single time point. At later study visits, some health factors used in the LS7 were unavailable. We were not able to assess the effects of the longitudinal change of LS7 on incident dementia. Second, whereas the GRS were generated based on the association statistics from large-scale GWAS of AD,^15,16 AD diagnosis incorporating biomarkers is not available in ARIC. We used all-cause dementia as our outcome. Given that AD is the most common form of dementia, and patients with AD often have mixed pathologies,^41,42 the GWAS of AD likely captured a large proportion of the genetic risk of dementia. Nevertheless, genetic risk specific for other forms of dementia could have been missed, resulting in an underestimation of the association of genetic risk with dementia, particularly among individuals of AA ancestry, in whom mixed pathology is more common.⁴² Third, the GWAS of AD used an additive genetic model and did not capture the genetic effects of other genetic models, particularly among AA individuals, if there are indeed substantial interacting genetic effects at the APOE region.⁴³ Fourth, the sample size of the GWAS in individuals of AA ancestry was much smaller than the GWAS in EA individuals and thus detected fewer significant signals. This could also contribute to underestimation of genetic risk among AA individuals. Fifth, it has been widely recognized that racism contributes to poorer health outcomes among AA individuals.^44-45 We observed lower LS7 scores and higher risk of incident dementia among AA participants, consistent with results from other studies.^14,47 Given that the ARIC study has limited data on participant experience of racism, we did not evaluate the relationship of racism with LS7 and incident dementia. Finally, in ARIC, there were fewer AA participants than EA participants and were largely recruited from one site. The wide CIs of the estimates among AA participants limit their interpretation and underscore the need for studies of larger sample size among AA individuals.

Higher LS7 scores, a metric for maintaining cardiovascular and brain health, are largely associated with lower risk of incident dementia across strata of genetic risk, supporting the use of LS7 for maintaining brain health and offsetting genetic risk. More research with larger study populations are needed to examine the effect of LS7 for dementia prevention among APOE ε4 homozygotes and AA individuals across genetic risk strata.

Glossary

AA

African American

AD

Alzheimer disease

AHA

American Heart Association

ARIC

Atherosclerosis Risk in Communities

BioLINCC

Biologic Specimen and Data Repository Information Coordinating Center

CVD

cardiovascular disease

DSM-5

Diagnostic and Statistical Manual of Mental Disorders, 5th Edition

EA

European American

FINGER

Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability

GRS

genetic risk scores

GWAS

genome-wide association studies

HR

hazard ratio

ICD-9

International Classification of Diseases–9

LS7

Life's Simple 7

PC

principal component

SBP

systolic blood pressure

SNP

single nucleotide polymorphism

TICS

Telephone Interview for Cognitive Status

Appendix. Authors

Study Funding

The Atherosclerosis Risk in Communities study has been funded in whole or in part with federal funds from the National Heart, Lung, and Blood Institute (NHLBI), NIH, Department of Health and Human Services contract numbers HHSN268201700001I, HHSN268201700002I, HHSN268201700003I, HHSN268201700004I, and HHSN268201700005I (R01HL087641, R01HL059367, and R01HL086694); National Human Genome Research Institute contract U01HG004402; and NIH contract HHSN268200625226C. Neurocognitive data were collected by U01 2U01HL096812, 2U01HL096814, 2U01HL096899, 2U01HL096902, and 2U01HL096917 from the NIH (NHLBI, National Institute of Neurological Disorders and Stroke, National Institute on Aging, and National Institute on Deafness and Other Communication Disorders). Infrastructure was partly supported by grant UL1RR025005, a component of the NIH and NIH Roadmap for Medical Research. This research was supported (in part) by the Intramural Research Program of the National Institute of Neurological Disorders and Stroke (NINDS), NIH.

Disclosure

The authors report no disclosures relevant to the manuscript. Go to Neurology.org/N for full disclosures.