Integrating contextual determinants and polygenic risk to examine dementia and late-life cognition in the Multi-Ethnic Study of Atherosclerosis.

Integrating contextual determinants and polygenic risk to examine dementia and late‐life cognition in the Multi‐Ethnic Study of Atherosclerosis - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Official websites use .gov A .gov website belongs to an official government organization in the United States. Secure .gov websites use HTTPS A lock ( Lock Locked padlock icon ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites. Search Log in Dashboard Publications Account settings Log out Search… Search NCBI Primary site navigation Search Logged in as: Dashboard Publications Account settings Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide PERMALINK Copy As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice Alzheimers Dement . 2026 Apr 16;22(4):e71384. doi: 10.1002/alz.71384 Search in PMC Search in PubMed View in NLM Catalog Add to search Integrating contextual determinants and polygenic risk to examine dementia and late‐life cognition in the Multi‐Ethnic Study of Atherosclerosis Diane Xue Diane Xue 1 Department of Genetics, University of Pennsylvania, Philadelphia, Pennsylvania, USA Find articles by Diane Xue 1, ✉ , Jana A Hirsch Jana A Hirsch 2 Urban Health Collaborative and Department of Epidemiology and Biostatistics, Dornsife School of Public Health, Drexel University, Philadelphia, Pennsylvania, USA Find articles by Jana A Hirsch 2 , Lilah Besser Lilah Besser 3 Comprehensive Center for Brain Health, Department of Neurology, University of Miami, Boca Raton, Florida, USA Find articles by Lilah Besser 3 , Timothy M Hughes Timothy M Hughes 4 Department of Internal Medicine, Section on Gerontology and Geriatric Medicine, Wake Forest University School of Medicine, Winston‐Salem, North Carolina, USA Find articles by Timothy M Hughes 4 , Elizabeth E Blue Elizabeth E Blue 5 Division of Medical Genetics, Department of Medicine, University of Washington, Seattle, Washington, USA 6 Institute for Public Health Genetics, University of Washington, Seattle, Washington, USA 7 Brotman Baty Institute, University of Washington, Seattle, Washington, USA Find articles by Elizabeth E Blue 5, 6, 7 , Annette L Fitzpatrick Annette L Fitzpatrick 8 Department of Family Medicine, University of Washington, Seattle, Washington, USA 9 Department of Epidemiology, University of Washington, Seattle, Washington, USA Find articles by Annette L Fitzpatrick 8, 9 , Alison E Fohner Alison E Fohner 9 Department of Epidemiology, University of Washington, Seattle, Washington, USA Find articles by Alison E Fohner 9 Author information Article notes Copyright and License information 1 Department of Genetics, University of Pennsylvania, Philadelphia, Pennsylvania, USA 2 Urban Health Collaborative and Department of Epidemiology and Biostatistics, Dornsife School of Public Health, Drexel University, Philadelphia, Pennsylvania, USA 3 Comprehensive Center for Brain Health, Department of Neurology, University of Miami, Boca Raton, Florida, USA 4 Department of Internal Medicine, Section on Gerontology and Geriatric Medicine, Wake Forest University School of Medicine, Winston‐Salem, North Carolina, USA 5 Division of Medical Genetics, Department of Medicine, University of Washington, Seattle, Washington, USA 6 Institute for Public Health Genetics, University of Washington, Seattle, Washington, USA 7 Brotman Baty Institute, University of Washington, Seattle, Washington, USA 8 Department of Family Medicine, University of Washington, Seattle, Washington, USA 9 Department of Epidemiology, University of Washington, Seattle, Washington, USA * Correspondence , Diane Xue, 3700 Hamilton Walk, Suite A301, Philadelphia, PA, 19104, USA. Email: diane.xue@pennmedicine.upenn.edu ✉ Corresponding author. Revised 2026 Jan 20; Received 2025 Oct 2; Accepted 2026 Mar 19; Collection date 2026 Apr. © 2026 The Author(s). Alzheimer's & Dementia published by Wiley Periodicals LLC on behalf of Alzheimer's Association. This is an open access article under the terms of the http://creativecommons.org/licenses/by-nc-nd/4.0/ License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non‐commercial and no modifications or adaptations are made. PMC Copyright notice PMCID: PMC13084535  PMID: 41988869 Abstract INTRODUCTION Alzheimer's disease and related dementias are influenced by genetic and environmental risk factors. We investigated the relationship between contextual exposures and cognitive outcomes, independent of and in interaction with polygenic risk. METHODS Using the Multi‐Ethnic Study of Atherosclerosis ( N = 5687), we assessed the associations of contextual determinants representing the social, chemical, and built environment with incident dementia and late‐life cognition using proportional hazards regression and generalized estimating equation models, then evaluated their joint effects stratified by genetic risk via Bayesian kernel machine regression. RESULTS Neighborhood disadvantage was associated with higher dementia risk and poorer cognitive scores after adjusting for genetic risk and other individual‐level covariates. Joint analysis of all contextual determinants indicated that more deleterious mixtures of contextual determinants are associated with lower late‐life cognition among apolipoprotein E ɛ4 non‐carriers with intermediate polygenic risk. DISCUSSION Contextual determinants are associated with dementia and late‐life cognition after adjusting for age, sex, education, and genetic risk. Keywords: Alzheimer's disease, gene by environment, social determinants of health, social exposome Highlights Neighborhood Deprivation Index (NDI) is associated with higher dementia risk and poorer cognitive scores after controlling for individual‐level covariates including sex, age, income, and genetic risk. Stratified analyses reveal the association between NDI and dementia is only significant among apolipoprotein E ɛ4 carriers. Bayesian kernel machine regression was used as an exploratory approach to evaluate the joint effects of multiple neighborhood‐level exposures on cognition. 1. BACKGROUND Alzheimer's disease (AD) and related dementias are a complex set of multifactorial disorders impacting cognitive and functional abilities in late life. 1 Delaying the progression of dementia or cognitive decline is crucial to alleviating the burden on patients, caregivers, and health‐care systems. A simulation study found that a 5‐year delay in age‐at‐onset of dementia would lower the prevalence of dementia in 2050 by 41% 2 . While recent advances in drug development hold promise, population‐level prevention strategies that address modifiable risk factors and delay symptom onset are also critical for impacting the burden of dementia. The 2024 Lancet Commission on dementia describes 14 modifiable risk factors for dementia prevention that span the life course and spectrum of actionability 3 . At one end of the spectrum are individual factors such as hearing loss, physical inactivity, and hypertension—conditions that can be directly addressed through personal health behaviors or medical treatment. At the other end are contextual factors—environmental or social determinants like air pollution, which require policy‐level interventions to address. The commission suggests that nearly half of dementia cases could be prevented by targeting these modifiable risk factors. At the same time, AD is highly heritable 4 , 5 . Modifiable determinants must also be investigated alongside genetic risk for a more complete understanding of disease etiology. Most studies of modifiable risk factors that include genetic information have only focused on apolipoprotein E ( APOE ) genotype 6 , 7 , 8 , 9 , 10 . The APOE ɛ4 allele is the strongest genetic risk factor for AD, but less than half of all patients carry a copy of the APOE ɛ4 allele 11 , and the effect of APOE ɛ4 may differ across populations due to local ancestry effects 12 , 13 . Beyond APOE ɛ4, dozens of disease‐associated loci have been identified from genome‐wide association studies (GWASs) 14 . Therefore, our understanding of modifiable risk factors in the context of genomic risk is incomplete. Using polygenic risk scores (PRSs) allows us to consider genetic risk among individuals who do not carry the APOE ɛ4 allele. PRSs summarize the overall genetic risk captured by GWAS signals. By aggregating the effects of single nucleotide polymorphisms (SNPs), PRSs provide a weighted genetic risk score for each individual 15 . The integration of genetic data with contextual factors remains underexplored, but the use of PRSs enables studies of gene–environment interactions that are more feasible than investigating individual SNP–environment interaction effects. Prior work using the UK Biobank has shown that the Townsend Deprivation Index is associated with incident dementia across levels of polygenic risk in a European ancestry sample 16 . A PRS‐based gene–environment approach has also been used to examine other complex traits including coronary heart disease and colorectal cancer 17 , 18 . Integrating PRSs and environmental data can reveal environmental factors that have more substantial effects within certain genetic subgroups—effects that may be masked or attenuated when analyzing the sample as a whole. In this study, we focus our analysis of modifiable risk factors on the role of meso‐ or neighborhood‐level contextual determinants of dementia incidence and cognition that are actionable at the population level. Rather than focusing on individual activity levels or diet, for example, we investigate the role of proximity to recreational space and favorable food markets. This approach responds to recent calls for more empirical evidence on neighborhood‐level determinants of dementia, including the food environment 19 . In addition to adjusting for APOE ɛ4 carrier status, we calculated PRSs that summarize the genetic effects of common AD‐associated variants 20 . We hypothesize that contextual determinants may have differing effects on dementia across groups differing in underlying genetic risk. This study is one of the first to examine neighborhood‐level modifiable risk factors of dementia in the context of genome‐wide genetic risk beyond APOE . 2. METHODS 2.1. Study population Our sample is derived from the Multi‐Ethnic Study of Atherosclerosis (MESA), a prospective cohort study that enrolled 6814 adults aged 45 to 84 years at baseline (2000–2002) 21 . Participants were recruited at one of six sites across the United States: Johns Hopkins University, Baltimore, Maryland; Northwestern University, Chicago, Illinois; Wake Forest University, Winston Salem, North Carolina; University of California Los Angeles, Los Angeles, California; Columbia University, New York, New York; and University of Minnesota, St. Paul, Minnesota. Recruitment was designed to target four racial/ethnic groups (Black/African American, Chinese, Hispanic/Latino, and White) at approximately equal sex ratios. Racial/ethnic classification was based on self‐identification. Informed consent was obtained from each participant at baseline and at each subsequent exam. Approval was received at each site from the local institutional review board for each examination. Participants missing dementia status, genotyping data, or information for every exposure variable were removed from the analysis. 2.2. Outcome assessment 2.2.1. Dementia diagnosis MESA participants were followed up every 9 to 12 months from baseline through 2018 to identify interim hospital admissions and deaths. Self‐reported hospitalizations were investigated by trained staff and data, including International Classification of Disease (ICD)‐9 and ICD‐10 discharge codes, were abstracted for each hospital stay. “Possible” clinical dementia cases were determined based on the following ICD codes: ICD‐9: 290, 294, 331.0, 331.1, 331.2, 331.82, 331.83, 331.9, 438.0, and 780.93; ICD‐10: F00, F01, F03, F04, G30, G31 (excluding G31.2), I69.91, and R41. Incident dementia was ascertained based on date of initial hospitalization and/or death. A physician blinded to the ICD codes validated diagnoses by analyzing text from electronic health records for language that pointed to significant cognitive decline and dementia 22 . This dementia outcome has been described in greater detail and used in previous MESA publications 23 , 24 . As part of the MESA MIND ancillary study, a subset of MESA participants completed detailed cognitive testing with the Uniform Data Set Neuropsychological Battery version 3 at Exam 6 (2016–2019), MESA MIND Exam A (2019–2021), and MESA MIND Exam B (2022–2024). Based on the cognitive scores, physical examination, and clinical data, participants were adjudicated by a consensus adjudication committee. The following cognitive status categories were used: non‐impaired, mild cognitive impairment, probable dementia, or cannot classify. For sensitivity analyses using binary dementia status, we defined cases as those who either met the criteria for possible dementia based on ICD codes from hospitalization or death records or those who were adjudicated to have probable dementia at Exam 6, MESA MIND Exam A, or MESA MIND Exam B. In cases of discrepancies between the ICD code and cognitive adjudication, the cognitive adjudication was used. The overlap between ICD‐based dementia classification and adjudicated cognitive status among MESA MIND participants is shown in Table S1 in supporting information. RESEARCH IN CONTEXT Systematic review : Previous studies have typically examined genetic variation and the effects of social determinants of health on dementia separately. Few studies have assessed the interactions between polygenic risk and neighborhood features on dementia and cognition. Interpretation : We found that greater neighborhood disadvantage is associated with both higher risk of incident dementia and lower cognitive scores, after adjusting for genetic risk and demographic features including income, race/ethnicity, age, and sex. When jointly modeling all exposures that capture the social, chemical, and built environment, we observed that the mixture of exposures had the strongest effect in the intermediate genetic risk group, with nitrogen dioxide emerging as a driving feature of the mixture. Future directions : Future studies will focus on testing causal pathways and uncovering the biological mechanisms underlying these associations. We will extend analyses to longitudinal cognitive decline, while incorporating endophenotypes, biomarkers, and neuropathological data will help clarify the biological mechanisms. 2.2.2. Global cognition The Cognitive Abilities Screening Instrument (CASI) score was measured at MESA Exam 5; 4475 MESA participants completed the CASI exam. The CASI has a score range of 0 to 100 and provides a global cognitive assessment covering the following domains: attention, concentration, orientation, short‐term memory, long‐term memory, language abilities, visual construction, list‐generating fluence, abstraction, and judgement 25 . 2.3. Exposure assessment We selected seven contextual determinants to capture the social environment, chemical environment (traffic‐ and fossil fuel–related pollution), and built environment (proximity to resources and facilities). Descriptive statistics for the exposures are included in Table S2 in supporting information. 2.3.1. Neighborhood disadvantage Neighborhood social and economic conditions are captured by the MESA area‐level Neighborhood Deprivation Index (NDI). NDI was available at the US Census Bureau tract level and was based on 2010 to 2011 American Community Survey data. The NDI used here was previously derived from a principal components analysis based on seven neighborhood characteristics (created from MESA Neighborhood Study): 26 (1) percent of neighborhood residents with a managerial occupation, (2) percent of residents with a bachelor's degree, (3) percent of residents with a high school degree, (4) percent of residents with an annual household income > $50,000, (5) median household income, (6) percent rental income, and (7) median home value. NDI data are available for MESA participants from Exams 1 through 5. Higher NDI should be interpreted as greater disadvantage. 2.3.2. Ambient air pollutants We assessed the role of long‐term exposure levels to nitrogen dioxide (NO 2 ) and particulate matter < 2.5 µm in aerodynamic diameter (PM 2.5 ). Funded by the ancillary MESA Air Study, measurement of long‐term individual‐level exposure to ambient air pollutants has been previously described 27 . Briefly, MESA Air used national universal kriging models to estimate ambient concentrations of PM 2.5 and NO 2 in 2‐week intervals, localized to each participant's residential address. The concentrations used here are 3‐year averages of the 2‐week measurements spanning from 1999 to 2002. All pollutant measures were converted to an interquartile range (IQR) scale. 2.3.3. Built environment features The built environment measures were chosen to represent the food and physical activity environment for each participant. Classification of business data for food stores and recreational facilities are derived from the National Establishment Time Series (NETS) data for all zip codes within a 5‐mile buffer of MESA address of record. Complete details of the classifications are described elsewhere 28 . The data are linked to annual MESA residential locations provided by participants and available for Exams 1 through 5. For each of the following built environment variables, we used the Euclidean distance to measure proximity to the nearest establishment: favorable food stores, fast‐food chains, alcohol outlets (liquor stores and on‐site drinking places), and physical activity facilities (including indoor and outdoor recreational facilities). Euclidean distance was used rather than density/number of establishments within a specified mile buffer to maximize sample size for Bayesian kernel machine regression (BKMR), which only retains individuals with unique features in the mixture matrix. The distances were log2‐transformed due to the skewed distributions of the raw values. A 1‐unit increase in log2‐transformed variables represents a 2‐fold increase, allowing for greater spread in lower values. 2.3.4. Genetic risk We used two parameters to model genetic risk: APOE ɛ4 carrier status and a PRS excluding the APOE region. PRS calculation for MESA participants have been described previously 29 . Briefly, the PRS is derived from summary statistics from the largest GWAS of clinically adjudicated AD cases and controls (the International Genomics of Alzheimer's Project Stage I results 20 ). PLINK 1.9 was used to calculate the PRS using clumping and thresholding with a genome‐wide significant p value threshold (5E‐08) and r‐squared < 0.01. The PRS was adjusted for population structure, captured by the first three principal components, resulting in a residualized and standardized PRS. The genetic variants included in the PRS can be found in Table S3 in supporting information. 2.3.5. Covariates All models were adjusted for age, sex, site, income, education, smoking, marital status, and self‐reported race/ethnicity at baseline. While NDI captures neighborhood‐level information on education, occupation, and household incomes, these covariates represent individual‐level measures. Models of ambient air pollutants and built environment features were further adjusted for NDI because of its influence on these factors and direct association with dementia and cognition, independent of the pathways involving pollution or built environment characteristics. 2.4. Statistical analysis 2.4.1. Linear models to examine independent effects of each contextual determinant To examine the effects of contextual determinants on both incident dementia and cognition, we used two types of regression models: Cox proportional hazards models for incident dementia and linear regression models with generalized estimating equations (GEE) for cognition. The Cox proportional hazards models estimate the hazard ratio for exposures at baseline on time‐to‐incident dementia. The GEE models allow for adjustments for census tract–level clustering and repeated measures for MESA participants across five time points. For both incident dementia and late‐life cognition, we modeled each contextual exposure individually, adjusting for covariates, APOE ɛ4 carrier status, and PRS. For transparency and additional context of our findings, we referenced a Bonferroni‐corrected significance threshold for seven contextual determinants of 0.007 30 . Next, we examined linear and non‐linear interactions between the contextual exposures and genetic risk. For each of the contextual determinants, we fit models stratified by genetic risk group: (1) Group 1 is carriers of the APOE ɛ4 allele; (2) Group 2 is non‐carriers of the APOE ɛ4 allele with PRS in the top quartile; (3) Group 3 consists of non‐carriers of the APOE ɛ4 allele with PRS in the interquartile range (25th to 75th percentile); and (4) Group 4 is non‐carriers of the APOE ɛ4 allele with PRS in the lowest quartile. We referenced a Bonferroni‐corrected significance threshold of 0.0018 for stratified analyses. We tested for non‐linear interactions by fitting each contextual exposure as a cubic spline with two degrees of freedom and tested interaction with genetic risk modeled by genetic risk group. Likelihood ratio tests (LRTs) were used to assess the statistical significance of interactions between exposures and genetic risk category. As a sensitivity analysis, we used GEE logistic regression models to assess the association between contextual determinants and combined binary dementia classification based on MESA MIND cognitive adjudication. In these models, binary dementia status was treated as time invariant, avoiding time‐to‐event data that may inaccurately estimate time of disease onset. Statistical power may be improved by the additional cases and due to more precise phenotyping. The contextual exposures continue to be modeled as time varying, as the GEE model accounts for neighborhood‐level clustering and incorporates repeated measures across five MESA exams for the social and built environment exposures. We used the GEE models to assess the association of each exposure with dementia status in all MESA participants as well as in the four genetic risk groups. We conducted additional sensitivity analyses for models with statistically significant associations. First, we applied inverse probability weighting (IPW) to assess the potential impact of selective attrition over follow‐up on estimated associations. Second, given that dementia cases occurring before age 65 are less likely to reflect late‐onset AD, we repeated analyses excluding participants with dementia onset prior to age 65. 2.4.2. BKMR to examine joint effects The effects of contextual determinants are not experienced individually. BKMR is an approach for modeling the joint effects of a mixture of exposures, allowing for non‐linear effects and interactions among mixture components 31 . As an exploratory, hypothesis‐generating analysis, we used BKMR to jointly model the effects of the seven contextual determinants on cognition (insufficient power to use BKMR for dementia status). BKMR estimates the effect of the mixture of contextual determinants at different quantiles compared to the median value of the mixture. We fit separate BKMR models for each of the four genetic risk categories to explore whether patterns of joint exposure effects differ qualitatively across genetic risk groups. All seven contextual determinants were standardized. Standardized distances to alcohol establishments and fast‐food chains were flipped based on our hypothesis that proximity to these establishments has opposing effects to the other features in the mixture. Each BKMR model is estimated using a Markov chain Monte Carlo algorithm with 10,000 iterations. Like the linear models, all BKMR models controlled for covariates: age (years), sex, site, income, education, smoking, marital status, and self‐reported race/ethnicity. Only complete cases were retained for the BKMR analysis. Posterior inclusion probabilities (PIPs) were calculated for each contextual determinant as a measure of the determinant's relative importance to the model. 3. RESULTS 3.1. Study population The final analytical sample consisted of 5687 individuals (Figure 1 ). Ambient pollutant models included 5330 individuals. In the total analytical sample, participants had a mean age of 62 years at baseline (standard deviation [SD] = 10.1), and 52% were women. Twenty‐seven percent were APOE ɛ4 carriers, with 2.5% homozygous for the APOE ɛ4 allele. Forty‐one percent of the population was White, 25% Black, 22% Hispanic and/or Latino, and 12% Chinese. After a median follow‐up time of 16.8 years, 500 (8.8%) MESA participants developed incident dementia according to hospital and death certificate ICD codes. Complete baseline demographics and case rates per genetic risk category are presented in Table 1 and Table S4 in supporting information, respectively. FIGURE 1. Open in a new tab Flowchart of participant exclusion. APOE , apolipoprotein E; MESA, Multi‐Ethnic Study of Atherosclerosis; PRS, polygenic risk score. TABLE 1. Baseline characteristics of participants (Exam 1: 2000–2002). Censored Affected Overall ( N = 5187) ( N = 500) ( N = 5687) Age (years) Mean (SD) 61.0 (9.83) 72.0 (7.69) 62.0 (10.1) Median [min, max] 61.0 [44.0, 84.0] 74.0 [45.0, 84.0] 62.0 [44.0, 84.0] Site of clinic visit WFU 792 (15.3%) 97 (19.4%) 889 (15.6%) COL 779 (15.0%) 85 (17.0%) 864 (15.2%) JHU 828 (16.0%) 69 (13.8%) 897 (15.8%) UMN 837 (16.1%) 95 (19.0%) 932 (16.4%) NWU 888 (17.1%) 79 (15.8%) 967 (17.0%) UCLA 1063 (20.5%) 75 (15.0%) 1138 (20.0%) Sex Female 2707 (52.2%) 245 (49.0%) 2952 (51.9%) Male 2480 (47.8%) 255 (51.0% 2735 (48.1%) Level of education < High school degree 853 (16.4%) 112 (22.4%) 965 (17.0%) High school degree 913 (17.6%) 120 (24.0%) 1033 (18.2%) Some college, no bachelor's degree 1474 (28.4%) 128 (25.6%) 1602 (28.2%) Bachelor's degree or higher 1931 (37.2%) 139 (27.8%) 2070 (36.4%) Missing 16 (0.3%) 1 (0.2%) 17 (0.3%) Marital status Divorced/separated/never married 1904 (36.7%) 244 (48.8%) 2148 (37.8%) Married/living as married 3268 (63.0%) 255 (51.0%) 3523 (61.9%) Missing 15 (0.3%) 1 (0.2%) 16 (0.3%) Household income < $30,000 1721 (33.2%) 242 (48.4%) 1963 (34.5%) > $30,000 3297 (63.6%) 229 (45.8%) 3526 (62.0%) Missing 169 (3.3%) 29 (5.8%) 198 (3.5%) Race/ethnicity Black 1295 (25.0%) 125 (25.0%) 1420 (25.0%) Chinese 665 (12.8%) 36 (7.2%) 701 (12.3%) Hispanic/Latino 1160 (22.4%) 104 (20.8%) 1264 (22.2%) White 2067 (39.8%) 235 (47.0%) 2302 (40.5%) Pack‐years of cigarette smoking Mean (SD) 10.9 (20.7) 12.6 (22.1) 11.1 (20.8) Median [min, max] 0 [0, 265] 0 [0, 140] 0 [0, 265] Missing 71 (1.4%) 4 (0.8%) 75 (1.3%) APOE genotype ε2/ε2 41 (0.8%) 2 (0.4%) 43 (0.8%) ε2/ε3 616 (11.9%) 54 (10.8%) 670 (11.8%) ε2/ε4 132 (2.5%) 13 (2.6%) 145 (2.5%) ε3/ε3 3180 (61.3%) 270 (54.0%) 3450 (60.7%) ε3/ε4 1100 (21.2%) 141 (28.2%) 1241 (21.8%) ε4/ε4 118 (2.3%) 20 (4.0%) 138 (2.4%) Missing Cognitive assessment score at exam 5 Mean (SD) 88.0 (8.61) 80.6 (12.1) 87.5 (9.03) Median [min, max] 89.5 [23.2, 100] 83.0 [28.1, 98.4] 89.0 [23.2, 100] Missing 1345 (23.7%) 259 (47.0%) 1604 (25.7%) Open in a new tab Note : Summaries provided for overall sample and by dementia status in 2018. Abbreviations: APOE , apolipoprotein E; COL, Columbia University; JHU, Johns Hopkins University; NWU, Northwestern University; SD, standard deviation; UCLA, University of California Los Angeles; UMN, University of Minnesota; WFU, Wake Forest University. 3.2. Linear models 3.2.1. Dementia risk Adjusting for APOE ɛ4 carrier status, PRS, age, sex, site, income, education, smoking, marital status, and self‐reported race/ethnicity, a one‐unit worse NDI was associated with a 14% greater hazard of dementia (HR NDI = 1.14; 95% confidence interval [CI]: 1.05–1.25; p = 0.004, Figure 2A and Figure S1 in supporting information). There was no evidence of association with dementia incidence for ambient pollutants or built environment features (Figure 2A ). FIGURE 2. Open in a new tab Forest plots for the association between contextual determinants and incident dementia/late‐life cognition. Each contextual determinant was modeled separately. All models adjusted for APOE ε4 carrier status, polygenic risk for Alzheimer's disease, sex, self‐reported race/ethnicity, study site, smoking, marital status, household income, and education. Regression models of pollutants (green) and built environment features (red) additionally controlled for NDI. Hazard ratios and estimates are reported per interquartile range difference for the pollutants and per 2‐fold change for the built environment features. A, Estimated hazard ratio of incident dementia from Cox proportional hazards regression. B, Estimated difference in cognition (CASI score) from linear regression using generalized estimating equations. APOE , apolipoprotein E; CASI, Cognitive Abilities Screening Instrument; CI, confidence interval; NDI, Neighborhood Deprivation Index. Cox proportional hazards models stratified by genetic risk category revealed that NDI was only associated with increased hazard of dementia among APOE ɛ4 carriers, where a one‐unit difference in NDI was associated with a 25% greater hazard of dementia (HR NDI_G1 = 1.25; 95% CI: 1.06–1.47; p = 0.01, Figure 3A ). While there was some evidence of a dose–response relationship between genetic risk category and NDI, as the estimated hazard of dementia associated with a one‐unit difference in NDI increased alongside increasing genetic risk, effect modification was not statistically significant (LRT p value > 0.5). The remaining contextual determinants were not associated with incident dementia in the genetically stratified groups (Figure 3A , Table S5 in supporting information). FIGURE 3. Open in a new tab Estimates of linear association between contextual determinants and incident dementia/cognition, stratified by genetic risk group. Forest plots for the association between contextual determinants and incident dementia/late‐life cognition stratified by genetic risk category. Multi‐Ethnic Study of Atherosclerosis participants were stratified based on their genetic risk for Alzheimer's disease. Each contextual determinant was modeled separately in each genetically stratified risk group. All models adjusted for sex, self‐reported race/ethnicity, study site, smoking, marital status, household income, and education. Regression models of pollutants and built environment features additionally controlled for NDI. Hazard ratios and estimates are reported per interquartile range difference for the pollutants and per 2‐fold change for the built environment features. A, Estimated hazard ratio of incident dementia from Cox proportional hazards regression. B, Estimated difference in cognition (CASI score) from linear regression using generalized estimating equations. APOE , apolipoprotein E; CASI, Cognitive Abilities Screening Instrument; CI, confidence interval; NDI, Neighborhood Deprivation Index. Sensitivity , analysis modeling dementia status as a binary trait revealed similar results to the survival models (Figures S2–S3 in supporting information). Among all assessed contextual determinants, only NDI was associated with odds of dementia (OR NDI = 1.1; 95% CI: 1.04–1.25; p = 0.002). When stratified by genetic risk, higher NDI was associated with increased odds of dementia only among APOE ɛ4 carriers (OR NDI_G1 = 1.38; 95% CI: 1.15–1.65, p = 0.001, Table S6 in supporting information). The results remained consistent after excluding 14 ICD‐ascertained dementia cases who had age‐at‐first‐occurrence prior to age 65 (Figure S4 in supporting information). Additionally, all contextual determinants were modeled with a two degree of freedom cubic spline to test for non‐linear interaction effects between contextual determinants and genetic risk group on incident dementia risk. We used LRT to test models that included interaction terms between the splines and four levels of genetic risk. There was no evidence of non‐linear interaction for any of the contextual determinants (LRT p value > 0.5; Figures S5–S11 in supporting information). 3.2.2. Global cognitive function (CASI) After controlling for APOE ɛ4 carrier status, PRS, age, sex, site, income, education, smoking, marital status, and self‐reported race/ethnicity, a one‐unit higher NDI was associated with lower CASI score (β NDI = −0.30, 95% CI: −0.49 to −0.11; p = 3.7E‐05; Figure 2B , Figure S1 ). After adjusting for NDI in addition to the covariates, for every 2‐fold greater distance from alcohol establishments, participants had a 0.17 higher CASI score (β ALC = 0.17, 95% CI: 0.02–0.33; p = 0.049). There was no evidence of a difference in CASI score for other contextual determinants. Models stratified by genetic risk revealed that greater NDI was associated with worse cognition among APOE ɛ4 carriers and those with moderate polygenic risk (β NDI_ɛ4 = −0.37, 95% CI: −0.70 to −0.05; β NDI_G3 = −0.45, 95% CI: −0.78 to −0.12, Table S7 in supporting information). Results from IPW analyses were not meaningfully different from those of the primary analyses (Figure S12 in supporting information). There was no evidence of linear or non‐linear effect modification between the contextual determinants and genetic risk group on cognition (LRT p value > 0.5). 3.3. BKMR We used BKMR to model the joint effects of all seven contextual determinants on cognition, stratified by genetic risk category. Only complete cases were retained. In total, 960 individuals were in Group 1 (ɛ4 carriers), 676 individuals were in Group 2, 1368 individuals in Group 3, and 683 individuals in Group 4. BKMR estimates the overall effect of the mixture, resulting in comparisons of CASI score when all exposures are fixed at a particular percentile compared to when all exposures are fixed at their median. There was an overall downward trend of CASI scores as the percentile of the mixture increased (Figure 4 ). For the APOE ɛ4 carriers, non‐ɛ4 carriers with high PRS, and non‐ɛ4 carriers with low PRS, however, the confidence intervals included the null across nearly all percentiles compared to the median (with slight exceptions at the top 20 th percentile in the low PRS group). For non‐ɛ4 carriers with intermediate PRS, mixture percentiles above the median showed a monotonic decrease in CASI score (Figure 4 ). Among all contextual exposures in the mixture, NO 2 had the strongest support for association with CASI score in the intermediate PRS model (PIP = 0.42), while no other exposures had PIPs > 0.10 (Table S8 in supporting information). Visual inspection of non‐linear interactions among the contextual determinants suggests that there may be interactions between NO 2 and NDI, with NO 2 appearing to have a weaker association with CASI score in groups at the tenth percentile of NDI compared to the median (Figure S13 in supporting information). However, this interaction result should be interpreted with caution and further investigation in larger samples are needed. FIGURE 4. Open in a new tab BKMR plots of overall mixture effect. The estimated change in CASI scores across quantiles of the contextual determinant mixture is shown separately for each genetic risk subgroup: non‐ε4 carriers with low PRS, intermediate PRS, high PRS, and APOE ε4 carriers. Estimates were obtained using BKMR models. Points indicate the estimated difference in CASI scores at each quantile of the contextual mixture compared to the median exposure (quantile = 0.5), with vertical bars representing 95% confidence intervals. APOE , apolipoprotein E; BKMR, Bayesian kernel machine regression; CASI, Cognitive Abilities Screening Instrument; PRS, polygenic risk score. 4. DISCUSSION This is one of the first studies to integrate PRS and contextual determinants to evaluate the associations among social, chemical, and built environmental context and late‐life dementia and cognition. Beyond linear models testing the association of one contextual determinant at a time, we also examined how the association between the joint effect of environmental context and cognition varies by genetic risk. We chose to examine the role of neighborhood disadvantage, ambient pollutants, and food/alcohol/physical activity environment based on previous literature on modifiable risk factors for AD and dementia across the life course. Living in socioeconomically disadvantaged neighborhoods has been consistently associated with cognition, dementia incidence, and AD pathology 32 , 33 . The evidence of association between ambient pollutants, especially fine particulate matter, and dementia is increasing, with multiple large cohort studies finding positive associations between pollution exposure or proximity to heavy traffic and dementia or cognition 7 , 34 , 35 . Aside from neighborhood disadvantage and pollution, diet, physical activity, and alcohol consumption are also modifiable risk factors for AD and dementia, but most studies have focused on individual‐level behavior. However, there is a growing body of work focused on neighborhood‐level determinants. A recent cohort study found that living in low food access and low‐income neighborhoods was associated with faster rate of cognitive decline in older adults 36 . In contrast, another study found no association between neighborhood density of grocery stores and cognitive function but found that higher density of fast‐food establishments was associated with lower levels of cognitive function, though effects were small compared to social organizations and highways 37 . A study examining neighborhood physical activity facilities suggests that neighborhoods supportive of physical activity may influence dementia risk via vascular mechanisms, showing an association between the number of activity facilities and dementia with vascular or mixed pathology, but no relationship with AD 38 . Studies on the alcohol neighborhood environment are more limited, but one recent study found that additional alcohol drinking places per square mile was nominally associated with improved memory function 39 . In line with previous studies, we found that higher neighborhood disadvantage is associated with both higher incident dementia and lower cognitive scores, after adjusting for individual‐level genetic risk and demographics including income, race/ethnicity, age, and sex. We also observed that the strength of association between NDI and dementia outcomes and cognition are strongest among APOE ɛ4 carriers, although the difference of effect across groups is not statistically significant. This null finding could reflect insufficient power to detect interaction effects due to sample size and a low case rate. In addition to the tests of individual contextual determinants, we used BKMR to model the joint effects of the contextual determinants on cognition, allowing for a more holistic model of the impact of the environment, rather than isolating the linear effects of one contextual determinant at a time. The joint model suggests that the effects of neighborhood context are not limited to those with high genetic risk. The plots of the overall effect of contextual determinants reveal that deleterious neighborhood context (percentiles of all environmental features above median) is associated with lower CASI scores in the non‐ɛ4 group with intermediate PRS. Further exploration of driving features of the mixture in this subgroup show that NO 2 is the most important determinant in the mixture. Although NO 2 was not associated with dementia or cognition in linear models, this may reflect the inability of linear approaches to capture non‐linear interaction effects with other environmental factors. Possible biological explanations of the association between the social environment and cognitive outcomes include the role of inflammation and oxidative stress, where neighborhood disadvantage contributes to greater risk for diseases that dysregulate inflammatory responses such as cardiovascular disease and hypertension 40 , 41 , 42 . Exposure to pollution may also act on similar pathways, with elevated exposure to combustion‐related pollutants including NO 2 shown to be associated with neuroinflammation, vascular risks, and oxidative stress. The inflammatory pathway also aligns with findings from genetic studies, as many implicated genes for AD are primarily expressed in immune cells 43 , 44 , 45 . Future studies must continue to leverage both genetic and contextual approaches to identify specific mechanisms that explain why contextual determinants are consistently linked to brain aging and dementia outcomes. Including both genetic and environmental risk for the same biological pathways can also improve risk prediction and, eventually, precision treatments. There are several limitations to this study. First, as this is an observational study, the results should not be interpreted as establishing causal relationships. Additionally, the Cox proportional hazards models use dementia diagnoses based on ICD codes, which typically underestimate dementia cases 46 , 47 . However, we were able to supplement these models with logistic regression models using the additional cognitive adjudication. It should also be noted that while Bonferroni corrections can be overly conservative, only the NDI association with incident dementia, probable dementia, and cognition among all participants remained significant after using a Bonferroni‐adjusted p value. Additionally, we did not include biomarkers of various forms of dementia‐related pathology and were unable to differentiate between dementia etiologic subtypes. There may be varying effects between contextual determinants on various forms of dementia pathology, for example, AD versus cerebrovascular disease. Our PRS is based on a genome‐wide association study meta‐analysis of AD. While there may be some overlap in genetic risk factors across dementia subtypes, we may be missing genetic risk factors for other forms of dementia than AD. Nonetheless, the most common form of dementia results from AD pathology or mixed AD and vascular pathology 48 . Furthermore, because our built‐environment and ambient‐pollution measures are assigned from participants’ residential addresses, they may not capture the full range of locations where participants spend time. 5. CONCLUSION Living in disadvantaged neighborhoods was consistently associated with higher dementia incidence and poorer late‐life cognition, with the strongest associations observed among individuals at elevated genetic risk. More replication is needed to understand the potential gene–environment interaction, but our exploratory joint analysis indicates that interactions between traffic‐related pollution and neighborhood disadvantage are associated with cognition in those with intermediate genetic risk, who represent most of the population. Our results suggest that socially and structurally patterned determinants can act on biological pathways similar to genetic factors, which may explain why the strongest associations between neighborhood disadvantage and dementia or cognition are among those at highest genetic risk. Future studies are needed to investigate the underlying biological mechanisms of these associations, but in addition to individual‐level biological interventions, our findings support the potential of policy‐level solutions to promote healthier cognitive aging. CONFLICT OF INTEREST STATEMENT The authors have no conflicts of interest to disclose. Author disclosures are available in the Supporting Information . CONSENT STATEMENT MESA activities include central institutional review board (IRB) and local IRB oversight. Informed consent was obtained from each participant at baseline and updated at each examination. Approval was received at each site from the local IRB for each examination. Supporting information Supporting Information ALZ-22-e71384-s002.docx (3.9MB, docx) Supporting Information ALZ-22-e71384-s001.pdf (409.4KB, pdf) ACKNOWLEDGMENTS The authors thank the other investigators, the staff, and the participants of the MESA study for their valuable contributions. A full list of participating MESA investigators and institutions can be found at http://www.mesa‐nhlbi.org. This paper has been reviewed and approved by the MESA Publications and Presentations Committee. This publication was developed under the Science to Achieve Results (STAR) research assistance agreements, No. RD831697 (MESA Air) and RD‐83830001 (MESA Air Next Stage), awarded by the U.S Environmental Protection Agency (EPA). It has not been formally reviewed by the EPA. The views expressed in this document are solely those of the authors and the EPA does not endorse any products or commercial services mentioned in this publication. Special thanks to Dr. Joel Kaufman for being an adviser on this project. This research was supported by NIA K00AG079792 and NIA K01AG071689. Funding for SHARe genotyping was provided by NHLBI Contract N02‐HL‐64278. Genotyping was performed at Affymetrix (Santa Clara, California, USA) and the Broad Institute of Harvard and MIT (Boston, Massachusetts, USA) using the Affymetrix Genome‐Wide Human SNP Array 6.0. This research was additionally supported by contracts 75N92025D00022, 75N92020D00001, HHSN268201500003I, N01‐HC‐95159, 75N92025D00026, 75N92020D00005, N01‐HC‐95160, 75N92020D00002, N01‐HC‐95161, 75N92025D00024, 75N92020D00003, N01‐HC‐95162, 75N92025D00027, 75N92020D00006, N01‐HC‐95163, 75N92025D00025, 75N92020D00004, N01‐HC‐95164, 75N92025D00028, 75N92020D00007, N01‐HC‐95165, N01‐HC‐95166, N01‐HC‐95167, N01‐HC‐95168, and N01‐HC‐95169 from the National Heart, Lung, and Blood Institute, and by grants UL1‐TR‐000040, UL1‐TR‐001079, and UL1‐TR‐001420 from the National Center for Advancing Translational Sciences (NCATS). REFERENCES 1. Jack CR Jr., Andrews JS, Beach TG, et al. Revised criteria for diagnosis and staging of Alzheimer's disease: Alzheimer's Association Workgroup. Alzheimers Dement. 2024;20:5143‐69. doi: 10.1002/alz.13859 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 2. Zissimopoulos J, Crimmins E, St Clair P. The value of delaying Alzheimer's disease onset. Forum Health Econ Policy. 2014;18:25‐39. doi: 10.1515/fhep‐2014‐0013 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 3. Livingston G, Huntley J, Liu KY, et al. Dementia prevention, intervention, and care: 2024 report of the Lancet standing Commission. Lancet. 2024;404:572‐628. doi: 10.1016/S0140‐6736(24)01296‐0 [ DOI ] [ PubMed ] [ Google Scholar ] 4. Gatz M, Reynolds CA, Fratiglioni L, et al. Role of genes and environments for explaining Alzheimer disease. Arch Gen Psychiatry. 2006;63:168‐174. doi: 10.1001/archpsyc.63.2.168 [ DOI ] [ PubMed ] [ Google Scholar ] 5. Liu S, Bush WS, Akinyemi RO, et al. Alzheimer disease is (sometimes) highly heritable: drivers of variation in heritability estimates for binary traits, a systematic review. PLoS Genet. 2025;21:e1011701. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 6. Besser L, Galvin JE, Rodriguez D, et al. Associations between neighborhood built environment and cognition vary by apolipoprotein E genotype: Multi‐Ethnic Study of Atherosclerosis. Health Place. 2019;60:102188. doi: 10.1016/j.healthplace.2019.102188 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 7. Shaffer RM, Blanco MN, Li G, et al. Fine particulate matter and dementia incidence in the Adult Changes in Thought Study. Environ Health Perspect. 2024;129:087001. doi: 10.1289/EHP9018 [ Google Scholar ] 8. Eid A, Mhatre I, Richardson JR. Gene‐environment interactions in Alzheimer's disease: a potential path to precision medicine. Pharmacol Ther. 2019;199:173‐87. doi: 10.1016/j.pharmthera.2019.03.005 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 9. Langella S, Barksdale NG, Vasquez D, et al. Effect of apolipoprotein genotype and educational attainment on cognitive function in autosomal dominant Alzheimer's disease. Nat Commun. 2023;14:5120. doi: 10.1038/s41467‐023‐40775‐z [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 10. Kolli A, Zhou Y, Chung G, Ware EB, Langa KM, Ehrlich JR. Interactions between the apolipoprotein E4 gene and modifiable risk factors for cognitive impairment: a nationally representative panel study. BMC Geriatr. 2022;22:938. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 11. Emrani S, Arain HA, DeMarshall C, Nuriel T. APOE4 is associated with cognitive and pathological heterogeneity in patients with Alzheimer's disease: a systematic review. Alzheimers Res Ther. 2020;12:141. doi: 10.1186/s13195‐020‐00712‐4 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 12. Blue EE, Horimoto ARVR, Mukherjee S, Wijsman EM, Thornton TA. Local ancestry at APOE modifies Alzheimer's disease risk in Caribbean Hispanics. Alzheimers Dement. 2019;15:1524‐32. doi: 10.1016/j.jalz.2019.07.016 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 13. Hohman TJ, Cooke‐Bailey JN, Reitz C, et al. Global and local ancestry in African‐Americans: implications for Alzheimer's disease risk. Alzheimers Dement. 2016;12:233‐243. doi: 10.1016/j.jalz.2015.02.012 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 14. Andrews SJ, Renton AE, Fulton‐Howard B, Podlesny‐Drabiniok A, Marcora E, Goate AM. The complex genetic architecture of Alzheimer's disease: novel insights and future directions. EBioMedicine. 2023;90:104511. doi: 10.1016/j.ebiom.2023.104511 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 15. Choi SW, Mak TS‐H, O'Reilly PF. Tutorial: a guide to performing polygenic risk score analyses. Nat Protoc. 2020;15:2759‐2772. doi: 10.1038/s41596‐020‐0353‐1 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 16. Klee M, Leist AK, Veldsman M, Ranson JM, Llewellyn DJ. Socioeconomic deprivation, genetic risk, and incident dementia. Am J Prev Med. 2023;64:621‐30. doi: 10.1016/j.amepre.2023.01.012 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 17. Hasbani NR, Ligthart S, Brown MR, et al. American Heart Association's Life's Simple 7: lifestyle recommendations, polygenic risk, and lifetime risk of coronary heart disease. Circulation. 2022;145:808‐18. doi: 10.1161/CIRCULATIONAHA.121.053730 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 18. Ren J, Zhang P, Li Z, et al. Association of screening status, polygenic risk score and environmental risk factors with colorectal cancer incidence and mortality risks. Int J Cancer. 2023;152:1778‐88. doi: 10.1002/ijc.34407 [ DOI ] [ PubMed ] [ Google Scholar ] 19. Walsh S, Klee M, Hui EK, et al. Social determinants of dementia: a scoping review. Alzheimers Dement. 2025;21:e70524. doi: 10.1002/alz.70524 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 20. Kunkle BW, Grenier‐Boley B, Sims R, et al. Genetic meta‐analysis of diagnosed Alzheimer's disease identifies new risk loci and implicates Aβ, tau, immunity and lipid processing. Nat Genet. 2019;51:414‐30. doi: 10.1038/s41588‐019‐0358‐2 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 21. Bild DE, Bluemke DA, Burke GL, et al. Multi‐ethnic study of atherosclerosis: objectives and design. Am J Epidemiol. 2002;156:871‐81. doi: 10.1093/aje/kwf113 [ DOI ] [ PubMed ] [ Google Scholar ] 22. Fujiyoshi A, Jacobs DR Jr, Alonso A, Luchsinger JA, Rapp SR, Duprez DA. Validity of death certificate and hospital discharge ICD codes for dementia diagnosis: the Multi‐Ethnic Study of Atherosclerosis. Alzheimer Dis Assoc Disord. 2017;31:168‐172. doi: 10.1097/WAD.0000000000000164 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 23. Fujiyoshi A, Jacobs DR Jr, Fitzpatrick AL, et al. Coronary artery calcium and risk of dementia in MESA (Multi‐Ethnic Study of Atherosclerosis). Circ Cardiovasc Imaging. 2017;10:e005349. doi: 10.1161/CIRCIMAGING.116.005349 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 24. Hughes TM, Tanley J, Chen H, et al. Subclinical vascular composites predict clinical cardiovascular disease, stroke, and dementia: the multi‐ethnic study of atherosclerosis (MESA). Atherosclerosis. 2024;392:117521. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 25. Teng EL, Hasegawa K, Homma A, et al. The Cognitive Abilities Screening Instrument (CASI): a practical test for cross‐cultural epidemiological studies of dementia. Int Psychogeriatr. 1994;6:45‐58. doi: 10.1017/S1041610294001602 [ DOI ] [ PubMed ] [ Google Scholar ] 26. Diez Roux AV, Mujahid MS, Hirsch JA, Moore K, Moore L V. The impact of neighborhoods on CV risk. Glob Heart. 2016;11:353‐63. doi: 10.1016/j.gheart.2016.08.002 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 27. Kaufman JD, Adar SD, Allen RW, et al. Prospective study of particulate air pollution exposures, subclinical atherosclerosis, and clinical cardiovascular disease: the Multi‐Ethnic Study of Atherosclerosis and Air Pollution (MESA Air). Am J Epidemiol. 2012;176:825‐37. doi: 10.1093/aje/kws169 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 28. Hirsch JA, Moore KA, Cahill J, et al. Business data categorization and refinement for application in longitudinal neighborhood health research: a methodology. J Urban Health. 2021;98:271‐84. doi: 10.1007/s11524‐020‐00482‐2 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 29. Xue D, Blue EE, Sofer T, Hughes TM, Rotter JI, Fohner AE. Polygenic risk scores for incident dementia in the Multi‐Ethnic Study of Atherosclerosis. MedRxiv. Published online March 7, 2025. doi: 10.1101/2025.03.05.25323412 [ Google Scholar ] 30. Wasserstein RL, Lazar NA. The ASA statement on p‐values: context, process, and purpose. Am Stat. 2016;70:129‐33. doi: 10.1080/00031305.2016.1154108 [ Google Scholar ] 31. Bobb JF, Claus Henn B, Valeri L, Coull BA. Statistical software for analyzing the health effects of multiple concurrent exposures via Bayesian kernel machine regression. Environ Health. 2018;17:1‐10. doi: 10.1515/fhep‐2014‐0013 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 32. Hunt JF V, Buckingham W, Kim AJ, et al. Association of neighborhood‐level disadvantage with cerebral and hippocampal volume. JAMA Neurol. 2020;77:451‐60. doi: 10.1001/jamaneurol.2019.4501 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 33. Dintica CS, Bahorik A, Xia F, Kind A, Yaffe K. Dementia risk and disadvantaged neighborhoods. JAMA Neurol. 2023;80:903‐9. doi: 10.1001/jamaneurol.2023.2120 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 34. Chen H, Kwong JC, Copes R, et al. Exposure to ambient air pollution and the incidence of dementia: a population‐based cohort study. Environ Int. 2017;108:271‐7. doi: 10.1016/j.envint.2017.08.020 [ DOI ] [ PubMed ] [ Google Scholar ] 35. Shi L, Zhu Q, Wang Y, et al. Incident dementia and long‐term exposure to constituents of fine particle air pollution: a national cohort study in the United States. Proc Natl Acad Sci. 2023;120:e2211282119. doi: 10.1073/pnas.2211282119 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 36. Kim B, Thorpe RJ, Szanton SL, Adkins‐Jackson PB, Samuel LJ. Disparities in neighborhood food environment and cognitive decline among US older adults: a cohort study. BMC Med. 2025;23:259. doi: 10.1186/s12916‐025‐04091‐1 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 37. Finlay J, Esposito M, Langa KM, Judd S, Clarke P. Cognability: an ecological theory of neighborhoods and cognitive aging. Soc Sci Med. 2022;309:115220. doi: 10.1016/j.socscimed.2022.115220 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 38. Moored KD, Desjardins MR, Crane BM, et al. Neighborhood physical activity facilities predict risk of incident mixed and vascular dementia: the Cardiovascular Health Cognition Study. Alzheimers Dement. 2025;21:e14387. doi: 10.1002/alz.14387 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 39. Chaar DL, Tu L, Moore K, et al. Neighborhood environment associations with cognitive function and structural brain measures in older African Americans. BMC Med. 2025;23:15. doi: 10.1186/s12916‐024‐03845‐7 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 40. Yaffe K, Falvey C, Harris TB, et al. Effect of socioeconomic disparities on incidence of dementia among biracial older adults: prospective study. BMJ. 2013;347:f7051‐f7051. doi: 10.1136/bmj.f7051 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 41. Faraco G, Iadecola C. Hypertension: a harbinger of stroke and dementia. Hypertension. 2013;62:810‐7. doi: 10.1161/HYPERTENSIONAHA.113.01063 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 42. Luchsinger JA, Mayeux R. Cardiovascular risk factors and Alzheimer's disease. Curr Atheroscler Rep. 2004;6:261‐6. doi: 10.1007/s11883‐004‐0056‐z [ DOI ] [ PubMed ] [ Google Scholar ] 43. Bertram L, Lange C, Mullin K, et al. Genome‐wide association analysis reveals putative Alzheimer's disease susceptibility loci in addition to APOE. Am Hum Genet. 2008;83:623‐32. doi: 10.1016/j.ajhg.2008.10.008 [ Google Scholar ] 44. Jonsson T, Stefansson H, Steinberg S, et al. Variant of TREM2 associated with the risk of Alzheimer's disease. N Engl J Med. 2013;368:107‐16. doi: 10.1056/NEJMoa1211103 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 45. Romero‐Molina C, Garretti F, Andrews SJ, Marcora E, Goate AM. Microglial efferocytosis: diving into the Alzheimer's disease gene pool. Neuron. 2022;110:3513‐33. doi: 10.1016/j.neuron.2022.10.015 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 46. Schliep KC, Ju S, Foster NL, et al. How good are medical and death records for identifying dementia? Alzheimers Dement. 2022;18:1812‐23. doi: 10.1002/alz.12526 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 47. Chow EPF, Hsu B, Waite LM, et al. Diagnostic accuracy of linked administrative data for dementia diagnosis in community‐dwelling older men in Australia. BMC Geriatr. 2022;22:858. doi: 10.1186/s12877‐022‐03579‐2 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 48. Kapasi A, DeCarli C, Schneider JA. Impact of multiple pathologies on the threshold for clinically overt dementia. Acta Neuropathol. 2017;134:171‐86. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Supplementary Materials Supporting Information ALZ-22-e71384-s002.docx (3.9MB, docx) Supporting Information ALZ-22-e71384-s001.pdf (409.4KB, pdf) Articles from Alzheimer's & Dementia are provided here courtesy of Wiley ACTIONS View on publisher site PDF (1.3 MB) Cite Collections Permalink PERMALINK Copy RESOURCES Similar articles Cited by other articles Links to NCBI Databases Cite Copy Download .nbib .nbib Format: AMA APA MLA NLM Add to Collections Create a new collection Add to an existing collection Name your collection * Choose a collection Unable to load your collection due to an error Please try again Add Cancel Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top

Related documents

• Associations between indicators of self-rated health, cognitive function, risk of cognitive impairment, and dementia: A coordinated data analysis.#1175400
• A generator-matrix causal-inference framework separates measurable aging biomarkers from mortality-driving latent dynamics in humans#435644
• Can Sensor Technologies Accurately Detect and Monitor Behavioural and Psychological Symptoms of Dementia? A Systematic Review.#9363
• Driving knowledge assessment in cognitively impaired older adults: evidence from geriatric day hospitals#280481
• Development and temporal validation of a prediction model for cognitive impairment in older adults with hearing loss based on the population health risk management framework.#418229