The contribution of Apoliproprotein E genetic variation to dementia risk in British South Asians
1Wolfson Institute of Population Health, Queen Mary University of London, London, UK
2Department of Neurology, Royal London Hospital, Barts Health NHS Trust, London, UK
3Institute for Behavioural Genetics, University of Colorado Boulder, United States
4Centre for Neuroscience, Surgery & Trauma, The Blizard Institute, Queen Mary University of London, London, UK
5Division of Psychiatry, University College London, UK
6Unit for Lifelong Health & Ageing, University College London, London, UK
#Corresponding author; Email: charles.marshall@qmul.ac.ukAbstract
INTRODUCTION
Understanding the genetic basis of dementia in diverse populations is essential to ensure that efforts to predict, prevent, and treat dementia are equitable. The strongest genetic risk factor for dementia - APOE genotype - has not been assessed in population-scale cohorts of South Asian ancestry.
METHODS
We analysed data from 51,104 volunteers in the Genes & Health study - a cohort study of British South Asians - who have undergone genotyping and consented for linkage to healthcare records. All-cause dementia was defined using electronic healthcare records. APOE genotypes were defined using phased, imputed genotype data. Cox proportional hazards models were used to assess the relationship between APOE genotype and dementia. Population attributable fractions were calculated for each APOE genotype.
RESULTS
APOE ε4 was associated with all-cause dementia in a dose-dependent fashion (Case N = 614, Control N = 50,490; APOE ε4/ε4: Hazard Ratio 2.7, P < 0.0001; APOE ε4/ε3: Hazard Ratio 1.5, P < 0.001). The overall proportion of dementia cases attributable to this allele was 14.2% (95% CI -2.8% - 25.0%). APOE ε4 was also associated with elevated triglycerides and Low Density Lipoprotein (LDL) cholesterol.
DISCUSSION
APOE ε4 - the major genetic risk factor for sporadic dementia in European-ancestry populations - has a similar impact on dementia risk in British South Asians.
Article notes
Competing Interest Statement
The authors have declared no competing interest.
Funding Statement
BMJ is supported by a Guarantors of Brain post-doctoral fellowship award and an NIHR Clinical Lectureship. AC is funded by the National Institute for Health and Care Research (NIHR - ref: NIHR203373). This work was supported by an Alzheimers Research UK (ARUK) London Network Centre Pump Prime funding award. SW is supported by the Global Parkinsons Genetics Program (GP2). GP2 is funded by the Aligning Science Across Parkinsons (ASAP) initiative and implemented by The Michael J. Fox Foundation for Parkinsons Research (https://gp2.org).
Research in context
Systematic review
We reviewed the literature using traditional sources (e.g. Pubmed) for articles containing terms ‘APOE’, ‘dementia’ or ‘Alzheimer/s’, and ‘ancestry’ or ‘South Asian’ in the title. Although APOE has been extensively studied in European-ancestry cohorts, and more recently ancestrally diverse cohorts such as AllOfUs and Million Veterans Programme, few studies have explored the impact of APOE on dementia risk in biobank-scale cohorts of South Asian ancestry.
Interpretation
We show that the common APOE alleles E4 and E3 are associated with increased dementia risk in a British South Asian cohort. At a population-level, these alleles are likely to explain a comparable overall proportion of dementia cases.
Future directions
Prediction and prevention studies aiming to use APOE to identify people at high risk of dementia should be inclusive of all ancestries. Genetic analysis of dementia risk in diverse populations, including South Asian cohorts, is essential for ensuring that the downstream benefits - for prediction, prevention, and drug discovery - are shared equitably.
Introduction
Genome-wide association studies (GWAS) of sporadic Alzheimer’s disease (AD) have demonstrated that AD has a strong genetic component driven by common variation in at least 75 risk loci1,2. Variation at the APOE locus (encoding apolipoprotein E; APOE) alone is estimated to account for over one third of cases3, however this is based almost entirely on evidence from those of European ancestry.
Common coding variants in APOE define three common APOE protein isoforms - E2, E3, and E4 - of which E4 is associated with increased risk of AD in a dose-dependent fashion and E2 with decreased risk with respect to the most common allele, E34,5. The association between haplotypes at the APOE locus and AD has been consistently demonstrated across cohorts1,6,7, including populations of European, recently admixed (African American), Hispanic, and East Asian ancestry. The association between APOE genetic variation and dementia risk has not been evaluated at scale in a population of South Asian ancestry, although small studies in Pakistani and Indian populations have suggested an effect similar to that seen in European populations8–12. British South Asians are at higher dementia risk and tend to be diagnosed with dementia at a younger age 13,14, experience mortality sooner after a dementia diagnosis13, and have higher prevalence of dementia risk factors including deprivation, dyslipidaemia and type 2 diabetes mellitus15,16. Understanding the genetic basis of dementia in this population is essential for prediction and prevention efforts, especially given a surge in forecasted dementia cases in coming decades17. To date, few studies have examined the prevalence of APOE variants and the magnitude of their associations with dementia in a population-scale cohort of South Asian ancestry.
We therefore aimed to consider whether the major European-ancestry genetic risk factor for sporadic dementia - APOE genotype - was associated with dementia and other health-related phenotypic outcomes in a cohort of ∼70,000 British South Asian participants in the Genes & Health study. We hypothesised that both allele frequencies and effects on dementia risk might vary relative to European ancestry populations.
Methods
Cohort
Genes & Health (G&H) is a longitudinal cohort study of self-identified British Bangladeshi and British Pakistani individuals with genetic data (array genotyping and exome sequencing) and linked electronic healthcare records for over 70,000 healthy volunteers, with recruitment expected to expand to over 100,000 individuals16.
Genetic data & Definition of APOE haplotypes
G&H volunteers were genotyped from saliva using the Illumina Global Screening Array version 3.0. Genotype quality control procedures have been previously described18–20. APOE haplotypes were determined from phased genotype data imputed to the Topmed-r3 reference21. Genotypes at two coding variants - rs429358 and rs7412 - define the common APOE haplotypes22 as follows: rs429358T - rs7412T (E2), rs429358T - rs7412C (E3), rs429358C - rs7412C (E4). We used data from the October 2023 data freeze, comprising imputed genetic data for 51,166 volunteers and excluded 62 volunteers of ambiguous genetic ancestry, giving a final sample size of 51,104.
Phenotype definitions
All-cause dementia cases were identified using a custom codelist applied to multiple sources of linked electronic healthcare records, including NHS Digital Hospital Episode Statistics, primary care data, and Barts Health and Bradford Teaching Hospitals secondary care records. All-cause dementia was used as the primary outcome as this both maximises the sample size (and thus statistical power) and because more precise, clinically-subtyped diagnoses of dementia phenotypes are generally unreliable in electronic healthcare records23. In secondary sensitivity analyses we used Alzheimer’s Dementia (ICD code G30 or F00), unspecified dementia (ICD code F03) and vascular dementia (F01) as the outcome measure. Phenotypes were curated as part of a bespoke Python pipeline. All code and codelists used for phenotype generation are available at https://github.com/genes-and-health/BI_PY.
Statistical analysis
Demographics
Demographic data are presented with medians and interquartile ranges for continuous variables and as numbers and percentages for categorical variables.
Allele frequency comparison
We tested for statistical enrichment/depletion of the APOE E4 allele in the Genes & Health cohort using a two-tailed binomial test, comparing the proportion of E4 carriers and non-carriers with the proportion of carriers observed in the subset of ∼340,000 White British UK Biobank volunteers24.
Association of APOE haplotypes with dementia
Survival analysis was conducted using Cox proportional hazards models. As individuals were recruited from the age of 16 or over, the ‘time-to-event’ was defined as the time from the age of 16 to the age of dementia diagnosis. For controls, the censoring time was defined as December 2024 when the last phenotype data extract was performed. Models were adjusted for age, gender, the first ten genetic principal components, and binary genetic ancestry (inferred Pakistani or Bangladeshi ancestry). For models, the E3/E3 allele was used as the reference as this is the most common allele. Each pair of alleles was considered separately (i.e. a ‘genotypic’ model). Hazard ratios are expressed with 95% confidence intervals and P values for the null hypothesis that the ratio = 1. Models were inspected for the proportional hazards assumption using Schoenfeld residuals and for linearity. As sensitivity analyses, we repeated the analysis using fewer covariates (age and gender alone), using logistic regression models instead of Cox models (adjusted for identical covariates), we used gender-stratified models, we repeated the analysis using coded ‘Alzheimer’s dementia’ rather than all-cause dementia, and we restricted the analysis to those recruited over the age of 60 and excluded cases diagnosed prior to 60 (note for these models age 60, rather than age 16, was used as time zero for survival analyses).
Population attributable fraction (PAF)
We estimated the PAF for all-cause dementia by using the stratified exposure methods described in Hanley (2001)26 and used in Williams et al. (2025)3. To do so we first used a logistic regression model of the form dementia ∼ APOE genotype + covariates, using the low-risk, rare ε2/ε2 genotype as reference, to derive as estimate of the Odds Ratio for dementia given each combination of APOE alleles. This means individuals with intermediate risk genotypes (carriers of ε3) are not subsumed into the reference group, which would be expected to bias downwards proportions attributable to risk alleles. We calculated the ‘case fraction’ i.e. the proportion of all cases accounted for by each genotype (i.e. N cases / total cases), and estimated the genotype- specific PAF as (OR - 1) / (OR x case fraction). The overall PAF attributable to either ε3 or ε4 was estimated by summing the genotype-specific PAFs overall all genotypes containing ε3 or ε4 (ε3/ε3, ε3/ε4, ε3/ε2, ε4/ε2, ε4/ε4). To estimate the PAF attributable to ε3 or ε4 alone, we decomposed the relative contributions of ε3 and ε4 to the effect estimate for the ε3/ε4 genotype using the approach described in Williams et al. (2025)3; specifically, the ratio of the Odds Ratios for ε4/ε2 and ε3/ε2 was used to derive a weight (ratio / ratio + 1 for ε4; 1 - this weight for ε3) quantifying the proportion of the PAF for ε4/ε3 attributable to each allele individually.
Confidence intervals were calculated by calculating the genotype-specific PAF for each stratum using the lower or upper bound of the effect estimate (OR +/- 1.96 x standard error), and summing these PAFs together. The confidence estimates for the decomposed effect of ε3/ε4 were derived using the upper and lower bounds of the ratio.
Phenome-wide association study
We analysed the association of APOE alleles relative to ε3/ε3 with all available binary (disease) traits with at least 500 cases and all quantitative traits with N > 1000. We used logistic (for binary traits) and linear (for quantitative) traits adjusted for the same covariates as in the primary Cox models (age, gender, first ten PCs, and ancestry). In the text we report associations surpassing a study-wide Bonferroni threshold correction to maintain an alpha of 5% (i.e. P = 0.0002 = 0.05 / 236 tests).
Results
APOE ε4 prevalence in British South Asians
We analysed genetic and healthcare data from 51,104 British South Asian participants in the Genes & Health study (methods). Demographic characteristics of identified dementia cases and controls are shown in the table (table 1). Both variants used to define APOE alleles were common and well-imputed in G&H: rs429358 (chr19:44908684:T:C, Minor allele frequency [MAF] = 0.10, Imputation quality score [INFO] = 0.96); rs7412 (chr19:44908822:C:T, MAF = 0.05, INFO = 0.92). For both of these APOE variants, the allele frequency was consistent with estimates from reference populations of South Asian (SAS) ancestry but lower than in Non-Finnish European (NFE) reference samples (rs429358: MAFG&H = 10%, MAFgnomAD-SAS = 10.1%27, MAFgnomAD-NFE = 15.1%; rs7412 : MAFG&H = 5%, MAFgnomAD = 4.2%, MAFgnomAD-NFE = 7.8%). The overall allele frequencies for APOE haplotypes in G&H were 10.4% for APOE ε4, 85.1% for APOE ε3, and 4.5% for APOE ε2. The allele frequency of APOE ε4 was 11.3% in the participants of inferred Bangladeshi ancestry and 9.1% in participants of inferred Pakistani ancestry.
Dementia cases (n = 614, of which 451 [73.5%] were incident) were recruited at an older age (median 71.3 vs 39.0), were more likely to be male (58.6% vs 44.5%), and were more likely to carry at least one APOE ε4 allele (25.9% vs 19.5%) than controls without dementia (n = 50,490). Overall, 19.6% of the population carried at least one APOE ε4 allele (18.4% heterozygous, 1.2% homozygous), and 97.5% carried at least one APOE ε3 allele (24.9% heterozygous, 72.6% homozygous). The proportion of British South Asians who carry at least one APOE ε4 allele is therefore somewhat lower than White British people (19.6% vs 28.8% in UK Biobank24, two-tailed Binomial test P < 2x10-16), but consistent with prior estimates from cohorts of South Asian ancestry (e.g. 19.4% in an Indian population8 and 18.4% among the few self-identified British South Asians in UK Biobank28).
Association of APOE alleles with dementia in Genes & Health
Carriage of the APOE ε4 allele was associated with dementia diagnosis in a dose-dependent fashion (figure 2; APOE ε4/ε4: Hazard Ratio [HR] 2.7, 95% CI 1.7 - 4.2, P < 0.0001; APOE ε4/ε3: Hazard Ratio [HR] 1.5, 95% CI 1.2 - 1.8, P < 0.001; Cox multivariable regression models adjusted for age at recruitment, gender, PCs 1 - 10, and genetic ancestry; all models used APOE ε3/ε3 as reference).
We observed weaker (i.e. P > 0.01) evidence for a risk-increasing association of APOE ε4/ε2 (HR 1.6, 95% CI 0.8 - 3.2, P = 0.2) and protective association of the rarer ε2 genotypes APOE (ε2/ε2: HR 0.5, 95% 0.1 - 3.8, P =0.53; ε2/ε3: HR 0.8, 95% 0.6 - 1.2, P =0.3). Each copy of the APOE ε4 allele was associated with a 55% increase in the hazard of all-cause dementia (HR 1.56, 95% CI 1.34 - 1.82, P = 1.1x10-8, additive genetic model). In a dominant model, carriage of the APOE ε4 allele was associated with a similar degree of risk increase (HR 1.58, 95% CI 1.32 - 1.90, P = 7.3x10-7). The magnitude of this association strengthened when considering Alzheimer’s disease rather than all-cause dementia as the outcome (APOE ε4/ε4: Hazard Ratio 7.4, 95% CI 3.1 - 17.6, P < 0.0001; APOE ε4/ε3: Hazard Ratio 2.2, 95% CI 1.3 - 3.7, P = 0.002) although these estimates were imprecise owing to the smaller number of cases (NAD = 82).
The association between APOE status and all-cause dementia was similar in men and women: in gender-stratified models, APOE ε4 homozygosity was associated with a substantially elevated risk in both males (HR 2.73, 95% CI 1.55 - 4.80) and females (HR 2.95, 95% CI 1.38 - 6.33). We observed a similar magnitude of association using logistic regression models (ε4/ε4: OR 3.03, 95% CI 1.8 - 5.2, P = 7.3x10-7; ε4/ε3: OR 1.47, 95% CI 1.2 - 1.8, P = 0.0004). As expected, considering the rare ε2/ε2 genotype as the reference genotype enhanced the effect sizes of genotypes containing ε3 and ε4 alleles, while reducing the precision of the estimates (e.g. ε4/ε4: OR 7.46, 95% CI 0.86 - 65.0, P = 0.07) due to the rarity of this genotype.
To account for bias due to the younger age of the control cohort, we performed a sensitivity analysis restricting to participants recruited over the age of 60 and dementia cases diagnosed >60, yielding a subset of 484 dementia cases (median age at recruitment 74.8, IQR 11.8; median age at diagnostic code report 76.6, IQR 12.7; 39.5% female) and 5,015 controls (median age at recruitment 66.0, IQR 9.0; 45.8% female). The association results were similar to the primary analysis (APOE ε4/ε4: Hazard Ratio [HR] 2.8, 95% CI 1.7 - 4.6, P < 0.0001; APOE ε4/ε3: HR 1.6, 95% CI 1.3 - 2.0, P = <0.0001; APOE ε4/ε2: HR 2.1, 95% CI 1.04 - 4.27, P = 0.04; APOE ε3/ε2: HR 0.9, 95% CI 0.6 - 1.2, P = 0.42; APOE ε2/ε2: HR 0.6, 95% CI 0.1 - 4.6, P = 0.66).
Population attributable fraction
To quantify the population-level significance of APOE variants for dementia in this population we estimated the population attributable fraction (PAF) for all-cause dementia (figure 3). We estimated the PAF for all-cause dementia contingent on the presence of any risk-increasing APOE allele (i.e. ε3 or ε4) as 62.6%, however due to the scarcity of ε2/ε2 genotypes th confidence intervals were broad (95% CI -206% - 95.4%). We estimated that approximately 14.2% of all-cause dementia cases could be attributed to the presence of APOE ε4 (95% CI - 2.8% - 25.0%), and 48.4% (95% CI -203% - 70.3%) to APOE ε3.
Phenome-wide associations of APOE
Dementia-associated APOE haplotypes (APOE ε4 and ε3) were associated with higher triglycerides, Low Density Lipoprotein (LDL) cholesterol, and total cholesterol, and with lower CRP (Bonferroni-adjusted P value < 0.05, total number of traits tested = 236). Other than for these traits and dementia, no other associations persisted at phenome-wide significance (figure 4).
Discussion
Here we demonstrate that the major genetic risk factor for sporadic Alzheimer’s Dementia - APOE variation - is associated with all-cause dementia in a biobank-scale cohort of South Asian ancestry, Genes & Health. Specifically, we show that the dementia-associated APOE ε4 allele i associated with all-cause dementia in a dose-dependent manner, with homozygotes almost 3 times as likely to develop dementia as those with the most common ε3/ε3 genotype. The population attributable fraction for dementia contingent on the presence of either APOE ε3 or ε4 - the proportion of cases which could theoretically be prevented by neutralising the impact of these variants - was estimated at 62.6%, analogous to estimates from predominantly White British UK Biobank participants and other predominantly European-ancestry cohorts3.
In line with the slightly lower frequency of the ε4 allele in this population than in Europeans, the estimated PAF for ε4 alone is slightly lower than European-ancestry estimates (14.2% in Genes & Health vs 29.3% in UK Biobank - unpublished data, Dylan WIlliams), although the broad confidence intervals mandate caution in interpreting this difference. Furthermore, we show that the association between APOE variants and dementia does not differ substantially by gender, and we replicate the association of APOE ε3 and ε4 with elevated triglycerides and LDL cholesterol as expected from previous work24.
The major limitations of this work are the relatively young age of the cohort and hence risk of incomplete ascertainment (many participants who will develop dementia in the future have not yet developed it) and the phenotyping of dementia cases, which relies on electronic healthcare records and may therefore mix various pathological subtypes of dementia, including misdiagnoses, and does not allow for a detailed examination of dementia phenotypes. Our use of imputed genetic data rather than sequencing is a limitation, but the very high imputation quality and the fact that both markers used are common is reassuring. Further work with sequencing data will be required to evaluate APOE haplotypes beyond the three two-SNP haplotypes examined in this paper, as there may be population-enriched variants seen only in this population which modulate dementia risk in addition to the SNPs encoding the major APOE isoforms29,30.
Our findings confirm APOE ε4 and ε3 as genetic risk factors for dementia in people of South Asian ancestry and suggest that at the population-level, these genetic factors alone account for a sizeable proportion of dementia cases. However, ε4 prevalence, effect size and PAF all showed a trend towards being lower in this British South Asian population than in those of European ancestry, suggesting that APOE variation is does not fully account for the excess dementia risk that we have previously described in this population14.
Supporting information
Data availability statement
All code used to produce these results is available at https://benjacobs123456.github.io/apoe_gh/.
Funding information
BMJ is supported by a Guarantors of Brain post-doctoral fellowship award and an NIHR Clinical Lectureship. AC is funded by the National Institute for Health and Care Research (NIHR; ref: NIHR203373). This work was supported by an Alzheimer’s Research UK (ARUK) London Network Centre Pump Prime funding award. SW is supported by the Global Parkinson’s Genetics Program (GP2). GP2 is funded by the Aligning Science Across Parkinson’s (ASAP) initiative and implemented by The Michael J. Fox Foundation for Parkinson’s Research (https://gp2.org).
Data Availability
All code used to produce these results is available at https://benjacobs123456.github.io/apoe_gh/.
Acknowledgements
Genes & Health is/has recently been core-funded by Wellcome (WT102627, WT210561), the Medical Research Council (UK) (M009017, MR/X009777/1, MR/X009920/1), Higher Education Funding Council for England Catalyst, Barts Charity (845/1796), Health Data Research UK (for London substantive site), and research delivery support from the NHS National Institute for Health Research Clinical Research Network (North Thames). We acknowledge the support of the National Institute for Health and Care Research Barts Biomedical Research Centre (NIHR203330); a delivery partnership of Barts Health NHS Trust, Queen Mary University of London, St George’s University Hospitals NHS Foundation Trust and St George’s University of London.
This research was conducted as part of the ‘Brain Consortium’ approved project (ID S00015).
Genes & Health is/has recently been funded by Alnylam Pharmaceuticals, Genomics PLC; and a Life Sciences Industry Consortium of AstraZeneca PLC, Bristol-Myers Squibb Company, GlaxoSmithKline Research and Development Limited, Maze Therapeutics Inc, Merck Sharp & Dohme LLC, Novo Nordisk A/S, Pfizer Inc, Takeda Development Centre Americas Inc.
We thank Social Action for Health, Centre of The Cell, members of our Community Advisory Group, and staff who have recruited and collected data from volunteers. We thank the NIHR National Biosample Centre (UK Biocentre), the Social Genetic & Developmental Psychiatry Centre (King’s College London), Wellcome Sanger Institute, and Broad Institute for sample processing, genotyping, sequencing and variant annotation. This work uses data provided by patients and collected by the NHS as part of their care and support. This research utilised Queen Mary University of London’s Apocrita HPC facility, supported by QMUL Research-IT, http://doi.org/10.5281/zenodo.438045
We thank: Barts Health NHS Trust, NHS Clinical Commissioning Groups (City and Hackney, Waltham Forest, Tower Hamlets, Newham, Redbridge, Havering, Barking and Dagenham), East London NHS Foundation Trust, Bradford Teaching Hospitals NHS Foundation Trust, Public Health England (especially David Wyllie), Discovery Data Service/Endeavour Health Charitable Trust (especially David Stables), Voror Health Technologies Ltd (especially Sophie Don), NHS England (for what was NHS Digital) - for GDPR-compliant data sharing backed by individual written informed consent.
Most of all we thank all of the volunteers participating in Genes & Health.
A favourable ethical opinion for the main Genes & Health research study was granted by NRES Committee London - South East (reference 14/LO/1240) on 16 Sept 2014. Queen Mary University of London is the Sponsor, and Data Controller.
Competing interests
The authors declare no relevant competing interests.