Genetic Associations Between Modifiable Risk Factors and Alzheimer Disease

Key Points Question What are the genetic associations between modifiable risk factors and Alzheimer disease (AD)? Findings In this genetic association study using a mendelian randomization framework with the largest genomic data sets to date, including 39 106 participants with clinically diagnosed AD and 401 577 control participants without AD, genetically determined increased high-density lipoprotein cholesterol and increased systolic blood pressure were associated with higher risk of AD. Meaning These findings suggest that genetically determined increased high-density lipoprotein cholesterol and systolic blood pressure may be involved in AD pathogenesis, which may thus inspire new drug targeting and improved early dementia prevention.


eAppendix 1. Supplementary Results
In the CAUSE method analysis, the causal models generally trended towards a better fit than the sharing models for both HDL cholesterol, SBP, and DBP; the differences were however not statistically significant (eTable 5 in Supplement 1). This is likely due to the fact that this method uses the entire genome-wide summary statistics of both the exposure and outcome, and thus includes large amounts of insignificant genetic variants, resulting in substantially reduced power compared to the IVW and MR PRESSO methods.
Results after removing SNPs on chromosome 19 remained similar to the main analyses (eFigures 5 to 6 in Supplement 1). After adjusting for sample overlap in the EADB-diagnosed dataset using cross-trait linkage disequilibrium, the corrected estimates did not differ substantially; the ORs were 1.08 (1.02-1.14) for high HDL cholesterol, 0.84 (0.80-0.88) for high educational attainment, 0.95 (0.88-1.01) for smoking initiation, and 0.94 (0.86-1.02) for high BMI, respectively. Furthermore, genetic predisposition to higher odds of AD were not associated with educational attainment, smoking, alcohol consumption, or BMI (eTable 6 in Supplement 1).

eAppendix 2. Supplementary Discussion
First, high SBP may lead to brain vascular abnormalities such as atherosclerotic plaques, contributing to brain ischemia and infarctions. Second, the opposite directions of the SBP and DBP associations may lie in that long-term high SBP is correlated with left ventricular mass and wall thickness 4 which is linked to diastolic dysfunction characterized by lower DBP 5 . Particularly, left ventricular mass increase is associated with cognitive decline and dementia independent of BP 6 . Third, arterial stiffness, characterized by higher SBP and lower DBP, leads to the reduced cerebral blood flow that is a main caveat for cerebral hypoperfusion and associates with neurodegenerative damage. Additionally, high SBP, might be directly involved in AD pathologies, including brain atrophy of abnormal white matter lesion volumes 7 and cerebral amyloid burden 8 , whereas low DBP may play a role in cognitive impairment and dementia mediated by tau-pathologies 9 .

eFigure 1. The Concept of Mendelian Randomization Design
In the Mendelian randomization (MR) design, genetic variants associated with different modifiable risk factors at a genome-wide significant level from the genome-wide association studies were exploited as instrumental variables for these risk factors. Since genetic variants are randomly allocated at conception, the MR strategy is largely free of confounding factors and reverse causationtwo common culprits in conventional observational analyses. MR builds on three principal assumptions: the instrumental variables should firstly be associated with different modifiable risk factors; secondly not be associated with confounding factors in the relation between different modifiable risk factors and Alzheimer's disease (AD); thirdly affect AD exclusively via different modifiable risk factors, but not via other pathways. The potential causal effect of different modifiable risk factors on AD could be estimated by dividing gene-AD associations by gene-risk factor associations.
Proxy-AD cases in the UK Biobank were identified via questionnaire data asking if parents had AD (arrow①). Despite the fact that longer educational attainment is a protective factor for developing AD, people with longer educational attainment are more prone to get diagnosed compared with those with shorter educational attainment (arrow②). Participants with longer educational attainment may have a tendency to report their parental AD status more precisely (arrow③). Longer educational attainment of the parents, often representing a high social-economic status, is possibly associated with the educational attainment of the participants (arrow④). Therefore, the backdoor path from the participants' educational attainment to AD status (proxy) opens via the parental educational attainment and parents' AD status. In addition, the genetic variants associated with the participants' educational attainment may also be associated with parental educational attainment through the parental genetic variants associated with parental educational attainment. This violates the assumption of Mendelian randomization, and consequently leads to a spurious association between educational attainment of participants and risk of the proxy-AD endpoint. AD: Alzheimer's disease.