Modifiable risk factors and polyexposure scores for dementia

Evaluating accessible risk scores and causal evidence for dementia-prevention targets

Polyexposure scores

Many clinical, lifestyle, and environmental factors contribute to dementia risk, but their individual effects are often small and interrelated. Polyexposure scores combine these factors into an overall measure of modifiable risk that can be calculated without expensive imaging or laboratory tests. My work has evaluated whether these scores predict cognitive impairment, whether they reflect Alzheimer’s disease biology, and whether scores developed in one population generalize to others.

In a population-based cohort followed for 12 years, I showed that the Australian National University Alzheimer’s Disease Risk Index predicted progression from normal cognition to mild cognitive impairment. I subsequently found that this index was broadly associated with general cognitive ability, dementia-related cognitive variation, and performance across multiple cognitive domains. More recent work demonstrated that risk scores do not perform uniformly across populations: the original CAIDE score was associated with dementia in Asian, Latinx, and non-Hispanic White participants but not Black participants, whereas a modified version showed broader validity. In a diverse community cohort, four polyexposure scores were associated with dementia across racial and ethnic groups, but CogDrisk showed the most consistent associations across diagnosis, cognition, plasma biomarkers, and neuroimaging. These findings support polyexposure scores as accessible tools for risk assessment while showing that their validity must be established in the populations in which they will be used.

Mendelian randomization

Risk scores are most useful for prevention when the factors they include contribute causally to disease, rather than simply appearing to predict it because of confounding or changes that occur during the early disease process. I use Mendelian randomization, which employs inherited genetic variants as proxies for an exposure, to test whether reported risk factors are likely to influence Alzheimer’s disease and related outcomes.

I led a comprehensive analysis of 22 modifiable factors across the “Alzheimer’s phenome,” including diagnosis, age at onset, brain structure, cerebrospinal-fluid biomarkers, neuropathology, and vascular brain injury. The results supported causal roles for education, blood pressure, cholesterol, smoking, and diabetes, while also showing that a factor can affect one component of the disease process without affecting all others. Subsequent studies applied this framework to specific, sometimes conflicting observational associations. We found no evidence that alcohol consumption lowers Alzheimer’s disease risk and instead found evidence that greater consumption may bring forward age at onset. Apparent protection associated with snoring was most consistent with reverse causation and weight loss during prodromal disease. Sleep apnea increased cardiovascular risk but was not directly associated with Alzheimer’s disease in genetic analyses, and the protective association of education remained after accounting for differences in research participation. Together, this work separates promising prevention targets from associations that may reflect bias, reverse causation, or vascular pathways, providing a stronger evidence base for the factors included in dementia risk-reduction strategies.

Selected publications

  • Andrews SJ, Boeriu AI, Belloy ME, et al. (2024). “Dementia risk scores, apolipoprotein E, and risk of Alzheimer’s disease: One size does not fit all.” Alzheimer’s & Dementia. doi:10.1002/alz.14300.
  • Okorie M, Jiang X, Yaffe K, Yokoyama JS, Andrews SJ. (2026). “Associations of dementia polyexposure scores to Alzheimer’s disease endophenotypes in a diverse population.” Alzheimer’s & Dementia. doi:10.1002/alz.71567.
  • Andrews SJ, Eramudugolla R, Velez JI, et al. (2017). “Validating the role of the Australian National University Alzheimer’s Disease Risk Index (ANU-ADRI) and a genetic risk score in progression to cognitive impairment.” Alzheimer’s Research & Therapy. doi:10.1186/s13195-017-0240-3.
  • Andrews SJ, McFall GP, Dixon RA, et al. (2019). “Alzheimer’s Environmental and Genetic Risk Scores are Differentially Associated With General Cognitive Ability and Dementia Severity.” Alzheimer Disease & Associated Disorders. doi:10.1097/WAD.0000000000000292.
  • Andrews SJ, Fulton-Howard B, O’Reilly P, Marcora E, Goate AM. (2021). “Causal Associations Between Modifiable Risk Factors and the Alzheimer’s Phenome.” Annals of Neurology. doi:10.1002/ana.25918.
  • Andrews SJ, Goate A, Anstey KJ. (2020). “Association between alcohol consumption and Alzheimer’s disease: A Mendelian randomization study.” Alzheimer’s & Dementia. doi:10.1016/j.jalz.2019.09.086.
  • Gao Y, Andrews SJ, et al. (2024). “Snoring and risk of dementia: a prospective cohort and Mendelian randomization study.” Sleep. doi:10.1093/sleep/zsae149.
  • Chatterjee A, Cavaillès C, Davies NM, Yaffe K, Andrews SJ. (2025). “Disentangling the Causal Effects of Education and Participation Bias on Alzheimer Disease Using Mendelian Randomization.” Neurology: Genetics. doi:10.1212/NXG.0000000000200307.
  • Cavaillès C, et al. (2024). “Causal Associations of Sleep Apnea with Alzheimer’s Disease and Cardiovascular Disease: a Bidirectional Mendelian Randomization Analysis.” Journal of the American Heart Association. doi:10.1161/JAHA.123.033850.
  • Ferguson EL, et al. (2024). “Relationships of visual impairment and eye conditions with imaging markers, cognition, and diagnoses of dementia: a bi-directional Mendelian randomization study.” JAMA Network Open. doi:10.1001/jamanetworkopen.2024.24539.