APOE isoforms ε2/ε3/ε4 and their effect sizes
A single gene variant shapes Alzheimer risk more than any other common one.
APOE (apolipoprotein E) is the strongest common genetic risk factor for late-onset Alzheimer's; three isoforms (ε2, ε3, ε4) arise from two amino-acid positions (residues 112/158). Relative to the ε3/ε3 reference, one ε4 allele raises risk roughly 3–4-fold and two ε4 alleles roughly 9–15-fold, each with earlier mean onset; ε2 is protective, roughly halving risk. Effect sizes vary by ancestry — the ε4 association is weaker in African- and Hispanic-ancestry populations and stronger in East-Asian and European ancestry — and a large share of patients carry no ε4 allele. APOE is a risk modifier, not deterministic.
The molecular mechanism of APOE4
A small structural change in the protein disrupts several of the brain's defence systems.
A single amino-acid change drives a "domain interaction" that alters lipid and receptor binding. Consequences include impaired amyloid-β clearance and enhanced aggregation, dysregulated lipid/cholesterol homeostasis (astrocytic cholesterol transport, the ABCA1 axis, myelin), microglial dysfunction and neuroinflammation via the TREM2–APOE axis, and blood-brain-barrier compromise via the cyclophilin-A–MMP9 pathway, alongside facilitated tau hyperphosphorylation. APOE2's protective biology includes reduced Aβ aggregation and favourable lipid handling.
Polygenic risk scores beyond APOE
Many small genetic effects together form an individual risk profile.
Genome-wide association work has identified dozens of common risk loci beyond APOE, implicating amyloid processing, tau, immune/microglial response, lipid metabolism and endocytosis. A polygenic risk score (PRS) sums risk alleles weighted by effect size. A global PRS predicts diagnosis most strongly, because loci act synergistically.
Pathway-specific polygenic scores
Instead of one number, they show which biological mechanism dominates.
Pathway-specific PRS partition the genetic burden into biological processes — for instance an immune, a lipid or an amyloid score. Evidence suggests immune-pathway scores associate more with cognitive decline, amyloid-pathway scores with amyloid positivity, and lipid-pathway scores with vascular dementia. Their value lies more in mechanistic understanding than in raw predictive power.
Multidomain lifestyle prevention & gene-by-intervention effects
Several levers at once — and the question of who benefits most.
Structured multidomain programmes — physical activity, healthy diet, vascular/cardiometabolic risk control, and cognitive-social engagement — can improve cognition in at-risk individuals, with structured programmes outperforming self-guided ones over about two years. Central is the gene-by-intervention interaction: some analyses show benefit consistent across ε4 status, pooled analyses a larger effect in ε4 carriers — an unresolved but actionable tension. Pharmacological adjuncts via metabolic-inflammatory pathways (e.g. incretin/GLP-1-receptor mechanisms) are being explored; the fact that such an agent reduced inflammatory markers in established disease yet did not slow clinical decline shows that mechanism-positive is not clinically effective — and that timing (prevention vs treatment) matters.
Modifiable risk factors across the life course
A substantial share of dementia cases could, in theory, be prevented.
Estimates suggest around 45% of dementia worldwide is potentially preventable by addressing 14 modifiable risk factors across the life course (a population-attributable-fraction concept). These range from low early-life education through midlife hearing loss, hypertension, obesity, alcohol, traumatic brain injury and elevated LDL cholesterol, to late-life smoking, depression, inactivity, diabetes, social isolation, air pollution and untreated vision loss.