Patient Guide · APOE ε4/ε4 · Understanding Your Risk
I Have Two Copies of APOE4. Am I Going to Get Alzheimer’s?
By Brian Paquette, DO, MPH
You will find two confident answers online, and they contradict each other. One says two copies makes Alzheimer’s effectively certain. The other says plenty of homozygotes never develop dementia. Both are drawing on real data. They disagree because the word “Alzheimer’s” is being used to mean two different things.
The short answer: Not certainly, and not on any schedule anyone can give you. Two copies of ε4 makes Alzheimer’s biology in the brain close to universal with age. It does not make dementia universal. How likely dementia is depends heavily on how far out you look: to the mid-eighties, most estimates land around a third to a half; followed all the way to 95, the best prospective cohort puts it at 59%. The gap between the biology and the symptoms is where your life actually happens, and it is measured in years to decades.
First: how uncommon is this?
Roughly 2% of people of European ancestry carry two copies — a figure both the major lifetime-risk analysis and the 2024 biomarker study converge on, with real variation by ancestry and geography (Genin et al., Molecular Psychiatry, 2011;16(9):903–907, DOI).
That has a practical consequence worth understanding before you read another number. In the four large cohorts assembled specifically to estimate risk for prevention trials, ε4 homozygotes numbered 158, 156, 67 and 11 participants. Every precise-looking figure you will be quoted was computed from groups that size. The confidence intervals around them are wide, and the estimates move substantially depending on who was studied (Qian et al., PLOS Medicine, 2017;14(3):e1002254, DOI). Precision in the presentation is not precision in the underlying number.
The 2024 finding that changed the conversation
If you have read anything alarming recently, this is probably its source. Fortea and colleagues pooled 3,297 brain donors with autopsy data and 10,039 people across five clinical cohorts with biomarker measurements, and asked a specific question: what happens to the brains of ε4 homozygotes as they age? (Fortea et al., Nature Medicine, 2024;30(5):1284–1291, DOI)
| What was measured | Finding in ε4 homozygotes |
| Alzheimer’s pathology at autopsy | Present in nearly all, at any age at death (about 50% in ε3/ε3) |
| Biomarker levels vs ε3/ε3 | Consistently higher from age 55 |
| Abnormal spinal-fluid amyloid by age 65 | Nearly all |
| Positive amyloid scan by age 65 | 75% |
| Ages at each clinical milestone | First symptoms 65.6 · MCI 71.8 · dementia 73.6 — about 7–10 years earlier than ε3/ε3 |
The authors concluded that the pattern in ε4 homozygotes resembles the inherited forms of Alzheimer’s disease more than it resembles a graded risk factor, and proposed that ε4/ε4 be treated as a distinct genetic form of the disease. That is a serious claim from serious investigators, and it is the strongest statement in this literature.
Why that is not the same as “you will get dementia”
Read the first four rows of that table again and notice what they measure. Plaques. Spinal fluid. Scan results. Every one describes biology in the brain. None of them says how many homozygotes lost their memory, stopped driving, or could no longer manage their own affairs.
This is the single most important distinction on this page. Having Alzheimer’s pathology and having Alzheimer’s dementia are different events. The first can be present for many years — often decades — before the second appears, and in some people the second never arrives. Headlines routinely collapse the two, which is how “nearly all homozygotes show the pathology” becomes “nearly all homozygotes get dementia.” The second sentence is not what the study found. The authors were explicit about the limit: their design was cross-sectional, and they state that they were unable to calculate the incidence of dementia for any genotype.
The same authors also name the reason the two numbers diverge, and it is worth sitting with: competing causes of death often arrive before the symptoms do. A brain can carry the pathology for a long time while the person carrying it dies of something else entirely, at a normal age, never having developed dementia.
So what are the actual odds of becoming symptomatic?
A range, not a number — and a wide enough range that you should be suspicious of anyone who gives you one figure. Note as you read that these estimates measure the emergence of symptoms. None of them contradicts the near-universal penetrance of the underlying biology; they describe how often, and how soon, that biology becomes a clinical illness.
| Source | What it estimated for ε4/ε4 |
| Genin 2011 Modelled from case–control data plus population incidence |
Risk of Alzheimer’s by age 85: 51% in men (95% CI 41–70), 60% in women (47–84). The same method gave 11% and 14% without reference to genotype, and 23–30% for a single ε4 copy. Substituting French incidence data raised the figure for homozygous women to 68%. |
| Qian 2017 Prospective population cohorts |
Among homozygotes still cognitively normal at 60–75, cumulative incidence by age 80–85 was 31–40% for dementia and 37–47% for mild cognitive impairment or dementia in the Framingham and Rotterdam cohorts. |
| Fang 2025 Prospective community cohort, followed to age 95 |
Lifetime risk of dementia from age 55 to 95: 59% with two ε4 copies (95% CI 53–65), 48% with one and 39% with none — against 42% for the whole cohort. Median age at diagnosis 79 in homozygotes vs 82 in non-carriers. |
| Reiman 2020 Autopsy-confirmed case–control |
Odds ratio for Alzheimer’s dementia versus ε3/ε3 of 31.2 when the diagnosis was confirmed at autopsy, and 10.7 when it was not. This is an odds ratio, not your probability. |
These estimates look like they disagree, and the reason matters more than the disagreement. All three accounted for the fact that people die of other causes. What differs is design and, above all, how far out each one looks. Genin modelled risk from genotype frequencies combined with published incidence rates and stopped at 85. Qian followed community-dwelling people forward but also stopped at 80–85. Fang followed an American community cohort for a median of 23 years and kept counting to 95.
Fang’s own paper makes the reconciliation explicit: its figures are consistent with the earlier population studies, which reported 16–38% cumulative incidence by age 80–85. Nothing contradicts anything. Dementia is overwhelmingly a disease of the late eighties and nineties, so the longer you follow anyone — carrier or not — the higher the number climbs. In Fang’s cohort only 3.9% of everyone had developed dementia by 75.
What a trial actually tells homozygotes: the Generation Study, a prevention trial that enrolled ε4/ε4 individuals and therefore had to decide what to say to them at genetic disclosure, settled on a lifetime risk of mild cognitive impairment or dementia of 30–55% for ε4/ε4, against 20–25% for one copy and 10–15% for none. Round numbers, deliberately broad, chosen to reflect real uncertainty rather than to sound authoritative. If you want one answer to take to a conversation with your own physician, that is the most defensible one available.
About that terrifying multiplier
An odds ratio of 31 does not mean a 31-fold chance of anything happening to you. It compares two groups against each other in a study, and it moves sharply with how those groups were assembled. In Reiman’s data the same genotype yielded 31.2 when cases and controls were sorted by autopsy and 10.7 when they were sorted clinically — and the authors argue the autopsy figure is the more accurate one, with clinical misdiagnosis diluting the other. Both are correct answers to a question that is not the question you are asking.
Here is the number almost nobody quotes. Those large multipliers compare you to people with the lowest-risk common genotype. Compared with the general population — your friends, your neighbors, people of unknown genotype — Qian’s regression models put the adjusted relative risk of mild cognitive impairment or dementia for an ε4 homozygote at about threefold (2.4 to 3.4 across cohorts). Their recommendation was that a threefold comparison is the honest one to disclose. Three is a serious number. It is not thirty-one.
And measured as absolute lifetime risk, the gap is smaller still. Over a full lifetime to 95, Fang found dementia in 59% of homozygotes against 39% of non-carriers and 42% of the cohort as a whole — a difference of roughly twenty percentage points. That is a large and real difference. It is not a multiple. Ratios shrink as the horizon lengthens, because lifetime risk is capped and everyone’s risk rises steeply with age; a hazard ratio and a lifetime risk ratio are answering different questions, and only the second one is about your life.
The question underneath the question is usually “when”
Most people asking whether they are going to get Alzheimer’s are really asking how much time they have. Here the literature is more consistent, and the answer arrives earlier than most people expect.
In Fortea’s data, homozygotes reached first symptoms at 65.6, mild cognitive impairment at 71.8 and dementia at 73.6. In Reiman’s autopsy-confirmed cases, mean age at dementia onset was 69.9 years, against 79.3 in those carrying an ε2 allele without an ε4 (Reiman et al., Nature Communications, 2020;11(1):667, DOI). Different designs and different milestones, but the same picture: for homozygotes who do become symptomatic, it happens roughly seven to ten years earlier than it otherwise would.
The practical consequence is not despair. It is timing. Fortea’s curves show spinal-fluid amyloid already diverging in participants in their late forties and tau markers in the early fifties — earlier even than the age-55 figure usually quoted. If the biology is measurably underway that early and symptoms cluster around 65, then the years that matter for anything you might do are your forties and fifties, not the year you first notice a problem.
Does it matter whether I’m a man or a woman?
Less than you may have read, at least for homozygotes specifically. A meta-analysis of nearly 58,000 participants detected no significant difference between men and women with two ε4 copies — though homozygotes are the smallest group in that analysis, so this is a failure to find a difference rather than proof there is none. The much-discussed female excess appeared in single-copy carriers, and was confined to a window: roughly ages 65 to 75 for Alzheimer’s, and 55 to 70 for mild cognitive impairment (Neu et al., JAMA Neurology, 2017;74(10):1178–1189, DOI).
Genin’s lifetime figures do show a gap for homozygotes — 51% in men against 60% in women by 85 — but both were calculated to the same fixed age, so this is not women simply having more years in which to develop the disease. It tracks the higher age-specific incidence of Alzheimer’s in women generally: the same gap appears in the 11%-versus-14% figures with no reference to genotype at all.
Does my ancestry change these numbers?
Yes — though the ancestry estimates are themselves imprecise. Nearly every figure on this page comes from participants of European ancestry, and Fortea’s authors say so directly about their own work. In the foundational pooled analysis, the ε4/ε4 odds ratio was 14.9 (95% CI 10.8–20.6) in clinic-based samples of European ancestry against 33.1 (13.6–80.5) in Japanese participants, with a weaker association in African American and Hispanic cohorts. Those confidence intervals overlap heavily, so the contrast is real but not sharply quantified. This is covered properly in APOE ε4 Risk Is Not the Same in Every Population.
What I would want said to me
1. This is a probability, not a sentence — but the horizon matters. Followed to the mid-eighties, a majority of ε4 homozygotes are still free of dementia. Followed to 95, a majority are not: 59% developed it in the best prospective cohort, against 39% of non-carriers. Both statements are true, and which one is relevant to you depends on how long you live. Nobody can tell you today which group you are in, and anyone who claims to can be safely ignored.
2. The number you were quoted is probably too confident in one direction or the other. Fatalism and false reassurance are both departures from the data. The defensible answer is a range, and it is a wide one.
3. Attend to the modifiable things in midlife — while knowing what that has and has not been shown to do. Blood pressure, lipids, glucose, sleep, hearing and fitness are worth managing on their own merits, and the large lifestyle trials found their benefit was preserved in ε4 carriers rather than blunted. But no intervention has been shown to prevent Alzheimer’s pathology or normalize dementia risk in carriers, and at least one large cohort found the protective association with a favorable risk profile absent in the highest genetic-risk group. This is covered honestly in Can APOE4 Risk Be Normalized?
4. Knowing the biology starts early is not an instruction to go get scanned. Routine biomarker testing in asymptomatic carriers is not established clinical care, and a test should answer a question that changes what you do. See Alzheimer’s Blood Tests vs. APOE Genetic Testing.
5. Do not let the result colonize the next thirty years. If the genotype is producing persistent anxiety or preoccupation, that is worth raising with your physician. It is a common and treatable response to genetic risk information, not a character flaw — and thirty anxious healthy years is a real cost, paid in advance, for a disease you may not get.
One further point matters specifically for homozygotes. There is one setting in which ε4/ε4 changes clinical decisions today rather than someday: the anti-amyloid antibody therapies. In the pivotal lecanemab trial, amyloid-related imaging abnormalities with brain swelling or effusion occurred in 12.6% of treated participants overall, and APOE genotype is a formal part of how candidacy and monitoring are decided (van Dyck et al., New England Journal of Medicine, 2023;388(1):9–21, DOI). If that conversation is ever on the table for you, your genotype belongs in it.
What to actually do — which measurements are worth having, what the lifestyle trials really showed, and where the supplement evidence collapses — is set out in I Have APOE4. What Do I Do Now? And if the question you really want answered is why some carriers never become symptomatic at all, that is taken apart in Why Do Some APOE4 Carriers Never Develop Dementia? — with the clinician-level version of the same evidence in The ε4/ε4 Carriers Who Do Not Develop Alzheimer’s: Resistance and Resilience, which makes the important point that escaping the pathology is rare, while remaining well despite it is the phenomenon actually worth studying.
Limitations
- Every risk figure here rests on small numbers of homozygotes. Confidence intervals are correspondingly wide — Genin’s 60% estimate for women carries limits of 47–84% — and estimates shift with the cohort studied.
- The Genin and Qian estimates stop at ages 80–85 and say nothing about those who live beyond that; Fang’s runs to 95. Comparing them without noticing the different endpoints is the single easiest way to misread this literature.
- Fang’s cohort contained only 402 people with two ε4 copies, and about half its dementia cases were identified by telephone interview rather than expert adjudication. Under a more conservative case definition the overall lifetime figure fell from 42% to 36%.
- Genin’s figures are modelled rather than observed; Qian’s are observed but begin only with people who were still cognitively normal in their sixties or seventies, which selects for those already doing well.
- Odds ratios from case–control studies cannot be converted into individual probabilities. Reiman’s are included here only to show how sensitive such figures are to study design.
- Fortea’s study reports symptom onset as 65.1 in its abstract and 65.6 in its results; the figure quoted above is the one in the results text. Its cohorts were convenience samples with opposing biases, and the design was cross-sectional.
- Nearly all of this evidence comes from participants of European ancestry.
- Nothing here accounts for family history, other genetic variants, education, or cardiovascular and metabolic health — all of which shift an individual’s position within these ranges by amounts genotype alone cannot capture.
References
- Fortea J, Pegueroles J, Alcolea D, et al. APOE4 homozygosity represents a distinct genetic form of Alzheimer’s disease. Nat Med. 2024;30(5):1284–1291. doi:10.1038/s41591-024-02931-w (Publisher Correction: Nat Med. 2024;30(7):2093. doi:10.1038/s41591-024-03127-y)
- Genin E, Hannequin D, Wallon D, et al. APOE and Alzheimer disease: a major gene with semi-dominant inheritance. Mol Psychiatry. 2011;16(9):903–907. doi:10.1038/mp.2011.52
- Qian J, Wolters FJ, Beiser A, et al. APOE-related risk of mild cognitive impairment and dementia for prevention trials: an analysis of four cohorts. PLoS Med. 2017;14(3):e1002254. doi:10.1371/journal.pmed.1002254
- Fang M, Hu J, Weiss J, et al. Lifetime risk and projected burden of dementia. Nat Med. 2025;31(3):772–776. doi:10.1038/s41591-024-03340-9
- Reiman EM, Arboleda-Velasquez JF, Quiroz YT, et al. Exceptionally low likelihood of Alzheimer’s dementia in APOE2 homozygotes from a 5,000-person neuropathological study. Nat Commun. 2020;11(1):667. doi:10.1038/s41467-019-14279-8
- Neu SC, Pa J, Kukull W, et al. Apolipoprotein E genotype and sex risk factors for Alzheimer disease: a meta-analysis. JAMA Neurol. 2017;74(10):1178–1189. doi:10.1001/jamaneurol.2017.2188
- Farrer LA, Cupples LA, Haines JL, et al. Effects of age, sex, and ethnicity on the association between apolipoprotein E genotype and Alzheimer disease: a meta-analysis. JAMA. 1997;278(16):1349–1356. doi:10.1001/jama.1997.03550160069041
- van Dyck CH, Swanson CJ, Aisen P, et al. Lecanemab in early Alzheimer’s disease. N Engl J Med. 2023;388(1):9–21. doi:10.1056/NEJMoa2212948
EDITORIAL NOTE: Every figure quoted above was checked against the full text of the primary source, not the abstract. Where sources conflict, the conflict is shown rather than resolved. Brian Paquette, DO, MPH is a double board-certified neurologist and pain medicine physician, a former US Naval Flight Surgeon, and an APOE ε4 carrier. APOE4 Insights accepts no industry sponsorship and has no supplement affiliations. Educational content only; not a substitute for individual clinical judgment or genetic counseling.