The largest study of how long people live looked at a million parents. What came out top was not a repair gene or a telomere gene — it was Alzheimer's risk, then smoking, then cholesterol, then heart disease, then blood pressure. There is no longevity gene in the results. There are disease genes, and not dying of those things is most of what a long life is.
Everybody wants there to be a longevity gene. The largest studies ever run have looked, and what they find instead is worth understanding properly — because it is more useful than the thing people hoped for, and it contradicts most of what is sold under the word.
The 2019 study did something clever to get the sample size. Asking people how long they live means waiting for them to die; asking people how long their parents lived is a question you can put on a questionnaire today. From about 500,000 genotyped people, that yields a million parental lifespans.
Here is what came out, strongest first.
Not one of them is about ageing as a process. Every one is about a specific way of dying earlier: dementia, smoking, cholesterol, blood pressure, heart disease.
Lifespan genetics is mostly disease genetics. At the level of a population, living a long time consists largely of not getting the common fatal illnesses — and the genetic variation in how long people live is, to a large extent, the genetic variation in susceptibility to those illnesses.
That is a real finding, not a disappointing one. It says the lever is the diseases, and it is the same lever medicine has been pulling for a century.
It also disposes of a common intuition. If there were a general ageing dial, we would expect to find variants that shift everything at once. What we find is a list of specific hazards.
FOXO3 is the most-cited longevity gene in the world. It appears in hundreds of papers, in supplement marketing, and in every popular article about the genetics of ageing.
Search the public GWAS Catalog for its best-known variant, rs2802292, and the only genome-wide significant association recorded is with platelet component distribution width. Nothing about lifespan.
That is not proof it does nothing — the million-lifespan study lists FOXO3 among previously reported loci it validates, and gene-level analyses are a different test from single-variant ones. But the honest statement is this: the most famous longevity gene does not appear in the association table of the largest study of longevity, and we are not going to publish it as though it did.
We would rather show you an absence than fill it.
Beyond the individual positions, the study looked at which tissues and pathways the signals cluster in. Two answers stand out.
Brain tissue — expression in foetal brain cells and in adult dorsolateral prefrontal cortex is enriched among lifespan-associated variants. Some of that is APOE and dementia. Some of it is not, and it is not fully explained.
Lipid handling, vesicle transport and synaptic function — the pathways that come up are about moving fats and moving cargo inside cells, which is the same machinery that keeps turning up in cardiovascular and neurodegenerative disease.
Less than people assume. Estimates from twin and family studies have historically run around 25%, and more careful work using large pedigrees has argued that even that is inflated — because people who marry each other resemble each other, and that resemblance gets counted as genetic when it is not.
So the honest framing is: how long you live is mostly not written in your genome. The part that is, is largely about which diseases you are prone to. And the part that is neither is where almost all the actual difference lives — smoking, blood pressure, weight, income, medical care, and luck.
It is not a predictor. Nobody can read a lifespan off a genotype, and the effect sizes here are shifts in an average across half a million families.
It is not a shopping list. No supplement on the market has been shown to extend human lifespan, and none of these positions is a reason to take one.
And it is not a reason to worry. If you want to act on anything in this page, act on the things it points at — the same ones a doctor would name without looking at your genome at all.
The primary source. Timmers et al. (eLife 2019) performed a genome-wide association analysis of 1 million parental lifespans in at least 500,193 European-ancestry offspring (GWAS Catalog GCST009890). The study validated previously unreplicated findings near CDKN2B-AS1, ATXN2/BRAP, FURIN/FES, ZW10, PSORS1C3 and 13q21.31, and identified and replicated novel associations near ABO, ZC3HC1 and IGF2R. It validated earlier findings at 5q33.3/EBF1 and FOXO3 while reporting contradictory evidence at other previously claimed loci. Gene-set and cell-specific analyses showed enrichment for expression in foetal brain cells and adult dorsolateral prefrontal cortex, and for pathways involving lipid proteins and homeostasis, vesicle-mediated transport and synaptic function.
The variants recorded here. rs8042849 (CHRNA3/CHRNA5, P = 2e-26), rs10455872 (LPA, P = 9e-25), rs11065979 (ATXN2/BRAP, P = 1e-12), rs1556516 (CDKN2B-AS1, P = 7e-11), rs6224 (FURIN/FES, P = 1e-10) and rs61348208 (near HTT, P = 6e-9). The strongest signal in the study, rs429358 in APOE (P = 3e-83), is not duplicated as a separate row: this site already carries the APOE e2/e3/e4 haplotype built from rs429358 and rs7412 read together, and listing one of those two positions again under a second story would put the same variant on the site twice.
The strict-longevity phenotype. Deelen et al. (Nature Communications 2019) took a different approach: rather than parental age at death, they defined cases as individuals surviving at or beyond the age corresponding to the 90th survival percentile (11,262 cases) or the 99th (3,484), against 25,483 controls whose age at death or last contact was at or below the 60th percentile. rs429358 (APOE e4) associated with lower odds of surviving to both thresholds; rs7412 (e2) showed the opposite direction; rs7676745 near GPR78 associated with lower odds at the 90th percentile. The concordance of APOE across both study designs is the most robust single finding in the field.
FOXO3. rs2802292 is the most frequently cited longevity-associated variant in the literature. In the GWAS Catalog its only genome-wide significant association is with platelet component distribution width (P = 2e-21); no lifespan or longevity association is recorded for it. Timmers et al. list FOXO3 among validated loci, and gene-level tests differ from single-variant tests, so this is not evidence of absence — but it is not the evidence that is usually implied, and the variant is therefore not published on this site as a longevity marker.
Clinical boundary. No genotype on this page is used in any clinical assessment of life expectancy, and no guideline references these positions. Reported effect sizes are differences in mean parental lifespan across a population of half a million families; they have no interpretation at the level of an individual. Nothing here supports the use of any supplement or intervention marketed for longevity.
What a 23andMe/AncestryDNA export or raw VCF can and can't tell you about Human Lifespan and Longevity comes down to these specific, well-studied positions — not a diagnosis.
CDKN2B-AS1 · rs1556516
See detailed info →CHRNA3 / CHRNA5 · rs8042849
See detailed info →Not in the results. The largest study of human lifespan found APOE first, then a smoking variant, then lipoprotein(a), then heart disease, then blood pressure. Every one is a specific way of dying earlier rather than a dial on ageing itself.
Its best-known variant, rs2802292, has one genome-wide significant association in the public GWAS Catalog — with platelet component distribution width. Nothing about lifespan. The million-lifespan study does list FOXO3 among loci it validates, and gene-level tests are not the same as single-variant ones, so this is not proof it does nothing. But it is not the evidence people imply, and we would rather show you the gap.
Less than most people think. Twin and family estimates cluster around 25%, and work with large pedigrees suggests even that is inflated because people partner with people like themselves. Most of the difference between lives is not in anyone's genome.
No. What was measured is a difference in average parental lifespan across half a million families — months, not years, and invisible in any one family. No clinical assessment anywhere uses these positions.
No. No supplement has been shown to extend human lifespan, and nothing on this page is a reason to take one. The things this page actually points at — smoking, blood pressure, cholesterol — are the ones a doctor would name without looking at your genome at all.
It is a common variant near HTT, not the CAG repeat expansion that causes Huntington's disease, and it has nothing to do with that diagnosis. What it is doing in this list is not established — which is why it is here with that said plainly rather than with a story attached.
Free to reuse. This page's text is original writing from freely-available research, licensed CC BY 4.0 — reuse it, including commercially, with attribution to MyGeneLog. It's general research-derived information, not medical advice or a diagnosis — see Terms of Use.