Blue Zones: What Survives the Data Audit

Five places are said to produce old people at a rate the rest of the world does not: a mountain district in Sardinia, Okinawa, the Greek island of Ikaria, the Nicoya peninsula in Costa Rica, and a town in California. The habits credited for it are famous — beans, walking, family, purpose. How the places themselves were found is less famous. Somebody counted what fraction of the people born in each district around 1890 reached one hundred, then shaded the districts where that fraction looked extraordinary. A Blue Zone is a shape drawn on a map out of birth and death records. The beans and the walking came afterwards, inferred from the populations that shape had already selected. So the quality of the paperwork is not a footnote to the argument. It is the argument.

Fourth in a series on the biology of ageing: gene editing, telomeres, epigenetics, the Blue Zones, and cellular reprogramming.

Where the zones came from

The term belongs to Michel Poulain and Giovanni Pes, from a 2004 paper in Experimental Gerontology. They worked from Sardinian municipal records — the registers each town keeps of births and deaths.

Their measure was one number per municipality, an extreme longevity index: the percentage of Sardinians born there between 1880 and 1900 who became centenarians. One village with two very old residents can make a small population look spectacular, so they smoothed the map with a Gaussian kernel: each district’s value pulled towards its neighbours’, near neighbours counting for more. What survived was the mountainous centre-east, across Ogliastra and Barbagia, which they drew in blue pen.

The headline count was not the anomaly. The sex ratio was.

Male centenarians are scarce almost everywhere: reach one hundred and the odds are strong that you are a woman. In Sardinia outside the zone, 2.43 women for every man; inside it, 1.35 to 1. A district that close to parity at one hundred needs explaining.

Dan Buettner extended the label to four more places — Okinawa, Ikaria, Nicoya, and the Adventist community of Loma Linda in California — and distilled a habit list from all five.

Eating largely plant-based, with legumes as the staple and meat roughly five times a month.
Activity built into terrain and work, not performed in a gym.
Stopping eating at about eighty per cent full.
Multigenerational households, faith communities, a sense of purpose.

Which habits stand on their own

Three items on that list have an evidence base that does not depend on the Blue Zones existing at all.

How to read these numbers

A hazard ratio compares how fast people are dying in two groups over the same stretch of time. 1.00 means both are dying at the same rate; 0.69 means the first is dying at 69 per cent of the second’s rate, roughly a third fewer deaths per unit of time. Lower is better. A relative risk (RR) is a different quantity but reads the same way in this range.

An odds ratio does similar work with different arithmetic, and its direction depends on what is counted. An odds ratio of 1.50 for survival means higher is better: the odds of still being alive at the end of follow-up are half again as high. As a ratio for dying, the same finding would come out below 1.00.

The 95 per cent confidence interval is the pair of numbers in brackets after the estimate — the range of values the data are reasonably consistent with. Check whether it crosses 1.00: if it does, the result is compatible with no effect at all.

The three that hold

Legumes. The Food Habits in Later Life cohorts followed 785 people over seventy, in Japan, Sweden, Greece and Australia, for up to seven years. Each additional 20 g of legumes a day was associated with a 7 to 8 per cent reduction in the mortality hazard ratio: RR 0.93, 95% CI 0.87–0.99. No other food group predicted survival consistently across all four cohorts.

Read the interval and the finding is more modest than the headline: the upper end, 0.99, sits only just below 1.00. And people who eat more legumes differ from those who eat fewer in ways an observational design cannot fully account for. Established as an association, not proof that the beans do the work.

Social connection. A meta-analysis pooled 148 studies covering 308,849 people, mean follow-up 7.5 years. Those with stronger social relationships had an odds ratio for survival of 1.50, 95% CI 1.42–1.59. The interval is narrow and nowhere near 1.00. Established, and large.

Observational too, and here reverse causation is a live worry, not a formality: illness produces isolation as readily as isolation produces illness — the already ill withdraw. The design cannot separate the two directions.

Physical activity. A pooled analysis of 661,137 adults recorded 116,686 deaths over a median of 14.2 years, and found a clean dose-response. The yardstick is the recommended weekly minimum of leisure-time activity; the hazard ratios fall as people clear more of it.

0.80 for any leisure activity at all, even below the recommended minimum.
0.69 at one to two times the minimum.
0.61 at three to five times it. That is the floor.

Past three to five times the minimum the benefit stopped accruing — more exercise bought nothing further, and cost nothing either: no excess risk appeared even at ten times the minimum.

None of these three needs Sardinia. They would survive intact if every Blue Zone turned out to be a clerical accident. Which is what the audit alleges.

The audit

An age is not a measurement. Nobody observes a person being 110. What exists is a chain of paperwork from a birth to a published record, and every link can fail without anything happening in the biology.

Read down the left column for how a claimed age of 110 gets onto the record, and across for what can go wrong at each step. Steps 2 to 5 are the failure points Newman documents; the 18% is his (bioRxiv preprint 10.1101/704080, 2019, revised 2024 — not peer reviewed).

Saul Justin Newman, a demographer at UCL, won the first Ig Nobel Prize in Demography on 12 September 2024, for work arguing that extreme old-age data are likely dominated by errors. The Ig Nobels are satirical in form; the citation was not.

His central result is about registration, not biology, and rests on an accident of American administration: birth certificates arrived in the United States state by state, at different dates. That gives the same population before and after documentation existed, and lets you ask what the documents did to the numbers.

They removed them. Introducing birth certificates in a state is associated with a 69 to 82 per cent fall in that state’s supercentenarian records. Roughly three quarters of the extremely old disappeared — not from the population but from the files, at the moment the files became checkable.

The second piece is what remarkable ages track where registration is uniform. In Italy, England and France they are predicted by poverty, low income, higher crime, worse health and fewer 90-year-olds. Every one points the wrong way for a longevity signal, the last most sharply: districts reporting the most people at extreme ages have fewer at ninety, an age you have to get past first. In England and France, old-age poverty alone predicts more than half the regional variation in reaching a remarkable age.

The third piece is the documents. Only 18 per cent of exhaustively validated supercentenarians hold a birth certificate, and in the United States none do. Their birthdates cluster on days divisible by five, the signature of a number recalled rather than recorded. Sardinia, Okinawa and Ikaria all sat below their national averages on income, literacy and life expectancy, and above on crime.

Put the three together and the argument reduces to a sentence: the best available predictor of a claimed extreme age is the absence of a document that could check it. On Newman’s reading, that is not the shape of a biological signal but of a record-keeping artefact.

Four numbers from the audit, all on the same 0–100 per cent scale, all Newman’s (bioRxiv preprint 10.1101/704080, 2019, revised 2024 — not peer reviewed). Violet on the top bar marks the upper end of the 69–82% range.

Two caveats, and both matter. The first is status: this is a preprint, posted in 2019 and revised as recently as 2025, not peer reviewed, whatever the Ig Nobel implies about its reach. The second is scope: the claim is not that poverty manufactures centenarians, but that missing documents and pension incentives inflate the records — that many of these people were alive in government files and dead in fact.

Okinawa, and the registers that burned

Okinawa is the awkward case. Its civil registers, the koseki, were destroyed in the 1945 battle and rebuilt afterwards: the documents an Okinawan age would be checked against were themselves reconstructed.

Michel Poulain — the same demographer who drew the first zone — reviewed Okinawan longevity in Demographic Research in 2011 and came back with a tie. The mortality pattern fits two explanations equally well: a genuine cohort effect, something real about that generation, or bad data. Reconstruction of the koseki is a plausible source of age misreporting, but no proof of invalid data exists either. He recommended in-depth validation, not a verdict.

Newman adds a detail that sits badly with the diet story: Okinawans eat the least vegetables in Japan and carry the highest body mass index — weight scaled against height — in the country.

The other side of the ledger

In December 2025 Steven Austad and Giovanni Pes replied in The Gerontologist, and the reply is not a brush-off.

Their case is that the ages in the original zones were never taken on trust: each was cross-checked against several independent traces of the same life:

civil records
ecclesiastical archives
genealogical reconstruction of the family line
interviews

That is the prevailing standard of gerontological demography, and the team applied it itself, striking out fraudulent claims as it found them.

They concede a great deal. Most extreme age claims anywhere are false. Zones are not permanent either: Okinawa’s has faded, and a new one has been described in the Netherlands.

Worth holding in mind while reading Pes defend the Sardinian zone: he is a co-discoverer of it — which also means he handled the parish books himself.

The two sides are not quite arguing about the same object. Newman’s evidence is a statistical pattern across whole populations of records; Austad and Pes defend a particular set of individuals whose documents they say were checked one at a time.

That is what contested looks like: two competent parties, the same parish books, no resolution.

What to take away

Two claims are bundled inside the phrase Blue Zone, and they are not equally strong.

The demographic claim is that certain districts produce centenarians at rates demanding a local explanation. It is contested, and the case that it is largely an artefact of missing paperwork won the discipline’s first Ig Nobel.

The lifestyle claims do not fall with it. Legume intake, social connection and physical activity each carry their own large-cohort evidence. None of it depends on Sardinia, and all of it is observational — association, not demonstrated cause. It is a natural mistake to think the audit threatens the beans; what it threatens is the story that a particular village proves them.

The honest version is unglamorous. Eat the legumes, walk up the hill, answer the phone — not because a mountain district in Ogliastra proves it works, but because several hundred thousand ordinary people already did.