The Broad Prevalence of Bad Epidemiological Data for Exceptional Human Life Expectancy
In recent years, greater attention has been given to efforts that push back against the present broad acceptance of established data on human life expectancy, particularly for the oldest surviving cohorts. It has been suggested, and the evidence for this assertion is broadly supportive, that the published data for exceptional longevity is largely of poor quality, and much of what has been hyped over the years (such as Blue Zones or Jeanne Calment's alleged life span of 122 years) is simply not real.
What is observed in the data is a selection effect for error, fraud, and outright falsehood that grows stronger at advancing ages. We should be quite confident that a small number of humans can survive into their 110s, as individual cases have been well vetted, but we should be much less confident about the accuracy of demographics of survival much past age 90.
Does any of this really matter? From the perspective of building therapies to treat aging, I think probably not. It doesn't affect the need for better ways to measure biological age than exist at present, and it doesn't change the list of programs and targets that should be undertaken to produce potential rejuvenation therapies. People do get somewhat up in arms about the demographics of aging, but it seems a tempest in a teacup to me, somewhat irrelevant to the real issue of making progress in the treatment of aging as a medical condition. Other people may see it differently, of course.
How long can humans live? We simply don't know
Many errors are undetectable and, therefore, we do not know their underlying frequency. This has prompted a rather absurd response from demographers, who say that, sure, some errors occasionally escape detection, but these errors must be rare. I usually ask them: if you cannot detect particular errors, how do you know that they are rare? The core problem is that age relies on one measurement system: paperwork. If a person's paperwork is consistent but wrong, there is no reproducible way of knowing. You often see a famous case discussed, the details exhaustively validated and all of the paperwork examined. But after decades, the case turns out to be false. It has passed every test that demography has, and it is still wrong.
I did not just observe this in individual cases. I found it in entire populations. In Greece, for example, at least 72% of centenarian records were cases of pension fraud. The person was left alive on paper while their younger relatives collected the pension cheques. That was the secret to longevity in Greece, and nobody in demography saw it for decades.
There are several overlapping error processes. Pension fraud is one. Clerical error is another, and that can be undetectable. People who have paperwork with incorrect details often do not know, because literacy rates a century ago were low. Some people purposefully increase their age to escape military service, others to marry or work earlier when they are young, and some just inherited paperwork from older relatives because it was easier than travelling or paying to register a new birth. Then there are identity substitutions. Imagine a room with 100 people over 100, all holding valid paperwork. Replace one of them with a younger sibling. How do you detect the swap? The paperwork is real. The person knows enough about their sibling to answer questions.
Even if you understand the social and administrative context, there is still no reproducible method to test whether the age on a person's paperwork is correct. That is the central issue. There are also broader patterns. Extreme longevity often appears in places with weak record systems, low incomes and low historical levels of birth certification. That pattern runs against expectations if the signal were biological.
The mathematical process for small errors to dominate at very old ages is counter-intuitive but simple. Normally, rare errors can be ignored. But in this case, they grow non-linearly. Take a large population at age 50. Introduce a small number of people whose true age is younger than this recorded age. These individuals are biologically younger than the rest, so they die at lower rates as the cohort ages. Each year, the proportion of people with an error in their records increases because people with an inflated age are more likely to survive than are people with accurate data. Even with tiny starting error rates, you can end up with a population that has a 100% rate of errors at very old ages.
This is a universal problem. Five to ten per cent of people in the United States misstate their age in the census. Often, they simply do not know. Nearly one-quarter of the world's children still do not receive a birth certificate. Add that to the slow historical roll-out of birth registration and you get widespread uncertainty. There has been a 40-year debate about whether there is a limit to human lifespan. Both sides seem to be wrong, and the data seem to be junk. Demographers have been drawing shaky inferences from bad data for decades.