Research on people who live past a hundred is intuitively appealing, since they represent the outcome everyone is interested in. The design carries specific limitations that shape what can be concluded from it.
Survivor selection is built into the sample
A study of centenarians examines only those who reached that age, and everyone who lived similarly but did not is absent from the data.
Characteristics found in survivors may have caused survival, may be consequences of it, or may be unrelated traits that happened to accompany it.
Without the comparison group that did not survive, the design cannot separate these possibilities, which is the central difficulty.
Recall reaches back a century
Information about earlier life is usually gathered by asking, and asking someone about diet and activity in their twenties depends on memory across many decades.
Memory is reconstructive and shaped by current beliefs, so a healthy centenarian may recall a healthier past than the records would show.
Contemporaneous records from that period are rarely available, and where they exist they cover a small number of people in specific places.
The environment being described has gone
Someone reaching a hundred today grew up in a world with different food, different infectious disease exposure, different work and different medicine.
Whatever contributed to their survival operated in conditions that cannot be recreated, and its relevance to someone growing up now is uncertain.
This is a general limitation of studying extreme longevity: the exposures of interest occurred long before the outcome could be observed.
What the design does establish
Comparing centenarians with their own siblings and children, who share genes and early environment, allows some separation of inherited from acquired factors.
Genetic studies of large numbers of very old people have identified variants appearing more often among them, which points to biological pathways worth investigating.
Studies of how these individuals aged, rather than why, also show that many compressed their illness into a short final period, which is a finding about pattern rather than cause.
Why families are studied instead
Long-lived families offer a stronger design, since multiple related individuals reaching old age is less likely to reflect chance than a single case.
Following such families prospectively, measuring rather than asking, addresses the recall problem at the cost of taking decades to produce results.
That trade-off between speed and quality of evidence is why conclusions in this field move slowly and why individual claims should be read cautiously.