A clock read from blood predicts a person's death and disease better than their age does
A trial that slows ageing cannot wait thirty years to count deaths; it needs a reading, taken at the start and the end, that stands in for them. The best candidates read the chemical marks on DNA from a blood sample. The first such clock, in 2013, matched a person's calendar age to within a few years from any tissue; the useful ones came later, trained not on age but on outcomes. In 2019 a Los Angeles group published GrimAge, trained on smoking and blood proteins, which predicted who would die, and when, better than age in several large groups of people, and later clocks (DunedinPACE, 2022) read the pace of ageing rather than its level. No regulator treats any of them as a medical test, and none is used in routine care, though several are sold to the public.
What would count
A clock read from a blood or saliva sample predicts time to death or to an age-related disease better than calendar age, in two or more independent groups of at least a thousand people followed for years, in peer-reviewed reports. For the clinic rung: a regulator clears or approves such a clock as a medical test, or a named health system uses it in routine care and publishes that practice. For wide use: offered in routine care in several countries.
Who settles it: Peer-reviewed validation reports; the regulator's clearance record, or a named health system's published practice, for the clinic rung.
Where it stands
First human. Reached 21 January 2019, on the record of doi.org. Next rung within 5 to 20 years.
When it reaches the clinic: the project's estimate
This is the one thing on this page that is a judgment rather than a record. The project's estimate of the year it reaches the clinic is 2034, no earlier than 2028 and no later than 2060; the model draws from that range every run, with the likely year the most probable draw. Set 16 October 2026, a first pass awaiting an outside review.
| Rung | No earlier than | Most likely | No later than | |
|---|---|---|---|---|
| In clinical use | 2028 | 2034 | 2060 | feeds the year at the top |
| In wide use | 2032 | 2042 | 2080 |
Why. Clocks are improving fast and several large health systems were piloting them in 2026; the likely year assumes a health system publishes routine use, or a regulator clears one as a test, around 2034. The earliest assumes a pilot already running becomes routine within two years; the latest allows for the clocks staying research and consumer tools because no regulator sees a treatment decision that depends on them.
Changes to this record
None since it was written on 16 October 2026. Every change to the stage or the estimate lands here with its date, the old value, the new one and the reason, and a change to the estimate never happens because a number looked wrong: it needs a dated reason and a source.
The last readings
- 17 October 2026 Quiet
Nothing dated in the window 2026-10-10 to 2026-10-17 bears on the test. The nearest items are older: a Nature Aging report that changes over time in several clocks predict death in the InCHIANTI group (699 people, up to 24 years of follow-up; listed as about 218 days old, so roughly 2026-03, and below the test's thousand-person floor), a 2026 HUNT study (PubMed 42163406, date not seen) reporting the pace of ageing read from blood predicting death in two Norwegian groups, and an annotated bibliography of the Dunedin pace-of-ageing measure dated 2026-10-01, none of which could be opened. The leaf stays at first human, whose test is already met, and nothing seen touches the clinic rung (a regulator's clearance or a health system's published routine use).source 1, source 2, source 3
What the story so far says about it
On the record
- 21 October 2013DNA methylation age of human tissues and cell types
Notes
The ladder fits a measuring tool loosely, and the test says how it is read here: first human is a validated clock in people, the clinic rung is a regulator or a health system treating it as a test, wide use is routine care. Clocks sold to the public (several companies since 2019) are not a rung, because a sale is not a validation. A clock used as the thing a trial must move is the leaf beside this one.
Not yet verified
These facts in the record were written from memory or from a secondary source and have not yet been checked against the paper, the regulator's record or the registry. Nothing here is presented as settled until it is cleared, by checking it and either deleting the line or correcting the record.
- That the 2015 report (Marioni and colleagues) was the first to show a clock read from DNA marks predicting death in large groups, and that Horvath's 2013 clock was the first multi-tissue clock, are priority claims not settled by any source read; the 2015 paper's four groups (446, 920, 2,635 and 657 people) include only one of over a thousand and show a link to death independent of age rather than a better prediction than age, so by the test as worded the stage date stays at 2019.