Fifty-one longitudinal intervention studies. Sixteen epigenetic clocks and 94 other DNA methylation biomarkers, calculated across 3,128 paired blood samples taken before and after.
Nineteen interventions significantly lowered the methylation ageing measures. Five raised them. The rest moved nothing that reached significance.
Those are the odds behind the biological age panel you are being sold in Singapore.
The work is Sehgal, Borrus, Armstrong and colleagues in Nature Medicine, run on a harmonised database called TranslAGE.1 It asks a narrower question than most longevity coverage does: which clocks respond to treatment at all.
DunedinPACE showed the greatest overall responsiveness. PCGrimAge gave the strongest statistical evidence, and the dependable set ran GrimAgeV2, PCGrimAge, SystemsAge, DunedinPACE and PCPhenoAge.
Every clock on that list is trained on mortality or on the pace of ageing.
The first-generation clocks, Horvath and Hannum, are trained to predict chronological age, and they responded least. A number that barely shifts when the behaviour shifts cannot grade the behaviour.
Singapore prices this across two orders of magnitude. A nine-marker blood check runs S$69; saliva methylation tests list from about S$160 to S$260; one clinic’s longevity programme sits at S$4,250.
None of the local providers in that survey names the clock its report runs on.
You are buying a number without being told which algorithm produced it.
You are buying a number without being told which algorithm produced it.
The harder finding is about who was measured. Population health status drove the responsiveness, and several biomarkers fell further in studies of people with disease than in studies of healthy participants.
Therapies targeting tumour necrosis factor moved almost all the second-generation biomarkers among people living with arthritis or inflammatory bowel disease.
The strongest signal in the dataset came from fixing something that was already broken.
The strongest signal in the dataset came from fixing something that was already broken.
Lifestyle and pharmacological interventions both lowered the scores, with the drugs producing larger average effects. Metformin was among the strongest, showing its largest changes in the inflammatory, brain and metabolic system scores.
Five senolytic studies gave divergent results across multiple biomarkers.
Study duration mattered as much as the compound. The paper names duration and population characteristics as the two factors driving whether a biomarker responded at all, which puts a panel taken six weeks after the last one outside the conditions where these clocks were shown to move.
Grade the evidence for what it is. This is a meta-analysis of longitudinal intervention studies rather than a trial, and the authors state that responsiveness alone does not establish a DNAm biomarker as a valid surrogate endpoint.
A clock that moved has still not shown that a life got longer.
What the paper hands you before the next panel is a question with a checkable answer: which clock does the report run, and does its name appear on that list of five.
The same market is scaling on the other side of the region. Fudan University’s Zhongshan Hospital has opened a longevity and anti-aging clinic, inside a silver economy the government put at RMB 7 trillion by the end of 2024.
The clocks arrived years before the grading. The grading is public now, and it names five.
Footnotes
-
The database is far larger than the slice this paper used. TranslAGE precalculates 41 clocks and 1,694 biomarker proxies across more than 34,000 people, which is a great many ways to be told your age. ↩