This review asks whether DNA-methylation “biological age” clocks are accurate and useful enough to be used in life insurance underwriting. Working through a case of a fit 50-year-old told he has a biological age of 46 and is aging at 0.7 years per year, the author walks through three generations of epigenetic clocks and concludes that these tests measure something real at the population level but are too noisy, poorly defined, and population-dependent to be trusted for any single individual. The verdict is that clocks add nothing beyond the traditional risk factors insurers already use, so they are not ready for prime time.
The longevity industry has a compelling product. Send in some saliva, receive a number that says your body is younger than your birth certificate, and receive the implicit promise that with the right supplements and habits you can push that number lower still. The author, a medical director at a large reinsurer, sets out to test whether that number means what buyers think it means, and whether an insurer should pay attention to it.
The number almost always comes from an epigenetic clock, an algorithm that reads DNA methylation patterns at a few hundred to a few thousand sites in the genome and maps them onto an age. The first of these, the Horvath clock from 2013, predicted chronological age with striking accuracy across a whole population. The catch is buried in the statistics. That accuracy is a population-level property. Applied to one person, the clock carries an error of roughly four years in either direction, which is often as large as the difference the test is claiming to detect.
Later clocks changed the target. Second-generation tools such as DNAmPhenoAge and DNAmGrimAge were trained on mortality and disease rather than birthdays, so they report a risk estimate dressed up as an age. Third-generation clocks such as DunedinPACE report a “pace of aging,” a velocity rather than a position. Each is intuitively appealing, and each was validated at the level of large cohorts, not individuals.
That distinction is the heart of the argument. A clock built to minimize average error across thousands of samples is not the same instrument as a reliable personal biomarker. Batch effects, sample handling, the time of day of the draw, sleep, stress and exercise all move the reading, and there is still no agreed definition of biological age or any gold-standard measure to validate a clock against. Predicted age and true biological age can, as the author bluntly puts it, be unassociated.
For an insurer the practical question is whether a clock adds information beyond age, sex, smoking, blood pressure, body mass index and medical history. The answer here is no. The very things a clock indirectly senses, such as inflammation and metabolic dysfunction, are already captured by existing underwriting. Clocks may eventually earn a role, particularly for young or seemingly healthy applicants where traditional markers are silent, but only after prospective studies prove they add independent predictive value.
Insights
The honest take-home is a caution rather than a protocol. If you have paid for a biological age or pace-of-aging test, treat the single number as entertainment, not a health readout. The paper’s central practical point is that individual-level error swamps the signal, so a reported result of “biological age 46” should really read “epigenetic age 46, range 42 to 50,” and a change between two tests within that band tells you nothing.
To illustrate the real-world magnitude, take the case subject: a chronological 50-year-old reported as biologically 46, meaning four years “younger.” Using the DNAmPhenoAge effect size the paper cites, each one-year reduction corresponds to a hazard ratio of 0.955 (the reciprocal of HR 1.045 per year). Four years compounds to roughly 0.955 to the fourth power, about 0.84, implying a 16 percent lower all-cause mortality hazard if the estimate were both real and causal. It is almost certainly neither at the individual level, because the roughly four-year measurement error means the entire four-year “benefit” sits inside the noise band. Framed as standardized effect, DNAmGrimAge carries the strongest signal in the literature the paper cites (HR 1.47 to 1.8 per standard deviation), yet even that is a population statistic.
Context and Source
- Open access Paper: Biological Clocks: Ready for Prime Time? by Timothy Meagher, MB, FRCP(C).
- Institution and country: Munich Re (Reinsurance Company of America) and McGill University, Montreal, Quebec, Canada. The author is Vice-President and Medical Director at Munich Re and Associate Professor of Medicine at McGill.
- Journal: Journal of Insurance Medicine, 2026, volume 53, pages 221 to 225.
- Impact evaluation: The Journal of Insurance Medicine carries a Journal Impact Factor of approximately 0.16 (SCImago Journal Rank 0.272, Q3). The impact score of this journal is 0.16, evaluated against a typical high-end range of 0 to 60-plus for top general science journals, therefore this is a Low impact journal.