Loading
Old version — revision 1
This is a fixed snapshot of Biological age, saved by Import as part of the initial corpus import. It is not edited and it is not updated; the article may have changed since.
Edit summary: Initial import of content/biological-age.md — the filesystem corpus, unchanged. Not an edit.
An estimate of how far an individual has progressed along the trajectory of age-related decline, expressed in years and compared against their chronological age.
Biological age is an estimate of where a person sits on the trajectory of age-related functional decline, expressed on the scale of years so that it can be compared with the date on their birth certificate. Two seventy-year-olds can differ enormously in organ function, disease burden and remaining life expectancy, and the concept exists to name that difference. It is used constantly in aging research and has no agreed definition, no gold-standard measurement, and no accepted regulatory status.
Chronological age is the best single predictor of death and of most chronic disease, and it is measured perfectly. The reason to want something else is that it is uninformative about individuals: it cannot distinguish a robust eighty-year-old from a frail one, and it cannot respond to an intervention. A measure that could do both would let researchers test geroprotectors in years rather than decades, which is the practical motivation described under Aging biomarkers.
The underlying assumption is that there exists a single latent quantity, the extent of accumulated aging, that many measurable things reflect imperfectly. This assumption is doing a great deal of work and may not be true. If the processes catalogued in the Hallmarks of aging proceed at partly independent rates in different tissues, then a person does not have one biological age but many.
Gerontologists have sought a functional measure of aging since at least the 1960s. Alex Comfort argued in 1969 for a battery of tests that could establish an individual's rate of aging, and set out roughly what such a battery would need to do.1 Efforts through the following decades combined physiological measures such as vital capacity, blood pressure, reaction time and visual accommodation into composite indices, generally by regressing them against chronological age.
The statistical framework most often used today comes from Klemera and Doubal, who showed that regressing biomarkers on age and then inverting the relationship produces a better estimate than the obvious approach of regressing age on biomarkers.2 A pattern runs through the whole history. Each generation of measures was assembled from whatever assay had just become cheap and high-throughput: physiological test batteries, then routine blood chemistry, then methylation arrays, then mass-spectrometry proteomics. The selection of inputs has been driven by instrument availability rather than by any theory of what aging is, which is one reason the resulting indices agree with each other so poorly.
Clinical composites combine routine laboratory values, blood pressure and spirometry. PhenoAge was derived this way from a national health survey before being ported onto methylation data.
Methylation clocks dominate current research, following the multi-tissue predictor Steve Horvath published in 2013. See Epigenetic clocks for the generations, their training targets, and their reliability problems.
Deficit accumulation takes a different route entirely. A frailty index counts the proportion of a long list of health deficits that a person has, without weighting them, and predicts mortality and institutionalisation robustly. It requires no laboratory at all.
Proteomic measures read plasma protein levels. Work from Tony Wyss-Coray's laboratory used organ-enriched plasma proteins to estimate the age of individual organs, finding that a substantial fraction of people show one organ aging markedly faster than the rest, with disease risk tracking that organ.3 Immune clocks built from cytometry and cytokine profiles capture a related axis, linked to Inflammaging.
Haematological measures exploit the fact that blood is the most accessible aging tissue: clonal expansions in the marrow, a marker of Stem cell exhaustion, become common after middle age and predict cardiovascular as well as haematological outcomes.
Imaging measures estimate brain age from MRI and retinal age from fundus photographs. Functional measures such as gait speed, grip strength, chair-rise time and VO2max are the oldest approach and among the best validated for predicting outcomes; they are also the endpoints that exercise moves most reliably.
The measures disagreeWhen multiple biological-age metrics are computed on the same cohort, they correlate with each other only weakly, often more weakly than each correlates with chronological age. Belsky and colleagues found this across telomere length, several methylation clocks and clinical composites.4 Either they are measuring different things, or most of them are measuring noise.
The weak inter-measure agreement admits two readings. On the optimistic reading, aging is multidimensional and each measure captures a genuine, partly independent component, in which case the right response is a panel rather than a number. On the pessimistic reading, most of the variance in these measures is technical and cohort-specific, and the shared signal is small.
Organ-specific results support the multidimensional reading. So does the observation that different measures predict different outcomes: methylation clocks predict mortality, frailty indices predict disability and institutional care, and cardiorespiratory fitness predicts cardiovascular events. Aggregating them into one number discards that structure.
Settings that accelerate aging-like change provide a partial test. Long-duration spaceflight produces bone loss, immune dysregulation and cardiovascular change on a compressed timescale, discussed under Space medicine, and cancer survivors treated with cytotoxic chemotherapy show elevated readings on several measures. If competing metrics agreed about who is aging fast, these are the populations where the agreement should be easiest to see. It is not consistently observed.
Interventions are routinely reported to have "reversed biological age by X years". Almost always the claim rests on a single measure, usually a methylation clock, in a small and often uncontrolled study. The pattern recurs across Caloric restriction trials, supplement studies, Metformin and the TAME trial cohorts and early Senolytics work. Three cautions apply. First, the measure has test-retest variation on the order of the reported effect. Second, a measure trained to predict an outcome is not thereby a cause of that outcome, so moving it need not move the outcome. Third, several of the interventions that shift clock readings, notably Epigenetic reprogramming, act directly on the molecular substrate the clock reads, which makes a clock a particularly poor referee for them.
Regression to the mean compounds the problem in consumer testing. A person who tests high, changes their habits and retests will frequently see a lower number regardless of what they did.
The field lacks an accepted ground truth. Measures are validated against mortality, which is the outcome they are supposed to predict, so validation is partly circular and cannot distinguish a measure of aging from a measure of current illness. Nothing yet distinguishes measures that track a causal process from measures that track its consequences, which is the same difficulty that dogs Cellular senescence markers and telomere length.
Standardisation is beginning through consortium work, and prize competitions such as XPRIZE Healthspan have sidestepped the problem by specifying functional restoration in muscle, cognition and immunity as the endpoint rather than any biomarker. That choice is a judgement about the state of the field: the organisers concluded that no biological-age measure is yet trustworthy enough to award a nine-figure prize on. Whether the geroscience programme can obtain a validated surrogate before it needs one for Healthspan-extension trials is the practical form of the question.
paperComfort, A. "Test-battery to measure ageing-rate in man." The Lancet, 1969. ↩
paperKlemera, P. & Doubal, S. "A new approach to the concept and computation of biological age." Mechanisms of Ageing and Development, 2006. ↩
paperOh, H.S.-H. et al. "Organ aging signatures in the plasma proteome track health and disease." Nature, 2023.↩Organ ages are estimated from plasma proteins enriched in each organ, not from measuring the organs themselves.
paperBelsky, D.W. et al. "Eleven telomere, epigenetic clock, and biomarker-composite quantifications of biological aging: do they measure the same thing?" American Journal of Epidemiology, 2018.↩Run within a single birth cohort at one age, so it tests agreement between the measures rather than their ability to rank people of different ages.