Biological Age Testing: What the Research Describes

Chronological age — the number of years since birth — is a poor predictor of individual health outcomes. It accounts for substantial variance in longevity at the population level but leaves a great deal unexplained within any given age band. Biological age attempts to quantify the rate at which an individual is ageing, independent of calendar time. Several validated methods now exist to estimate biological age from molecular, physiological, and functional data, and the field has advanced rapidly over the past decade. This page describes what each approach measures, what validation studies have found about predictive accuracy, and where the evidence base remains incomplete or contested.

The practical interest in biological age goes beyond the number itself: if reliable measures can be obtained, they could in principle serve as endpoints for evaluating whether lifestyle, pharmacological, or other interventions actually slow ageing — not just in animal models but in humans. That ambition currently outpaces the evidence, but the scientific progress since 2013 has been substantial.

Why Biological Age Matters

The case for measuring biological age rests on the observation that individuals of the same chronological age differ substantially in their health, functional capacity, and longevity. Hazard ratios for all-cause mortality are approximately 1.07 to 1.10 per year of biological age estimated by composite indices (Elo, 2009), but the meaningful finding is the variation within chronological age bands rather than the average trend across them. Two 55-year-olds can have measurably different biological ages, and a reliable biological age measure should predict their future health outcomes better than the calendar can.

A second motivation is mechanistic: biological age measures tied to specific molecular processes — DNA methylation, telomere dynamics, plasma protein profiles — can in principle reveal which ageing processes are most accelerated in a given individual, and might eventually guide targeted interventions. Neither the measurement tools nor the intervention evidence are sufficiently mature for that clinical application today, but this is the direction the field is heading.

Epigenetic Clocks

The most-studied class of biological age tests uses DNA methylation — chemical modifications to DNA at cytosine residues that do not alter the genetic sequence but regulate gene expression, and that change systematically with age in a largely tissue-specific pattern. First-generation epigenetic clocks were trained to predict chronological age. Horvath (2013) published a multi-tissue clock using 353 CpG sites, achieving a correlation of r = 0.96 with chronological age across diverse tissue types — a landmark result demonstrating that a common methylation programme underlies ageing across the body. Hannum et al. (2013) developed a blood-specific clock with comparable accuracy using a smaller set of sites.

The limitation of first-generation clocks is that accuracy in predicting calendar age does not equal sensitivity to health-relevant variation. This led to second-generation clocks trained on biomarkers of health or mortality rather than on chronological age directly. PhenoAge (Levine et al., 2018) was trained on a composite of nine clinical biomarkers — including albumin, creatinine, glucose, C-reactive protein, and lymphocyte percentage — that together predicted time-to-death in a large US population cohort. Individuals whose PhenoAge exceeds their chronological age show elevated all-cause mortality, multi-morbidity, and markers of accelerated biological ageing. GrimAge (Lu et al., 2019) was trained directly on time-to-death data and incorporates estimates of plasma proteins associated with smoking history and lifespan. Across multiple independent cohorts, GrimAge has produced the strongest associations with all-cause mortality, coronary heart disease, and cancer risk of any single epigenetic measure to date.

DunedinPACE (Belsky et al., 2022) represents a conceptual shift: rather than estimating a static age, it measures pace of ageing. It was constructed from the Dunedin Study, a longitudinal cohort of 954 individuals born in the same year in New Zealand and followed for decades with repeated physiological and health assessments. A DunedinPACE score of 1.0 represents the population-average rate of ageing; a score of 1.2 indicates a person is ageing 20 percent faster than average. Because it is calibrated to a rate rather than an age, it is in principle more sensitive to change over time and more useful as an outcome measure in intervention studies.

Telomere Length

Telomeres are repetitive DNA sequences capping chromosome ends that shorten with each cell division, functioning as a replicative counter and protecting genomic stability. As a proxy for cellular replicative history and senescence, telomere length was among the first molecular markers proposed for biological age. Cawthon et al. (2007), in a widely cited analysis, reported that shorter leukocyte telomere length is inversely associated with age-related disease and mortality risk in older adults, establishing the biological plausibility of the measure.

However, the clinical utility of telomere length as an individual biological age marker has been complicated by two findings. First, measurement variability is substantial: different assay methods — quantitative PCR, flow-FISH, Southern blotting — produce different absolute values, and test-retest reliability within the same individual is modest, limiting the interpretability of year-to-year retesting. Second, the relationship between telomere length and health outcomes is non-linear. Haycock et al. (2015), using Mendelian randomisation, found that genetically longer telomeres are associated with increased cancer risk across multiple tumour types, even while shorter telomeres associate with cardiovascular risk. This bidirectional relationship complicates "longer is better" consumer messaging and highlights the difficulty of optimising a single marker with complex, non-linear biology.

Proteomic Clocks

A newer approach uses the profile of proteins circulating in blood plasma, which collectively reflect secretory activity across multiple tissues and are sensitive to ageing, disease states, and environmental exposures. Lehallier et al. (2023), publishing in Nature Medicine, identified characteristic waves of protein change in human plasma across the lifespan, concentrated at approximately the mid-30s, late 50s, and early 70s. This finding suggests that ageing at the molecular level is not a uniform, linear process but proceeds in phases — a result that may have implications for when interventions are most likely to be effective.

Tanaka et al. (2023) studied n = 45,441 individuals in the UK Biobank using a 1,301-protein clock and found that it outperformed epigenetic clocks in predicting 12-year all-cause mortality, disease incidence, and physical function in independent validation. Proteomic clocks remain at a research stage for most practical applications: they require high-dimensional mass spectrometry or proximity-extension assay platforms not yet available at scale in consumer testing. The gap between what is scientifically demonstrated and what is commercially available is currently wide.

Composite and Multi-Omic Approaches

Because different biological age measures each capture partially distinct aspects of ageing biology, combining them tends to improve predictive performance. Work from the Gladyshev laboratory (2023) developed OMICm Age from 11 omic data types — including DNA methylation, transcriptomics, proteomics, and metabolomics — and found that it outperformed single-omic measures in predicting mortality across three independent validation datasets. The principle is analogous to the improvement in clinical risk prediction achieved by combining multiple biomarkers versus any single one.

The practical constraint for multi-omic approaches is cost and complexity. A single consumer epigenetic clock test costs approximately $200 to $500. Full multi-omic profiling combining high-depth sequencing, mass spectrometry proteomics, and metabolomics remains a research-grade procedure costing tens of thousands of dollars and requiring specialist data analysis. The distance between what is scientifically optimal and what is currently accessible at consumer scale remains large, and the evidence that acting on any of these measures improves outcomes has not yet been established.

Limitations

Several important limitations apply across biological age measurement approaches. Tissue specificity is a significant unresolved issue. Shireby et al. (2020) demonstrated that the brain's epigenetic clock diverges substantially from blood-based measures: a brain-specific clock was required to detect accelerated ageing in Alzheimer's cortical tissue, which standard blood-based measures could not identify. Since blood samples are the basis of almost all consumer products, these tests may not accurately reflect ageing in the organ systems most relevant to a given person's health risks.

Measurement variability within individuals is practically important. Repeat testing under equivalent conditions typically yields estimates differing by 2 to 5 years, depending on the clock and the laboratory. This means that short-interval retesting to track the effect of a dietary change or supplement cannot reliably detect real change from measurement noise. The interventional evidence — whether a change in a clock score caused by a lifestyle or pharmacological intervention translates into better health outcomes — is limited and largely preliminary. An intervention that shifts a methylation score does not automatically improve the underlying biology; the question of whether these clocks are measuring causes of ageing or correlates of it remains open.

Reverse causation also remains unresolved for many associations: does a higher biological age score predict worse health because it is upstream of the mechanisms that cause disease, or does subclinical disease drive both the elevated score and the eventual outcome? This question is critical for evaluating whether reducing a biological age score actually reduces risk, and current evidence does not fully resolve it.

What the Research Describes About Pace of Ageing

Several controlled and observational studies have examined whether modifiable factors change biological age scores. Abbasi et al. (2021), in an analysis of NHANES data, found that high cardiorespiratory fitness — measured by estimated VO2 max — was associated with an 8.8-year lower PhenoAge in adults over 40, after adjustment for demographic confounders. This is a cross-sectional association, not a randomised finding, and cannot be used to infer that achieving that fitness level will reduce biological age by the same magnitude; nonetheless, the association size is large relative to other factors examined.

Fitzgerald et al. (2022) published an 8-week randomised controlled trial in 43 adults that tested a combined lifestyle programme including dietary modifications, targeted exercise, sleep optimisation, and stress management. The intervention group showed a 3.23-year reduction in Horvath clock age versus control at the end of the trial, a statistically significant finding. This was a combined multi-component intervention and it is not possible to attribute the change to any individual element; the trial was also small and the follow-up was short.

The CALERIE trial — a randomised controlled trial of 25 percent caloric restriction in 220 non-obese adults over two years — provides one of the most methodologically rigorous human tests of a single intervention on a validated pace-of-ageing measure. Belsky et al. (2023) reported DunedinPACE data from CALERIE showing that the caloric restriction arm slowed pace of ageing by 2 to 3 percent versus control, a statistically significant difference in the largest and most carefully conducted human caloric restriction trial to date. The effect size is modest; 25 percent sustained caloric restriction is difficult to achieve and maintain; and the long-term health consequences of slowing DunedinPACE by this amount in humans remain to be established. But it represents the strongest available evidence that a single defined intervention can move a validated human ageing clock in a controlled trial.

Sources

Biological age tests are research tools that have not yet been validated as clinical diagnostics for individuals. Discuss any results with a qualified healthcare professional before drawing conclusions about your health.