Biological age tests (GrimAge, PhenoAge, Horvath): what they measure and don't

Epigenetic clocks, biological age estimators based on DNA methylation patterns, have become one of the most-marketed longevity diagnostics. The underlying research is real and interesting. The consumer products oversell what the measurements can tell an individual.

What epigenetic clocks measure#

DNA methylation is one of the major mechanisms of epigenetic regulation. Specific methylation patterns at certain CpG sites change predictably with age across many tissues.

Steve Horvath's 2013 work showed that combining methylation values at 353 specific sites could predict chronological age with remarkable accuracy across tissues, typically within 3–5 years. This was the original "Horvath clock."

Subsequent clocks have been built for different purposes:

• Horvath 2013: multi-tissue age predictor
• Hannum 2013 — blood-specific age predictor
• PhenoAge (Levine 2018): designed to predict mortality and morbidity, not just chronological age
• GrimAge (Lu 2019): designed to predict time to death; the strongest mortality predictor among epigenetic clocks
• DunedinPACE (Belsky 2022): designed to measure rate of ageing rather than absolute age

These are research tools developed in academic labs from large cohorts.

What the research actually shows#

The clocks predict population-level outcomes well:

• Mortality. GrimAge in particular shows strong associations with all-cause mortality in cohort studies
• Age-related disease incidence. Cardiovascular events, cancer, dementia all correlate with accelerated epigenetic age
• Effect of lifestyle. Smoking, BMI, exercise, and other factors associate with epigenetic age in cross-sectional data
• Some interventions appear to slow epigenetic aging. Caloric restriction (CALERIE-2 substudy), some exercise interventions show modest effects on PhenoAge/DunedinPACE

The associations are robust. They're also imperfect; a meaningful fraction of variation isn't captured.

The consumer products#

Consumer biological age tests vary widely in what they actually measure:

• TruDiagnostic / TruAge: uses peer-reviewed clocks (Horvath, GrimAge, PhenoAge, DunedinPACE); £200–£500 per test
• Elysium Index — proprietary epigenetic test; £200
• MyDNAge: older product based on Horvath clock
• Various biological age "calculators" based on biomarkers (not DNA methylation), easier and cheaper but measure something else
• Wearable-based "biological age" estimates: vendor constructs with minimal validation

The quality varies dramatically. Products using peer-reviewed clocks on properly-processed methylation data are more reliable than vendor constructs.

What the consumer tests can't reliably tell you#

Individual reproducibility is poor. Test-retest variability for some clocks is large. A test result of "biological age 47" might come back "biological age 52" three months later from the same person without meaningful biological change. This is a serious problem for tracking individual change.

Short-term interventions don't reliably move the dial. A 6-week intervention rarely produces measurable epigenetic age change above the test-retest noise.

The clocks measure different things. GrimAge and Horvath can give very different results for the same person because they're optimised for different outcomes.

The clocks were developed for population research, not individual diagnosis. Population-level associations are robust; individual interpretation is much more uncertain.

What gets oversold#

• "This test shows your biological age accurately." With substantial noise, not as precisely as the number suggests
• "Watch your biological age drop with our intervention." Most consumer interventions don't reliably move epigenetic age above the noise floor in months
• "This is your true age." It's an estimate based on methylation patterns; "true biological age" is a marketing construct
• "Take this supplement to reverse biological age." Almost universally unsupported

The Sinclair / Lifespan-related claims#

David Sinclair's 2019 Cell paper proposed that ageing is fundamentally an epigenetic phenomenon and that interventions can reverse it. The supplement marketing built on this has dramatically outpaced the human evidence.

The Sinclair work on partial reprogramming (Yamanaka factors) in mice is interesting science. It is not a basis for current consumer interventions.

The legitimate use cases#

Research participation. Contributing methylation data to longevity research is valuable.

Population-level monitoring. Healthcare systems may eventually use epigenetic clocks as risk stratification tools. This is research-stage.

Pharmaceutical development. Drug companies use these clocks to screen interventions.

These aren't individual consumer use cases.

What we'd actually say#

Epigenetic clocks are real biology with genuine predictive power at the population level. They are noisier and less actionable at the individual level than the consumer market implies.

For someone wanting to know their "biological age" out of curiosity, the consumer tests will give you a number. The number has more uncertainty than the marketing suggests and shouldn't drive significant lifestyle or financial decisions.

For tracking the effect of personal interventions, the test-retest noise often exceeds the expected intervention effect in short timeframes. You can't reliably show whether your supplement stack "reversed your biological age" in a 6-month period.

The money spent on consumer epigenetic testing might be better spent on the lifestyle interventions that actually have evidence, sleep, exercise, dietary patterns, social connection.

For people genuinely committed to the longevity science, contributing to clinical trials using these clocks may be more valuable than buying consumer tests.

When to talk to a clinician#

• Significant lifestyle changes prompted by a biological age test; the validity of the prompt matters
• Considering high-cost or risky interventions because of a biological age result
• Pre-existing cardiovascular or metabolic risk factors; these are more actionable than epigenetic age
• Cancer history + considering interventions targeting epigenetic age
• Family history of age-related conditions; clinical risk stratification is more actionable