How Often Should You Retest Biological Age? The Slope Matters More Than the Number
Rate of epigenetic age change predicts mortality independent of baseline. What that means for your retest interval, clock generation, and sample type.
How Often Should You Retest Biological Age? The Slope Matters More Than the Number **For most people, once a year is the practical ceiling, and a single retest will tell you very little on its own. The strongest longitudinal evidence to date is a Nature Aging analysis of 699 adults in the InCHIANTI cohort, published in the March 2026 issue (volume 6, pages 534-540, DOI 10.1038/s43587-026-01066-6), which found that the rate at which epigenetic clocks changed over time was significantly associated with mortality independent of baseline epigenetic age and other confounders. Those rates were calculated from DNA methylation sampled at two or three timepoints across 1998, 2007, and 2013, with vital status followed to the end of January 2024, which is to say from samples taken years apart, not months. The takeaway for anyone deciding when to retest is that the informative quantity is the slope of your own line across several widely spaced, identically collected samples, and what any of it means for you is a conversation for your clinician.** Biological age testing is sold as a number. You send a sample, you get a report that says your body looks like it is a certain age, and the implicit promise is that you can buy the test again in a few months and watch that number move. The 2026 research literature has quietly moved somewhere else. It is now much more interested in the derivative than the value, in how fast the number is changing rather than where it currently sits. That shift changes almost everything about how a retest should be scheduled, which test should be bought in the first place, and what a coach or a client should expect to learn. This article walks through the evidence, the serious criticism attached to it, and how to turn a one-off purchase into a measurement protocol. MyProtocolStack is a tracking and education tool. Nothing here diagnoses anything or tells you to start, stop, or change any protocol, and the decision to test at all belongs with you and a qualified clinician. One framing note before the evidence. Everything described below is human data from observational cohorts and population studies. None of it comes from an animal model, and none of it is a randomized trial testing whether deliberately changing your measured slope changes any health outcome. That trial has not been run.
Level Versus Slope: The Distinction That Decides Your Retest Interval
A consumer epigenetic report usually gives you a level: an estimated biological age, or a pace value, at one moment. A level is a snapshot. It answers the question "where am I now," and it is the thing every marketing page leads with.
A slope is different. It is the change in that level between two or more comparable samples, divided by the time between them. It answers the question "which direction am I moving, and how fast." You cannot compute a slope from one test, and you cannot compute a trustworthy one from two tests taken close together, because the measurement noise between them can easily exceed the real biological change.
That single distinction is what sets the retest interval. If the informative quantity is the slope, then the interval has to be long enough that genuine change is large relative to assay and collection variability. If the informative quantity were the level, you could retest whenever you liked and simply accept a noisy reading. The 2026 evidence points firmly at the slope, which means it also points at longer, more disciplined intervals.
The InCHIANTI Analysis: What Rate of Change Actually Predicted
The Nature Aging paper (Kuo PL, Moore AZ, Tanaka T, and colleagues, with Luigi Ferrucci as senior author) analyzed 699 adults from the InCHIANTI cohort in Tuscany, Italy. Average baseline chronological age was 63 years, with a standard deviation of 16. The analysis drew on 1,721 samples: 376 participants were measured at two timepoints and 323 at three, from collections in 1998, 2007, and 2013. Vital status was tracked through linkage with municipal registry records to the end of January 2024, a span of up to 24 years.
The team evaluated seven epigenetic clocks spanning all three generations: the Hannum and Horvath clocks (first generation), DNAmPhenoAge, DNAmGrimAge, and DNAmGrimAge version 2 (second generation), and DunedinPOAm38 and DunedinPACE (third generation).
The central result: people whose clocks sped up faster over time had significantly higher mortality risk, and that association held independent of baseline epigenetic age and other confounders, including how old participants were when they entered the study. In other words, the rate carried information that the starting value did not.
Two further details matter for how you read your own reports. First, the trajectories were not uniform. DunedinPOAm38 showed a relatively stable pace of epigenetic aging, while the rate of increase in DunedinPACE accelerated as participants got older. That undercuts any assumption that you can draw a straight line through two points and extrapolate it. Second, models combining baseline plus rate of change reached the highest C-statistics in the study: 0.808 for DNAmGrimAge version 2, 0.806 for DNAmGrimAge, 0.801 for DNAmPhenoAge, and 0.800 for DunedinPACE, with measurable gains in discrimination and reclassification over baseline alone.
Keep those C-statistics in proportion. They describe how well a statistical model separated who died from who did not, across an entire cohort, over decades. They are not a statement about the accuracy of your individual report, and a C-statistic near 0.8 in a research cohort does not mean a consumer test can tell you anything specific about your own future. This was also an observational cohort, not an intervention trial, so nothing in it shows that changing your slope changes your risk.
Which Clock Generation You Are Actually Buying
Most people buying a biological age test do not know which generation of clock they are purchasing, and the generations were built to answer genuinely different questions. This is the single most useful thing a coach can clarify before a client spends money.
|---|---|---|---|
The head-to-head worth knowing about is "An unbiased comparison of 14 epigenetic clocks in relation to 174 incident disease outcomes," published in Nature Communications on December 16, 2025 (DOI 10.1038/s41467-025-66106-y) and announced by the University of Edinburgh's Institute of Genetics and Cancer on December 17, 2025. It compared 14 epigenetic clocks against 174 incident disease outcomes in 18,859 Generation Scotland participants with 10 years of linked clinical records. Second- and third-generation clocks substantially outperformed first-generation ones for disease prediction. The strongest associations were with respiratory and liver-related outcomes, including primary lung cancer and cirrhosis, along with diabetes. For 32 outcomes, adding an epigenetic clock to standard risk factors improved prediction accuracy. The researchers also found little evidence that the clocks behaved differently in men versus women, or in smokers versus non-smokers.
If a client is buying a first-generation clock and expecting a disease-risk readout, they have bought the wrong instrument for the question they are asking. That is an expectation a coach can set in one sentence, before any money moves.
Why Short Retest Intervals Mostly Sample Noise
The counterweight to all of this arrived on May 4, 2026, when Idan Shalev, Associate Professor of Biobehavioral Health at Penn State, and Abner Apsley, a postdoctoral researcher in molecular biology at the University of Illinois Urbana-Champaign, published a critique in The Conversation. Their argument is not that epigenetic clocks are worthless. It is that they are valuable research instruments at the population level and are not reliable enough at the individual level to guide personal health decisions.
Their specific concerns map directly onto retest design:
Separately, and worth knowing before you buy: as reported in an Advisory Board daily briefing on May 5, 2026, no direct-to-consumer biological age test has been approved by the FDA, and health experts quoted there caution that the tests may not be accurate or useful for individual health decisions.
Put that next to the InCHIANTI design and the retest logic becomes obvious. The evidence that slope predicts mortality was built on samples separated by roughly six to nine years. The confounders Shalev and Apsley describe operate on a timescale of hours to weeks. A retest three months after your baseline is measuring an interval where the noise sources are fully active and the biological signal is small. You are mostly buying variance.
This is worth stating carefully alongside our earlier post on [GLP-1s and biological aging](/blog/glp1-biological-aging-epigenetic-2026), which described a small randomized study reporting a DunedinPACE shift over roughly 32 weeks. That post is not contradicted here, but it is easy to misread. A randomized trial with a control group can detect a group-level difference over 32 weeks precisely because randomization and averaging across participants cancel out individual noise. One person retesting themselves at 32 weeks has no control group and no averaging, so the same interval that works for a trial does not work for an individual. Group means and personal readings are different things.
Building a Retest Protocol Instead of Buying a Number
The way to make this purchase worth anything is to stop treating it as a product and start treating it as a measurement series. That means deciding the rules once, up front, and then holding them constant for years.
|---|---|---|
Two points define a line but they do not establish a trend, especially given that DunedinPACE was observed to accelerate with age rather than move at a constant rate. Three or more comparable points, collected under identical conditions and spaced at least a year apart, is the minimum before the shape of your own curve is worth discussing with anyone.
There is an operational wrinkle worth planning for. The consumer market is consolidating. On April 29, 2026, Infinite Epigenetics, the parent company of TruDiagnostic, announced it had acquired the assets of Tally Health, the consumer epigenetic age company co-founded by David Sinclair, in what the announcement described as the largest acquisition in epigenetic age testing. Tally Health continues as a standalone consumer brand under CEO Melanie Goldey, while Matthew Dawson remains CEO of Infinite Epigenetics and TruDiagnostic. Financial terms were not disclosed in the announcement. For anyone committing to a multi-year measurement series, the practical implication is simply this: providers change hands, assays and reporting can change with them, and a break in methodology mid-timeline can quietly invalidate the comparison you were building. Keep your own copies of every raw report, including the sample type, collection date, and the exact clock name and version, so a future comparison is at least auditable.
What to Track: Markers Worth Logging Alongside a Clock
An epigenetic clock is expensive, slow to repeat, and noisy at the individual level. The blood markers underlying biological age estimates are none of those things, which makes them the better backbone of a longitudinal record. None of this is a recommendation to test or treat anything. It is a list of what is commonly tracked so you can organize your own data for a clinician conversation.
These are inexpensive, widely available, and far less sensitive to collection conditions than a methylation assay, which means a two-year trend line in hs-CRP or HbA1c is usually more legible than a two-year trend line in a clock. For background on how these fit together, see our guides on [biological age markers](/blog/biological-age-markers) and building a [longevity blood panel](/blog/longevity-blood-panel), or browse the full [biomarker library](/biomarkers).
[Log your epigenetic clock results and the blood markers behind them in one place with MyProtocolStack.](/auth/login?mode=signup)
The reason a structured record helps is not glamorous. It is that slope requires memory. Three years from now, the thing that determines whether your retest means anything is whether you can still find the sample type, the collection date, the provider, and the exact clock version from your first test. A folder of PDFs rarely survives that. MyProtocolStack stores and visualizes the series so the direction of travel is legible, and it does not interpret the result for you. Interpretation belongs with a licensed clinician who can weigh it against your full history.
Frequently Asked Questions
How often should you retest biological age?
Once a year is a reasonable practical ceiling, and the evidence supports treating even that as frequent. The 2026 Nature Aging analysis that established rate of change as a mortality predictor used DNA methylation samples taken years apart, drawn from 1998, 2007, and 2013 in the InCHIANTI cohort. Because epigenetic readings are sensitive to short-term factors like diet, illness, environmental exposures, and time of day, a retest a few months after baseline is dominated by measurement noise rather than real biological change. Plan for at least three comparable samples spaced a year or more apart before treating the pattern as a trend, and review any of it with your clinician.
Why does the rate of change matter more than the biological age number?
In the Nature Aging InCHIANTI analysis of 699 adults followed for up to 24 years, people whose epigenetic clocks sped up faster over time had significantly higher mortality risk, and that association held independent of baseline epigenetic age and other confounders, including age at study entry. That means the slope carried information the single starting value did not. Models combining baseline plus rate of change reached C-statistics of 0.800 to 0.808, topping out at 0.808 for DNAmGrimAge version 2. Those figures describe model performance across a research cohort, not the accuracy of any individual consumer report, and the study was observational rather than an intervention trial.
Can you mix saliva and blood samples when tracking biological age over time?
No. Idan Shalev and Abner Apsley noted in their May 2026 critique that testing epigenetic age in saliva versus blood can yield substantially different results for the same person. Because there is no established gold-standard method for generating epigenetic clocks across laboratories, a switch in sample type or provider mid-timeline can produce an apparent change that is entirely methodological rather than biological. Pick one sample type and one provider at the start and hold both constant for the whole series.
Which generation of epigenetic clock should you look for?
It depends on the question, and knowing the answer before purchase matters. First-generation clocks such as Horvath and Hannum were trained to estimate chronological age and remain useful for aging biology research, but in a Nature Communications head-to-head of 14 clocks against 174 incident disease outcomes in 18,859 Generation Scotland participants, they were substantially outperformed by second- and third-generation clocks for disease risk prediction. Second-generation clocks such as PhenoAge and GrimAge and third-generation pace measures such as DunedinPACE were the stronger performers, with respiratory and liver-related outcomes, including lung cancer and cirrhosis, among the strongest predictive domains.
Are consumer biological age tests reliable enough to make personal health decisions?
Shalev and Apsley argued in The Conversation on May 4, 2026 that epigenetic clocks are valuable research instruments at the population level but are not reliable enough at the individual level to guide personal health decisions. They cited sensitivity to short-term fluctuations in diet, environmental exposures, illness, and time of day, plus substantial differences between saliva and blood samples from the same person, and the absence of a gold-standard method for building clocks across laboratories. Separately, Advisory Board reported in May 2026 that no direct-to-consumer biological age test has been approved by the FDA. Treat a result as a data point to track and discuss with a qualified clinician, not as a finding to act on.
Sources
1. Nature Aging, Kuo PL, Moore AZ, Tanaka T, et al., "Longitudinal changes in epigenetic clocks predict survival in the InCHIANTI cohort," volume 6, pages 534-540, March 2026. https://www.nature.com/articles/s43587-026-01066-6
2. PubMed record for the InCHIANTI analysis, PMID 41826710, DOI 10.1038/s43587-026-01066-6 (listed publication date March 13, 2026). https://pubmed.ncbi.nlm.nih.gov/41826710/
3. Medical Xpress, "Changes in pace of epigenetic clocks over time may help predict mortality risk," March 17, 2026. https://medicalxpress.com/news/2026-03-pace-epigenetic-clocks-mortality.html
4. Nature Communications, "An unbiased comparison of 14 epigenetic clocks in relation to 174 incident disease outcomes," December 16, 2025. https://www.nature.com/articles/s41467-025-66106-y
5. University of Edinburgh Institute of Genetics and Cancer, "Second-generation epigenetic clocks show greater promise for disease risk prediction," December 17, 2025. https://institute-genetics-cancer.ed.ac.uk/second-generation-epigenetic-clocks-show-greater-promise-for-disease-risk-prediction
6. The Conversation, Idan Shalev and Abner Apsley, "Biological age tests reveal what slows or hastens aging, but they're useful only for researchers, not consumers," May 4, 2026. https://theconversation.com/biological-age-tests-reveal-what-slows-or-hastens-aging-but-theyre-useful-only-for-researchers-not-consumers-275974
7. Advisory Board, "What do 'biological age' tests really tell you?," May 5, 2026. https://www.advisory.com/daily-briefing/2026/05/05/biological-age
8. PR Newswire, "Infinite Epigenetics Acquires Tally Health to Accelerate the Future of Epigenetics-Driven Health and Longevity," April 29, 2026. https://www.prnewswire.com/news-releases/infinite-epigenetics-acquires-tally-health-to-accelerate-the-future-of-epigenetics-driven-health-and-longevity-302756729.html
*MyProtocolStack is a tracking and education tool, not medical advice, diagnosis, or treatment, and you should always consult a qualified healthcare professional before making any changes to your health protocol.*
Enter your blood work in MyProtocolStack, run StackAI analysis, and get personalized insights based on your actual numbers — not generic charts.
Start Free →