Raghav Sehgal and Albert Higgins-Chen on on why biological age clocks can shift daily – and what that means for longevity science.
Biological age clocks promise to measure how fast we age and accelerate the search for longevity therapies. But before they can guide medicine or clinical trials, the field must confront a fundamental question: are these clocks reliable and stable?
For most of human history, aging has been measured in exactly one way: time. Each year adds another candle to the cake, another number to the calendar. Yet anyone who has looked around knows that people do not age at the same rate. Some individuals remain healthy and active well into their seventies, while others begin developing chronic diseases decades earlier.
This gap between chronological age and physical condition is what scientists refer to as biological aging. Over the past decade, a new class of biomarkers known as biological age clocks has emerged as one of the most powerful tools for measuring it.
These clocks, most commonly built from patterns of DNA methylation, promise to quantify how fast we are aging. They are increasingly used to predict disease risk, evaluate lifestyle changes, and assess whether interventions might slow the aging process itself.
But as the field moves rapidly toward using these tools in clinical trials and longevity interventions, a fundamental question is becoming impossible to ignore.
How reliable are these clocks?
Dr Matt Kaeberlein raised this question at an individual level in his famous video I Took 4 Different Biological Age Tests & Compared the Results, and we reiterate this in our recent research paper evaluating the reliability of widely used epigenetic aging clocks across 1000+ samples, we found that the answer is more complicated than many assume.
The rise of biological age clocks
Epigenetic clocks estimate biological age by analyzing patterns of DNA methylation. These chemical modifications regulate gene expression and change systematically throughout the lifespan. By training statistical models on large datasets, researchers can predict age from these patterns with striking accuracy.
The first generation of clocks focused on predicting chronological age. Later models expanded their scope, attempting to capture mortality risk, physiological decline, or the pace at which individuals age.
Today dozens of clocks exist. Some attempt to summarize whole body aging, while others aim to capture specific aspects of biological decline. They have become central tools in the rapidly growing field of geroscience, where researchers seek to understand aging as a modifiable biological process.
Many studies now use these clocks to evaluate whether interventions such as dietary changes, pharmaceuticals, or lifestyle modifications can slow biological aging.
But all of these applications rely on one key assumption.
The measurement itself must be stable.
Reliability is not a simple concept
Reliability might sound like a straightforward property, but in practice it has multiple layers.
One layer is technical reliability. This asks whether the same biological sample produces the same result when measured repeatedly. If a DNA sample is processed multiple times using the same laboratory methods, a reliable clock should produce nearly identical age estimates.
Most clocks since 2022 perform well on this dimension [1]. When researchers analyze technical replicates, results are typically highly consistent.
However, technical consistency alone does not guarantee that a biomarker reflects stable biology.
The deeper challenge is biological reliability.
Biological reliability asks a different question: if we measure the same individual repeatedly across short time periods, do we obtain similar biological age estimates?
This is where things become far more complex.
When biological age fluctuates
In our study, we evaluated eighteen widely used epigenetic clocks using repeated biological samples collected from the same individuals. The goal was to understand how stable these clocks are when measuring real biological systems rather than identical laboratory replicates.
What we observed was substantial variability across many models. Biological age estimates could fluctuate across samples collected within hours or days as much as equivalent to 5-10 years within a day on average with up to 40 years of change as shown below. (As shown in the figure below for some of the famous clocks out there.) Short term physiological factors that ideally should not change biological age such as circadian rhythms or eating can influence methylation signals that clocks rely on.
This means that a person’s biological age estimate can shift even when the underlying aging process has not meaningfully changed.
We evaluated biological and technical reliability across four independent datasets using the Interclass Correlation Coefficient (ICC), a standard metric of measurement consistency. An ICC ≥ 0.90 is considered excellent and typically corresponds to ~0–1 year of variation in biological age between replicate samples. ICC values ≥ 0.75 are considered good, reflecting ~0–3 years of fluctuation. Moderate reliability (ICC ~0.50–0.75) translates to ~3–5+ years of variation, while poor reliability (ICC < 0.50) can result in substantial instability, often exceeding 5–10 years between replicates.
Across the clocks we evaluated, only a small subset demonstrated moderate biological reliability when assessed using repeated measurements. Perhaps the most surprising finding was that technical reliability did not predict biological reliability. Some clocks produced nearly identical results when measuring the same sample repeatedly in the laboratory but still fluctuated when measuring the same individual across time.
In other words, a biomarker can be technically perfect yet biologically noisy.
Why reliability matters
These findings carry important implications for the longevity field.
Many studies now interpret decreases in biological age as evidence that an intervention has slowed aging. But if biological age estimates fluctuate naturally, interpreting those changes becomes challenging.
Imagine a clinical trial where a participant’s biological age decreases by two years following a treatment. Without strong reliability, it is difficult to determine whether that shift reflects a genuine biological improvement or simply normal measurement variability.
The same challenge applies to studies linking biological age to disease risk. If the biomarker itself varies substantially across measurements, associations with health outcomes may appear weaker or more inconsistent, especially in small studies.
Reliability therefore places a fundamental limit on the usefulness of any aging biomarker.
A clock that is slightly less predictive but highly stable may ultimately be far more valuable than one that appears powerful but fluctuates unpredictably.
A turning point for aging biomarkers
The rapid growth of biological age clocks has been one of the most exciting developments in modern aging research. These tools provide a quantitative lens into aging biology that would have been unimaginable just two decades ago.
But the field is now entering a new phase.
For years the focus of clock development has been predictive accuracy. Researchers have tried to maximize correlations with chronological age, mortality risk, or disease outcomes.
Going forward, reliability must become an equally important priority.
Clocks need to be systematically evaluated across repeated biological samples, diverse populations, and intervention studies. Researchers must better understand how physiological states and environmental exposures influence clock measurements.
In short, aging clocks must increasingly be treated not only as statistical models but as clinical measurements.
Clinical measurements require stability. HbA1c, cholesterol, and hemoglobin are expected to produce consistent results under similar conditions. Aging biomarkers will need to meet similar standards if they are to guide medical decisions or therapeutic development.
Measuring aging is only the beginning
The idea that your biological age could change overnight without you aging a single day captures both the promise and the uncertainty of this emerging field.
Biological age clocks remain powerful tools for studying human aging. They offer the possibility of transforming aging from something we observe into something we can measure and eventually modify.
But measurement is only the first step.
If these biomarkers are going to guide clinical trials, accelerate longevity therapeutics, and shape the future of preventive medicine, the field must ensure that the measurements themselves are dependable.
Accuracy matters. Prediction matters.
But as our work suggests, reliability is an equally important challenge and is not to be overlooked.
About Raghav Sehgal
Raghav Sehgal is a Forbes 30 Under 30 honoree and Faculty at Yale University, where he applies artificial intelligence to longevity science. His work focuses on understanding and measuring human aging using multi-omic and multimodal data, with the goal of developing clinically actionable biomarkers.
He is the creator of SystemsAge, the first clinically deployed organ-specific aging biomarker, capable of quantifying aging across 11 organ systems from a single blood draw. He is also affiliated with the Yale Cancer Biology Program and serves as Director of Data Science at the Healthy Longevity Clinic, where he translates aging biomarkers into real-world clinical applications.
In parallel, Raghav is a Scientific Principal at LongevityTech.fund, advising on investments at the intersection of AI, digital health, and geroscience. With over a decade of experience across academia, industry, and startups, his work spans cancer prediction, metabolic health, and aging, with publications in leading journals including Nature.
About Albert Higgins-Chen
Albert Higgins-Chen is an Assistant Professor in the Department of Psychiatry at Yale University where he is a principal investigator. During his MD/PhD training at the University of Michigan, he worked with C elegans identifying genes regulating aging and longevity. He is a clinically trained psychiatrist and has applied aging biomarkers to investigate how mental health and treatment affect aging. He now develops new methods for measuring the biological aging process and the effects of aging interventions.
