How old are you? It’s a seemingly simple question, yet at the same time complex, with an air of intrusiveness and impropriety. Most of us will give a straightforward answer: the number of years that have passed since our birth.
In many ways, however, this question signifies something more than a query about your time on Earth. It elicits something about your state of being, your health, and unfortunately for some, even your status and value in society. We attach meaning to ages like 65 or 100 and even consider a person’s age when assessing their suitability for a job, like president of the United States.
In reality, time is not what gives age its significance. Instead, it is the underlying biological process that changes over time. That process can take a robust and resilient body and transform it into a state of frailty, dysfunction and, eventually, death.
To date, no human has ever escaped this process, yet some scientists and longevity gurus have suggested its inevitability is up for debate. Some scientists believe that we may one day discover an intervention capable of transforming our biology—allowing years to pass with wisdom gained and experiences lived but without bodily degeneration. This possibility assumes that time and the biological process we call aging are separable. To break the bonds that currently link time and aging, scientists first need new ways to measure the process of aging, rather than merely counting our trips around the sun.
That’s where biological age comes in. The concept dates to the 1950s when British biologist Peter Medawar declared the need for an agreed-upon system of measurements for what he, at the time, termed “senescence.” Nearly 70 years later, one might think we’ve achieved it. Many biologists, mathematicians and computer scientists have published papers detailing new models and equations claiming to quantify biological aging. Companies advertise their tests online, and influencers tout lifestyles and protocols that they claim to reverse biological age. The reality is less certain. While experts largely agree that the concept of biological aging is valid, they don’t agree on how best to measure it.
Much of the challenge lies in the fact that biological aging is what statisticians call a “latent variable,” which means it cannot be directly observed or measured. Instead, latent variables must be inferred mathematically from other measurements. This implies we will never measure it perfectly but instead will only generate fuzzy estimations of it. For example, we can precisely measure weight, blood pressure, chronological age and other observable variables. Latent variables such as intelligence, health and biological age are abstract concepts that we will never be able to perfectly capture. (This is why people still argue over whether IQ tests provide good estimates of intellect.)
Not all hope is lost. Biological age estimates exemplify the maxim of 20th century statistician George E.P. Box: “All models are wrong, but some are useful.” No model will ever provide an accurate read-out of biological age. That said, some current models can reveal important insights about your health and aging from a biological perspective.
So how do we judge the merit of something for which there is no ground truth to compare it? Our only solution is to come up with metrics, or benchmarks, to help us assess the degree to which a measure of biological age captures the underlying theoretical construct we are trying to estimate.
When assessing the validity of any model claiming to measure biological age, here are the criteria I and many other experts use:
Biological age should be loosely linked to chronological age. Because we can’t directly assess biological age, we need to use other observable metrics. And since no one has yet to discover a way to indefinitely halt biological aging in people, any measure of it should increase as a function of chronological age. Measures of biological age that have weak correlations with chronological age—where it wouldn’t be hard to find 80-year-olds with the same biological age estimate as 30-year-olds—are probably not valid. An example of a weakly correlated age estimate is measuring the length of telomeres (protective caps at the end of DNA strands that shorten over time) in white blood cells.
Conversely, measures that correlate so strongly with chronological age as to be indistinguishable are also not valid—we don’t expect all 30-year-olds to have the same biological age. Like Goldilocks, what we look for is a correlation that is strong, but not too strong. Most 60-year-olds, for instance, would have higher biological ages than most 50-year-olds, but they’d have a small range of predicted biological ages (say 10 years) among themselves.
Measures of biological age should signify something about the individual’s function, disease risk and remaining life expectancy. More so than chronological time, biological age should be associated with health and disease. Individuals with higher estimated biological age (compared to their chronological age) should be those who are indeed at greater risk of functional decline, adverse health events and lower life expectancies. If measures do not predict these outcomes, they are likely capturing characteristics that change over time yet have no actual bearing on health and well-being—for example, the graying of hair. Similarly, behaviors or interventions that we know alter health outcomes should also alter any good measure of biological age, and those that do not affect health outcomes should not. For instance, exercise and a healthy diet are thought to be beneficial to the aging process and thus should be reflected in estimates of biological age.
Measures of biological age should be stable, but not too stable. We do not expect someone’s biological age to fluctuate wildly over short periods. Moreover, if we were to estimate someone’s biological age twice in the same day, we would expect to get the same answer. Unfortunately, many of the tests currently being used to quantify aging are rife with technical artifacts and noise, making them ill-suited for use in individuals (though they may be useful in population studies). For instance, in 2022 my team and I discovered that epigenetic clocks, the most popular methods to date for estimating biological age, can produce estimates that differ by nearly nine years when measured from the same blood sample. (This method estimates biological age by interpreting methylation patterns in DNA from genes being turned on and off, which change as people age.) While we were able to develop statistical techniques to remove this noise, many estimates available don’t correct for these artifacts and are therefore still at risk of delivering biased results.
Measures of biological age should reflect the multifactorial nature of the aging process. I immediately dismiss any estimate of biological age based on one metric alone—for instance, VO2 max, glucose or lung capacity. Most lack the proper stability, or they fail to robustly predict bodily decline. Even if they do both, they are certainly painting an incomplete picture of an individual’s overall aging process. While each of us can be defined by a single chronological age, we can’t say the same about biological age. Every organ in our body, if not every cell, is likely aging at a different rate, and therefore has a different biological age. This explains why some people develop heart disease, yet maintain muscle strength, while others might experience cognitive decline, but enjoy prolonged metabolic health. The multifactorial nature of the biological aging process means that we will never adequately capture it using a single quantitative value. Instead, we need to work towards developing models that more fully encompass all the ways our bodies might fail us over time.
Science has shown that discovering biomarkers of biological aging is within our reach. In fact, with the arrival of advances in artificial intelligence and machine learning and the success of AI in domains like language and vision, many are now turning their sights to biology. Models are being developed to quantify and manipulate the states of molecules, cells, organs and organisms, and eventually these powerful tools may uncover the keys to programming health.
Despite AI advances, biological age estimates will always have shortcomings. We will still have to be cautious in our personal pursuits of health to avoid over-interpreting biological age changes. Claims to have reversed biological aging should be examined critically, and individuals employing such metrics for their own longevity journeys should be mindful of concepts like Goodhart’s law: “When a measure becomes a target, it ceases to be a good measure.”
Explore the emerging science of healthspan in other stories in this special report.
