Biological Age: The Science and History Behind the Biomarker of Health

The history, scientific basis of, and commercial applications of the biomarker, biological age.

This article is a deep dive into the science and history of the measurement of Biological Age. After a brief intro, check out the deep dive informed by DeepResearch.

The vast majority of biomarkers in healthcare are biomarkers of disease status, assigning a binary status of "Normal" or "Abnormal." While helpful for monitoring disease status, the "Normal" assignment serves little value for those trying to improve their health. Biological Age is an emerging biomarker that may hold promise for health-focused interventions.

This article reviews the history, scientific basis, and commercial applications of biological age (BA). BA was designed to capture a more accurate picture of physical health than chronological age (CA), or how old you are. With the rise in people’s interest in their own health data, there is a need for biomarkers that represent not only absence of disease, but markers that can guide health even when not ill.

Biological age is typically based on functional assessments or molecular markers of cell or tissue aging (or environmental exposures, like pollution). Companies like Function Health (and many more below) offer a BA score. 40 year olds are delighted to report that they have a BA of 32! Perhaps this provides reassurance to some that they are living a healthy lifestyle. 40 year olds with a BA of 45 may want to consider changing some habits.

But here’s the catch. What to do with this information?

While a BA might give you peace of mind, there is still work to do considering its accuracy, prediction of eventual disease risk, and more (e.g. do you want molecular features suggesting you’re a 10 year old, as an ideal metric of health?).

To be clear, I support companies developing biomarkers of health and BA is a good one. Some of the best markers have been tested and validated against prediction or mortality and morbidity. I also am interested in BA scores like DunedinPACE, a DNA methylation measure of current aging rate (years aged per year). An aging rate, based off of analysis of longitudinal data, is more likely to respond quickly and understandably to a health intervention. From a business model perspective, this is also attractive, as more regular measurements might tell the user whether or not they are on track.

But to understand the future of biological age, I wanted to get a comprehensive understanding of the start and the science behind it.

See below a comprehensive report on Biological Age I conducted with DeepResearch. It is an in depth read so if you don’t have much time, I recommend you skim the headers and review the two tables (1 on science, 2 on companies).

Also, let me know if you like or dislike me posting this level of detail with an AI research tool! References at the end

Photo by Ryunosuke Kikuno on Unsplash

1. Introduction & History

Concept of Biological Age: Biological age (BA) refers to the functional age of an organism’s cells or systems, as opposed to simple time-based chronological age (CA) . Unlike CA, which advances uniformly, BA aims to quantify the degree of aging – the accumulated damage and decline in function – which can vary widely between individuals of the same chronological age . Researchers sought better metrics for aging because CA alone poorly predicts health status in older adults . In gerontology, the idea of a “biological age” emerged in the mid-20th century as scientists observed that people age at different rates. Early attempts in the 1960s–1980s used batteries of physiological tests to derive an estimated BA. For example, Borkan and Norris (1980) combined physical fitness measures to compare active vs. inactive individuals’ “biological age” . These early studies introduced the notion that lifestyle factors could make one’s body younger or older than their birth age.

Early Biomarker Milestones: Over time, researchers identified specific biological markers of aging. Telomere length – the protective DNA end-caps of chromosomes – was one of the first molecular biomarkers proposed for BA. By the 1990s, studies showed telomeres shorten with age and individuals with shorter telomeres tend to have higher biological ages . However, telomere length alone proved to have high variability and limited accuracy for individuals . In the 2000s, scientists developed composite indices of BA by combining multiple biomarkers. Notably, Klemera and Doubal (2006) introduced a new computational method (KDM) to combine biomarker data with improved accuracy . Their approach addressed statistical issues in earlier linear models and yielded BA estimates that better predicted mortality than chronological age .

Epigenetic Clock Breakthrough (2011–2013): A major modern milestone was the discovery of epigenetic clocks. These are predictors of age based on DNA methylation levels at specific sites on the genome. In 2011, preliminary reports suggested DNA methylation patterns could act as an “aging clock” . The first robust epigenetic age predictor was published by Gregory Hannum’s group in 2013, using 71 methylation markers in blood to predict age with ~5 year error . Later in 2013, Steve Horvath published a multi-tissue epigenetic clock using 353 CpG sites, which became a landmark in aging research . Horvath’s epigenetic clock could estimate the ages of diverse tissues with a median error of only ~3.6 years . This high accuracy (correlation r≈0.96 with chronological age) clearly outperformed earlier biomarkers like telomere length . The publication of Horvath’s clock was heralded as “a milestone in epigenetics and aging research” . It demonstrated for the first time that a single biomarker model could robustly capture the biological aging process across the body.

Modern Developments: Since 2013, the field of biological age assessment has expanded rapidly. Researchers have created numerous next-generation “aging clocks” that incorporate different molecular data – including transcriptomic (gene expression) clocks, proteomic clocks, metabolomic clocks, immune biomarkers, and composite multi-omic clocks. For example, in 2018 Morgan Levine and colleagues developed DNAm PhenoAge, an epigenetic predictor trained not just to chronological age but to physiological health (morbidity/mortality risk) . In 2019, Lu et al. introduced GrimAge, an epigenetic clock that predicts remaining lifespan by integrating DNA methylation surrogates for risk factors (like smoking) . Beyond DNA methylation, Stanford researchers Lehallier et al. (2019) analyzed blood proteins and found a set of 373 proteins could predict age within a few years. Meanwhile, studies like the DunedinPoAm and DunedinPACE (2020–2021) created longitudinal BA measures to quantify the pace of aging in individuals by tracking multi-year changes in biomarkers. The concept of BA has thus evolved from crude physiological indices to sophisticated molecular clocks. Today, we have a toolkit of epigenetic clocks, proteomic and metabolomic signatures, and composite algorithms that collectively mark the modern era of aging research. Historically, doctors assessed aging by outward signs (frailty, gray hair), then by clinical measures (blood pressure, cholesterol), and now by “omics-based aging clock biomarkers” that integrate vast molecular changes into a single BA metric.

In summary, the quest to measure biological age has a rich history. Early insights (e.g. telomeres, frailty indices) set the stage, but modern molecular and multi-omic clocks represent a quantum leap. Especially since the 2010s, key discoveries – Horvath’s epigenetic clock, second-generation clocks (PhenoAge, GrimAge), and multi-omic integrations – have transformed BA from a gerontological concept into a measurable biomarker. This history underpins current efforts to apply BA in science and medicine.

2. Scientific Basis & Approaches

Biological age can be estimated through multiple approaches. These include single-biomarker methods (like telomere length), composite indices of clinical measures, and data-driven multi-“omic” clocks using DNA methylation, transcript levels, proteins, metabolites, or combinations thereof. Different statistical techniques – from simple regression to machine learning – are used to derive BA from these biomarkers. Below we outline the major approaches, the mathematical models behind them, and compare their performance.

2.1 Molecular and Multi-Omic Biomarkers of Aging

• Telomere Length: Telomeres are repetitive DNA sequences capping chromosomes, which shorten with each cell division. Telomere length was one of the earliest molecular indicators proposed for BA. Shorter telomeres correlate with older age and age-related diseases. For example, leukocyte telomere length has a weak inverse correlation with chronological age (people roughly lose 20–40 base pairs per year) and individuals with unusually short telomeres can be biologically older . However, telomere length as a BA biomarker has significant limitations. It has high inter-individual variability and is strongly influenced by inherited factors and early life environment. Thus, its predictive accuracy for individual aging is low – Horvath noted that DNA methylation clocks “clearly outperformed…telomere length” as age predictors . Telomere tests often have error margins of ±5–10 years. They also primarily reflect cellular replicative history, neglecting other aging processes. As a result, telomere length is now seen as one piece of the aging puzzle rather than a comprehensive BA measure. It’s most useful in specific contexts (e.g. diagnosing telomere syndromes, or as one input among many in composite BA algorithms).

• DNA Methylation (Epigenetic) Clocks: DNA methylation (DNAm) clocks are currently the leading BA biomarkers. They rely on epigenetic markers – specifically the methylation status of CpG sites in the genome – which change in reproducible ways with age. The Horvath and Hannum clocks introduced in 2013 demonstrated that a weighted combination of methylation values at specific genomic sites can closely predict chronological age . Statistically, these clocks are built using penalized linear regression models (e.g. LASSO or elastic net) on large training datasets . The model outputs a formula:

where each Xj is the methylation level at one CpG site and Bj its weight . Horvath’s multi-tissue clock, for instance, uses m=353 CpG sites with learned coefficients Bj. The clock’s prediction is in units of years and was found to correlate ~0.96 with actual age (error ~3.6 years) . Another blood-based clock by Hannum et al. uses 71 CpGs (trained via elastic net) with ~4.9 year error. Advantages: Epigenetic clocks are highly accurate for age estimation – often with Pearson correlations 0.90–0.99 to CA and median errors of 3–5 years . They also capture aspects of biological aging: deviations between DNAm age and actual age (known as “epigenetic age acceleration”) have been linked to lifestyle, disease, and mortality risk . Indeed, DNA methylation age has been shown to predict all-cause mortality better than chronological age . Epigenetic clocks can be measured in saliva, blood, or tissue and tend to be robust and repeatable for a given individual. Limitations: Early DNAm clocks were trained to maximize correlation with chronological age, which means they excel at age prediction but do not inherently indicate health status. Someone with DNAm age 5 years above their CA may have higher risk, but one 5 years below isn’t necessarily “disease-proof.” Newer clocks address this by training on functional outcomes (e.g. GrimAge predicts mortality risk rather than age). Another limitation is tissue-specificity – some clocks work only in blood, for example – though Horvath’s was deliberately pan-tissue . Finally, DNAm analysis requires lab assays (bisulfite sequencing or arrays), which are relatively expensive (a few hundred dollars) and require DNA, though these costs are dropping. Despite limitations, epigenetic clocks are considered the gold standard of BA estimation today due to their accuracy and proven links to aging phenotypes.

• Transcriptomic (Gene Expression) Clocks: Aging affects gene expression patterns, so researchers have built clocks from transcript levels (e.g. RNA sequencing data). The first major transcriptomic clock (Peters et al. 2015) used blood RNA data from multiple cohorts . However, gene expression clocks have shown lower accuracy in predicting chronological age compared to DNAm clocks . Peters’ blood transcript clock had highly variable error across cohorts and significantly worse age correlation than DNA methylation clocks . The heterogeneity of expression data (influenced by cell type proportions, circadian rhythms, recent stimuli, etc.) and technical noise (different array/sequencing platforms) contributed to this variability . Subsequent efforts improved performance – for example, an ensemble of multiple models by Fleischer et al. (2018) in skin fibroblasts slightly outperformed elastic net in that setting . More recently, researchers showed that clever pre-processing (e.g. binarizing expression as high/low) plus elastic net can achieve R² ~0.92 and ~6.6-year error in human fibroblasts . Unique value: Transcriptomic clocks may capture aspects of aging not seen in methylation. In Peters’ study, the RNA clock was associated with certain biomarkers (like systolic blood pressure) that the Horvath/Hannum clocks did not pick up . This suggests gene expression could reflect dynamic physiological changes (e.g. inflammation, stress responses) relevant to aging. Limitations: Expression data are inherently noisy and cell-type specific. Most transcriptomic clocks to date have been pilot studies on limited samples . They also lack extensive validation in independent cohorts . Thus, while promising, transcript-based BA measures are less established. They may ultimately be used in combination with other omics for a fuller picture, rather than as stand-alone age predictors (except perhaps in model organisms where they’ve shown excellent results ).

• Proteomic Clocks: Systemic aging causes changes in circulating protein levels (e.g. hormones, cytokines, structural proteins). Recent advances in proteomics (such as aptamer-based assays measuring thousands of proteins) enabled “proteomic clocks.” In 2018, Tanaka et al. developed a plasma protein clock that predicted age and whose deviation (Proteomic Age Gap) correlated with clinical aging traits . In 2019, Lehallier et al. analyzed ~3,000 proteins in plasma and found a subset that could predict age with high accuracy across independent cohorts . Both studies observed that proteomic BAs were associated with functional outcomes – e.g. worse physical and cognitive function, and multimorbidity, corresponded to older proteomic age . Tanaka later showed the proteomic clock’s age acceleration predicts mortality and frailty, similar to epigenetic clocks . Advantages: Proteins are the effectors of biology, so proteomic clocks can be biologically interpretable. Many age-associated proteins have known roles in aging (e.g. inflammatory cytokines, growth factors). In fact, over 200 proteins identified in plasma aging signatures have direct links to organ health or lifespan regulation . For instance, these clocks often highlight immune and metabolic pathways – consistent with known Hallmarks of Aging. An advantage over DNA markers is that proteins change more rapidly in response to interventions (diet, drugs, etc.), so proteomic BA might be more sensitive to short-term changes. Also, blood proteome reflects contributions from many organs, offering a window into whole-body aging . Limitations: Proteomic technologies are newer and less standardized than DNA assays. Not all proteins can be measured reliably yet (current platforms detect a few thousand of the >10,000 proteins in plasma) . Results can be confounded by factors like kidney function (which alters concentrations of many proteins) . Proteomic clocks have not (as of yet) achieved the same predictive accuracy for chronological age as top DNAm clocks; typical performance is r ~0.9 and ~5–7 year error, although improving . They also require expensive equipment (affinity arrays or mass spectrometry). Nonetheless, proteomic clocks add valuable information – often complementing epigenetic clocks by capturing current physiological state (e.g. levels of inflammatory proteins) rather than stable epigenetic “memory” of aging.

• Metabolomic and Other Clocks: Metabolites (small molecules like lipids, amino acids, etc.) also change with age. Metabolomic clocks are under development – e.g. a 2020 study by Robinson et al. built a metabolomic age predictor and found its age gap had low correlation with epigenetic clocks, indicating it measured different aging signals . Like proteomics, metabolomics can reflect real-time metabolic health (insulin sensitivity, oxidative stress, etc.). However, metabolite profiles are highly dynamic (affected by diet, time of day), making stable age prediction challenging. Other niche approaches include immune age (e.g. the “iAge” clock by Furman et al. 2021 uses inflammatory cytokines to predict immune system aging ) and microbiome age. The gut microbiome’s composition shifts with age, and AI models have predicted age within ~5–6 years using microbiome data . For instance, a 2023 study used stool metatranscriptomic profiles from 90,000 individuals to predict age with R²=0.53 (i.e. ~73% correlation) . Interestingly, microbiome-based BA was associated with lifestyle – e.g. people on high-fiber vegetarian diets had younger microbiome ages, while those with unhealthy diets or IBS had older microbiome ages . While intriguing, these “non-traditional” clocks (metabolites, microbiome, etc.) are generally less validated and less accurate for age prediction than DNA or protein clocks. Their real strength may lie in capturing specific aspects of aging (e.g. immune system inflammation or gut health) that general clocks might miss. Going forward, they are likely to be used in multi-omic composite clocks rather than alone.

• Phenotypic/Clinical Markers: Before the molecular era, researchers derived BA from clinical and functional markers (blood pressure, lung capacity, grip strength, blood biomarkers like glucose, cholesterol, etc.). These phenotypic age models remain relevant, especially when one wants a practical health-based age. A prominent example is Phenotypic Age (PhenoAge) by Levine et al. (2018), which combines 9 common clinical lab values (e.g. albumin, blood sugar, C-reactive protein) and chronological age into a score that corresponds to mortality risk . The PhenoAge formula was published along with an equation to calculate a person’s BA in years . Such clinical indices often use weighted sums or principal components. Another example is the Frailty Index, which can be interpreted as a BA: it sums up health deficits (symptoms, diseases, functional impairments) and can be mapped to an equivalent “frailty age” by comparing to population averages. Advantages: Phenotypic BA measures use readily available data (clinical tests or exam findings), so they are easy to implement in healthcare. They also directly relate to health outcomes – PhenoAge, for instance, was shown to better predict mortality and comorbidities than chronological age . Limitations: These measures are generally less precise as clocks of aging. Their correlation with chronological age is modest (often R≈0.7–0.8) because they capture age-related decline only when it is advanced enough to affect clinical markers. They may also be confounded by acute illness (e.g. an infection will raise inflammation markers and phenotypic age temporarily). In addition, they provide limited insight into underlying biology compared to molecular clocks. Nonetheless, clinical/phenotypic BA indices are useful as prescriptive tools (they translate easily into medical advice, see Section 3) and can be combined with molecular data to enhance predictions.

Notes: Accuracy is described in terms of correlation (r) with chronological age and typical median error. “Multi-omic” clocks are an emerging category – research suggests different omics clocks capture distinct aspects of aging , so integration could improve overall fidelity. For instance, one study found epigenetic vs. proteomic age measures were uncorrelated in the same individuals, each reflecting different phenotypes . This supports the idea that combinations of markers will give the most robust biological age.

2.2 Statistical Methods for Deriving Biological Age

Regardless of data type, constructing a BA estimator requires statistical modeling. Early approaches used simple multiple linear regression (MLR): select a set of biomarkers and fit a linear equation to chronological age. While intuitive, MLR can suffer issues like overfitting (especially if biomarkers are numerous and inter-correlated). Modern clocks therefore use more advanced methods:

• Penalized Regression: Most first-generation clocks (Hannum, Horvath, etc.) used regularized regression (e.g. LASSO or elastic net). These add a penalty for including too many variables, effectively performing feature selection and reducing multicollinearity . For example, Horvath’s clock started from 21k CpGs and via elastic net narrowed to 353 CpGs with nonzero weights. The resulting linear model is simple to apply and interpretable: positive weight CpGs mean higher methylation -> older age, etc. . An equation from a generic linear clock is given above. Advantages: well-understood, fast, yields an explicit formula. Disadvantages: may miss nonlinear relationships and interactions.

• Principal Component Analysis (PCA): An alternative is to reduce biomarkers to principal components and use the leading component(s) as BA. Some early BA models did this, essentially defining BA as the first principal component of a set of age-related biomarkers . PCA can mitigate multicollinearity and capture the largest variance in one score. However, the resulting “age score” is a unitless composite that then must be mapped to years (often by regression to chronological age). PCA was popular in older studies, but it still assumes linear combinations and doesn’t necessarily align with optimal prediction of outcomes .

• Klemera-Doubal Method (KDM): KDM is a specialized algorithm introduced in 2006 to improve BA estimation from multiple biomarkers. It treats true BA as a latent variable and derives a formula that minimizes error in predicting both age and each biomarker . Without going into derivation, KDM produces a weighted average of biomarkers where weights depend on each marker’s correlation with age and with each other. It “solves the paradox” that simply regressing on age may overweight some biomarkers . KDM-based BA has been found to track mortality risk better than naive methods. Many later studies (including Belsky et al. 2015) adopted KDM for combining clinical biomarkers. Pros: lower estimation error and more biologically consistent BA (doesn’t give absurd values at extremes). Cons: more computational steps and less transparency than a plain regression.

• Machine Learning & AI: In recent years, non-linear models and AI techniques have been applied to aging clocks. For example, deep neural networks (DNNs) were used by Putin et al. (2016) to predict age from basic blood tests in >60k people . That work showed DNNs slightly outperformed linear models and could handle complex patterns. Deep models have also been applied to imaging (photo-based age estimators) and multi-omics. A benefit of DNNs is the ability to capture interactions and non-linear effects automatically. For instance, DeepMAge (2018) and others used autoencoders on transcriptomic data to boost signal-to-noise . Another novel idea is using Generative Adversarial Networks (GANs) to simulate a “digital twin” that can be aged forward or backward by changing inputs – though this is still experimental. Pros: When large training data are available, ML/DL can improve accuracy and incorporate diverse data types simultaneously . For example, one can train a model on combined clinical, methylation, and proteomic data to predict age or health status, which would be hard to do with standard regression. Cons: These models are “black boxes” – less interpretable – and risk overfitting if data are limited (the “curse of dimensionality” is acute for omics: thousands of features vs. limited samples ). Indeed, a 2018 study by Pyrkov et al. noted that extremely flexible models can overfit noise if not carefully controlled . Additionally, deep models typically still need to be calibrated to output an age in years (often by training to chronological age).

• Longitudinal vs Cross-Sectional Models: A methodological distinction: most clocks are trained cross-sectionally (different people of various ages). But some models use longitudinal data – tracking individuals’ biomarker changes. The Pace of Aging measure (Belsky et al.) is derived from slopes of biomarkers over decades. Such longitudinal approaches require repeat measurements, but they directly capture aging rate. They are valuable for detecting change within an individual (e.g. in an intervention trial) even if their absolute BA calibration is less emphasized. There is growing interest in clocks like DunedinPACE, a DNA methylation measure of current aging rate (years aged per year).

In practice, many BA models blend techniques. For example, PhenoAge first created a mortality score via Cox regression of clinical markers, then fitted that to chronological age via linear model to get an age estimate . GrimAge combined a nested approach: it built surrogate DNA methylation biomarkers for plasma proteins and smoking pack-years, then summed them to predict lifespan . Such multi-step models still ultimately output a single BA number.

Accuracy and Validation: To evaluate a BA model, researchers use metrics like correlation with chronological age, mean absolute error (MAE) in years, and test-retest reliability. For clinical utility, more important is whether BA predicts hard outcomes (mortality, disease) better than chronological age. Epigenetic clocks, especially second-generation ones, have consistently shown added predictive value . For example, persons whose DNAm age is 5 years > their CA have significantly higher mortality over follow-up . Conversely, if an intervention slows biological aging, one hopes the BA measure will reflect that change.

In summary, the science of BA involves both the biology (what biomarkers to use) and the analytics (how to combine them). Table 1 highlighted differences in biology; on the analytics side, approaches range from simple to sophisticated. Traditional linear modeling has given way to regularized regression and now to machine learning as data become high-dimensional. Each method has trade-offs in interpretability and performance . Ensuring that BA models are not just accurate but meaningful (tied to health) is an ongoing challenge. Continued research compares these methods – for instance, a 2023 review noted that first-generation epigenetic clocks mostly captured chronological time, while newer clocks and other omics capture more “physiologically relevant” aging signals. The next frontier is likely multi-modal models that use both biological insight and AI to integrate many markers, hopefully yielding an even more accurate and predictive measure of biological age.

3. Commercial and Clinical Uses

Biological age has moved from research into practical use. In this section, we explore how BA is being used in health and longevity contexts – as a prescriptive biomarker to guide lifestyle changes, as a potential diagnostic/prognostic tool in emerging “Medicine 3.0” paradigms, and in personalized longevity and preventive medicine.

BA as a Health Indicator: Many longevity experts argue that “your biological age matters way more than your chronological age” for evaluating health. In clinical or wellness settings, measuring BA provides an accessible summary of an individual’s healthspan status. For instance, if a 60-year-old has a BA of 50, it suggests they are physiologically younger than peers – often due to healthier lifestyle or genetics. Conversely, a BA higher than CA can be a warning sign. Healthcare providers are beginning to include BA testing in preventive health assessments. It serves as a motivational metric for patients to improve their lifestyle: seeing one’s BA drop after weight loss or better diet can reinforce those changes. Some companies call this “bio-age tracking,” analogous to tracking cholesterol or blood pressure.

Prescriptive Use and Lifestyle Coaching: Biological age is increasingly used in personalized wellness programs to prescribe interventions. Companies and clinics analyze an individual’s BA and then recommend targeted strategies (nutrition, exercise, supplements, etc.) to reduce it. For example, an elevated “inflammatory age” (like iAge) might prompt an anti-inflammatory diet and omega-3 supplements. If someone’s BA is much higher than their CA, a clinician might prescribe more aggressive interventions – stricter exercise regimen, caloric restriction or fasting, or pharmacological therapies (like statins or metformin) – aiming to slow down the clock. The idea is that BA provides a quantitative goal for healthy behavior: e.g. “let’s try to get your biological age down by 5 years over the next 6 months through lifestyle changes.” Some consumer testing companies include follow-up coaching: for instance, TruDiagnostic’s service pairs epigenetic age results with a consult to develop a personalized plan to slow aging. There is evidence this approach can work. A small 2019 trial (FAHY et al.) found that diet, exercise, and a drug cocktail reduced DNAm age by ~2 years in 12 months, suggesting BA is modifiable. Larger studies like CALERIE and others are underway to see if lifestyle interventions consistently lower biological age.

Medicine 3.0 and Diagnostics: Medicine 3.0 is a term popularized by Dr. Peter Attia for a preventive, personalized approach to healthcare. In Medicine 3.0, doctors focus on anticipating and forestalling chronic diseases rather than reacting after they occur. Biological age fits naturally into this paradigm as a preventive diagnostic tool. Rather than waiting for a patient to develop diabetes or heart disease, a physician might monitor the patient’s BA and related biomarkers; if their BA is climbing faster than ideal, that may signal underlying issues (e.g. rising insulin resistance, vascular aging) warranting early intervention. Attia notes that optimizing healthspan requires using advanced metrics like BA, not just traditional risk factors . Already, some forward-looking practices incorporate BA: for example, a concierge clinic might test epigenetic age and find a 45-year-old client has a BA of 55 – prompting advanced screening for arterial plaque or cognitive testing, on the premise that their body shows signs of accelerated aging. In geriatric medicine, BA could help identify “biologically frail” patients who are in need of closer monitoring irrespective of chronological age. In insurance and underwriting, there is interest in using aging biomarkers to refine risk estimates (though ethical concerns exist). Notably, an NIH workshop and the FDA are discussing use of epigenetic clocks as clinical trial endpoints for anti-aging therapies, effectively treating BA as a surrogate marker of efficacy. This is still in early stages, but if validated, a drug that lowers BA could be approved for “aging-related indication” in the future.

Longevity and Personalized Optimization: A burgeoning field is longevity medicine – medical specialty focusing on extending healthy lifespan. Longevity clinics often measure a panel of biomarkers (genomic, epigenomic, microbiome, imaging, etc.) and among these, biological age is key. Longevity physicians consider the BA gap (BA minus CA) as an indicator of whether a patient is on track or needs aggressive interventions. For example, Dr. Evelyne Bischof, a longevity physician, has noted that they combine genetic, physiological, imaging, and epigenetic data to define the optimal biological age for a patient and tailor interventions to achieve it. BA is also used to track response to novel therapies: if a patient engages in a 3-month rapamycin trial or a specialized diet, measuring their epigenetic age before and after helps gauge impact. In the biohacker community, people self-experiment and use commercial BA tests (Section 5) to see if supplements (like NAD boosters, senolytics) or lifestyle hacks are “making them younger.” While results are anecdotal, this practice underscores the demand for BA as a feedback metric for personal optimization.

Patient Motivation and Psychology: Importantly, BA can have a psychological impact that traditional metrics lack. Telling a person their “biological age” often resonates more than saying “your cholesterol is high.” It converts abstract health data into an age equivalence that people intuitively understand. A Guardian article described an individual who gifted his 59-year-old mother an epigenetic test; finding out she had a BA closer to 40 greatly relieved her anxiety about “entering her 60s”. On the other hand, someone who appears outwardly healthy might be spurred into action if their BA comes back higher than expected. Thus, BA testing can galvanize individuals to make positive changes, functioning as a behavior change tool.

Use in Monitoring Interventions: In preventive cardiology or endocrinology, doctors might use BA alongside other markers to monitor progress. For example, after 6 months of optimized blood pressure, glucose, and exercise, an improvement in BA (say from 60 to 55 in a 58-year-old) would confirm holistic benefit. BA can also integrate the net effect of multiple interventions: instead of looking at many separate risk factor improvements, the clinician sees the bottom-line impact on the patient’s aging trajectory. This is analogous to how “vascular age” is used in cardiology to summarize risk factors as one age number; here it’s just broader to overall aging.

Challenges and Cautions: While promising, using BA clinically comes with caveats. There is person-to-person variability – not everyone with an elevated BA will immediately experience adverse outcomes, so it’s a probabilistic indicator. There’s also a risk of misinterpreting the results: e.g. a patient might be demoralized by an “old” BA and view it as immutable fate, which must be countered with education that BA can change. Additionally, since multiple BA algorithms exist, a clinician must choose the appropriate one (a DNAm clock? a phenotypic age?). Right now, these tests are mostly in the wellness domain rather than standard medical practice. However, as evidence builds and tests become cheaper, it’s easy to envision BA being recorded in medical charts alongside other vital signs.

Medicine 3.0 Vision: In Medicine 3.0 – the preventative, data-driven model – biological age might eventually be treated as another vital sign or lab result to manage. Attia notes that over 90% of longevity is determined by lifestyle and environment, not genetics . Therefore, tracking a metric that encapsulates the cumulative effect of lifestyle (which BA largely does ) is extremely valuable. Some futurists imagine continuous or frequent BA monitoring (for instance, using epigenetic markers from wearables or blood drops) guiding real-time adjustments to one’s health regimen . While that is still speculative, it underlines the central role BA could play in next-generation healthcare.

In summary, biological age is increasingly used as a prescriptive and prognostic biomarker. It serves as a catalyst for preventative action in Medicine 3.0, a personalized target in longevity programs, and a means to measure the efficacy of interventions aimed at extending healthspan. Both consumer and clinical domains are adopting BA – from wellness coaching dashboards to integrative longevity clinics. The ultimate hope is that by managing one’s biological age, one can delay or prevent age-related diseases, effectively treating “aging” as a modifiable risk factor. Early adopters are optimistic: as one longevity biotech CEO put it, “AI can show you to what extent you can revert closer to your optimal biological age”, empowering individuals to take control of their aging process .

4. Limitations & Optimal Biological Age

Despite the excitement around biological age, it’s critical to recognize the limitations of current models and to ask: Is there an optimal or lowest possible biological age for a given person? This section discusses conceptual and practical limitations, and whether we can (in theory or practice) reach an “ideal” biological age.

Inter-individual Variability: Aging is a highly individual process – genetics, early development, lifestyle, and diseases all influence it. Biological age models attempt to reduce this complexity to one number, which inevitably has limitations. People can have discordant aspects of aging: for instance, a person might have youthful cardiovascular markers but accelerated immune aging. A single BA metric may not capture such nuance. Indeed, studies find low correlations between different types of biological age measures in the same person . One analysis showed epigenetic age acceleration was uncorrelated with proteomic or metabolomic age acceleration in individuals . This means a person could be “older” by one clock and “younger” by another. Thus, an important limitation is that current BA models reflect only portions of the aging process. We might say everyone has multiple “biological ages” (heart age, brain age, immune age, etc.), and an aggregate BA tries to summarize them. If someone’s BA is labeled 50, it might hide internal divergences – e.g. their brain might function like age 40 but kidneys like age 60. This variability complicates the idea of a singular optimal BA.

Optimal Biological Age: Is there an optimal or minimum biological age one should aim for? Intuitively, the “best” scenario would be to have the biology of a young adult (when peak physiological function is attained). Indeed, researchers have suggested that the optimal biological state corresponds to roughly the 20s to early 30s, when organ function, strength, and cognitive performance are near maximum . An AI-guided longevity study noted the goal of interventions is to keep an individual’s biological network state as close as possible to that of “the age group with optimal performance (20–30 years)” . In other words, returning an older person’s BA to that of a 25-year-old is a theoretical ideal. Beyond young adulthood, even teenagers or children are not “better” in terms of health – for example, a 5-year-old has a young immune system but also not a fully developed one. So having a BA of a child is not a meaningful target for an adult (and obviously not realistically attainable without reverting development). This suggests the floor for optimal BA in mature adults is around the mid-20s. Many clocks also implicitly consider 20–30 as the baseline, since they often set BA≈CA for young adults in training data.

Can BA be too low? Generally, a lower BA (relative to CA) is considered positive – it indicates slower aging. There isn’t evidence of harm in, say, a 60-year-old having biomarkers akin to a 50-year-old. However, one could speculate about imbalances: if someone artificially resets certain markers to very youthful levels, could it have negative effects? One example is that extremely low cholesterol (typical of a teenager) in a 70-year-old might cause hormone issues. But these are specific cases. For composite BA, having a much younger profile is usually beneficial. A more salient concern is whether it’s possible to substantially reduce BA in all aspects. Some biological changes from aging might be irreversible with current technology. For instance, cross-linking of collagen that stiffens tissues or accumulation of certain cellular waste (lipofuscin) aren’t easily undone by lifestyle changes. So even if one improves many markers, some residual “age” may remain. This may impose a practical lower bound on BA higher than the theoretical 25-year optimal. It’s been observed that even the fittest elderly athletes cannot achieve the blood markers of a 20-year-old in everything – some metrics inevitably track with age.

Technical Limitations of Models: Each BA model has inherent error and uncertainty. Epigenetic clocks, while precise, have an error of ~3 years . This means if you “reverse” your BA by 2 years, it might be within the noise. Also, most clocks are population-calibrated – by design, average BA = CA in the training population and about half of people will have BA > CA and half < CA. Extremely low BA (far below one’s age) might simply be an artifact (falling at tail of distribution or outside training range). For example, Horvath’s clock occasionally assigns a 70-year-old an age of 20 if they are an outlier in methylation – but that doesn’t necessarily mean their whole body is like 20. It could be model extrapolation error. Mathematically, clocks are not well-validated outside the age ranges they were trained on. So if someone 60 tries to drive their epigenetic age down to 25, we should be cautious in interpreting that – the clock may not have had many 25-year-old reference points for a 60-year-old’s data (since certain methylation values might not normally be seen in older folks). In short, the accuracy of extreme BA values is limited.

Another limitation is that BA models generally cannot distinguish all causes of aging. If someone has a high BA due to a treatable condition (say chronic inflammation from periodontal disease), lowering that inflammation will lower BA. But some BA components might reflect past history (e.g. lifelong exposure or early developmental factors) that can’t be fully undone. Epigenetic marks, for instance, carry memory of lifetime exposures – some theorize there’s a portion of DNAm age that is “entrenched” and not reversible without epigenetic reprogramming. This raises the question: can interventions actually reset BA to youthful levels, or do they just slow further increase? The TRIIM trial and anecdotal reports suggest partial reversal is possible, but likely there’s a ceiling (or rather a floor).

Biological Constraints: Aging is multi-factorial and some processes might only be safely reversible to a point. Complete cellular rejuvenation (like turning an adult cell into a pluripotent stem cell) resets epigenetic age to zero , but you obviously wouldn’t do that to an entire person because it would erase cell identity (essentially creating embryonic cells, which in vivo would cause cancer or loss of function). Hence the challenge: to rejuvenate markers of age without undoing necessary differentiation. The emerging field of partial cellular reprogramming tries to dial cells back to a younger state without erasing their identity. In mice, partial reprogramming reversed epigenetic age and improved tissue function – raising hope that one day we might significantly restore youthful BA in humans. Until then, fundamental constraints are that some aging damage (scar tissue, certain neuron loss, etc.) cannot be fully reversed, setting a practical limit on how “young” one’s BA can get with current methods.

Additionally, there’s the concept of biological trade-offs. Some genetic variants that promote longevity come at a cost (e.g. lower fertility). So the “optimal” state might depend on context. For an individual, the goal is to minimize aging damage while maintaining necessary functions. It might not be ideal to push certain biomarkers to extremely youthful levels if it unbalances homeostasis. Fortunately, most BA interventions (exercise, diet, etc.) naturally tend to restore a healthy balance without overshooting.

Precision and Reliability: Current BA tests are not 100% precise. Day-to-day or lab-to-lab variation can occur. For example, one study found a test-retest reliability of around r=0.96 for DNAm GrimAge – excellent, but still a test could differ by ~2 years due to technical noise or transient changes. Other clocks like metabolomic or microbiome have lower reliability (they can fluctuate if you catch a cold or change diet). Thus, a limitation is measurement noise. Clinicians will want to see a substantial BA change (beyond error margins) to conclude real improvement.

No Unified Standard: There is not yet a consensus on which BA model to use for what purpose. This can limit adoption – a physician might be unsure whether to trust an epigenetic age vs. a phenotypic age. Efforts are underway to standardize methodologies and create reference databases. Until then, different tests might give different BA readouts for the same person, which can be confusing. Ideally, future composites or multi-omic clocks will reconcile these differences.

Is There an Ideal BA Target?: Rather than an absolute “optimal” BA, it may be more meaningful to speak of biologically normal age for a given individual. Some people may genetically run “hotter” or “cooler” on certain biomarkers. Instead of chasing an absolute number, physicians often focus on slowing the pace of aging – i.e., keeping BA increase <1 year per chronological year. In fact, the DunedinPACE measure explicitly quantifies that rate . If one’s pace is <1.0 (meaning your BA is increasing slower than calendar time), that’s ideal. If >1.0, it’s a sign of accelerated aging. So an optimal scenario is to consistently maintain a pace at or below 1, which over decades will yield a BA lower than CA.

In conclusion, current BA models have important limitations: they capture only parts of aging biology, have measurement error, and may not fully reverse to youthful states due to biological constraints. There likely is an optimal biological age range (young adulthood) that represents peak health, and the realistic aim is to approach and maintain that state. We should be cautious not to overinterpret small changes or believe we can turn a 70-year-old into a teenager. As one paper noted, “the ageing rate is not constant through life” and differs by organ – thus, one number will always be an approximation. Rather than an “ideal BA” for everyone, the focus is on minimizing biological aging given one’s circumstances, and slowing the trajectory of age-related decline. Fundamental constraints (like irreversible damage and safety limits of reprogramming) mean we may never achieve a BA of 20 at chronologically 80 without advanced therapies. But if we can get that 80-year-old to have the health profile of, say, an average 60-year-old, that is a huge win for longevity and quality of life.

5. Companies & Industry Landscape

The growing interest in biological age has spurred numerous biotech startups and commercial companies offering BA testing or related services. These companies range from direct-to-consumer testing services to biotech firms using aging biomarkers for drug discovery. Table 2 provides an overview of key players in the industry, including their founding year, location, focus, company type, size, and funding/revenue (where available).

Notes on the Landscape: The companies above illustrate different segments. Consumer-facing startups (Elysium, Tally, Viome, InsideTracker) emphasize accessible testing and actionable recommendations. They often bundle BA tests with supplements or coaching, targeting health enthusiasts and early adopters. Prices range a few hundred dollars per test. Biotech and B2B firms (TruDiagnostic, Zymo, Life Length) cater to physicians, researchers, or provide laboratory services. They seek to be gold-standard providers of accurate assays (e.g. TruDiagnostic building the largest DNA methylation database ). Tech-driven companies (Deep Longevity, Gero) focus on algorithms and drug discovery – using BA as a tool rather than selling tests to consumers. Notably, many of these companies collaborate: for example, some consumer brands rely on lab analysis by TruDiagnostic or Zymo in the back-end.

The industry is relatively young – most entrants formed in the last ~5–7 years as the science matured. Funding levels vary: a few have significant venture capital (Viome’s $175M shows interest in multi-omics), while others grew with modest funding but high revenue (TruDiagnostic’s rapid growth without big VC injection). Geographically, it’s a global field: USA has many frontrunners, but Europe (Spain’s LifeLength, UK’s GlycanAge/Chronomics) and Asia (Hong Kong’s Deep Longevity, Singapore’s Gero) also contribute. This reflects worldwide interest in longevity tech.

We also see established life sciences companies dipping in: e.g. Illumina invested in epigenetic clock development and labs like Quest Diagnostics recently launched “biological age” panels . Even traditional insurance companies are experimenting with wellness programs around aging biomarkers .

In summary, the BA industry landscape is dynamic and growing. There’s a mix of direct-to-consumer services, clinical testing providers, and research-focused biotech. Each is carving out a niche – whether it’s providing the most user-friendly epigenetic age report, the most clinically validated test for physicians, or using aging clocks to guide therapeutics development. In the coming years, we can expect consolidation and maturation: perhaps standardization of tests, more regulatory oversight if tests are used medically, and continued influx of capital as longevity tech proves its market.

6. Conclusion

Key Takeaways: Biological age has emerged as a powerful integrative biomarker that condenses complex aging processes into a single metric. Historically, the concept evolved from rudimentary beginnings – recognizing that individuals age at different rates – to cutting-edge molecular clocks. The advent of epigenetic clocks in the 2010s was a watershed moment, providing accurate and biologically meaningful measures of aging . Since then, diverse approaches (telomere, proteomic, metabolomic, composite clinical) have been developed, each illuminating different facets of aging. Scientifically, we now understand that aging is multi-dimensional, and no one measure captures it all . However, DNA methylation-based BA has proven especially robust, correlating with functional decline and mortality risk . The rise of machine learning has further refined these models, enabling integration of multi-omics and potentially increasing predictive power.

Applications: In practical terms, biological age is transitioning from the lab to the clinic and consumer space. It is being used as a prescriptive tool – an actionable metric for individuals to monitor and improve their health. The philosophy of Medicine 3.0, which prioritizes prevention and personalization, aligns closely with using BA as a “North Star” for health optimization . Early adopters in the medical community use BA to identify patients at risk of accelerated aging and to tailor interventions accordingly. In the longevity industry, BA is both an endpoint and a motivator: it helps test anti-aging interventions in trials, and it motivates patients to adhere to lifestyle changes by providing concrete feedback. The commercial landscape has grown rapidly, indicating confidence that consumers value this information. Companies are packaging sophisticated science into relatively accessible tests, although care must be taken in interpretation.

Current Limitations: Notwithstanding the excitement, current biological age measures have limitations. They are imperfect proxies of the complex aging process. One must remember that a BA number is a model output – it comes with statistical error and may not capture everything important about an individual’s health. There is ongoing debate and research on which BA measure is most relevant for a given purpose (e.g. epigenetic vs phenotypic vs combined). Many models still need validation in diverse populations; most clocks were developed on European-ancestry cohorts, and their performance in other ethnic or genetic backgrounds is being studied. Technically, issues like batch effects, differences in lab methods, and cost/access are hurdles to broader implementation (though costs are dropping steadily). On the biological front, it remains challenging to determine causal relationships – does reducing BA (say via a drug) truly reduce disease risk, or is it just cosmetic in the biomarkers? This is a critical question to answer through clinical studies.

Optimal BA & Future Directions: An intriguing question is how far we can modulate biological age. There likely is an optimal window of BA corresponding to peak physiological function (young adulthood). The consensus is that keeping one’s biological age as low as possible (without causing other issues) is desirable for longevity. Achieving that in practice means starting prevention early in life – slowing aging from mid-life onward so that one’s BA rises more slowly than chronological years. In the future, emerging therapies like senolytics (drugs that remove senescent cells), epigenetic reprogramming, advanced gene therapies, etc., could dramatically impact biological age. For instance, partial reprogramming techniques in animals have shown actual reversal of epigenetic age and restoration of function, which if translated to humans might allow resetting the clock in ways currently not possible. The field of aging biomarkers will continue to evolve alongside therapies: more sensitive clocks may be needed to detect subtle rejuvenation effects, or conversely, entirely new biomarkers (like proteomic “aging clocks” for specific organs) might complement existing ones.

We can also expect standardization and integration. It’s plausible that future BA assessments will involve a composite score that weights multiple clocks (epigenetic, proteomic, etc.) to give a more comprehensive picture. Initiatives like the NIH’s biological aging clock consortium are working to compare and unify these measures. In clinical practice, guidelines may eventually emerge on how to use biological age in decision-making (for example, using BA to adjust screening recommendations or drug dosages in older adults – treating patients by biological vs chronological age).

Challenges ahead: There are challenges to resolve. Ethically, widespread BA testing raises questions – could it be used in ways that discriminate (e.g. insurers or employers favoring “younger” biomarker profiles)? Ensuring informed and appropriate use will be important. Scientifically, a major challenge is to pin down the drivers of these biomarkers: for instance, what mechanisms cause epigenetic age acceleration, and can we target them directly? This goes hand in hand with the broader quest to understand aging biology (e.g. the Hallmarks of Aging ) and interventions to modify it.

In conclusion, biological age is a transformative concept that encapsulates the progress in aging research. It provides a quantifiable link between the bench (molecular changes) and the bedside (clinical outcomes). Key milestones have brought us to a point where we can measure BA with reasonable accuracy and use it to gauge interventions. The commercial sector has taken notice, bringing these measures to consumers and doctors and thus generating real-world data and feedback. While current methods are not without flaws, they are continually improving, and their predictive value for healthspan is compelling. The future of biological age assessment is likely to involve more precise, multi-dimensional clocks integrated into routine health monitoring, guiding both individual choices and medical practice. Achieving consensus on optimal BA targets and effectively reversing BA are on the horizon but will require further scientific breakthroughs. Overall, treating biological age as an important vital sign is a cornerstone of the emerging paradigm of longevity-focused medicine, offering hope that we can not only add years to life, but life to years.

7. References

1. Horvath, S. (2013). DNA methylation age of human tissues and cell types. Genome Biology, 14(10): R115. (Introduced the first multi-tissue epigenetic clock; ~353 CpG-based age estimator with ~3.6 year accuracy) .

2. Hannum, G. et al. (2013). Genome-wide methylation profiles reveal quantitative views of human aging rates. Molecular Cell, 49(2): 359-367. (First blood-based DNA methylation age predictor using 71 CpGs; showed methylome can measure aging) .

3. Liu, Z. et al. (2018). A new aging measure captures morbidity and mortality risk across diverse subpopulations from NHANES IV: a cohort study. PLoS Medicine, 15(12): e1002718. (Developed DNAm Phenotypic Age clock combining clinical biomarkers with methylation; better predicts mortality than chronological age) .

4. Tanaka, T. et al. (2018). Plasma proteomic signature of age in healthy humans. Aging Cell, 17(5): e12799. (One of the first plasma proteome clocks; identified proteins whose levels predict age and age-related outcomes) .

5. Lehallier, B. et al. (2019). Undulating changes in human plasma proteome profiles across the lifespan. Nature Medicine, 25(12): 1843-1850. (Revealed nonlinear waves of protein changes with age; defined a 373-protein clock that predicts age and correlates with health measures) .

6. Jylhävä, J., Pedersen, N. L., & Hägg, S. (2017). Biological age predictors. EBioMedicine, 21: 29-36. (Review of different biomarkers and methods for estimating biological age; discusses DNA methylation, telomeres, composites, etc.) .

7. Zhavoronkov, A. et al. (2019). Deep biomarkers of aging and longevity: from research to applications. Aging, 11(22): 9391-9410. (Overview of AI-based aging clocks; discusses deep learning models on various data types and their applications in insurance, health, etc.) .

8. Field, M. G. et al. (2023). Measuring biological age using omics data. Nature Reviews Genetics, 24(6): 329-346. (Recent comprehensive review on multi-omic aging clocks; compares clocks from epigenome, transcriptome, proteome, metabolome and their overlap) .

9. Belsky, D. W. et al. (2020). Quantification of the pace of biological aging in humans through a blood test: the DunedinPoAm DNA methylation algorithm. eLife, 9: e54870. (Presented a DNA methylation algorithm to measure the pace of aging – how fast someone is aging biologically – based on longitudinal cohort data) .

10. Lu, A. T. et al. (2019). DNA methylation GrimAge strongly predicts lifespan and healthspan. Aging, 11(2): 303-327. (Developed GrimAge clock by incorporating DNA methylation proxies for smoking and plasma proteins; shown to predict time-to-death and disease with high accuracy) .

11. Fahy, G. M. et al. (2019). Reversal of epigenetic aging and immunosenescent trends in humans. Aging Cell, 18(6): e13028. (Small pilot trial showing that a combination of growth hormone, DHEA, and metformin reduced participants’ DNAm age by ~2.5 years over 1 year – first reported epigenetic age reversal in humans).

12. The Guardian (June 13, 2022). “Real age versus biological age: the startups revealing how old we really are.” (Interview with Dr. Steve Horvath; discusses consumer DNA methylation tests like Elysium’s, and the concept of using biological age to motivate lifestyle changes) .

13. Atanasova, T. (2024). “GlycanAge raises $4.2M Seed to Revolutionize Health and Longevity with Biological Age Testing.” (The Recursive, Feb 19, 2024) – News on GlycanAge startup: glycan-based BA test and plans to develop diagnostic biomarkers .

14. Fitt Insider (2023). “Tally Health Lands $10M for Personalized Longevity Platform.” (Mar 7, 2023) – Press coverage of Tally Health launch: co-founded by David Sinclair, offering cheek-swab epigenetic clock and membership model .

15. Duke University OTC (2020). “TruDiagnostic Signs License for DunedinPoAm Aging Algorithm.” (July 2020) – Press release highlighting TruDiagnostic’s lab and its exclusive license to the Dunedin pace of aging clock .

16. Hone Health (2024). “How to Find a Medicine 3.0 Doctor Like Peter Attia.” (Feb 5, 2024) – Article explaining Attia’s Medicine 3.0 principles (preventative focus, using advanced testing) and emphasizing biological age as more meaningful than chronological age .

17. Segterra Inc. InsideTracker – Company Info. (2023). (InsideTracker’s background: founded 2009 by aging and genetics scientists; platform to track “InnerAge” and other wellness metrics) .

18. Centro Nacional de Investigaciones Oncológicas (CNIO) – Spin-offs. (2010). – Profile of Life Length: founded 2010 by María Blasco’s team to commercialize telomere length analysis as a measure of biological age .

19. Kim, S. et al. (2023). “An accurate aging clock developed from large-scale gut microbiome and blood transcriptome data.” Cell Reports Medicine, 4(8): 101965. (Describes Viome’s development of dual microbiome and transcriptome-based aging metrics; found ~0.7 correlation and associations with lifestyle) .

20. Lopez-Otín, C. et al. (2013). “The Hallmarks of Aging.” Cell, 153(6): 1194-1217. (Seminal paper outlining key mechanisms of aging; provides conceptual framework that inspires biomarker selection for BA models) .