Phenotypic Age Explained

Phenotypic Age, often shortened to PhenoAge, is a mortality-risk estimate expressed in years. The original clinical version combines chronological age with nine routine blood measurements. It does not measure the literal age of your organs, diagnose disease or prove how long you will live. A result above your calendar age means that, in the model’s reference population, your combination of age and laboratory values resembled the risk pattern seen at an older age. Levine and colleagues developed the measure in about 10,000 US adults. Humanic Blue is tracing what that model-based age means without turning it into a personal prognosis.

That makes Phenotypic Age potentially useful for research and for noticing which ordinary lab values deserve attention. It does not make the composite score a treatment target. If the goal is living longer and better, the underlying measurements—and the reasons they changed—matter more than winning an argument with a calculator.

What is Phenotypic Age?

Clinical Phenotypic Age is a statistical model. Its developers began with 42 clinical measurements in the U.S. National Health and Nutrition Examination Survey III and selected nine biomarkers, along with chronological age, that jointly predicted mortality. The development sample included about 10,000 adults, and the model was calibrated against long-term mortality follow-up. The researchers then converted estimated mortality risk into an age-like number. The development methods and coefficients are reported in the original 2018 paper.

The nine measurements are:

• albumin
• creatinine
• glucose
• C-reactive protein, or CRP
• lymphocyte percentage
• mean cell volume, or MCV
• red-cell distribution width, or RDW
• alkaline phosphatase
• white-blood-cell count

Each measurement already has its own clinical context. For example, creatinine is commonly used when assessing kidney function, but it is also affected by muscle mass, diet, age and activity. CRP reflects inflammation and can rise with an infection or injury. A complete blood count supplies several of the blood-cell measures used in the formula. MedlinePlus explains the clinical context and limitations of creatinine, CRP and the complete blood count.

The output is therefore an age-shaped summary of a risk equation, not a direct reading of biological wear. If a 50-year-old receives a Phenotypic Age of 55, the defensible interpretation is that this combination of age and lab values mapped to the model’s mortality-risk level for age 55. It does not mean every cell or organ has become five years older.

Clinical PhenoAge and DNAm PhenoAge are not the same test

Two related measures share almost the same name:

1. Clinical Phenotypic Age uses nine blood measurements plus chronological age.
2. DNAm PhenoAge reads chemical marks on DNA, called methylation, at 513 locations to estimate the clinical Phenotypic Age pattern.

The DNA-methylation model was trained in about 450 participants from the InCHIANTI study. In the developers’ comparisons, the clinical measure predicted illness and death more strongly than DNAm PhenoAge. Those training details and comparative results appear in Levine et al.

That distinction matters when reading headlines about an intervention changing “PhenoAge.” A standard blood-panel calculation and a methylation test are related, but they are not interchangeable. A study of one does not automatically validate a claim about the other.

What could get lost when two different PhenoAge measures share one name?

What does Phenotypic Age acceleration mean?

Researchers often analyze Phenotypic Age acceleration, which asks whether a person’s score is higher or lower than expected for their chronological age. A positive value indicates an older-than-expected risk pattern; a negative value indicates a younger-than-expected pattern. For DNAm PhenoAge, acceleration is likewise calculated relative to chronological age rather than against a universal biological standard. The original clinical validation defines the clinical measure, while the methylation paper describes DNAm PhenoAge acceleration.

A “+5” result is therefore a model comparison, not a diagnosis and not a prediction that five years have been removed from your remaining lifespan. The number depends on the model, its reference data and the measurements entered.

How strong is the human evidence?

The evidence comes mainly from studies that followed groups of people and looked for links between the score and later health outcomes.

In a 2018 analysis of NHANES data, researchers studied about 11,400 adults ages 20 to 84, plus a smaller group of older adults. During a median 12.6 years of follow-up, about 1,000 participants died. After accounting for chronological age, each additional year of Phenotypic Age was associated with a 9% higher rate of all-cause mortality during follow-up; the statistical range around that estimate was 8% to 10%. The association was also present in a subgroup of about 1,900 participants classified as disease-free. Liu and colleagues reported the population, follow-up and results in PLOS Medicine.

This supports Phenotypic Age as a population risk marker. It does not show that the model will forecast an individual person’s outcome with the same precision, and it does not establish that the score itself causes disease.

Newer comparisons also argue against treating any biological-age measure as the single correct clock. A 2025 analysis evaluated 14 biological-aging measures in about 19,000 people in Scotland and examined associations with 174 diseases over roughly ten years. Different clocks captured different aspects of risk; the investigators did not identify one universal best performer across diseases. Mavrommatis and colleagues reported the comparison in Nature Communications.

Have randomized trials shown that PhenoAge can be slowed?

Not in a way that establishes better health or longer life.

The CALERIE trial assigned healthy adults without obesity, ages 21 to 50, at random to a calorie-restriction intervention or an unrestricted control group for two years. In a later methylation analysis, calorie restriction was associated with a small change in DunedinPACE, another aging measure, but not with a significant change in DNAm PhenoAge or GrimAge. The study measured biomarkers, not reductions in disease or mortality. The trial analysis reports the sample, duration and clock-specific findings.

A 2024 fasting-mimicking-diet analysis is sometimes described online as evidence that PhenoAge fell. It used a different seven-marker biological-age composite, not the nine-marker clinical PhenoAge formula or DNAm PhenoAge. The paper combined data from two trials and reported biological-age estimates for about 90 participants; the intervention food was supplied by L-Nutra, and the paper disclosed that USC had licensed relevant intellectual property to the company. The methods, sample and competing-interest statement are in the primary paper.

The careful conclusion is not that diet cannot affect health. It is that these trials do not prove that lowering clinical Phenotypic Age improves health outcomes or helps people live longer.

Which health outcome would a lower PhenoAge score still need to be linked to?

Why can one result change?

The formula incorporates measurements that can move for reasons unrelated to a durable change in aging.

• Inflammation: CRP rises and falls with inflammation, including responses to infection or injury. MedlinePlus describes what CRP measures and why levels change.
• Creatinine: muscle mass, diet, age and activity can influence the level, and eating meat can temporarily raise it. MedlinePlus details these influences.
• Glucose and testing conditions: food intake and collection conditions can affect results; a comprehensive metabolic panel may require fasting. MedlinePlus advises that fasting may be required for a CMP.
• Technical variation: methylation clocks can give different results when the same sample is measured again. In a study of six clocks, some original implementations differed by as much as nine years between repeat measurements; redesigned versions using a noise-reducing statistical method called principal components repeated more consistently. Higgins-Chen and colleagues evaluated technical reliability.
• Tissue and population: methylation results can differ by tissue, and clock performance may not transfer evenly across populations. A study comparing tissues from about 80 people found substantial differences between mouth and blood samples for some clocks. A 2026 eLife analysis found that some clocks performed inconsistently in populations of people with mixed genetic ancestry. Apsley et al. studied cross-tissue differences; Cruz-Gonzalez et al. examined population generalizability.

One surprising result can be a reason to inspect the inputs. It is not, by itself, proof of accelerated aging.