PhenoAge vs KDM biological age: why calculators disagree

You enter the same age and the same laboratory results into two biological age calculators. PhenoAge says 49. KDM says 43. Your chronological age is 46.

Which number is correct?

The most useful answer is: the calculators are answering different statistical questions. PhenoAge and the Klemera–Doubal Method (KDM) can use overlapping blood biomarkers, but they were not designed to produce identical results. A gap between them is therefore not automatically an error.

This guide explains why PhenoAge and KDM disagree, what each result actually means, and how to choose one for consistent tracking. If you first need the formulas, biomarker lists, and basic interpretation, start with our complete PhenoAge and KDM blood-test guide.

Quick answer: PhenoAge is calibrated to mortality risk and expresses that risk as an age equivalent. KDM is primarily calibrated to the way biomarkers vary with chronological age in a reference population. Different targets, biomarker panels, coefficients, and reference samples can produce different biological ages from the same person.

Quick answer: PhenoAge and KDM answer different questions

PhenoAge estimates a mortality-risk-equivalent age. KDM estimates the age that best fits a multivariate biomarker profile. Because their targets and implementations differ, the same blood results can legitimately produce two different ages. The correct comparison is each score against the same model over time—not PhenoAge against KDM as if both were measuring one fixed quantity.

Key facts

• PhenoAge and KDM do not measure the same construct. PhenoAge is mortality-oriented; KDM estimates multivariate physiological age.
• “KDM” is a method, not one universal calculator. Its output depends on the biomarkers and reference population used to train it.
• PhenoAge is more standardized. The commonly used clinical version has a published set of nine biomarkers plus chronological age and fixed coefficients.
• A result should identify its exact implementation. A label such as “KDM age” is incomplete without the panel, units, training population, and formula.
• Do not average the two ages. Choose the model that fits your question and compare future results only with the same implementation.
• Neither score is a diagnosis. Use abnormal laboratory values and clinical risk with a qualified healthcare professional.

The one-table comparison

Feature	PhenoAge	KDM biological age
Main training target	Mortality risk	Chronological-age pattern across biomarkers
Typical inputs	9 blood biomarkers + chronological age	A configurable set of clinical or functional biomarkers; chronological age may also enter the estimator
Output meaning	The age in the reference population associated with a similar estimated mortality risk	The age that best fits the person’s multivariate biomarker profile under that fitted model
Standardization	Common published clinical formula	Many valid implementations
Main strength	A fixed, risk-oriented benchmark	Flexible integration of multiple physiological systems
Main limitation	Linear fixed formula can react strongly to particular markers and does not capture every nonlinear clinical relationship	Result depends heavily on panel choice and population calibration
Best tracking rule	Same formula, units, lab conditions, and biomarker definitions	Same trained model, panel, units, reference population, and test conditions

Both approaches are population models. Neither directly measures the number of years your tissues have existed, and neither proves that an intervention has changed the fundamental rate of aging.

What PhenoAge is designed to estimate

Clinical Phenotypic Age was developed from U.S. National Health and Nutrition Examination Survey data. Researchers screened dozens of routine clinical markers and used a penalized survival model to select chronological age plus nine biomarkers that predicted mortality.

Those inputs are:

1. albumin;
2. creatinine;
3. glucose;
4. C-reactive protein;
5. lymphocyte percentage;
6. mean corpuscular volume;
7. red cell distribution width;
8. alkaline phosphatase;
9. white blood cell count;
10. chronological age.

The model first estimates mortality risk. It then converts that risk into the age at which an average person in the reference population would have a comparable risk. This second conversion is why the output looks like a familiar age in years.

A PhenoAge of 52 does not mean every organ is biologically 52 years old. It means the model assigned a mortality-risk profile comparable to its reference value for that age. The related value, PhenoAge acceleration, asks whether the result is older or younger than expected after accounting for chronological age.

In the published validation, Phenotypic Age predicted morbidity and mortality across multiple demographic and health subgroups. That makes it useful as a compact risk-oriented summary. It does not turn the score into a clinical diagnosis or an individually guaranteed forecast.

What KDM biological age is designed to estimate

KDM takes a different route. It examines how each selected biomarker changes with age in a reference sample, how much unexplained variation remains, and how the markers relate to one another. It then finds the biological age that best reconciles the person’s full biomarker pattern.

In plain language, KDM asks:

At what age would this combination of physiological measurements be most plausible under this reference model?

Markers that track age more strongly and precisely can contribute more information. Markers with a weak age relationship or high residual noise generally contribute less. The method also accounts for the fact that a biological age estimate itself has uncertainty.

This is a powerful framework, but it creates an important consequence: there is no single universal KDM number.

One study may build KDM from albumin, creatinine, glucose, blood pressure, lung function, and grip strength. Another may use a different laboratory panel. One model may be sex-specific; another may pool sexes. One may be trained in NHANES; another in UK Biobank or a national cohort. All can be described as KDM, yet they need not return the same result.

Why the same blood work produces different ages

1. The models optimize different targets

PhenoAge selected and weighted variables for their ability to predict mortality. KDM traditionally weights variables according to their relationship with chronological age and their residual variation.

That distinction matters. A biomarker can be strongly related to mortality without changing smoothly with age. Another can track chronological age well but contribute less independent information about near-term risk. The algorithms reward different properties, so they assign different weights.

2. The biomarker panels may only partly overlap

The standard clinical PhenoAge formula is fixed at nine biomarkers plus age. KDM is configurable. Even when an online calculator asks for a familiar subset, it may omit, add, or transform variables compared with another KDM implementation.

A result based on 10 laboratory markers is not directly interchangeable with KDM built from 18 laboratory and organ-function measures. Panel composition changes what “older” and “younger” mean inside the model.

3. Chronological age enters differently

Chronological age is explicitly included in PhenoAge’s mortality model. In KDM, chronological age can act as an additional source of information in the estimation formula, depending on the implementation.

This tends to pull both outputs toward age-related population patterns, but not with the same force or logic. Two people with identical biomarker values but different chronological ages therefore need not receive the same acceleration result.

4. KDM depends on its reference population

KDM coefficients are estimated from a dataset. Change the age range, health profile, sex distribution, ethnicity, country, laboratory methods, or calendar period, and the biomarker-age relationships can change.

This is not merely a technical footnote. A model trained in relatively healthy adults aged 40–70 may behave differently at age 25 or 85. Applying any model far outside its training range is extrapolation, where apparently precise years can become misleading.

PhenoAge also reflects its U.S. NHANES development population, but its widely used clinical implementation usually preserves published coefficients. KDM calculators are more likely to embody different calibrations while sharing the same label.

5. Units and transformations can change the score

PhenoAge requires exact units. C-reactive protein is a classic source of mistakes because the published formula uses a transformed value and calculators may request mg/L while a report uses mg/dL. Creatinine, glucose, white blood cell count, and alkaline phosphatase also require careful unit matching.

KDM implementations may standardize, log-transform, winsorize, or otherwise preprocess inputs according to their training pipeline. If a calculator does not document these steps, you cannot reliably reproduce or compare its output.

Before interpreting a surprising result, run the unit and input audit in our guide to PhenoAge changes and laboratory noise.

6. A single unusual marker may matter more in one model

Imagine a person with excellent glucose, albumin, kidney markers, and blood counts but temporarily elevated C-reactive protein after an infection.

PhenoAge may move noticeably because its fixed mortality-oriented equation gives CRP a defined contribution. A KDM model with many additional measures may dilute that one signal, or it may react strongly if CRP has a reliable age gradient in its reference data. The answer depends on the fitted KDM version.

Now imagine consistently lower grip strength and lung function with otherwise favorable blood chemistry. A function-rich KDM model can register that pattern, while standard clinical PhenoAge cannot because those measures are not inputs.

Neither output must be “wrong.” Each sees only the variables it was built to see.

Three disagreement patterns and what they suggest

Pattern	Plausible explanation	Best next check
PhenoAge older, KDM near chronological age	One or more mortality-weighted PhenoAge markers are unfavorable, but the broader KDM pattern remains age-typical	Inspect individual PhenoAge contributors, units, inflammation, and recent illness
KDM older, PhenoAge near chronological age	The KDM panel includes age-sensitive functional or clinical measures absent from PhenoAge	Identify the exact KDM panel and the markers driving its result
Both older or both younger	Multiple physiological signals may align across models	Confirm standardized conditions and persistence on a repeat measurement

These are interpretation hypotheses, not diagnoses. A calculator’s aggregate age should never override the clinical interpretation of glucose, kidney function, inflammation, blood counts, blood pressure, or any other input.

For a wider framework covering blood, methylation, and wearable scores, see why biological age tests disagree.

Which calculator should you use?

Choose based on the decision you want to support.

Use PhenoAge when you want a reproducible mortality-oriented benchmark

PhenoAge is the cleaner choice when you have the required laboratory values and want to compare your result with a widely published fixed formula. It is relatively easy to reproduce, audit, and calculate again.

It is especially useful when:

• all nine required markers are available in the correct units;
• you want continuity with the published clinical Phenotypic Age literature;
• you can repeat the same panel under similar conditions;
• you understand that the age is a risk-equivalent summary, not a literal tissue age.

Use KDM when you want a broader physiological-age model

KDM can be more suitable when a well-documented implementation combines multiple relevant systems and is calibrated for a population similar to yours.

It is especially useful when:

• the exact biomarker panel and coefficients are disclosed;
• the reference population and valid age range are documented;
• the same trained model will remain available for future measurements;
• you care about multivariate physiological deviation rather than one fixed mortality-oriented score.

Do not choose by whichever result is younger

Calculator shopping creates a false sense of precision. If you test enough formulas, one will eventually produce a flattering number. That does not make it more valid for your question.

Instead, document the model before looking at the outcome. Then keep the method fixed.

A five-point calculator audit

Before trusting a PhenoAge or KDM result, verify:

1. Identity: Does the calculator name the exact method and source paper?
2. Inputs: Does it list every required biomarker and accepted unit?
3. Implementation: For KDM, does it disclose the panel, coefficients or software, training population, and valid age range?
4. Reproducibility: Can you enter the same values twice and export or record them?
5. Interpretation: Does it describe uncertainty and avoid diagnostic or anti-aging treatment claims?

If a tool fails the identity or implementation checks, treat its result as an opaque app score rather than a research-grade PhenoAge or KDM estimate.

How to compare results over time

The safest longitudinal question is not “What is my true biological age?” It is:

Has the same model moved consistently under comparable measurement conditions?

Use this protocol:

• keep the calculator version and formula unchanged;
• use the same laboratory where practical;
• match fasting status and time of day;
• avoid testing during acute illness or immediately after unusually hard exercise;
• record medication and supplement changes;
• save the raw laboratory values, not only the final age;
• repeat an unexpected result before changing your plan;
• inspect which inputs moved and whether they changed in a clinically meaningful direction.

Small differences should be interpreted in light of measurement error and model uncertainty. A calculator displaying one decimal place is not necessarily accurate to one tenth of a year. Our guide to biological age confidence intervals explains why precision on the screen can exceed certainty in the model.

Do not average PhenoAge and KDM. An average of two differently defined constructs has no established clinical meaning and makes future changes harder to diagnose.

How SuperAge handles PhenoAge and KDM

SuperAge is designed to keep the final score connected to its evidence. Rather than presenting a biological-age number as a verdict, it helps organize the underlying blood markers, trends, and calculation context on your iPhone.

That distinction is crucial when PhenoAge and KDM diverge. The useful information is often not the gap between “49” and “43,” but:

• which biomarkers drove each estimate;
• whether the calculation used the correct units;
• whether the same method was used last time;
• whether the change is distributed across several systems;
• whether an acute event could explain the movement.

For a broader method decision, compare epigenetic clocks with blood-biomarker age. Whichever method you use, keep medical decisions anchored to validated clinical measures and professional care.

Frequently asked questions

Why is my PhenoAge different from my KDM biological age?

They use different mathematical targets and may use different biomarkers. PhenoAge converts estimated mortality risk into an age equivalent. KDM estimates the age that best fits a multivariate biomarker pattern in a particular reference population.

Is PhenoAge more accurate than KDM?

There is no universal winner for every purpose. Performance depends on the outcome, cohort, implementation, and biomarker availability. Comparative research shows both can predict important health outcomes. Choose the model whose target and validation fit your question.

Do all KDM calculators give the same result?

No. KDM is a modeling method that can be fitted with different biomarkers and populations. Two KDM calculators may both be legitimate yet produce different ages. A trustworthy tool should document its model.

Can I compare PhenoAge with KDM acceleration?

You can examine whether both point broadly older or younger than expected, but their year values are not directly interchangeable. For longitudinal tracking, compare PhenoAge with the same PhenoAge formula and KDM with the same KDM implementation.

Should I average PhenoAge and KDM?

No. Averaging differently trained scores does not create a validated “truer” biological age. It can also hide the biomarker or model responsible for a disagreement.

Which result should I discuss with my doctor?

Discuss the underlying laboratory or functional values, especially anything outside a clinical reference range or changing persistently. The aggregate biological-age result can provide context, but it is not a diagnosis or treatment target.

The bottom line

PhenoAge and KDM disagree because they define biological age differently.

PhenoAge asks what chronological age corresponds to a similar modeled mortality risk. KDM asks what age best explains a multivariate physiological profile in its reference population. Once you know that, the disagreement stops looking like a broken calculator and starts revealing what each model emphasizes.

Pick one implementation for a defined purpose. Preserve the raw inputs. Standardize repeat testing. Investigate the biomarkers behind a change. And never let a single calculated age outweigh sound clinical interpretation.

Track the evidence behind the age

Want to keep your blood markers, biological-age context, and longitudinal trends together? Download SuperAge and preserve the inputs behind each result—not just the headline number.

References

1. Levine ME, Lu AT, Quach A, et al. An epigenetic biomarker of aging for lifespan and healthspan. Aging. 2018;10(4):573–591.
2. Liu Z, Kuo PL, Horvath S, Crimmins E, Ferrucci L, Levine M. A new aging measure captures morbidity and mortality risk across diverse subpopulations from NHANES IV. PLOS Medicine. 2018;15(12):e1002718.
3. Klemera P, Doubal S. A new approach to the concept and computation of biological age. Mechanisms of Ageing and Development. 2006;127(3):240–248.
4. Levine ME. Modeling the rate of senescence: can estimated biological age predict mortality more accurately than chronological age?. The Journals of Gerontology: Series A. 2013;68(6):667–674.
5. Kwon D, Belsky DW. A toolkit for quantification of biological age from blood chemistry and organ function test data: BioAge. GeroScience. 2021;43(6):2795–2808.
6. Parker DC, Bartlett BN, Cohen HJ, et al. Association of blood chemistry quantifications of biological aging with disability and mortality in older adults. The Journals of Gerontology: Series A. 2020;75(9):1671–1679.
7. Liu Z, Kuo PL, Horvath S, et al. No winners or losers: clinical chemistry-based biological aging metrics perform similarly across cohorts and health outcomes. The Journals of Gerontology: Series A. 2025;80(11):glaf179.

This article is for educational purposes only and does not provide medical advice, diagnosis, or treatment.