The Klemera-Doubal method: biological age from blood biomarkers
Prepared by: Research Desk
Prepared by: Research Desk. Sources are linked for verification.
Long before saliva-kit epigenetic clocks, researchers were computing biological age from ordinary lab results. One of the most widely used approaches is the Klemera-Doubal method (KDM), published in 2006 [1]. It still appears in large cohort studies today, often next to PhenoAge, and it is the engine behind many "blood-based biological age calculators." This explainer covers how the Klemera-Doubal maths works, what data it uses, how cohorts have applied it, and where it breaks down.
Where the method came from
Petr Klemera and Bohumil Doubal started from a complaint. Biological age (BA) had no exact definition, so setting the results of various published BA methods side by side "makes little sense," and proving any of them optimal was impossible [1]. Their answer was to state simple assumptions about how BA relates to chronological age (CA), then derive the estimator mathematically instead of choosing one ad hoc.
According to their abstract, the result is a method that:
is presented as an optimum estimator of BA under its stated assumptions, and can be applied "even in nonlinear cases" [1];
lets researchers evaluate the precision of each BA estimate rather than reporting a bare number [1];
includes tools to check whether its own assumptions hold in a dataset [1];
uses chronological age as a standard biomarker inside the model. The authors report that this improves the precision of the BA estimate and clarifies what they call the relativity of the known "paradox of biomarkers" [1].
The theory was checked with a dedicated simulation program. The authors also argued that many earlier published BA results were affected by the problems their framework exposed [1].
How a KDM biological age is built
In practice, a KDM age is built in three stages.
Choose a reference population. Each candidate biomarker is modelled against chronological age in a training dataset. For example, Hu and colleagues developed their KDM biological age in the China Health and Nutrition Survey (CHNS) [5].
Combine the biomarkers. Each marker's relationship with age is used to infer the age that most plausibly explains a person's full set of values. Markers that track age more tightly carry more weight. Chronological age enters as one extra "biomarker," as Klemera and Doubal proposed [1].
Derive acceleration. Because KDM age correlates with chronological age, studies usually analyse the gap between them. Chen and colleagues defined KDMAge acceleration as the residual from regressing KDMAge on chronological age. Values above 0 counted as accelerated ageing; values at or below 0 did not [4]. The epigenetic age acceleration explainer covers the same residual logic for methylation clocks.
The inputs are routine measurements, not DNA. The BioAge R toolkit describes its scope as biological age "from blood chemistry and organ function test data" [2]. In a large NHANES analysis, twelve blood chemistry parameters were used to compute four biological age indicators, including KDM [3].
How KDM differs from epigenetic clocks
KDM and epigenetic clocks answer a similar question with different raw material:
Klemera-Doubal (KDM) | Epigenetic clocks | |
|---|---|---|
Input | Blood chemistry and organ-function tests [2] | DNA methylation at CpG sites [10] |
Anchoring | Chronological age as a standard biomarker [1] | Trained on age (Horvath) or on phenotypes linked to mortality [10][8] |
Software | BioAge R package [2] | Many public methylation packages [2] |
Kwon and colleagues note that blood-chemistry measures were introduced at about the same time as methylation measures. In multiple studies, they performed better at predicting healthy lifespan, yet they spread more slowly because fewer software tools existed [2]. The BioAge package was built to close that gap. It implements three published methods for quantifying biological aging based on chronological age and mortality risk [2]. For the methylation side, the epigenetic clock guide explains how clocks such as Horvath's multi-tissue estimator [10] are constructed. The PhenoAge explainer covers the clinical-biomarker score that later became DNAm PhenoAge [8].
How cohort studies have used KDM
KDM's main role today is as an exposure or outcome in observational research. Several recent human studies show the range:
Diet patterns (NHANES 1999-2018, 16,666 participants). Researchers computed four biological age measures: homeostatic dysregulation, allostatic load, KDM and phenotypic age. They reported that all 4 dietary metrics studied were significantly associated with biological age acceleration or deceleration [3]. This was an association analysis, not an intervention trial.
Rheumatoid arthritis (NHANES and UK Biobank). KDMAge and PhenoAge acceleration were used as surrogate measures of biological ageing. The authors examined their associations with disease risk, genetic susceptibility, and life expectancy in people with the condition [4].
Obesity and cardiovascular disease (multi-cohort). A KDM biological age was developed in CHNS. Validation analyses were run in UK Biobank and the Hongguang Elderly Health Examination Cohort. The main analysis used CHARLS participants aged over 45 years with 3 waves of data, testing whether biological age acceleration mediated links between obesity indices and incident cardiovascular disease [5].
Periodontitis (NHANES 2009-2014, 9803 participants). KDM, PhenoAge and homeostatic dysregulation were analysed cross-sectionally alongside Mendelian randomization [6].
Plasma proteomics (UK Biobank, 48,728 participants). Researchers analysed 2920 plasma proteins against several aging phenotypes, including KDM-BA acceleration, PhenoAge acceleration, frailty index, leukocyte telomere length and healthspan. They used Mendelian randomization with replication in FinnGen [7]. Related approaches appear in our page on proteomic aging clocks.
The common thread is that KDM serves as a standardised summary of many lab values. It is not a diagnostic test, and the associations above do not show that changing a KDM score changes outcomes.
Limits worth knowing
Reference population dependence. A KDM age only means something relative to the dataset used to fit it. Hu and colleagues trained in one Chinese cohort and validated in UK and Chinese cohorts precisely for this reason [5]. A calculator fitted to NHANES may not transfer cleanly to other groups.
Biomarker choice varies. Studies differ in which and how many markers they include. The NHANES diet study used twelve [3]. Two "KDM ages" from different papers are therefore not strictly the same measure.
Assumptions can fail. Klemera and Doubal built in tools to test their presumptions [1]. Applied papers do not always report those checks.
Lab variation. Single blood results fluctuate. The accuracy and repeat-results guide explains why one reading can mislead.
Regulation covers labs, not scores. Blood panels are typically run in laboratories certified under CLIA, the federal program administered by CMS for clinical laboratory testing [9]. That certification concerns the lab's testing, not whether a downstream age score has been validated. The regulatory explainer covers how aging tests fit into U.S. rules.
Further reading on the statistics
Reading KDM papers critically depends mostly on regression. That means understanding residuals, measurement error, and how a model fitted in one population behaves in another. A good biostatistics text is the most efficient way to build that skill. When weighing books, look for worked examples on cohort or survey data, coverage of survival analysis and Mendelian randomization basics, and reproducible code in R, since the BioAge toolkit is an R package [2]. Also check whether the edition is recent, what the price is across print and e-book formats, and whether example datasets are freely downloadable.
An introductory text that pairs theory with runnable code usually serves quantified-self readers better than a purely mathematical reference. A second, more advanced volume on longitudinal data adds depth for anyone who wants to replicate cohort analyses.
Frequently asked questions
Is the Klemera-Doubal method an epigenetic clock?
No. KDM uses blood chemistry and organ-function test results, not DNA methylation [2]. It predates most epigenetic clocks in its 2006 form [1] and is often reported alongside them.
What does a KDM age acceleration above zero mean?
In studies such as Chen and colleagues', a value above 0 means the person's KDM age was higher than predicted for their chronological age [4]. It is a research classification used to study associations, not a diagnosis.
Why do KDM and PhenoAge give different numbers?
They are built differently. KDM anchors on chronological age as a standard biomarker [1], while PhenoAge-type measures draw on mortality-linked phenotypes [8]. Their biomarker sets and reference populations also differ, so divergence is expected [3][6].
Can any blood panel produce a KDM age?
Only if it includes the markers a given KDM model was fitted on and those values can be mapped to the same reference population. Results from different calculators are not interchangeable [3][5].
Related on this site
Biological age tests: how epigenetic clocks work and what they mean
Are biological age tests FDA approved? How aging tests are regulated
PhenoAge explained: the clinical-biomarker and epigenetic versions
Epigenetic age acceleration: how it is calculated and its limits
Last updated: 26 October 2026.
Not medical advice. This page reports published research and regulatory records for information only.
Research-only notice: Content on this site is provided for research and educational purposes only and is not intended for human use. It is not medical advice.
Translation notice: Some content on this site may have been machine-translated using Google Translate and may contain errors.