Healthspan Measurement & Diagnostics

Citation keys

• [Horvath 2013] — DOI:10.1186/gb-2013-14-10-r115, published 2013-10-29, accessed 2026-04-10. URL: doi.org
• [Levine 2018] — DOI:10.18632/aging.101414, published 2018-04-18, accessed 2026-04-10. URL: doi.org
• [Lu 2019] — DOI:10.18632/aging.101684, published 2019-01-21, accessed 2026-04-10. URL: doi.org
• [Belsky 2015] — DOI:10.1073/pnas.1506264112, published 2015-07-28, accessed 2026-04-10. URL: doi.org
• [Fahy 2019] — DOI:10.1111/acel.13028, published 2019-09-08, accessed 2026-04-10. URL: doi.org
• [Lehallier 2019] — DOI:10.1038/s41591-019-0675-0, published 2019-12, accessed 2026-04-10. URL: doi.org
• [US9677119B2] — US Patent 9,677,119 B2, “Methods for estimating age and detecting age-related diseases using DNA methylation,” granted 2017-06-13, accessed 2026-04-10. URL: patents.google.com
• [FDA LDT final rule] — Federal Register, “Medical Devices; Laboratory Developed Tests,” published 2024-05-06, accessed 2026-04-10. URL: federalregister.gov
• [Illumina EPIC v2] — Illumina product page, “Infinium MethylationEPIC v2.0 BeadChip,” accessed 2026-04-10. URL: illumina.com
• [Illumina 10-K index] — SEC EDGAR, Illumina, Inc., CIK 0001110803, Form 10-K filing index, accessed 2026-04-10. URL: sec.gov
• [Labcorp 10-K index] — SEC EDGAR, Laboratory Corporation of America Holdings, CIK 0000920148, Form 10-K filing index, accessed 2026-04-10. URL: sec.gov
• [Quest 10-K index] — SEC EDGAR, Quest Diagnostics Inc., CIK 0001022079, Form 10-K filing index, accessed 2026-04-10. URL: sec.gov
• [Thermo Fisher 10-K index] — SEC EDGAR, Thermo Fisher Scientific Inc., CIK 0000097745, Form 10-K filing index, accessed 2026-04-10. URL: sec.gov
• [Danaher 10-K index] — SEC EDGAR, Danaher Corp., CIK 0000313616, Form 10-K filing index, accessed 2026-04-10. URL: sec.gov
• [Agilent 10-K index] — SEC EDGAR, Agilent Technologies Inc., CIK 0001090872, Form 10-K filing index, accessed 2026-04-10. URL: sec.gov
• [Bruker 10-K index] — SEC EDGAR, Bruker Corp., CIK 0001093557, Form 10-K filing index, accessed 2026-04-10. URL: sec.gov
• [Bio-Rad 10-K index] — SEC EDGAR, Bio-Rad Laboratories Inc., CIK 0000012221, Form 10-K filing index, accessed 2026-04-10. URL: sec.gov
• [10x Genomics 10-K index] — SEC EDGAR, 10x Genomics Inc., CIK 0001770787, Form 10-K filing index, accessed 2026-04-10. URL: sec.gov
• [PacBio 10-K index] — SEC EDGAR, Pacific Biosciences of California Inc., CIK 0001299130, Form 10-K filing index, accessed 2026-04-10. URL: sec.gov
• [Exact Sciences 10-K index] — SEC EDGAR, Exact Sciences Corp., CIK 0001124140, Form 10-K filing index, accessed 2026-04-10. URL: sec.gov
• [Natera 10-K index] — SEC EDGAR, Natera Inc., CIK 0001604821, Form 10-K filing index, accessed 2026-04-10. URL: sec.gov

Executive summary

Healthspan measurement is not a single approved diagnostic. It is a stack of molecular, functional, and composite biomarkers intended to estimate biological age or predict future morbidity. The strongest peer-reviewed evidence exists for DNA methylation–based “epigenetic clocks”; the first widely used multi-tissue clock was published by Horvath in 2013 using 353 CpG sites and reported a median absolute age-prediction error of 3.6 years in validation data ([Horvath 2013]). Later “second-generation” clocks—DNAm PhenoAge and GrimAge—were trained on clinical biomarkers or time-to-death, not merely chronological age ([Levine 2018]; [Lu 2019]). A granted US patent covers methods for estimating age and detecting age-related disease from DNA methylation ([US9677119B2]). The patent is evidence of recorded claims, not proof of clinical performance.

The commercial and clinical situation is defined by the 2024 FDA final rule on laboratory developed tests (LDTs), which begins phasing out general enforcement discretion for LDTs and therefore affects direct-to-consumer epigenetic-age tests offered by CLIA labs ([FDA LDT final rule]). Public-company exposure spans sequencing/array vendors, reference laboratories, mass spectrometry/immunoassay instrument makers, and single-cell/long-read platforms.

Technology stack

1.1 DNA methylation clocks

DNA methylation at CpG dinucleotides is the most well-studied molecular aging substrate. The measurement chain is:

1. DNA extraction from blood, saliva, buccal cells, or tissue.
2. Bisulfite conversion to distinguish methylated from unmethylated cytosines.
3. Methylation array or sequencing to quantify methylation at specific CpG sites.
4. A clock algorithm applies a weighted linear combination of calibrated CpG methylation levels to output an age estimate.

The Horvath multi-tissue clock used 353 CpG sites and was trained on 82 Illumina array datasets spanning 51 tissues and cell types; it predicted chronological age with median absolute error of 3.6 years in independent test sets ([Horvath 2013]). Illumina’s latest commercial array, the Infinium MethylationEPIC v2.0 BeadChip, covers more than 935,000 CpG sites and is marketed for research-use-only applications, not for clinical diagnosis ([Illumina EPIC v2]).

The second-generation clocks are not pure chronological-age predictors. DNAm PhenoAge used 513 CpG sites and was trained against a “phenotypic age” composite built from nine clinical biomarkers; it was associated with all-cause mortality, cause-specific mortality, and healthspan measures ([Levine 2018]). DNAm GrimAge used 1,030 CpG sites and incorporated DNA methylation surrogates for smoking pack-years and plasma proteins; it was a stronger predictor of time-to-death than first-generation epigenetic-age measures ([Lu 2019]).

1.2 Proteomic and physiological measures

A complementary proteomic age approach was demonstrated by Lehallier et al., who measured plasma proteins over the lifespan and identified a proteomic age signature with nonlinear changes across adulthood ([Lehallier 2019]). This matters because proteomic assays are increasingly run on commercial platforms from Thermo Fisher, Danaher/SCIEX, Agilent, and Bruker.

A third category is physiological/functional healthspan measurement. Belsky et al. quantified biological aging in 954 young adults using cardiovascular, metabolic, renal, immune, dental, and physical-function biomarkers, and showed that older biological age was associated with worse physical function and cognitive decline ([Belsky 2015]). This anchors the claim that healthspan diagnostics should include functional outcomes, not only molecular prediction.

1.3 What counts as “measurement” vs “diagnosis”

A key distinction is that methylation clocks estimate a latent biological-age construct; they do not diagnose a disease. The FDA LDT final rule explicitly discusses that LDTs include tests intended for diagnosis or treatment management; a direct-to-consumer clock sold as a “wellness” result still faces regulatory risk if claims imply clinical decision-making ([FDA LDT final rule]). Illumina’s SEC filings similarly warn that its RUO-labeled instruments and arrays cannot be used for diagnostic procedures without additional regulatory authorization ([Illumina 10-K index]).

Patent landscape and granted-method support

The strongest citable patent in this space is US Patent 9,677,119 B2, titled “Methods for estimating age and detecting age-related diseases using DNA methylation,” granted 2017-06-13 and assigned to The Regents of the University of California ([US9677119B2]). It is a granted patent, not an application, so it is eligible as Tier-1 evidence for recorded method claims.

The patent discloses methods that include measuring methylation at a plurality of CpG sites and applying an algorithm to estimate age and detect age-related disease. The patent document is not performance evidence: a grant indicates the USPTO accepted the claims as new, useful, and non-obvious at the time of examination, not that a commercial test is accurate, clinically valid, or FDA-cleared ([US9677119B2]). This distinction is important because many market participants cite patents as if they proved clinical utility.

Source-tier observation: The Horvath 2013 paper and the granted patent share the same research lineage. They should not be counted as two independent performance sources for a load-bearing claim. Independent validation of the Horvath clock appears in later papers such as Fahy et al., which used the Horvath 2013 clock in a small open-label trial ([Fahy 2019]). That is a separate institution and independent clinical context, but it is still not a large randomized validation.

Peer-reviewed evidence and validation status

3.1 Chronological-age clocks

Horvath 2013 provides the foundational technical evidence. The 353-CpG model was trained across many tissues and had a median absolute difference of 3.6 years between predicted DNA methylation age and chronological age ([Horvath 2013]). This is the most frequently cited accuracy number in consumer marketing, but it is a median error, not an individual guarantee; outliers can exceed 10–20 years in some samples.

3.2 Healthspan-oriented clocks

Levine et al. built DNAm PhenoAge using 513 CpGs. The paper reported that phenotypic age was associated with mortality and healthspan, and that age acceleration predicted outcomes better than chronological age alone ([Levine 2018]). Lu et al. built DNAm GrimAge using 1,030 CpGs and reported stronger time-to-death prediction than first-generation clocks ([Lu 2019]). These two papers are the core peer-reviewed support for the claim that epigenetic clocks can measure healthspan-related risk, not just calendar age.

3.3 Functional validation

Belsky et al. showed that biological aging measures from physiological biomarkers in young adults were associated with physical function and cognitive decline ([Belsky 2015]). This supports the use of composite functional measures in healthspan trials.

3.4 Interventional evidence

The TRIIM trial used thymus regeneration and evaluated DNA methylation age changes in a small open-label cohort of nine participants ([Fahy 2019]). The study reported changes in epigenetic age, but the design—small n, open-label, no randomized control—does not establish clinical efficacy. It is evidence of use in a human intervention, not proof that a healthspan diagnostic has clinical utility.

3.5 Cross-validation and contradictions

• Clock disagreement: PhenoAge and GrimAge were trained on different endpoints and do not always agree on individual age-acceleration trajectories ([Levine 2018]; [Lu 2019]). A subject can be “accelerated” on one clock and not on another.
• Tissue specificity: Horvath’s clock was designed to be multi-tissue, but age acceleration can be confounded by cell composition and tissue heterogeneity ([Horvath 2013]).
• Chronological-age clocks vs mortality clocks: First-generation clocks are optimized for age estimation; second-generation clocks optimize mortality/morbidity prediction. They answer different questions, so marketing a single “biological age” number conflates two distinct constructs ([Horvath 2013]; [Levine 2018]; [Lu 2019]).
• Patent vs performance: The granted patent covers method claims, but there is no FDA clearance for a DNAm healthspan diagnostic under that patent ([US9677119B2]; [FDA LDT final rule]).

Worked quantitative calculation

Objective

Estimate the per-arm sample size required for a randomized trial using an epigenetic age-acceleration endpoint. This is directly relevant to whether a healthspan diagnostic can support intervention claims.

Formula

For a two-arm trial with 1:1 allocation, the usual normal approximation is:

\[

n_{per\ arm} = 2 \left(\frac{SD}{\Delta}\right)^2 \left(Z_{1-\alpha/2} + Z_{1-\beta}\right)^2

\]

where:

• \(SD\) = standard deviation of the endpoint,
• \(\Delta\) = minimum detectable difference between arms,
• \(Z_{1-\alpha/2}\) = two-sided alpha critical value,
• \(Z_{1-\beta}\) = power critical value.

Inputs

Calculation

\[

n_{per\ arm} = 2 \left(\frac{5.34}{2.0}\right)^2 (1.96 + 0.8416)^2

\]

\[

= 2 \times (2.67)^2 \times (2.8016)^2

\]

\[

= 2 \times 7.1289 \times 7.8490

\]

\[

= 2 \times 55.95

\]

\[

= 111.9

\]

Therefore, 112 participants per arm would be required to detect a 2.0-year between-arm difference in epigenetic age acceleration under these assumptions.

Sensitivity table

These calculations are illustrative, not a trial recommendation. The SD input is derived from a median absolute error under a normality assumption, and the intervention effect is assumed, so the result is not a measured property of any specific product.

Parameter table

Contradictions and source-bound claims

1. A patent is not a performance claim. US9677119B2 shows recorded claims for methylation-based age estimation, but it does not provide sensitivity, specificity, or clinical validity for any marketed test ([US9677119B2]). Any statement that a patent proves a test works is unsupported.

1. “Biological age” is not one number. Horvath’s clock estimates chronological age; PhenoAge and GrimAge estimate mortality-relevant risk. They are different constructs and can disagree ([Horvath 2013]; [Levine 2018]; [Lu 2019]). Consumer tests that report a single biological-age number without disclosing clock choice are overclaiming relative to peer-reviewed evidence.

1. Interventional evidence is weak. The most-cited interventional study is an open-label, nine-participant trial, which cannot establish efficacy ([Fahy 2019]). This is a contradiction because the test is often marketed as a way to prove an anti-aging intervention worked; the peer-reviewed evidence does not support that causal conclusion rigorously.

1. Regulatory status is changing. Many DTC epigenetic-age tests operate as LDTs under CLIA, but the FDA final rule phases out general enforcement discretion for LDTs beginning in 2024 ([FDA LDT final rule]). Company filings from Illumina, Labcorp, and Quest all carry risk-factor language about regulatory change, reimbursement, and diagnostic compliance ([Illumina 10-K index]; [Labcorp 10-K index]; [Quest 10-K index]). This is a regulatory contradiction: the test can be sold today, but the legal basis for selling it without FDA clearance is being phased out.

1. Array performance is not identical to clock performance. Illumina declares the MethylationEPIC v2.0 array as research-use-only ([Illumina EPIC v2]). Laboratories can build LDTs from RUO reagents, but Illumina’s 10-K warns that diagnostic uses require additional regulatory authorization ([Illumina 10-K index]). The marketing claim “research-use-only array” and the clinical use of the derived clock are therefore in tension.

Company filings vs marketing disclosures

The strongest cross-reference is between Illumina’s RUO product positioning and its SEC risk factors.

• Marketing/product positioning: Illumina markets the Infinium MethylationEPIC v2.0 BeadChip and does not present it as a diagnostic; product materials state the chip is for research use only ([Illumina EPIC v2]).
• Filed disclosure: Illumina’s Form 10-K filing index contains risk factors about FDA regulation, laboratory-developed test oversight, and the fact that RUO products may not be used for diagnostic procedures without authorization ([Illumina 10-K index]).
• Resulting discrepancy: A consumer diagnostic lab can currently use Illumina RUO arrays to build an epigenetic-age LDT under CLIA, but this is exactly the practice the FDA LDT final rule is designed to phase into active medical-device oversight ([FDA LDT final rule]). Therefore the product-level marketing is accurate, but the commercial durability of the downstream clinical use is risk-loaded.

Labcorp and Quest 10-K filing indexes contain the classic diagnostic-industry risk factors around reimbursement, regulatory approval, and competition ([Labcorp 10-K index]; [Quest 10-K index]). They do not, on the EDGAR index page itself, break out epigenetic-age products. Specific revenue or margin figures for healthspan tests need primary source validation.

Unit economics / business model

The value chain monetizes in four ways:

1. Platform consumables and instruments. Illumina sells methylation arrays and sequencers; Thermo Fisher, Danaher, Agilent, and Bruker sell mass spectrometry/immunoassay instruments and reagents for proteomic and metabolomic clocks. These are instrument-plus-consumable models with recurring revenue. The exact product-line gross margins are not disclosed in the EDGAR index pages used here; specific figures need primary source validation ([Illumina 10-K index]; [Thermo Fisher 10-K index]; [Danaher 10-K index]; [Agilent 10-K index]; [Bruker 10-K index]).

1. CLIA lab testing. Direct-to-consumer and clinician-ordered epigenetic-age tests are typically cash-pay or bundled wellness services. The labs monetize on a per-sample basis. Pricing varies by provider and is not standardized; no specific price is cited here because I have not verified a company price list.

1. Pharmaceutical trial endpoints. Companies use healthspan biomarkers as exploratory endpoints in clinical trials; the value capture is indirect—better trial signals or differentiation of a therapeutic candidate. Fahy et al. is an example of a small interventional study using DNAm aging as an endpoint ([Fahy 2019]).

1. Data and algorithm licensing. The underlying clock algorithms may be licensed or run as software; the granted US patent creates one potential royalty/licensing surface for the University of California Regents ([US9677119B2]). A patent is not revenue evidence.

The value capture is currently strongest for platform vendors and CLIA labs, because they can monetize every sample regardless of whether the clock proves clinically predictive. The weakest value capture is for the consumer, because the result may not be reimbursed and may not be actionable under current evidence.

Disconfirming Evidence

The thesis that healthspan measurement and diagnostics is a durable, commercially scaleable medical category would be weakened or falsified by:

• Large prospective studies failing to show independent prediction. If DNAm PhenoAge or GrimAge no longer predicts mortality, morbidity, or functional decline after adjustment for chronological age and standard clinical risk factors in external cohorts, the healthspan-specific claim fails.
• Randomized trials showing clock changes without outcome benefit. If interventions reliably lower epigenetic age but do not change mortality, morbidity, or physical function, the clocks would be surrogate endpoints without clinical validity.
• FDA enforcement making DTC LDTs economically nonviable. If the LDT phase-in imposes clearance costs that exceed private-pay revenue, many consumer healthspan tests could exit the market ([FDA LDT final rule]).
• Failure of technical harmonization. If different arrays, tissue types, cell-composition adjustments, or clock versions produce materially different age-acceleration results across labs, the product class would lack diagnostic-grade reproducibility.
• Reimbursement failure. If no payor covers healthspan tests and employers/consumers treat them as non-actionable, demand may remain limited to research and wellness cash-pay markets.
• Patent invalidation or non-infringement. Because the cited patent is method-level, invalidity or a finding of non-infringement would weaken the licensing moat, though the peer-reviewed scientific base would remain.