Critical ReviewBiology & Life Sciences
Epigenetic Clocks in the Clinic: Why Biological Age Has Not Yet Changed Medical Practice
Epigenetic clocks can predict mortality and disease risk in research cohorts, but clinical adoption faces barriers of cost, standardization, interpretability, and the absence of actionable thresholds β a gap between statistical association and medical utility that the field is only beginning to address.
By ORAA Research
This blog summarizes research trends based on published paper abstracts. Specific numbers or findings may contain inaccuracies. For scholarly rigor, always consult the original papers cited in each post.
Epigenetic clocks β algorithms that estimate biological age from DNA methylation patterns β have accumulated a strong epidemiological record. They predict all-cause mortality, cardiovascular events, cancer incidence, and cognitive decline with effect sizes that remain significant after adjusting for conventional risk factors. A physician reading these papers might reasonably ask: why are we not using biological age in clinical decisions today?
The answer is not that the science is wrong. It is that the gap between population-level prediction and individual clinical utility is wider than the literature sometimes suggests.
The Research Landscape: A Proliferation of Clocks
Since Horvath's original multi-tissue clock in 2013, the field has produced a succession of increasingly refined estimators. The current generation includes purpose-built clocks optimized for specific outcomes rather than chronological age itself.
GrimAge and GrimAge2 incorporate DNA methylation surrogates for plasma proteins (including PAI-1 and GDF-15) and smoking pack-years. Zhu et al. (2025) demonstrate in a retrospective cohort that GrimAge and GrimAge2 age acceleration predicts mortality risk with consistent hazard ratios across multiple validation datasets. GrimAge2 showed marginal improvements over the original, particularly in non-smoking populations.
DunedinPACE takes a different approach β rather than estimating a static biological age, it measures the pace of aging (years of biological aging per calendar year) using longitudinal data from the Dunedin birth cohort. This rate-of-change metric is conceptually attractive for monitoring interventions.
PhysAge, introduced by Arpawong et al. (2025), integrates DNA methylation with physiological system markers to create a "physiological health age" that captures multi-system decline. The authors demonstrate improved mortality prediction relative to earlier clocks, though validation remains limited to specific cohorts.
Kusters and Horvath (2024) provide the most comprehensive public health perspective, reviewing how epigenetic aging metrics might be used for early detection of chronic conditions and monitoring population health interventions. Their assessment is notably measured: the potential is clear, but "considerable work remains before routine clinical deployment."
Critical Analysis
<
| Claim | Evidence | Verdict |
|---|
| Epigenetic age acceleration predicts mortality independently of conventional risk factors | Multiple large cohort studies (MESA, HRS, Framingham) confirm independent associations | β
Supported β the epidemiological evidence is robust |
| GrimAge2 improves on GrimAge | Zhu et al. (2025) show modest improvement, especially in non-smokers | β οΈ Marginal β incremental rather than transformative improvement |
| DunedinPACE captures intervention effects | Some RCTs show pace changes with caloric restriction and lifestyle interventions | β οΈ Promising but limited β effect sizes are small and clinical significance uncertain |
| Epigenetic clocks are ready for clinical deployment | No regulatory-approved clinical test exists; no established clinical decision thresholds | β Not yet β multiple barriers remain |
| Newer clocks (PhysAge, EpiScore) outperform older ones | Limited head-to-head comparisons in diverse populations | β οΈ Unclear β cohort-specific validation does not guarantee generalizability |
The Standardization Problem
A fundamental barrier to clinical adoption is that different clocks give different answers for the same individual. Horvath's clock, Hannum's clock, PhenoAge, GrimAge, and DunedinPACE each use different CpG sites, different algorithms, and different training outcomes. They correlate with each other only moderately. For a clinician, the question "What is this patient's biological age?" does not have a single answer β it has five or more, depending on which algorithm is applied.
Yamada (2025) argues that this proliferation is not merely an inconvenience but a conceptual issue: each clock captures a different dimension of aging. GrimAge emphasizes inflammation and plasma protein proxies. DunedinPACE emphasizes rate of change in organ function. PhenoAge emphasizes metabolic and immune markers. They are measuring related but distinct biological processes, and collapsing them into a single "biological age" number obscures this complexity.
The Cost and Access Barrier
Clinical-grade DNA methylation profiling (Illumina EPIC array) costs $200β$400 per sample, requires specialized laboratory processing, and takes days to return results. This is not prohibitive for research studies but is impractical for population screening. Point-of-care testing does not exist. Until the assay becomes faster and cheaper β perhaps through targeted panels measuring only the CpG sites needed for a specific clock β routine clinical use is economically challenging.
The Actionability Gap
The most critical barrier may be the absence of established clinical decision thresholds. If a patient's GrimAge is 5 years accelerated, what does the physician do differently? Prescribe exercise? Adjust statin dosing? Refer for further testing? The epidemiological literature establishes that accelerated aging is associated with worse outcomes, but the intervention studies are sparse and the effect sizes of available interventions on epigenetic age are modest.
Srivatsa et al. (2025) address this partially in the cardiovascular domain, showing that epigenetic age acceleration adds predictive information beyond coronary artery calcium scores in the MESA cohort. But additive prediction does not automatically translate to improved clinical decision-making β that requires randomized evidence that acting on biological age information changes outcomes.
Open Questions
Which clock for which purpose? The field needs consensus on matching specific clocks to specific clinical contexts rather than treating them as interchangeable.Intervention responsiveness: Can lifestyle or pharmacological interventions reliably reduce epigenetic age acceleration, and does reduction translate to clinical benefit?Equity implications: DNA methylation patterns differ by race, socioeconomic status, and environmental exposures. Clocks trained on predominantly White populations may perform differently in other groups.Regulatory pathway: What evidence would a regulatory body require to approve an epigenetic clock as a clinical diagnostic? No precedent exists.Integration with multi-omics: Combining epigenetic clocks with proteomic, metabolomic, and imaging data may yield more clinically useful composite biomarkers than any single modality.Closing
Epigenetic clocks represent a genuine scientific advance in quantifying biological aging. The epidemiological evidence linking accelerated epigenetic aging to adverse health outcomes is consistent and well-replicated. But clinical translation requires more than prediction β it requires standardization, actionable thresholds, cost-effective assays, and evidence that using biological age information improves patient outcomes. The field is building toward these milestones, but the gap between research utility and clinical practice remains substantial.
λ©΄μ±
μ‘°ν: μ΄ κ²μλ¬Όμ μ 보 μ 곡μ λͺ©μ μΌλ‘ ν μ°κ΅¬ λν₯ κ°μμ΄λ€. νμ μ μλ¬Όμμ μΈμ©νκΈ° μ μ ꡬ체μ μΈ μ°κ΅¬ κ²°κ³Ό, ν΅κ³ λ° μ£Όμ₯μ μλ³Έ λ
Όλ¬Έκ³Ό λμ‘°νμ¬ κ²μ¦ν΄μΌ νλ€.
μμμμμ νμ±μ μ νμ μκ³: μλ¬Όνμ λμ΄κ° μμ§ μλ£ μ€λ¬΄λ₯Ό λ°κΎΈμ§ λͺ»ν μ΄μ
DNA λ©νΈν ν¨ν΄μΌλ‘λΆν° μλ¬Όνμ λμ΄λ₯Ό μΆμ νλ μκ³ λ¦¬μ¦μΈ νμ±μ μ νμ μκ³(epigenetic clock)λ μνμ κ·Όκ±°λ₯Ό ννν μΆμ ν΄ μλ€. μ΄ μκ³ λ¦¬μ¦μ κΈ°μ‘΄ μν μΈμλ€μ 보μ ν νμλ μ μλ―Έν ν¨κ³Ό ν¬κΈ°λ‘ μ μμΈ μ¬λ§λ₯ , μ¬νκ΄ μ¬κ±΄, μ λ°μλ₯ , μΈμ§ κΈ°λ₯ μ νλ₯Ό μμΈ‘νλ€. μ΄λ¬ν λ
Όλ¬Έλ€μ μ½μ μμμλΌλ©΄ ν©λ¦¬μ μΈ μλ¬Έμ κ°μ§ μ μλ€. μ μ°λ¦¬λ μ€λλ μμμ μμ¬ κ²°μ μ μλ¬Όνμ λμ΄λ₯Ό νμ©νμ§ μλκ°?
κ·Έ λ΅μ κ³Όνμ΄ νλ ΈκΈ° λλ¬Έμ΄ μλλ€. μ§λ¨ μμ€μ μμΈ‘κ³Ό κ°μΈμ μμμ μ μ©μ± μ¬μ΄μ κ°κ·Ήμ΄ λ¬Ένμμ μμ¬νλ κ²λ³΄λ€ ν¨μ¬ ν¬κΈ° λλ¬Έμ΄λ€.
μ°κ΅¬ νν©: μκ³μ μ¦μ
2013λ
Horvathμ μ΅μ΄ λ€μ‘°μ§(multi-tissue) μκ³ μ΄ν, μ΄ λΆμΌλ μ μ λ μ κ΅ν΄μ§ μΆμ λͺ¨λΈλ€μ μ°μ΄μ΄ μ μν΄ μλ€. νμ¬ μΈλμ λͺ¨λΈμλ μ°λκΈ°μ λμ΄ μμ²΄κ° μλ νΉμ κ²°κ³Όμ μ΅μ νλ λͺ©μ λ§μΆ€ν μκ³κ° ν¬ν¨λλ€.
GrimAgeμ GrimAge2λ νμ₯ λ¨λ°±μ§(PAI-1 λ° GDF-15 ν¬ν¨)κ³Ό ν‘μ° κ°λ
(pack-year)μ λν DNA λ©νΈν λ리 μ§νλ₯Ό ν΅ν©νλ€. Zhu λ±(2025)μ νν₯μ μ½νΈνΈ μ°κ΅¬μμ GrimAgeμ GrimAge2μ λμ΄ κ°μμ΄ λ³΅μμ κ²μ¦ λ°μ΄ν°μ
μμ μΌκ΄λ μνλΉ(hazard ratio)λ‘ μ¬λ§ μνμ μμΈ‘ν¨μ μ
μ¦νμλ€. GrimAge2λ νΉν λΉν‘μ° μ§λ¨μμ μλ λͺ¨λΈμ λΉν΄ λ―Έλ―Έν μμ€μ κ°μ μ 보μλ€.
DunedinPACEλ λ€λ₯Έ μ κ·Ό λ°©μμ μ·¨νλ€. μ μ μΈ μλ¬Όνμ λμ΄λ₯Ό μΆμ νλ λμ , Dunedin μΆμ μ½νΈνΈμ μ’
λ¨ λ°μ΄ν°λ₯Ό νμ©νμ¬ λ
Έν μλ(λ¬λ ₯ μ°λλΉ μλ¬Όνμ λ
Έν μ°μ)λ₯Ό μΈ‘μ νλ€. μ΄ λ³νμ¨ μ§νλ κ°λ
μ μΌλ‘ μ€μ¬ ν¨κ³Ό λͺ¨λν°λ§μ μ ν©νλ€.
Arpawong λ±(2025)μ΄ μκ°ν PhysAgeλ λ€μ€ μ²΄κ³ μ ν΄λ₯Ό ν¬μ°©νλ 'μ리μ κ±΄κ° λμ΄(physiological health age)'λ₯Ό ꡬμ±νκΈ° μν΄ DNA λ©νΈνμ μ리μ μ²΄κ³ μ§νλ₯Ό ν΅ν©νλ€. μ μλ€μ μ΄κΈ° μκ³λ€μ λΉν΄ ν₯μλ μ¬λ§λ₯ μμΈ‘ μ±λ₯μ μ
μ¦νμμΌλ, κ²μ¦μ νΉμ μ½νΈνΈμ κ΅νλμ΄ μλ€.
Kustersμ Horvath(2024)λ λ§μ± μ§νμ μ‘°κΈ° λ°κ²¬κ³Ό μ§λ¨ κ±΄κ° μ€μ¬ λͺ¨λν°λ§μ νμ±μ μ νμ λ
Έν μ§νλ₯Ό νμ©νλ λ°©μμ κ²ν νλ©° κ°μ₯ ν¬κ΄μ μΈ κ³΅μ€ λ³΄κ±΄ κ΄μ μ μ μνλ€. κ·Έλ€μ νκ°λ νμ ν μ μ λμ΄ μλ€. μ μ¬λ ₯μ λΆλͺ
νμ§λ§ "μΌμμ μΈ μμ λ°°μΉ μ΄μ μ μλΉν μμ
μ΄ λ¨μ μλ€."
λΉνμ λΆμ
<
| μ£Όμ₯ | κ·Όκ±° | νμ |
|---|
| νμ±μ μ νμ λμ΄ κ°μμ κΈ°μ‘΄ μν μΈμμ λ
립μ μΌλ‘ μ¬λ§λ₯ μ μμΈ‘νλ€ | 볡μμ λκ·λͺ¨ μ½νΈνΈ μ°κ΅¬(MESA, HRS, Framingham)κ° λ
립μ μ°κ΄μ±μ νμΈνλ€ | β
μ§μ§λ¨ β μνμ κ·Όκ±°κ° κ²¬κ³ νλ€ |
| GrimAge2λ GrimAgeλ₯Ό κ°μ νλ€ | Zhu λ±(2025)μ νΉν λΉν‘μ°μμμ λ―Έλ―Έν κ°μ μ 보μΈλ€ | β οΈ λ―Έλ―Έν¨ β λ³νμ κ°μ μ΄ μλ μ μ§μ κ°μ μ΄λ€ |
| DunedinPACEλ μ€μ¬ ν¨κ³Όλ₯Ό ν¬μ°©νλ€ | μΌλΆ RCTμμ μ΄λ μ ν λ° μν μ΅κ΄ μ€μ¬λ‘ μλ λ³νκ° κ΄μ°°λλ€ | β οΈ μ λ§νλ μ νμ β ν¨κ³Ό ν¬κΈ°κ° μκ³ μμμ μ μμ±μ΄ λΆνμ€νλ€ |
| νμ±μ μ νμ μκ³λ μμ λ°°μΉ μ€λΉκ° λμ΄ μλ€ | κ·μ μΉμΈμ λ°μ μμ κ²μ¬κ° μμΌλ©°, ν립λ μμμ μμ¬ κ²°μ μκ³κ°λ μλ€ | β μμ§ μλ β 볡μμ μ₯λ²½μ΄ λ¨μ μλ€ |
| μ ν μκ³(PhysAge, EpiScore)λ ꡬν μκ³λ₯Ό λ₯κ°νλ€ | λ€μν μ§λ¨μμμ μ§μ λΉκ΅ μ°κ΅¬κ° μ νμ μ΄λ€ | β οΈ λΆλͺ
ν β μ½νΈνΈ νΉμ΄μ κ²μ¦μ΄ μΌλ°ν κ°λ₯μ±μ 보μ₯νμ§ μλλ€ |
νμ€ν λ¬Έμ
μμ λμ
μ κ·Όλ³Έμ μΈ μ₯λ²½μ μλ‘ λ€λ₯Έ μκ³κ° λμΌν κ°μΈμ λν΄ μλ‘ λ€λ₯Έ λ΅μ μ μνλ€λ μ μ΄λ€. Horvath μκ³, Hannum μκ³, PhenoAge, GrimAge, DunedinPACEλ κ°κ° μλ‘ λ€λ₯Έ CpG λΆμ, μλ‘ λ€λ₯Έ μκ³ λ¦¬μ¦, μλ‘ λ€λ₯Έ νλ ¨ κ²°κ³Όλ₯Ό μ¬μ©νλ€. μ΄λ€μ μλ‘ μ€λ±λμ μκ΄κ΄κ³λ§μ 보μΈλ€. μμμ μ
μ₯μμ "μ΄ νμμ μλ¬Όνμ λμ΄λ μΌλ§μΈκ°?"λΌλ μ§λ¬Έμλ λ¨μΌν λ΅μ΄ μμΌλ©°, μ΄λ€ μκ³ λ¦¬μ¦μ μ μ©νλλμ λ°λΌ λ€μ― κ°μ§ μ΄μμ λ΅μ΄ μ‘΄μ¬νλ€.
Yamada(2025)λ μ΄λ¬ν μ¦μμ΄ λ¨μν λΆνΈν¨μ΄ μλλΌ κ°λ
μ λ¬Έμ λΌκ³ μ£Όμ₯νλ€. μ¦, κ°κ°μ μκ³λ λ
Ένμ μλ‘ λ€λ₯Έ μ°¨μμ ν¬μ°©νλ€. GrimAgeλ μΌμ¦ λ° νμ₯ λ¨λ°±μ§ λ리 μ§νλ₯Ό κ°μ‘°νκ³ , DunedinPACEλ μ₯κΈ° κΈ°λ₯μ λ³ν μλλ₯Ό κ°μ‘°νλ©°, PhenoAgeλ λμ¬ λ° λ©΄μ μ§νλ₯Ό κ°μ‘°νλ€. μ΄λ€μ μλ‘ κ΄λ ¨λμ΄ μμ§λ§ ꡬλ³λλ μλ¬Όνμ κ³Όμ μ μΈ‘μ νκ³ μμΌλ©°, μ΄λ₯Ό λ¨μΌν "μλ¬Όνμ λμ΄" μμΉλ‘ νμνλ©΄ μ΄λ¬ν 볡μ‘μ±μ΄ κ°λ €μ§λ€.
λΉμ© λ° μ κ·Όμ± μ₯λ²½
μμ λ±κΈμ DNA λ©νΈν νλ‘νμΌλ§(Illumina EPIC array)μ μνλΉ $200β$400μ λΉμ©μ΄ λ€λ©°, μ λ¬Έμ μΈ μ€νμ€ μ²λ¦¬κ° νμνκ³ κ²°κ³Όλ₯Ό λ°ννλ λ° μμΌμ΄ μμλλ€. μ΄λ μ°κ΅¬ λͺ©μ μμλ κ°λΉν μ μλ λΉμ©μ΄ μλμ§λ§, μΈκ΅¬ μ§λ¨ μ λ³κ²μ¬μλ λΉμ€μ©μ μ΄λ€. νμ₯ κ²μ¬(point-of-care testing)λ μ‘΄μ¬νμ§ μλλ€. νΉμ μκ³μ νμν CpG λΆμλ§μ μΈ‘μ νλ νμ ν¨λ λ±μ ν΅ν΄ λΆμλ²μ΄ λ λΉ λ₯΄κ³ μ λ ΄ν΄μ§κΈ° μ κΉμ§λ, μΌμμ μΈ μμ μ¬μ©μ΄ κ²½μ μ μΌλ‘ μ΄λ ΅λ€.
μ€ν κ°λ₯μ± κ²©μ°¨
κ°μ₯ ν΅μ¬μ μΈ μ₯λ²½μ ν립λ μμ μμ¬κ²°μ κΈ°μ€κ°μ λΆμ¬μΌ κ²μ΄λ€. νμμ GrimAgeκ° 5λ
κ°μλμ΄ μλ€λ©΄, μμ¬λ 무μμ λ€λ₯΄κ² ν΄μΌ νλκ°? μ΄λμ μ²λ°©ν΄μΌ νλκ°? μ€νν΄ μ©λμ μ‘°μ ν΄μΌ νλκ°? μΆκ° κ²μ¬λ₯Ό μλ’°ν΄μΌ νλκ°? μν λ¬Ένμ κ°μ λ
Ένκ° λ λμ κ²°κ³Όμ κ΄λ ¨μ΄ μμμ ν립νκ³ μμ§λ§, μ€μ¬ μ°κ΅¬λ λΆμ‘±νλ©° νμ±μ μ νμ λμ΄μ λν κ°μ©ν μ€μ¬λ€μ ν¨κ³Ό ν¬κΈ°λ λ―Έλ―Ένλ€.
Srivatsa λ±(2025)μ μ¬νκ΄ μμμμ μ΄λ₯Ό λΆλΆμ μΌλ‘ λ€λ£¨μμΌλ©°, MESA μ½νΈνΈμμ νμ±μ μ νμ λμ΄ κ°μμ΄ κ΄μλλ§₯ μΉΌμ μ μλ₯Ό λμ΄μλ μΆκ°μ μΈ μμΈ‘ μ 보λ₯Ό μ 곡ν¨μ 보μ¬μ£Όμλ€. κ·Έλ¬λ λΆκ°μ μμΈ‘μ΄ μλμ μΌλ‘ μμ μμ¬κ²°μ μ κ°μ μΌλ‘ μ΄μ΄μ§λ κ²μ μλλ©°, μ΄λ₯Ό μν΄μλ μλ¬Όνμ λμ΄ μ 보μ λ°λ₯Έ νλμ΄ κ²°κ³Όλ₯Ό λ³νμν¨λ€λ 무μμ κ·Όκ±°κ° νμνλ€.
λ―Έν΄κ²° μ§λ¬Έλ€
μ΄λ€ λͺ©μ μ μ΄λ€ μκ³λ₯Ό μ¬μ©ν κ²μΈκ°? μ΄ λΆμΌλ μνΌμ λ€ν± μκ³λ€μ μνΈ κ΅ν κ°λ₯ν κ²μΌλ‘ μ·¨κΈν기보λ€λ, νΉμ μκ³λ₯Ό νΉμ μμ λ§₯λ½μ λμμν€λ ν©μκ° νμνλ€.μ€μ¬ λ°μμ±: μν μ΅κ΄ λλ μ½λ¦¬νμ μ€μ¬κ° νμ±μ μ νμ λμ΄ κ°μμ μ λ’°μ± μκ² κ°μμν¬ μ μλκ°, κ·Έλ¦¬κ³ κ·Έ κ°μκ° μμμ μ΄λμΌλ‘ μ΄μ΄μ§λκ°?ννμ± ν¨μ: DNA λ©νΈν ν¨ν΄μ μΈμ’
, μ¬νκ²½μ μ μ§μ, νκ²½μ λ
ΈμΆμ λ°λΌ λ€λ₯΄λ€. μ£Όλ‘ λ°±μΈ μΈκ΅¬ μ§λ¨μ λμμΌλ‘ νλ ¨λ μκ³λ λ€λ₯Έ μ§λ¨μμ λ€λ₯΄κ² μλν μ μλ€.κ·μ κ²½λ‘: κ·μ κΈ°κ΄μ΄ μνΌμ λ€ν± μκ³λ₯Ό μμ μ§λ¨ λκ΅¬λ‘ μΉμΈνκΈ° μν΄ μ΄λ€ κ·Όκ±°λ₯Ό μꡬν κ²μΈκ°? μμ§ μ λ‘κ° μ‘΄μ¬νμ§ μλλ€.λ€μ€ μ€λ―Ήμ€μμ ν΅ν©: μνΌμ λ€ν± μκ³λ₯Ό λ¨λ°±μ²΄ν, λμ¬μ²΄ν, μμ λ°μ΄ν°μ κ²°ν©νλ©΄ λ¨μΌ λ°©μλ³΄λ€ μμμ μΌλ‘ λ μ μ©ν λ³΅ν© λ°μ΄μ€λ§μ»€λ₯Ό λμΆν μ μμ κ²μ΄λ€.λ§μΉλ©°
μνΌμ λ€ν± μκ³λ μλ¬Όνμ λ
Ένλ₯Ό μ λννλ λ° μμ΄ μ§μ ν κ³Όνμ μ§λ³΄λ₯Ό λνλΈλ€. νμ±μ μ νμ λ
Έν κ°μμ λΆμ μ κ±΄κ° κ²°κ³Όμ μ°κ²°νλ μνμ κ·Όκ±°λ μΌκ΄λκ³ μ μ¬νλμ΄ μλ€. κ·Έλ¬λ μμ μ μ©μ μμΈ‘ μ΄μμ μꡬνλ€. μ¦, νμ€ν, μ€ν κ°λ₯ν κΈ°μ€κ°, λΉμ© ν¨μ¨μ μΈ λΆμλ², κ·Έλ¦¬κ³ μλ¬Όνμ λμ΄ μ 보μ νμ©μ΄ νμ κ²°κ³Όλ₯Ό κ°μ νλ€λ κ·Όκ±°κ° νμνλ€. μ΄ λΆμΌλ μ΄λ¬ν μ΄μ νλ₯Ό ν₯ν΄ λμκ°κ³ μμ§λ§, μ°κ΅¬ μ μ©μ±κ³Ό μμ μ€λ¬΄ μ¬μ΄μ κ°κ·Ήμ μ¬μ ν μλΉνλ€.
References (5)
Kusters, C., & Horvath, S. (2024). Quantification of epigenetic aging in public health. Annual Review of Public Health, 45, 295β316.
Yamada, H. (2025). Epigenetic clocks and EpiScore for preventive medicine: risk stratification and intervention models. Journal of Clinical Medicine, 14(10), 3604.
Zhu, T., He, Y., & Wang, Y. (2025). GrimAge and GrimAge2 age acceleration effectively predict mortality risk. Epigenetics, 20(1).
Srivatsa, S., Rice, N., & Pike, J. R. (2025). Epigenetic aging clocks and incident cardiovascular outcomes: Results from the MESA. Journal of the American Heart Association, 14.
Arpawong, T., HernΓ‘ndez, B., & Potter, C. (2025). Physiological health Age (PhysAge): a novel multi-system molecular timepiece. GeroScience.