Physical activity and the epigenetic clock: what a major meta-analysis reveals about exercise and biological ageing

Table of Contents

1. Key Highlights
2. Introduction
3. What the review found: scope, main results and statistical takeaways
4. How epigenetic clocks differ and why that matters for interpreting results
5. Why some clocks might register physical activity and others do not
6. Interpreting the reported effect sizes: what does 0.03–0.09 SD mean in practical terms?
7. Measurement heterogeneity: self-report vs. objective activity measures
8. Confounding and reverse causation: interpreting cross-sectional evidence cautiously
9. Biological plausibility: mechanisms linking exercise to DNA methylation
10. Methodological limitations across the literature
11. What well-designed longitudinal studies and trials should look like
12. Real-world examples and parallels
13. Implications for public health, clinicians and individuals
14. Recommendations for researchers and funders
15. Practical guidance for people who want to use this evidence
16. Where uncertainty remains and how to interpret null findings
17. Next frontiers: single-cell methylation, tissue-specific clocks and personalized interventions
18. Ethical, social and commercial considerations
19. Final reflections on the state of evidence
20. FAQ

Key Highlights

• A systematic review of 44 studies (145,465 participants) found that higher physical activity is generally associated with lower biological age measured by DNA methylation clocks, with the strongest and statistically significant associations for Horvath and GrimAge measures.
• Meta-analysis of seven cross-sectional studies showed each 1 SD increase in MET-minutes/week corresponded to 0.03 SD lower Horvath epigenetic age acceleration (EAA) and 0.09 SD lower GrimAge EAA; no significant links were observed for Hannum or PhenoAge clocks.
• Evidence is dominated by cross-sectional data, varied activity measurements, and heterogeneous clock designs; randomized trials and longitudinal studies using standardized, objective activity metrics are needed to establish causality and dose-response.

Introduction

Physical activity lowers the risk of early death and a wide range of chronic diseases. Whether that protective effect is mirrored in molecular measures of ageing—specifically DNA methylation (DNAm) epigenetic clocks—has been a central question for researchers aiming to connect lifestyle to cellular ageing. A systematic review and meta-analysis published in The Lancet Healthy Longevity (May 2026) collated decades of observational work and provides the most comprehensive synthesis to date on the relationship between physical activity and epigenetic age.

The review finds a consistent, though modest, signal: people who report or demonstrate higher activity tend to show lower epigenetic age as measured by some DNA methylation clocks. However, the strength and consistency of that association depend heavily on which clock is used, how activity was measured, and whether analyses adjust for confounders such as smoking, BMI and socioeconomic factors. That pattern raises two parallel questions: what do different epigenetic clocks actually measure, and can increasing physical activity causally slow molecular ageing?

The evidence assembled and analyzed in this review illuminates where the field stands and where it must go next to convert associative signals into rigorous, actionable public health guidance.

What the review found: scope, main results and statistical takeaways

The authors identified 44 studies, together including 145,465 participants with mean ages ranging from about 24 to 78 years. Study designs were mostly observational, with a mixture of cross-sectional and longitudinal cohorts; only a minority of studies included objective activity measurement (accelerometers) rather than self-report.

Across the body of work, higher physical activity tended to correlate with lower DNAm age, but many individual studies did not reach statistical significance. The meta-analysis pulled together seven cross-sectional studies that reported comparable metrics. When physical activity was expressed in metabolic-equivalent-of-task (MET)-minutes per week—a common epidemiological unit—each one standard deviation higher MET-minutes was associated with:

• 0.03 standard deviations lower Horvath epigenetic age acceleration (EAA)
• 0.09 standard deviations lower GrimAge EAA
• No statistically significant association with Hannum EAA or PhenoAge EAA

Those effect sizes are small on an individual basis. A 0.09 SD reduction in GrimAge EAA implies a modest shift in the distribution of biological age across a population rather than a dramatic reversal in a single person’s molecular age. Nevertheless, even modest shifts at the population level could translate to measurable changes in morbidity and mortality if they reflect true biological change.

The authors emphasize that evidence is predominantly cross-sectional. Cross-sectional associations cannot establish temporality or causality: more active people could have lower epigenetic age because they were healthier to begin with, or because lower epigenetic age predisposes to being active. Longitudinal studies and randomized controlled trials (RCTs) with standardized, objectively measured activity are necessary to test whether physical activity causally modifies ageing trajectories.

How epigenetic clocks differ and why that matters for interpreting results

Not all epigenetic clocks are the same. They differ in how they were developed, what outcome they were trained to predict, and which CpG sites they use. Those differences matter when a behavioural exposure like physical activity shows an association with some clocks but not others.

• Horvath clock: Developed as a multi-tissue predictor of chronological age, Horvath’s clock uses many CpG sites across the genome and was trained on diverse tissues. When researchers measure epigenetic age acceleration (EAA), they typically compare predicted DNAm age to chronological age; a positive EAA indicates that DNAm age exceeds chronological age. Horvath EAA is a generalised marker of methylation-based deviation from calendar age.
• GrimAge: Designed to predict lifespan and time-to-death, GrimAge incorporates DNAm surrogates for plasma proteins and smoking history. It correlates strongly with mortality and healthspan outcomes. The meta-analysis found the strongest association between physical activity and GrimAge EAA, perhaps because GrimAge captures pathways—such as inflammation and vascular function—that respond to activity.
• Hannum clock: Trained on blood methylation data to predict chronological age, Hannum’s clock reflects age-related changes in blood methylation, but its relationship to morbidity and mortality is weaker compared with GrimAge.
• PhenoAge: Built to predict a composite phenotypic age metric derived from clinical biomarkers, PhenoAge aims to reflect biological status beyond chronological age and predict morbidity. The meta-analysis observed no significant association between physical activity and PhenoAge EAA.

Clocks differ not only in predictive goals but in sensitivity to systemic states. A clock optimized for mortality (GrimAge) may pick up signatures linked to behavioural risk factors and inflammatory status that respond to exercise. A clock trained only to mirror calendar age may be less tuned to those functional pathways.

Why some clocks might register physical activity and others do not

Several reasons explain why associations with physical activity vary by clock:

1. Biological pathways captured: GrimAge includes DNAm surrogates of proteins and smoking exposure—variables that mediate the effect of exercise on health (e.g., inflammatory cytokines, growth factors). Exercise influences inflammation, insulin sensitivity, lipids and vascular function; GrimAge may be more sensitive to those pathways than clocks designed to predict chronological age.
2. Cell-type composition: Most clocks are derived from bulk blood methylation. The proportion of immune cell subtypes changes with both age and exercise. Acute and chronic activity alter leukocyte distributions; changes detected by clocks may reflect shifts in immune cell composition rather than cell-intrinsic epigenetic reprogramming.
3. Tissue specificity: Exercise produces systemic effects, but some biological responses are tissue-specific (muscle, adipose, vascular endothelium). Blood-based clocks may miss beneficial changes occurring predominantly in muscle or brain tissue.
4. Time scale and reversibility: A single bout of exercise transiently modifies inflammatory markers and some epigenetic marks. Sustained training may be required to produce stable alterations in DNAm that clocks detect. Cross-sectional snapshots may capture long-term habitual activity better than short-term changes.
5. Measurement sensitivity and statistical power: Smaller studies and those with self-reported activity have less precision. Given the modest effect sizes, only well-powered studies with precise exposure measurement will reliably detect associations.

Understanding which clock best reflects modifiable, clinically relevant ageing processes remains an active research task. For interventions, the ideal clock would be sensitive to changes in physiology that translate into reduced morbidity and mortality with an intervention.

Interpreting the reported effect sizes: what does 0.03–0.09 SD mean in practical terms?

Standardized effect sizes provide a unitless way to compare associations across studies and scales. A 0.09 SD lower GrimAge EAA per 1 SD higher MET-minutes/week is modest. Translating that into years of biological age is challenging without additional calibration, because the mapping between standard deviations of EAA and years varies across datasets and clocks.

For context, epidemiological studies that link EAA to clinical endpoints typically find that each additional year of EAA increases risk of disease or death modestly. A small reduction in EAA across a whole population—if causal—can still yield meaningful public health benefits. Consider an analogy: a small shift in blood pressure distribution (for example, a 2 mm Hg population-level reduction) produces significant reductions in stroke and heart disease rates. Similarly, modest reductions in epigenetic age across many people could lower disease burden.

The effect sizes reported here are more useful for researchers designing trials and power calculations than as individual-level prognoses. They indicate that (1) physical activity associates with molecular ageing measures, and (2) studies seeking to detect these effects need adequate sample sizes and precise exposure measurement.

Measurement heterogeneity: self-report vs. objective activity measures

One major obstacle in this literature is heterogeneity in how physical activity is measured. Studies fall into two broad categories:

• Self-report instruments: Questionnaires are inexpensive and scalable but prone to recall bias, social desirability bias, and misestimation of intensity and duration. People often overestimate their moderate-to-vigorous activity and underestimate sedentary time.
• Objective measures: Accelerometers and heart-rate monitors offer minute-by-minute measures of movement and intensity. They provide more reliable estimates of MET-minutes, sedentary time, and bout patterns. Wearable devices have become commonplace in large cohorts (for example, many recent population studies have integrated wrist-worn accelerometer data).

The meta-analysis included both types, but studies using objective measures are likelier to detect smaller associations because of reduced measurement error. Researchers should standardize activity metrics: total MET-minutes/week, minutes of moderate-to-vigorous physical activity (MVPA) per day, and sedentary time are useful common denominators. Using device-derived measures wherever possible will increase sensitivity to detect associations with epigenetic endpoints.

Confounding and reverse causation: interpreting cross-sectional evidence cautiously

Cross-sectional designs dominate the literature. Such studies can reveal associations but cannot determine directionality. Two broad challenges arise:

• Confounding: Physical activity correlates with many variables that also influence DNAm and health—smoking, alcohol, diet quality, socioeconomic status, chronic disease burden and body composition. Studies that fail to adjust adequately for these variables risk attributing epigenetic differences to activity when they reflect these other factors.
• Reverse causation: People with lower biological age (or fewer health limitations) are more likely to be active. Without temporal separation of exposure and outcome or intervention assignment, distinguishing whether activity reduces epigenetic age or whether lower epigenetic age enables activity is difficult.

Analytical strategies can partially mitigate these problems: careful covariate adjustment, sensitivity analyses, stratification, and using objective measures help. Stronger evidence will come from longitudinal cohorts with repeated methylation and activity measures, and from intervention trials.

Biological plausibility: mechanisms linking exercise to DNA methylation

Several plausible biological mechanisms could link habitual physical activity to DNAm patterns captured by epigenetic clocks:

• Inflammation modulation: Exercise decreases systemic inflammation—a driver of ageing-related pathophysiology. Changes in inflammation-related DNAm signatures could explain why clocks that capture inflammatory pathways (such as GrimAge) are sensitive to activity.
• Metabolic regulation: Exercise improves insulin sensitivity, lipid profiles and mitochondrial function. DNAm marks associated with metabolic pathways might shift in response to sustained activity.
• Immune cell composition: Chronic activity influences the distribution and function of circulating leukocytes. Bulk methylation profiles partially reflect these composition shifts.
• Oxidative stress and DNA repair: Exercise modulates oxidative stress and stress-response signaling pathways, which can influence methylation enzymes and epigenetic maintenance.
• Muscle-specific signalling: Repeated muscle contraction induces local alterations (myokine release, chromatin remodeling) that could produce systemic effects detectable in blood through secreted factors.

These mechanisms are plausible and supported by experimental biology, but direct causal chains from exercise to specific CpG methylation changes to reduced morbidity remain under active investigation.

Methodological limitations across the literature

The review identifies recurring methodological challenges that limit causal inference and comparability:

• Predominance of cross-sectional designs: Few longitudinal or randomized studies are available, reducing the ability to infer causation.
• Measurement heterogeneity: Mixing self-report and device data, along with different activity metrics, complicates pooled analyses.
• Heterogeneity among clocks: Different clocks capture different biological signals; pooling results across clocks without accounting for their distinct constructs reduces clarity.
• Incomplete adjustment for confounders: Not all studies adjusted for smoking, adiposity, medication use, comorbidities or socioeconomic factors.
• Population diversity: Studies varied in age range, ethnicity, and health status. Lack of representation limits generalizability.
• Tissue limitations: Nearly all clocks applied to blood; activity-induced changes in other tissues may be missed.
• Publication bias: Studies with null results may be underreported, potentially inflating estimates of association.

Addressing these limitations should be a priority for future research.

What well-designed longitudinal studies and trials should look like

To move beyond association and toward causation, the field needs coordinated, rigorous studies with the following design elements:

1. Objective activity measurement: Use accelerometers or other validated devices to quantify activity intensity, duration, frequency and sedentary patterns. Standardize reporting in MET-minutes, MVPA minutes, and sedentary minutes.
2. Repeated measures: Collect DNAm profiles and activity data at multiple time points to model trajectories and test temporal precedence.
3. Sufficient duration and intensity: Intervention trials should last long enough to plausibly alter methylation (months to years), include dose variation, and have adherence monitoring.
4. Standardized epigenetic endpoints: Pre-specify which clocks will be primary outcomes and report other clocks as secondary, ensuring comparability across studies.
5. Comprehensive covariate collection: Capture smoking, alcohol, diet, sleep, BMI, medications, comorbidities and socioeconomic variables.
6. Power calculations: Use the small effect sizes reported here to plan adequately powered trials and cohort analyses.
7. Multi-omics integration: Combine DNAm with transcriptomics, proteomics, metabolomics and immune phenotyping to establish mechanistic links.
8. Diverse populations: Include younger and older adults, male and female participants, and varied ancestral backgrounds to assess generalizability and effect modification.
9. Tissue considerations: Where feasible, assay tissues beyond blood (e.g., muscle biopsies) to evaluate tissue-specific epigenetic responses to exercise.
10. Causal inference methods: Complement RCTs with Mendelian randomization, negative controls and other approaches to triangulate causal effects.

Several large cohorts and biobanks collecting accelerometry and methylation data (or capable of adding these measures) offer opportunities to implement these designs at scale.

Real-world examples and parallels

Several programmatic and research efforts illustrate how the field might progress:

• Population cohorts that combine wearables and methylation: Large epidemiological cohorts are increasingly embedding wrist-worn accelerometers alongside blood sampling. When those cohorts add methylation assays, they provide prospectively measured activity plus molecular endpoints—ideal for trajectory analyses.
• Exercise interventions with molecular outcomes: Trials that enroll sedentary adults into structured training programs and collect blood methylation pre- and post-intervention can test whether sustained increases in activity shift epigenetic age. Even small, well-controlled trials provide mechanistic insights, while larger multicenter trials offer robust effect estimates.
• Multicomponent lifestyle trials: Interventions combining exercise, diet, sleep hygiene and smoking cessation could plausibly produce larger, clinically meaningful changes in DNAm age. However, attributing effects to exercise per se requires factorial design or careful mediation analysis.
• Wearable-driven public-health programs: Integrating DNAm endpoints into real-world activity-promotion programs could evaluate whether scalable behavior change initiatives produce molecular benefits at the population level.
• Individual-level biofeedback and personalization: As wearables become more granular and accessible, investigators could explore whether personalized exercise prescriptions based on baseline epigenetic profiles produce optimized outcomes.

These approaches require coordination, standardization and sufficient funding, but the infrastructure exists to mount them.

Implications for public health, clinicians and individuals

The meta-analysis strengthens the case that regular physical activity correlates with younger epigenetic age on certain molecular clocks, adding molecular-level evidence to well-established epidemiological benefits of exercise. However, two caveats guide how this evidence should be used:

• For public health messaging: Physical activity remains a cornerstone of chronic disease prevention. The new molecular evidence provides supportive mechanistic plausibility but does not fundamentally change existing recommendations. Encouraging population-level increases in activity is justified by the large body of clinical and epidemiological evidence linking exercise to lower mortality, improved cardiometabolic health, and better mental health.
• For clinicians and patients: Clinicians can present epigenetic evidence as one more line of data supporting exercise, while avoiding overpromising effects on "biological age." Advice should remain practical: meet established activity guidelines (e.g., at least 150 minutes of moderate-intensity activity per week or equivalent), reduce sedentary time, and combine aerobic and resistance training as tolerated.

For individuals curious about biomarker testing: DNAm age tests are available commercially, but their interpretation for personal decision-making is limited. A single epigenetic age estimate does not capture the full complexity of health or guarantee future outcomes. Lifestyle change remains the most reliable approach to improving health metrics.

Recommendations for researchers and funders

The review’s findings create a focused research agenda:

• Fund longitudinal and interventional studies with objective activity measures and standardized epigenetic outcomes.
• Support harmonization efforts to align activity metrics, covariates, analytic pipelines and reporting standards across cohorts.
• Encourage multi-omics and tissue-specific work to untangle whether observed methylation changes reflect immune composition, systemic signaling, or cell-intrinsic epigenetic remodeling.
• Prioritize diversity in study samples to understand whether associations differ by age, sex, ancestry and baseline health.
• Explore dose-response relationships and whether particular activity modalities (aerobic vs resistance vs interval training) differentially affect epigenetic measures.
• Invest in methodological work to translate standardized-effect estimates into interpretable units (e.g., years of epigenetic age) and to connect epigenetic change to hard outcomes like disease incidence and mortality.

Coordinated consortia and pooled analyses will accelerate progress and make maximal use of existing cohort data.

Practical guidance for people who want to use this evidence

For individuals wondering whether to change behaviour based on epigenetic clocks, the guidance aligns with established public-health recommendations:

• Aim for regular aerobic activity: Aim for at least 150 minutes per week of moderate-intensity exercise, or 75 minutes of vigorous activity, distributed across most days.
• Incorporate strength training: Resistance exercises at least two days per week preserve muscle mass and function with age.
• Reduce sedentary time: Break prolonged sitting with light activity or standing; even low-intensity movement has measurable metabolic benefits.
• Choose activities you can sustain: Long-term adherence matters more than short bursts of intense exercise followed by relapse.
• Monitor objectively if possible: Wearable devices provide consistent feedback and better measurement for personal tracking and for research contexts.
• Combine behaviors: Smoking cessation, improved diet, adequate sleep and weight management enhance the benefits of physical activity.

Epigenetic age can be a motivating concept, but behaviour change should be guided by achievable, evidence-based steps that improve overall health.

Where uncertainty remains and how to interpret null findings

The lack of association between physical activity and some clocks (Hannum, PhenoAge) does not negate the overall evidence that exercise benefits health. Instead, these null results highlight what clocks measure. A clock trained specifically to emulate calendar age or a composite clinical phenotype might not be tuned to the molecular pathways most responsive to exercise.

Null findings may also reflect measurement limitations: underpowered studies, activity misclassification, inadequate adjustment, or short follow-up. When future studies use objective measurement, standardized clocks and longitudinal designs, some currently null associations may change.

Researchers and practitioners should interpret current findings as suggestive rather than definitive. The balance of epidemiology, physiology and molecular biology supports benefits of activity; the epigenetic evidence adds nuance regarding which molecular signatures are modifiable.

Next frontiers: single-cell methylation, tissue-specific clocks and personalized interventions

Emerging technologies could resolve key uncertainties:

• Single-cell methylation profiling: Bulk blood methylation mixes signals from different cell types. Single-cell approaches can distinguish cell-intrinsic methylation changes from shifts in cell composition induced by exercise.
• Tissue-specific clocks: Clocks developed for muscle, adipose, brain or vascular tissues could directly capture exercise-induced remodeling in those tissues, potentially revealing stronger associations than blood-based clocks.
• Dynamic clocks: Epigenetic marks that change over weeks may be most relevant for short-term interventions; researchers can develop clocks optimized to detect dynamic responses rather than long-term chronological drift.
• Integration with wearables and longitudinal phenotyping: Dense time-series data on activity, physiology and molecular markers could permit causal inference and individualized response profiling.

These approaches require investment but offer pathways to determine whether exercise can be used specifically to modulate biological ageing trajectories.

Ethical, social and commercial considerations

As epigenetic clocks and activity tracking converge, several non-scientific issues emerge:

• Commercial tests: Direct-to-consumer methylation assays purporting to measure biological age have proliferated. Without clear standards and validated interventions, claims about reversing biological age are premature and risk misleading consumers.
• Equity: Wearable-driven and molecularly informed interventions could widen disparities if access is uneven. Research and policy should prioritize equitable access to measurement tools and intervention programs.
• Privacy: Combining molecular markers and continuous activity data raises privacy concerns. Data governance frameworks must protect participant confidentiality and consent.
• Motivational use: Epigenetic age estimates could motivate behaviour change for some individuals, but the psychological effects of receiving an “accelerated” or “younger” molecular age are not fully understood.

Researchers, clinicians and policymakers should develop guidelines that balance innovation with consumer protection.

Final reflections on the state of evidence

The 2026 systematic review and meta-analysis establishes a consistent but modest association between physical activity and certain epigenetic clocks—most clearly with GrimAge and to a lesser extent Horvath EAA. These findings align with biological plausibility and mechanistic pathways responsive to exercise. Yet the evidence remains primarily cross-sectional, measured heterogeneously, and limited by clock-specific sensitivity.

The path forward requires coordinated longitudinal work and interventional trials that use objective activity metrics, standardized epigenetic endpoints and comprehensive covariate control. Those studies will determine whether increasing physical activity can causally shift molecular ageing and, crucially, whether such shifts translate into reduced disease and extended healthspan.

For now, the molecular evidence supplements a robust epidemiological foundation: physical activity remains a low-cost, high-value intervention for health. The emerging epigenetic signals hint that this behavioural prescription may indeed align with measurable changes at the molecular level—but that promise must be tested with rigorous science before it becomes clinical practice.

FAQ

Q: Which DNA methylation clocks showed an association with physical activity? A: The meta-analysis reported statistically significant associations for Horvath epigenetic age acceleration and GrimAge epigenetic age acceleration. GrimAge showed the larger standardized effect (about 0.09 SD lower EAA per 1 SD increase in MET-minutes/week). No statistically significant associations were observed for Hannum or PhenoAge EAA in the pooled analysis.

Q: How large are the effects—are they clinically meaningful? A: The observed associations are small on an individual level. A 0.09 standard-deviation shift in GrimAge EAA suggests a modest change in molecular-age distribution at the population level. Whether these shifts produce clinically meaningful reductions in disease or mortality requires longitudinal and interventional evidence linking DNAm change to hard outcomes.

Q: Why do different clocks give different results? A: Clocks were developed for distinct purposes and trained on different targets (chronological age, mortality risk, composite clinical phenotypes). GrimAge incorporates DNAm proxies for plasma proteins and smoking exposure and was optimized for mortality prediction, making it potentially more sensitive to physiological pathways affected by exercise. Horvath’s clock predicts chronological age across tissues; Hannum and PhenoAge capture different aspects of ageing biology. Measurement properties and the specific CpG sites included determine each clock’s responsiveness to behavioural exposures.

Q: Can I use commercial epigenetic age tests to track the effect of my exercise program? A: Commercial tests offer a snapshot of DNAm-based age but have limitations. Single measurements are noisy and may reflect transient states or measurement variability. Current evidence does not support using these tests as definitive feedback on the success of an individual exercise regimen. If you pursue testing, combine it with validated clinical measures and interpret changes cautiously.

Q: Does this review prove that exercise slows biological ageing? A: No. The bulk of included studies were cross-sectional, which can show association but not causation. Longitudinal studies and randomized controlled trials with objective activity measurement are necessary to establish whether physical activity causally reduces epigenetic age and whether that leads to improved clinical outcomes.

Q: What should future studies measure and how should they be designed? A: Future studies should use objective activity tracking (accelerometers), collect repeated DNAm profiles, pre-specify epigenetic clocks, control comprehensively for confounders, include diverse populations, and integrate multi-omics when possible. Randomized trials of sustained, well-characterized interventions with sufficient duration and power will be crucial.

Q: Are there practical takeaways for clinicians and public health practitioners? A: Yes. The epigenetic evidence adds molecular plausibility to established recommendations: promoting regular aerobic and resistance activity, reducing sedentary time, and combining exercise with healthy diet and smoking cessation remain cornerstone strategies for healthy ageing. The molecular findings do not change clinical guidance but provide additional motivation for evidence-based activity promotion.

Q: Could wearable devices and large cohorts answer remaining questions? A: Yes. Cohorts that combine device-based activity monitoring with methylation assays and long-term follow-up are particularly well positioned to map trajectories and assess causality. Wearables improve exposure measurement, while large sample sizes increase power to detect modest effects.

Q: What ethical and practical concerns arise from using epigenetic clocks in public health? A: Concerns include potential commercial overreach by direct-to-consumer tests, inequitable access to testing and wearables, privacy risks around combining molecular and continuous behaviour data, and uncertain psychological impacts of sharing biological-age information. Policy and ethics frameworks must evolve alongside scientific advances.

Q: Where can I read the full systematic review? A: The review is published in The Lancet Healthy Longevity (May 2026) with DOI: https://doi.org/10.1016/j.lanhl.2026.100835.