Scientists compared 51 longevity studies to see which aging biomarkers respond most reliably

By Pooja Toshniwal PahariaReviewed by Susha Cheriyedath, M.Sc.Aug 25 2026

A large-scale comparison of human longevity studies reveals why some epigenetic clocks detect biological changes far more consistently than others.

Study: Responsiveness of epigenetic aging biomarkers to longevity interventions in humans. Image Credit: Lightspring / Shutterstock

In a recent study published in the journal Nature Medicine, researchers explored how epigenetic aging biomarkers respond to interventions aimed at promoting longevity or healthspan. Clocks trained to predict mortality or the pace of aging demonstrated the strongest responsiveness across interventions. DNAm biomarkers showed particularly strong responses to lifestyle and pharmacological interventions, while population health status also influenced biomarker responsiveness.

The findings could guide the choice of intervention and DNAm biomarkers for further trials, potentially reducing study duration and sample size while helping researchers select appropriate study populations. An improved understanding of biomarker responses to longevity interventions could also help identify DNAm markers that may ultimately be suitable for use as surrogate endpoints in aging research.

Using aging biomarkers, researchers could potentially evaluate the effects of longevity interventions more rapidly without requiring trials to continue for long periods, which can sometimes span decades. Nevertheless, researchers must assess how aging biomarkers, including epigenetic clocks, respond to these interventions before using them as surrogate endpoints in clinical trials.

Surrogate markers that use short-term measurements to predict longer-term outcomes could accelerate the testing and clinical translation of interventions. However, most previous studies have explored associations between biomarkers and risk factors, morbidity, and mortality, or have used inconsistent clock panels, which limit comparisons between studies.

About the study

In the present study, researchers assessed how epigenetic aging biomarkers respond to longevity interventions in humans by curating the TranslAGE database, which includes 51 longitudinal intervention studies. Using this database, they calculated 16 epigenetic clocks across the included studies. They also assessed 94 DNAm biomarkers to examine biological processes that may underlie the changes observed across epigenetic clocks.

The TranslAGE database included data from the European Molecular Biology Laboratory (EMBL), the Gene Expression Omnibus (GEO), and TruDiagnostic clinical trials. The epigenetic clock panel included first-generation clocks (Horvath and Hannum), second-generation clocks (PhenoAge, DunedinPoAm38, and GrimAgeV1), and their updated versions. The researchers also used explainable clocks (generation X) developed during testing, including OMICmAge, DNAmEMRAge, and SystemsAge.

From all included studies, the team curated metadata, including intervention type, sample population, comorbidities, age, and sex, and adjusted the DNAm biomarkers for chronological age. To allow meaningful comparisons between the different trials and DNAm biomarkers, they standardized each biomarker's age-adjusted values.

The interventions were classified into lifestyle, pharmacological, supplement, and medical-procedure interventions. Using paired t-tests, the researchers compared the pre- and post-intervention blood samples. In addition, they used linear regression models to explore associations between the clinical covariates and DNAm effect sizes.

Results

The team discovered patterns of responsiveness across various DNAm markers and longevity interventions. Clocks trained to predict mortality or the pace of aging demonstrated the strongest responses across interventions. Among interventions tested, 19 significantly decreased DNAm aging measures, whereas five significantly increased them. Most of the remaining interventions showed no significant effects.

Lifestyle and pharmacological interventions drove the strongest responses from DNAm markers. These two intervention groups significantly reduced DNAm aging biomarker scores, with pharmacological interventions producing larger average effects. Separately, population health status strongly influenced responsiveness, with several DNAm biomarkers showing greater decreases in studies involving people with disease than in healthy populations. In particular, therapies targeting tumor necrosis factor (TNF) modified almost all second-generation DNAm biomarkers to a similar extent among people living with arthritis or inflammatory bowel disease (IBD). TNF-targeting agents may therefore have considerable effects on epigenetic aging biomarkers among people with inflammatory conditions.

In addition, metformin was among the pharmacological interventions that produced strong changes in DNAm biomarkers. Explainable biomarkers indicated particularly large changes in inflammatory, brain, and metabolic system scores. More broadly, the authors suggested that pharmacological interventions may influence DNAm biomarkers through inflammatory and metabolic pathways involving AMP-activated protein kinase (AMPK), mechanistic target of rapamycin (mTOR), or TNF.

Two types of the Mediterranean diet decreased similar second-generation biomarkers in healthy individuals. Conversely, five senolytic studies showed divergent changes in multiple DNAm biomarkers, suggesting that the effects of senolytic interventions on epigenetic aging may not be consistent.

Reliable generation 2+ DNAm biomarkers, including GrimAgeV2, PCGrimAge, SystemsAge, DunedinPACE, and PCPhenoAge, showed significant decreases across several interventions. DunedinPACE showed the greatest overall responsiveness to intervention-associated changes, while PCGrimAge provided the strongest statistical evidence of responsiveness. The study duration and sample population characteristics were important factors influencing the responsiveness of DNAm biomarkers to interventions.

Explainable clocks designed using multiple subscores provided greater specificity and biological insight into intervention responses than single-score epigenetic clocks. The epigenetic proxy for triglycerides was significant in six of seven dietary interventions, while the glucose proxy was significant in four of seven, suggesting that they may be useful for assessing the effects of dietary interventions.

Conclusion

The findings suggest that biomarkers predicting the pace of aging or mortality may be promising candidates for further evaluation as surrogate endpoints in clinical trials. Pharmacological interventions produced the largest effects on DNAm biomarkers.

While animal studies suggest that TNF-targeting agents may extend lifespan in mice, corresponding biomarkers should continue to be tested as potential surrogate endpoints in long-term validation studies.

Since reliable generation 2+ clocks were the most responsive across interventions, including these clocks, rather than epigenetic clocks trained primarily to predict chronological age, could improve the ability of future trials to detect intervention-related changes.

However, responsiveness alone does not establish a DNAm biomarker as a valid surrogate endpoint or show that an intervention slows aging. The authors emphasized that biomarker changes still need to be linked with clinically meaningful outcomes and long-term effects on disease, healthspan, or lifespan.