Analysis Finds Some Epigenetic Aging Clocks Respond More Consistently to Interventions

An analysis of 51 longitudinal human intervention studies found that some epigenetic aging clocks respond more consistently to interventions than others, with clocks trained to predict mortality or the pace of aging showing the strongest and most aligned changes across the studies examined.

The analysis used TranslAGE, a harmonized database containing public and private clinical studies. Researchers calculated results for 16 prominent epigenetic clocks in each study and examined 94 additional DNA methylation biomarkers to help interpret the changes associated with each clock. DNA methylation refers to chemical markings on DNA that can change with age and other influences.

The interventions represented in the database included exercise, dietary approaches such as calorie restriction, metformin and therapeutic plasma exchange. The source describes these interventions as examples of approaches that have been assessed in humans and are known or hoped to affect late-life health or life expectancy. Their effects in the catalog were generally modest, reflecting the still limited amount of rigorous human intervention data collected since aging clocks became available.

Across the studies, pharmacological and lifestyle interventions produced the strongest responses from DNA methylation biomarkers. The characteristics of the study population and the length of the study were also important factors in determining whether a biomarker responded to an intervention.

The analysis found that clocks with multiple subscores—described as “explainable clocks”—provided more specificity and greater mechanistic insight than clocks producing a single score. These subscores may help researchers identify which biological components change during an intervention rather than relying only on an overall aging estimate.

Aging clocks are generally built with machine-learning methods using biological data that changes with age. They are calibrated to estimate age in a reference population, and clock ages that exceed chronological age tend to correlate across populations with mortality risk and age-related conditions. However, the source emphasizes that there is no established direct connection between the data used to construct a given clock and the underlying mechanisms of aging. As a result, a clock could underestimate or overestimate the effect of a particular intervention, and its reliability cannot currently be known in advance.

That limitation is central to the findings. The analysis may help guide the selection of biomarkers, study populations, intervention durations and sample sizes in future clinical trials, while reducing unnecessary testing of multiple measures. It does not establish that the clocks can already serve as surrogate endpoints for aging, rejuvenation or life expectancy. Validating that role would require determining whether changes in a clock reliably reflect meaningful long-term outcomes.