GeroScience · Nov 16, 2026

Epigenetic clock choice matters: systematic variation in age estimation and socioeconomic patterning of biological aging

By Rähse T, Großkopf A, Riedel RI et al.

Epigenetic clocks based on DNA methylation (DNAm) are common tools in aging research. While first-generation clocks were developed to predict chronological age, second- and third-generation clocks aim to capture aspects of biological aging. However, little is known about how different clocks compare in estimating deviations between chronological and DNAm age and their sensitivity to social and life stage disparities in population-based samples. This study examines delta age across epigenetic clocks from three generations (1st-Horvath, 2nd-PhenoAge, 3rd-DunedinPoAm), focusing on estimation patterns and socioeconomic gradients. Data stem from the UK Household Longitudinal Study, including DNAm profiles and socioeconomic measures. Delta age was computed as the difference between DNAm and chronological age. Linear mixed-effects models assessed delta age by clock type, age group, sex, and interactions with socioeconomic status (education, occupation, income), adjusting for batch effects and cell composition. Substantial variation in delta age emerged across clocks and age groups. First-generation clocks overestimated age in younger and underestimated it in older individuals. The second-generation clock showed consistent negative delta age. The third-generation clock displayed minimal deviation and stable calibration. While the Horvath clocks showed weak or inconsistent socioeconomic associations, PhenoAge and DunedinPoAm captured clearer gradients: Higher education and income were associated with lower delta age, indicating slower epigenetic aging. Epigenetic clocks differ in structure and sensitivity to socioeconomic factors. Next-generation clocks appear more suitable for research on health inequalities and their embodiment. Findings underscore the importance of clock selection and call for longitudinal validation across populations and life stages. Graphical Abstract

doi.org/10.1007/s11357-026-02630-zOpen accessPubMed 43037914Biomarkers

AgingResearch.News