The question as to why those of the same chronological age exhibit different disease susceptibility and health status has elicited the development of other forms of measurement better adapted to predict patterns of aging. Biological age extends beyond chronological age to represent accumulated deterioration of cell, tissue, and organ function. Elevated biological age compared to chronological age indicates accelerated aging, while reduced biological age compared to chronological age is suggestive of a slower pace of aging. Biological aging clocks based on proteomic, metabolomic, immune signatures, and DNA methylation have since been introduced to assess biological age. Epigenetic aging biomarkers are derived from DNA methylation patterns and purported to inform biological age. Epigenetic age is often viewed as an exposure or outcome that provides enhanced insight as compared to chronological age in reference to aging processes and potential health risks.
Several epigenetic clocks are currently in use, each implementing unique exposures and methods optimized to detect a particular outcome, with the final objective being to project biological or factual age from DNA methylation profiles. First generation chronological clocks (Horvath’s DNAmAge and Hannum clock) rely on penalized regression models, selecting and summing specific CpG sites to assess biological age. Their limitation lies in poor sensitivity to morbidity-associated biological processes. Second generation phenotypic clocks (PhenoAge, GrimAge, GrimAge2) are trained on clinical biomarkers, morbidity, and mortality risk to arrive at an approximate biological age. Such measures more closely correlate with age-related disease pathways and are more informative as to mechanistic links between exposures and health outcomes. Third generation pace-of-aging clocks (DunedinPoAm and DunedinPACE) estimate speed of aging of organs and physiological systems based on longitudinal cohort data. An advantage of this method is that it helps to illuminate whether an exposure might be linked to an increased or decreased rate of biological aging. Finally, specialized clocks (Intrinsic Capacity and Cortical clocks) are trained on specific scenarios, measuring environmental exposure effects or examining organ age. In general, epigenetic age is arrived at via a weighted linear sum of particular CpG sites, comparable to conventional risk prediction models. For accurate interpretation of these clocks, it is subsequently an imperative to understand which outcomes and types of specimens are studied, the total number of measurements made of the DNA methylation array, the statistical algorithms applied, and the quantity of CpG sites analyzed.
To date, most research evaluating the quality of epigenetic age clocks has focused predominantly on exposures related to lifestyle factors, serum, or urinary biomarkers, and mortality and CVD as outcomes. Such a narrow variable data set creates considerable gaps, especially that of the aggregate of exposures that occur across the lifespan. The current generation of clocks fail to relate accelerated epigenetic aging to cumulative exposures on health outcomes. Future clocks should aim to incorporate more comprehensive lifestyle, environmental, metabolic, and pharmacological exposures to better capture other key determinants of longterm health. Several additional issues exist in the scientific literature covering the topic of epigenetic age clocks that should not be ignored. These include inadequate reporting of causal models and methods, and neglecting to describe the causal models and strategic analytical decisions. Very few papers provide visual causal diagrams explicitly demonstrate assumed causal, mediator, and confounder variables. Several studies use relatively small sample sizes (N < 500) which prohibits the determination of a mediation effect and leads to results that could be mere statistical coincidence. The prevalent use of cross-sectional designs prevents assessment of causation which necessitates a strict timeline. Furthermore, a cross-sectional study design hinders the ability to control for baseline levels of the mediator and outcome, allowing for introduction of measurement errors and bias. Finally, authors frequently fail to highlight timing violations or cite consequent biases as a limitation of the study.
Beyond the concerning lack of methodological rigor of many studies, most research to date has centered on those of European ancestry. Asian and African ancestries are minimally represented, creating a severe imbalance due to divergent environmental exposures, genetic backgrounds, and social determinants of health between populations. While methylation clocks are becoming more accessible and informative, their critical limitations should be observed with a keen eye in order to avoid overconfident interpretation the information they provide.
Fujii R, Tsuboi Y, Cardenas A, et al. Do epigenetic aging biomarkers mediate the association between exposure and health outcomes? A scoping review of mediation analysis. Environ Res. 2026;302:124600. doi:10.1016/j.envres.2026.124600