Mavrommatis · Nature communications 2025 · prospective cohort study · n=18,859

An unbiased comparison of 14 epigenetic clocks in relation to 174 incident disease outcomes.

Cited 25 times in the scientific literature.

Level 3 - non-randomized controlled study

Prospective longitudinal cohort study assessing biomarker prediction of incident disease outcomes.

PubMed 41402269 · doi:10.1038/s41467-025-66106-y · record verified 2026-08-31

What was done

Researchers evaluated 14 widely used epigenetic clocks across 18,859 individuals to predict all-cause mortality and 174 incident disease outcomes over a 10-year follow-up period. Associations were assessed using fully adjusted Cox proportional hazards regression models controlling for lifestyle and socioeconomic factors, alongside classification models comparing clocks against traditional risk factor baselines.

What was found

Second- and third-generation epigenetic clocks significantly outperformed first-generation clocks. There were 176 Bonferroni-significant associations (P < 0.05/174); in 27 diseases (including primary lung cancer and diabetes), the hazard ratio for the clock exceeded its association with all-cause mortality. Adding clock metrics to traditional risk factors significantly increased classification accuracy by >1% for 32 of the 176 associations. Interactions between clocks and sex or smoking status were minimal. Clocks performed best for respiratory and liver-related conditions.

Why it matters

This large-scale benchmark demonstrates that newer epigenetic clocks capture specific disease risks beyond general biological aging and traditional risk factors, while confirming that first-generation clocks have limited utility in clinical disease prediction.

Limits

The abstract does not report population demographics, geographic origin, or ethnic diversity of the cohort. Exact hazard ratios and confidence intervals are not provided in the abstract, and the incremental improvement in classification accuracy was modest (>1% in only 32 of 176 significant associations).

Cited by