In the longevity community, DNA methylation-based epigenetic clocks have been hailed as highly promising biomarkers of aging. Biohackers and researchers alike frequently use these tests to measure biological age and track accumulated aging-associated damage. However, a recent short communication published in Aging Cell throws a critical lens on the clinical utility of these commercial clocks when compared to standard doctor-office measurements.
According to the study, simple and highly affordable traditional disease risk factors—specifically body mass index (BMI), waist-to-hip ratio (WHR), smoking status, and alcohol consumption—outperform state-of-the-art epigenetic clocks in predicting the incidence of non-communicable chronic diseases.
The Study: Pitting Clocks Against Traditional Risk Factors
Researchers from Tampere University and other Finnish institutions analyzed data from a middle-aged population cohort participating in the Young Finns Study. The study included 1,108 individuals who were aged 34 to 49 years at baseline and were entirely free from the studied chronic diseases. Over a 7-to-9-year follow-up period, 20.0 percent of these individuals were diagnosed with at least one non-communicable chronic disease or condition, such as hypertension, cardiometabolic disease, cancer, or steatotic liver disease.
The team tested the predictive value of several prominent epigenetic clocks, including:
- First-generation clocks like Hannum and Horvath.
- Second-generation clocks like PhenoAge and GrimAge.
- The pace-of-aging clock, DunedinPACE (Belsky et al., 2022).
- Principal component derivatives of these epigenetic clocks.
The Findings: Attenuation by Lifestyle
When looking at minimally adjusted models that only accounted for age, sex, and testing array version, advanced clocks like GrimAge, PhenoAge, and DunedinPACE successfully predicted disease incidence. This finding aligns with established literature showing that second- and third-generation clocks are superior to their predecessors as predictors of future health outcomes.
However, the narrative shifted dramatically when researchers adjusted the statistical models for easy-to-measure lifestyle variables.
- In fully adjusted models that included smoking, alcohol consumption, WHR, and BMI, none of the analyzed epigenetic clocks remained statistically significant predictors of disease incidence during the follow-up.
- The attenuation of the epigenetic clocks’ predictive power was largely driven by anthropometric variables, because when models were adjusted only for smoking and alcohol, GrimAge and DunedinPACE remained statistically significant.
- A statistical model relying exclusively on traditional risk factors (age, sex, smoking, alcohol, WHR, and BMI) demonstrated better discriminative performance than any model that included an epigenetic clock.
- The differences in the Area Under the Curve (AUC) between the traditional risk factor model and the minimally adjusted epigenetic clock models were statistically significant.
What This Means for Longevity Enthusiasts
These findings raise an important question: do epigenetic clocks provide any unique predictive value beyond what we can learn from basic, practically free measurements?
The study suggests that epigenetic clocks and traditional risk factors actually capture overlapping information. In this cohort, the individuals who remained healthy during the follow-up were simply younger and had a lower WHR and BMI at baseline. While epigenetic DNA methylation tests can be costly, obtaining your BMI and waist circumference is easy and affordable.
The authors caution that if epigenetic clocks are to be utilized in clinical settings or marketed as personal health monitoring tools, their added value over simple and affordable traditional risk factors must be clearly established. While these epigenetic metrics remain incredibly valuable for studying the biological aging process itself, a simple cost-effective approach using traditional disease risk factors currently outperforms them in predicting individual future health outcomes.
Reference: Kostiniuk et al., 2026. @adssx