Basic risk factors beat epigenetic clocks

DNA methylation-based epigenetic clocks have been promoted as biomarkers of ageing, and they have repeatedly been shown to predict morbidity and mortality. What the literature has largely skipped is the harder question: do they add anything beyond the information already available from cheap, routine measures such as age, smoking status, drinking and body composition? Incremental value, not raw predictive ability, is what determines whether a test is worth running in a clinic or selling to consumers. This study set out to make that comparison formally.

The researchers used a middle-aged Finnish population cohort of 1108 participants, aged 34 to 49 years at baseline, and tracked incidence of ageing-associated non-communicable chronic disease over a 7-to-9-year follow-up. They calculated the most commonly used epigenetic clocks from DNA methylation data. They then built statistical prediction models based on traditional risk factors - age, sex, smoking, alcohol consumption, waist-hip ratio and body mass index - and compared these against models that added or substituted an epigenetic clock. The comparison was framed explicitly around added value rather than whether the clocks predicted disease at all.

The combination of traditional risk factors outperformed any model that included an epigenetic clock for predicting incident chronic disease in this cohort. In other words, the methylation measures did not improve on what age, sex, smoking, alcohol, waist-hip ratio and body mass index already conveyed. The authors do not claim the clocks are uninformative in isolation; their point is that the incremental gain over simple and affordable measures was absent here. They conclude that added value should be clearly established before clocks are used clinically or as tools of personal health monitoring.

This complicates rather than contradicts the existing evidence that epigenetic clocks predict morbidity and mortality: predicting an outcome and improving on established predictors are different achievements. The cohort is relatively young and the follow-up covers seven to nine years, so it is possible clocks perform differently in older people, over longer horizons, or for mortality specifically. It is also a single cohort, and prediction comparisons can be sensitive to which clocks and which risk factors are chosen. Replication across ages, populations and outcome types is the obvious open question.

This is a reason to be sceptical of consumer biological-age tests that promise insight beyond the basics, at least for predicting chronic disease in middle age. The unglamorous variables - your age, whether you smoke, how much you drink, and your body composition - carried the predictive weight in this cohort. That does not mean epigenetic clocks are meaningless, only that their extra contribution has not been demonstrated where it matters commercially. Treat a clock result as a research output rather than a personal health verdict.