How reliable are epigenetic clocks for measuring your biological age?

Epigenetic clocks measure biological age using chemical markers on your DNA (known as methylation patterns). They predict chronological age remarkably well across virtually every tissue and in different populations, outperforming most other molecular ageing markers in the process. That is quite an achievement for a measurement taken from a drop of blood or a saliva swab.

Particularly interesting is that the clocks measure more than just how old you are. People whose biological age is higher than their chronological age also face an increased risk of death, cardiovascular disease and cancer, even after correcting for other risk factors. The most advanced version, GrimAge, combines seven blood-protein markers with a smoking indicator and predicts mortality and heart disease with statistically exceptionally strong signals in validation studies involving thousands of people. It is worth noting that the developers have a commercial interest (patent), which is a caveat for this study.

Not all clocks perform equally well in all cell types. Older versions fell short with skin cells and certain other cell types. Newer clocks partly address this: they show high correlations in skin, blood, saliva, neurons, liver and bone. Nevertheless, DNA methylation profiles are tissue-specific, and how well a clock performs outside the tissue it was trained on has not yet been fully worked out.

A serious caveat: a critical analysis shows that many clocks have methodological problems. These include training groups that are too small or too homogeneous, so-called data leakage (where test data inadvertently influences the training process), and insufficient independent validation. This can cause a clock to perform worse in practice than the publication suggests. This does not apply equally to all clocks, but the problem is widespread.

Can a clock also demonstrate that a treatment makes you biologically younger? A small pilot study with nine participants suggested an average decrease of 1.5 years in epigenetic age after one year of hormone treatment. But without a control group and with only nine participants, that is far from proof. It is an indication that calls for larger, independent studies.