Overall, this is a study using buccal tissue (skin from the inside of the cheek) to look at DNA methylation marks associated with age. Previous studies have identified the DNA methylation differences in blood and developed a machine learning “age prediction” biological clock (an example is the well-known Horvath Clock). This study developed their “CheekAge clock” to identify regions that undergo methylation changes during ageing and then examined these regions in a longitudinal population (in this case the Lothian Birth Cohorts (LBC)) and used machine learning approaches to identify which of these regions could predict mortality – literally by examining the survival of individuals at various ages with and without these changes. 

However, I couldn’t see anything in the manuscript that said the CheekAge clock will be able to forecast the day (or even year) that someone will die. Epigenetic changes are theoretically reversible, so while it is possible that one can use the data to predict retrospectively when someone was going to die, based on the epigenetic changes that had accumulated, prospectively it would be only be useful for warning people that they may be heading for an early demise based on the accumulated epigenetic changes.

Given these changes are as a result of lifestyle choices (smoking, weight gain, poor diet and lack of fitness) combined with underlying disease (cancer, obesity, diabetes etc), all factors that are evident without epigenetic testing, it's not yet clear whether patients will be more likely to change their lifestyle choices when confronted with the epigenetic clock data than when their GP warns them to do so.

Epigenetic clocks already predict biological age – this study claims to take it potentially further by using the data to predict mortality – but did this retrospectively – in other words looked at a population and measuring the survival of people with the epigenetic clock markers.  What is needed is to do a “prospective study” – i.e. do the predictions on people alive today and see whether they die at a predicted time. This study was very careful and did not associate any epigenetic clock marker with any death age, and the focus was more on the accuracy of using buccal tissue rather than blood.

While the development of CheekAge, a next-generation epigenetic oral clock, represents a significant advance in non-invasive biomarkers of ageing, the use of the term ‘predictive of mortality’ raises ethical concerns. I believe this phrase may be misleading, as it implies a deterministic ability to predict death, which is neither scientifically nor ethically appropriate in this particular case. In reality, such clocks provide probabilistic risk assessments rather than concrete predictions. Emphasising mortality in this context could cause unnecessary anxiety and foster a fatalistic mindset in some people, rather than promoting practical knowledge about the duration of health and well-being.

A more nuanced approach would be to present CheekAge as a tool for assessing biological age and associated health risks, encouraging interventions aimed at extending healthy ageing rather than focusing on mortality. Clear communication is essential to avoid misinterpretation and to maintain ethical standards in the application of these technologies.