Longevity Briefs: Could A CT Scan Tell You How Fast You Are Ageing?

Longevity briefs provides a short summary of novel research in biology, medicine, or biotechnology that caught the attention of our researchers in Oxford, due to its potential to improve our health, wellbeing, and longevity.

The problem:

Different people age at different rates, and those rates can be measured. The most common way to measure how quickly someone is ageing is to use epigenetic clocks. DNA is collected (usually from a blood sample) and a machine learning algorithm uses the presence of methyl groups (molecular ‘tags’ that are added to or removed from the DNA molecule throughout life) in order to produce an age prediction. If that predicted age is greater than the person’s chronological age, it suggests that they may be ageing more rapidly than average.

Epigenetic clocks have been shown to be good predictors of disease and mortality but aren’t without their problems. Epigenetic age measurements can vary according to which organ or specific cell type the data comes from, or even according to the time of day. Here, researchers use an alternative source of data: abdominal computed tomography (CT) scans. CT scans provide information on the state of skeletal muscle, abdominal fat, blood vessel health and bone density, all of which change with age. What’s more, we know that changes have a causative relationship with age-related disease and death.

The discovery:

Researchers looked at data from 123,281 adults with a mean (average) age of 53.6 and who underwent abdominal CT scans. Based on the followup data (which extended to at least 5.3 years post-scan for at least half of the participants) researchers selected CT scan measurements that were best at predicting mortality. The best predictor was found to be skeletal muscle density, followed by calcification of the aortic artery, visceral fat density and bone density. They then used machine learning to produce a model that would predict CT-based biological age (CTBA).

Researchers then tested their model on a new group of 40,718 adults (mean age 53.9) to see if it would correlate well with mortality. They found that the model was significantly better at predicting mortality than a model using demographic predictors like sex and ethnicity. The first quartile (the quarter of participants with the lowest CT-based biological age) were almost all still alive 15 years after their CT scan. People in the next quartile up still had an approximately 95% probability of survival, while the third quartile had a survival rate of about 80%. The last quartile – the 25% of people with the highest CT-based biological age – had a 15-years survival rate of only 40%.

The implications:

This study suggests that there is untapped potential to use CT scan data to predict mortality and warn people who may not be aware that they are at risk, and to intervene to address any modifiable causes. This is appealing given that it is quite common for middle aged people to undergo CT scans for one reason or another, and advancements in machine learning mean that a biological age estimate could be produced from these scans without any additional work from a clinician.