Caristo Diagnostics is a spin-off from the University of Oxford
Current methods for predicting heart attacks are woefully outdated, says Cheerag Shirodaria, a cardiologist at the Oxford University Hospitals NHS Foundation Trust, UK. “We basically look at a patient’s age, obesity and whether they have diabetes. The most sophisticated we get is measuring their cholesterol,” he says.
Heart attacks are often caused when fatty accumulations inside arteries, known as plaques, result in the narrowing of blood vessels. These plaques then rupture and form blood clots. Age, being overweight or obese and having diabetes are all risk factors, and people in at-risk groups are closely monitored. “But 50% of heart attacks are in people without narrowing, and they’re being missed by current tests,” he says.
A desire to drag prognosis into the modern era is what pushed Shirodaria and a group of like-minded cardiologists to set up Caristo Diagnostics. Rather than blood-vessel narrowing, the team focuses on another driver of plaque ruptures — vascular inflammation. The company uses algorithms to analyse information that until now has been hidden in the noise of computed tomography (CT) scans. This enables it to assess inflammation and calculate the likelihood that a person will have a heart attack in a given time period.
“We’ve always been looking for something like this,” says Nehal Mehta, senior investigator at the US National Institutes of Health’s section of inflammation and cardiometabolic diseases in Bethesda, Maryland, and who is not affiliated with Caristo. “I always take new medical technologies with caution, but this is a big deal.”
Although Caristo is less than two years old, its technology is backed up by decades of research by its founders. “They’ve done their homework,” says Angela Kukula, director of enterprise at The Institute of Cancer Research in London. “It’s not something they’ve dreamed up on the back of an envelope.” Kukula was on the judging panel for the Nature Research Spinoff Prize, but stepped down after being offered a role at the University of Oxford.
The reason it’s taken so long to turn the research into a commercial enterprise is that the technology is based on findings that initially challenged the scientific status quo. “When you discover something that goes against the prevailing science, then you need to be 100% sure that the biology is correct before you move on,” says Charalambos Antoniades, a cardiologist at the University of Oxford and Caristo’s chief scientific officer.
Fat tissue bordering blood vessels, known as perivascular fat, had long been thought to be a cause of inflamed arteries. “There was this perception that perivascular fat was the bad guy,” says Antoniades. That’s why his research started out as a hunt for any potential chemicals produced and sent by the fat to the blood vessels — the working theory used to be that such chemicals could have been behind the problem.
“But the results always showed the opposite,” he explains. The signals were instead going the other way. Antoniades showed that when an artery is inflamed, it sends out SOS signals that are then picked up in the surrounding fat1. In response, the fat undergoes lipolysis — the lipids break down and the water content in the fat cells increases (see ‘Sensing a heart at risk’). That effectively means that diminished perivascular fat is an indicator of inflammation, and therefore also of an increased risk of heart attack. Other scientists have validated the link between perivascular fat abnormalities in CT scans and coronary inflammation2.
“It took seven years for that idea to mature and to convince the scientific community and ourselves that the biology is real,” says Antoniades. “Having to change the textbook is a big challenge.”
That is what impressed Kukula. “The innovation in my mind was them making the connection that differences in the fat cells means something,” she says. “That alone is interesting, but they’ve taken it a step further.”
At the core of Caristo’s innovation is its method for detecting and visualizing the water content of perivascular fat cells by looking for them on CT scans.
“Millions of scans happen each year, so why didn’t anyone notice this before?” says Antoniades. The reason, he says, is that manufacturers of CT scanners have, for many years, focused on “creating clear pictures for their scans by suppressing artefacts — and the fat has always been considered as an artefact.” To see the fat, Caristo’s technology reverse-engineers CT scans to bring perivascular fat to the surface of the image.
Caristo’s team tested its technology in a study of almost 4,000 people who had undergone a CT scan and were followed for up to nine years after the examination. The results showed that — independently of other risk factors — people with abnormal perivascular fat are six to nine times more likely to die of a heart attack3.
“They’ve shown in thousands of people that fat attenuation around the arteries corresponds with risk,” says Mehta. “It’s groundbreaking.”
Despite these encouraging results, there are still hurdles that Caristo must overcome. “The next challenge is going to be getting cardiologists to understand how much information is hidden in CT scans,” says Shirodaria. There’s also the issue of how to monetise the technology. Once it gets regulatory approval, Caristo intends to sell its software as a service — hospitals would send CT scans to Caristo, which would do the analysis. “It fits well with the cardiologist’s workflow,” he says.
But that model isn’t the norm, warns Kukula. “It’s not the way the NHS in the United Kingdom tends to work; they like to buy software,” she says, referring to the country’s health system. “There’s a question of how confident they can be in delivering it as a service.”
But Shirodaria says Caristo’s technology will ultimately save health services money. “You’re identifying at-risk patients and preventing the costs of a downstream heart attack,” he says. “Heart attacks are very expensive.”
Margaritis, M. et al. Circulation 127, 2209–2221 (2013).
Kwiecinski, J. et al. JACC Cardiovasc. Imaging 12, 2000–2010 (2019).
Oikonomou, E. K. et al. Lancet 392, 929–939 (2018).