“What does that mean for me?”

“When you start a conversation with a patient about the best treatment for cardiovascular disease – and in particular, how to prevent it – the medical advice is often quite invasive,” Visseren says. “Especially because a patient may not be feeling sick at all yet. You want to keep it that way, so you recommend a change in lifestyle and possibly prescribe cholesterol or blood pressure reducers or anticoagulants. In doing so, you can refer to large scientific studies and to the average effects for the average patient, but when someone asks in the consultation room, ‘What does that mean for me?’ you as a physician are at a loss for words.”

Predicting individual effects

“Everyone is different, but the way we practice medicine is still kind of one size fits all.” That observation by Visseren led him to ask whether it would be possible to estimate the effect of treatments on individual people. “Together with Jannick Dorresteijn [Internist-vascular medicine at UMC Utrecht, ed.] and colleagues from Harvard, we then developed a methodology to predict the effect of individual interventions and express it in the number of healthy years of life, free of cardiovascular disease, someone gains. We can do that for people who are healthy, people who have had a cardiovascular problem previously, and people with diabetes.”

This way of working – thinking not only in terms of risk, but in terms of solutions – has found a place in the new European directive that came out last August. “If you can express the health gains in terms of years, you get a very different conversation in the consultation room. Everyone understands that and can assess with their doctor on that basis whether it is of sufficient benefit, for example, to start treatment. It says much more than saying that a ten-year cardiovascular disease risk drops from, say, 18 to 16 percent if someone starts taking medication. Patients also watch the result on the screen. It is very motivating when you see the graphs about your future health, and the effects of lifestyle changes. In this way, patients themselves are more in the driver’s seat. Moreover, the result of this way of working is that you start treating people when it has an effect, and don't treat them when medication has little or no effect. All of the studies show that this approach is cost-effective, and it helps to live a healthy life for as long as possible. I invest in my retirement finances by building up a pension. You also want to invest in a healthy old age.”

Lifetime prediction

Without going too deep into the underlying methodology, Visseren explains how U-Prevent came about: “We have brought together clinical, epidemiological, and statistical knowledge. And above all a ton of data, from which we derive predictions that we then validate in independent cohorts. In personalized estimates of risk and of treatment effects, we use the best of two worlds, namely the world of prediction – we can predict lifetime – and the trial world in which the effects of drugs are investigated very carefully. Based on the gold trial standard – placebo-controlled, double-blind, and randomized – an average effect is obtained. You can link this average relative trial effect to an absolute individual risk of cardiovascular disease. The results produced then indicate how much absolute health gain you are achieving. Lifetime prediction has to be done based on solid data.” Visseren himself is not a big believer in Big Data in the sense that lots of data is always good, he says: “A lot of garbage is still garbage. If you put a lot of low-quality data together you still have low-quality data. Above all, I want good data, and preferably a lot of it. The trick is to gather this type of high-quality data. You need good baseline measurements and good follow-up of participants, because you can’t estimate the value of a given variable until you know how patients are doing. You won’t know for five or ten years whether a particular risk factor or biomarker actually increases the risk of cardiovascular disease. It would be great if we in the Netherlands, as in Scandinavia, had many more national registries, so that the Netherlands could be one big cohort. This can be done anonymously, with safeguards for privacy, and if you don’t want to you can indicate that, but essentially everyone participates. That would benefit us, as the Netherlands business community, a great deal."

Trusting in the ‘black box’

In addition to collaborating with universities and data centers worldwide, Visseren and his colleagues at U-Prevent have also partnered with ORTEC. Why is that? “As scientists, we identify a problem, try to contribute to the solution, validate it, and publish that. In the case of U-Prevent, we created a website on which we posted the algorithms to make them available to the rest of the world. But we soon realized that we needed a partner who knows how to handle data in a secure and reliable manner. ORTEC also brought explainable AI to our attention. For many people, prediction is still something of a ‘black box’: how can you trust an outcome if you don’t know what will happen? A CE certification is fine, but you also need to help people understand it. And if U-Prevent is to be successful in practice, it must be improved. The algorithms are based on relatively simple data such as a patient’s age, gender, cholesterol level, and blood pressure. If you can load this data automatically from the electronic patient record, it saves a lot of time and input errors: a doctor then does not have to spend time on this in a ten-minute consultation. We want to move towards that. The link with the EHR is already technically possible, it's just a matter of making good agreements.”

However, many people have cold feet, even among hospitals. “That is indeed the biggest hurdle we have to overcome. A copy of the data is used briefly outside “the hospital” to carry out calculations, results are shown, and then the data disappears from the Cloud. This can be done in a safe way, however, this is still unknown territory for many hospital administrators. You have to look for cloud solutions if you want to make this scalable and keep costs under control. Healthcare providers place a great deal of information into the EHR, but we then do little to use all that information to make the best decisions for that particular patient. You need calculations for that. There are so many variables to consider; this goes beyond your own calculating skills. The final decision about a treatment must be made by the healthcare professional together with the patient, and that conversation should be fueled by smart individual information. In that way, you make the decision-making process more personal.” Visseren would prefer to see U-Prevent become a self-learning system that will be used as a basis in various places in the Netherlands, but also worldwide, and would like to see the algorithm be fine-tuned based on data available locally. “To make even better predictions, you must adapt the prediction rule to the situation on a continent, in a country, or in a region. For example, particulate matter in exhaust gases may cause the condition of patients in urban and rural areas to differ. Even within the city of Utrecht – simply due to referral patterns – there can be a difference in patients at Diakonessenhuis and UMC, even though we are only a few kilometers apart.” In any event, in his view, U-Prevent should be widely implemented in the coming years: “This is the way to go. Everyone thinks this is a good idea, and U-Prevent is included in the 2021 European Directive for Cardiovascular Prevention. Good for patients, good for health care providers, and good for health insurance companies.”

About Frank Visseren

Frank Visseren is an internist, epidemiologist and professor of vascular medicine at Utrecht University Medical Center. As head of the vascular medicine department, he is responsible for patient care, research, and education. Visseren is particularly involved in research on patients who are at increased risk of cardiovascular disease and/or insulin resistance. Together with national and international partners, he translates the results of large, clinical trials to individual patients. In addition, Visseren is Principal Investigator of the SMART (Second Manifestations of ARTerial disease) cohort, with over 14,000 patients with vascular disease, diabetes, or other serious risk factors. Visseren is the author and co-author of more than 400 publications, and associate editor of two scientific journals.

About U-Prevent

U-Prevent is an innovative application of UMC Utrecht and ORTEC in the field of cardiovascular risk management. U-Prevent calculates the 10-year and lifetime risk of cardiovascular disease of individual patients and estimates individual treatment effects. In this way, U-Prevent supports the conversation between healthcare provider and patient to reach the best decision, and the healthcare provider and patient work together to prevent cardiovascular disease. ORTEC owns U-Prevent and the platform under U-Prevent for risk prediction models, including in other medical domains. ORTEC has completely re-engineered the current U-Prevent to meet the requirements and regulations for use in clinical practice.