Patients undergoing weight loss (bariatric) surgery can rely on surgeons and medical professionals to explain the exact process of the operation from beginning to end. However, the post-surgery journey to achieving their ideal weight is completely unknown. Although some tools have tried to estimate post-surgery weight loss, these tools only show the total number of kilos a patient can expect to lose, without showing the journey.
“It's not a linear trajectory,” says Francois Pattou, professor of surgery at the University of Lille. He explains that patients losing weight after surgery may experience periods of fast or slow weight loss, or even slight weight gain in places.
“When it’s non-linear, it’s difficult to predict. It was always difficult,” he says. This in turn can lead to patients feeling unmotivated to lose weight, or feel that the surgery was ineffective.
Prof. Pattou and others in the Innovative Medicines Initiative project SOPHIA felt that a better tool was needed to help patients predict their weight loss trajectory ahead of bariatric surgery.
The AI-driven tool they developed now allows patients and their doctors to accurately predict weight loss over five years. The tool is freely available online for everyone, and its development is detailed in a recent a paper in The Lancet Digital Health. By filling in information such as the patient’s height, weight, age and type of surgery, the tool can generate a graph predicting their weight trajectory over five years following surgery.
The unique trajectory curve the tool generates means that, for example, patients may see that their post-surgery weight loss is first dramatic, then levels off as they reach the ideal weight loss goal.
Gathering information
To collect the data to design the tool, the project used information volunteered by patients from 10 cohorts across 8 counties in Europe, America, and Asia. These patients had undergone some type of bariatric surgery and were monitored through check-ups over five years. Patient data from two randomised trials acted as a control. In total, information from 10 231 patients was used to find a way to predict post-surgery weight loss over five years.
“The final tool is simple, but the information behind was quite a lot of work,” said Prof. Pattou. “You had to validate everything globally across 10 cohorts.” He explains how their first analysis generated over 400 baseline characteristics that affected weight loss, though creating a tool with hundreds of variables would be unusable for both doctors and patients.
However, by using a machine learning algorithm called LASSO they were able to reduce the amount of viable variables to just seven: height, weight, type of operation, age, diabetes status, diabetes duration, and smoking status.
While other factors do affect weight loss over these five years, Prof. Pattou says these seven variables strike the right balance between making the tool usable and accurate.
“You cannot have a precise metric, because it's always subject to the environment and many other variables,” he said. “But it’s at least efficient enough to be to be clinically viable and precise enough for the patient.”
Peace of mind for patients
The SOPHIA project also includes patient groups who helped design the ‘front end’ of the tool, for example suggesting to remove or clarify medical jargon. The end result is a tool that the authors claim offers the most accurate prediction of weight loss for bariatric surgery patients.
There are two ways to use it with patients, says Prof. Pattou; one was expected and one was discovered later on.
The first is for patients who have not yet had surgery a model showing the trajectory of their weight loss.
“As a patient when you see a trajectory, that makes it much more concrete,” said Prof. Pattou. It also allows healthcare providers to clearly explain any anticipated periods of slower weight loss or even weight gain. “This type of discussion you can have with your patients, where you're seeing the same curve at the same time, makes it easy and concrete.”
The second benefits was discovered post-surgery. The tool helps reassure those who might notice weight gain during post-surgery, and also help clinicians spot patients whose weight falls out of the curve – for example patients who lose too much weight. Primary care clinicians can also use the tool to easily monitor the patient’s progress.
“When you have these tools that any one of us can just take two minutes to understand, you become an actor,” he said, “you have something to discuss with them.”
The tool has just launched in the past two months as the project is approaching the finish line. Prof. Pattou says that they are now working on validating the tool by contacting 10 other cohorts in the SOPHIA project. In the end, he hopes that clinicians using the tool to help their patients lose weight can become the legacy of the SOPHIA project.
SOPHIA is supported by the Innovative Medicines Initiative, a partnership between the European Union and the European pharmaceutical industry.