The project? ML-Cardiotox. The objective? Predict the risk of cardiotoxicity for a given patient with breast cancer in order to implement personalized monitoring. The two drugs concerned by the study are anthracyclines and anti-HER2, known to be cardiotoxic. They are mainly responsible for heart failure, and widely used in the treatment of breast cancer.
Thousands of women affected each year
Do we have the figures for patients who present with cardiotoxicity? “These figures are relatively little known, because generally the studies which validate anticancer drugs primarily focus on the effectiveness of the drug. Cardiotoxicity is less documented and there is less information on this subject. Thus, to know which patients are experiencing cardiotoxicity, we are forced to turn to real-life data which are more difficult to retrieve, unlike the very formal monitoring implemented in a study. The data to which we have access are therefore very fragmented,” specifies Dr Legallois.
According to these data, the specialist estimates between 1 to 17% the number of women affected by cardiotoxicity one year after treatment. Thus, of the 60,000 women who are diagnosed with breast cancer each year, approximately two thirds will be treated with one of the two drugs. Taking the high range, more than 6,500 women could be affected. “Knowing that part of this cardiotoxicity can occur 5, 10 or even 15 years after treatment,” the cardiologist further specifies.
Risk factors invisible to the human eye
Nearly 600 women will be recruited for this study. They will be followed over a year – according to the usual protocol. The data will then be processed by artificial intelligence algorithms called machine learning. “It is an artificial intelligence tool, increasingly developed in medicine, in order to capture links between different pathologies and risk factors which are generally difficult to identify using traditional techniques,” explains in a video captured by the Heart and Research Foundation Dr Trecy Gonçalves, from the cardio-oncology unit at Lariboisière-Saint-Louis hospital, member of a team specializing in machine learning.
Currently, reinforced monitoring is already recommended for women who present a risk factor for cardiotoxicity; age, hypertension, diabetes, tobacco, etc. Thus, a young woman of 20 years old, with no history or risk factors, will be offered a very light monitoring protocol because her cardiotoxic risk is considered low. “But this dichotomy, overall, doesn’t work very well. The mesh of the sieve for individualized monitoring is quite coarse, and there are many parameters that are not taken into account, regrets Dr. Legallois. We want to develop a tool to better stratify risk and focus our resources on patients who need it, what we call risk personalization. »
Thanks to this study, the cardiologist and his team hope to pinpoint new risk factors for cardiotoxicity. “For example, a sign on a normal electrocardiogram (ECG), invisible to the human eye, could be identified as a risk factor by the AI which will have found this same sign on hundreds of other ECGs”, illustrates the doctor . Appropriate monitoring and treatment to protect the hearts of women at risk can then be put in place even before heart damage occurs.
Giving scratch games to a child: why is it a bad idea?
Source: Destination Santé
Related News :