Investigators have used machine learning to identify 8 primary factors that increase a patient’s risk of bleeding in the upper gastrointestinal (GI) tract following a heart attack in a study published in the European Heart Journal – Cardiovascular Pharmacotherapy. Though some of these factors were already identified, the study found additional predictors, including smoking, blood pressure, and blood glucose.
Bleeding in the upper GI tract is among the most common bleeding complications following acute myocardial infarction, causing considerable suffering and increasing the risk of death. Bleeding complications also limit the use of antithrombotics, which can worsen the cardiovascular prognosis.
“If we can identify patients at high risk of upper gastrointestinal bleeding following heart attack, doctors will be able to take prophylactic measures to mitigate this risk,” said Moa Simonsson, deputy consultant at Karolinska University Hospital and doctoral student at Karolinska Institutet’s Department of Clinical Sciences, Danderyd Hospital, in a press release. “There are, for instance, drugs that combat bleeding complications, gut bacteria tests that can be used on risk groups and other possibilities for personalized treatment for heart attack patients at high risk of bleeding complications.”
The investigators analyzed data from approximately 150,000 patients with acute myocardial infarction between 2007 and 2016 from the Swedish SWEDEHEART registry. Among study participants, approximately 1.5% suffered GI bleeding within a year of their heart attack and had an increased risk of death and stroke.
The study confirmed several known factors that increase the risk of upper GI tract bleeding, including low levels of hemoglobin, previous upper GI tract bleeding, age, and intensive antithrombotic treatment. It also identified new risk factors using an algorithm, including smoking, blood pressure, blood glucose and previous treatment for stomach disorders, such as ulcers and acid reflux.
“If you combine traditional statistical models with machine learning methods, you can create unique opportunities to find key risk factors for previously unknown cardiovascular events,” said Philip Sarajlic, doctoral student at the Department of Medicine, Solna, Karolinska Institutet, in the release. “This makes it possible for us to make effective use of valuable data from the medical quality registry by taking account of complex relationships between risk factors and outcomes in order to further optimize the current recommendations for patient care.”
The investigators plan to start a major clinical study to investigate the significance of diagnosis and treatment of a common infection in the upper GI tract later this year.
Eight predictors of upper gastrointestinal bleeding after heart attack [news release]. EurekAlert; August 23, 2021. Accessed August 23, 2021. https://www.eurekalert.org/news-releases/925988