LAAO or DOAC? Mayo Clinic AI helps guide treatment of AFib patients
Researchers have developed a new artificial intelligence (AI) algorithm that can identify atrial fibrillation (AFib) patients who may benefit from left atrial appendage occlusion (LAAO). The group shared its findings in JACC: Clinical Electrophysiology.[1]
“Lifelong oral anticoagulation is recommended for stroke prevention in most patients with AFib, and clinical guidelines recommend direct oral anticoagulant (DOAC) in preference to warfarin in most patients,” wrote first author Che Ngufor, PhD, an AI specialist with Mayo Clinic, and colleagues. “However, oral anticoagulation increases the risk of bleeding and the adherence to lifelong drug therapy is poor, leaving many patients undertreated. Transcatheter LAAO offers an attractive alternative to lifelong drug therapy, but how to select the best candidates for LAAO remains unclear in everyday clinical practice.”
Ngufor et al. started with data from 744,000 AFib patients who underwent treatment from 2015 to 2019, focusing on the 1.9% who were treated with LAAO due to a heightened stroke risk. Propensity score matching was used to land on two pairs of more than 14,000 patients; one pair was treated with LAAO, while the other pair was treated with DOAC. The mean patient age was 76.8 years old.
The group used these data to develop several causal forest (CF) models for estimating the best treatment option when patients present with AFib. The CF model that performed the best was able to accurately determine when LAAO would have a positive impact on the patient, a neutral impact or a negative impact.
For external validation of that model, the researchers then explored data from 377 AFib patients who underwent LAAO and more than 26,000 AFib patients prescribed DOAC therapy from 2016 to 2021. The group then used propensity score matching once again to land on two comparable pairs of 371 patients.
External validation confirmed the AI model’s effectiveness, showing that patients “deemed more likely to benefit by the algorithm indeed exhibit higher risks.” Older patients and those who present with more comorbidities—particularly dementia, pneumonia or respiratory failure—were linked to increased benefits from LAAO compared to DOAC.
“The current study is the first to apply a novel causal machine learning framework to a large national database to develop algorithms to predict heterogeneous treatment effect in patients managed in routine clinical practice,” the authors wrote. “The algorithm developed from this study can help clinicians, especially those out of subspecialty care, to select patients to refer for further examination and discussion of LAAO.”
They also emphasized that predicted treatment effect is “just one piece of the information that needs to be discussed” when making care decisions.
“Other factors, such as outcomes, timeframes, patients’ values, preferences, and costs, also play crucial roles,” Ngufor et al. wrote.
Click here to read the full analysis in JACC: Clinical Electrophysiology, an American College of Cardiology journal.
Michael has more than 18 years of experience as a professional writer and editor. He has written at length about cardiology, radiology, artificial intelligence and other key healthcare topics.