AI-ECG strategy identifies patients at risk for AF

According to one of the first prospective studies to evaluate such an approach, artificial intelligence (AI) guided screening for atrial fibrillation (AF) based on ECGs can successfully identify patients who are more likely to have the disease. arrhythmia detected on subsequent monitoring.

Patients labeled by the AI-ECG algorithm as high risk versus low were significantly more likely to have an episode of AF lasting at least 30 seconds identified during subsequent ambulatory monitoring (7, 6% versus 1.6%; OR 4.98; 95% CI 2.22-11.75), with similar results when longer duration AF episodes were used, report researchers led by Peter Noseworthy , MD (Mayo Clinic, Rochester, MN).

The study, published online this week in the Lancetbuilds on the group’s previous work by leveraging millions of ECGs stored at the Mayo Clinic to derive AI-ECG algorithms for various disease states such as long QT syndrome and COVID-19[feminine]. What makes this study unique is its prospective design.

“It’s one thing to say that we can distinguish patients who appear to have a propensity for atrial fibrillation from those who don’t on the 12-lead ECG. It’s another thing to say that we can identify high-risk patients and they actually have a higher rate of incident atrial fibrillation diagnosed during prospective cardiac monitoring and follow-up,” Noseworthy said. at the TCTMD.

The study was designed to evaluate the AI-ECG approach as it would be used if it were eventually adopted in clinical practice, he said, noting that participants had already undergone ECGs for many reasons. not necessarily related to the FA and were then contacted to participate. in screening with telemonitoring.

“We were able to take a population of patients with existing cardiovascular risk factors, who are likely already at relatively high risk for atrial fibrillation, and then demonstrate that we can further stratify that population using ECG,” explained Noseworthy. “So not only is the ECG a marker of these comorbidities that drive the risk of AF, but it also offers an additional prognostic benefit on top of existing clinical risk factors. This means we can identify patients who have risk factors for stroke should they be diagnosed with atrial fibrillation and target screening to those who we believe would benefit the most.

We can identify patients who have risk factors for stroke if they are diagnosed with atrial fibrillation and target screening to those we think would benefit the most. Pierre Noseworthy

Noseworthy and his colleagues at the Mayo Clinic reported about 3 years ago that their AI algorithm could identify signs of AF on ECGs taken during normal sinus rhythm, potentially finding a subset of the population that should be targeted for AF screening. But it was unclear whether the AI-ECG approach improved risk stratification relative to clinical factors alone and whether it uncovered cases of AF that otherwise would not have been detected by routine clinical practice.

To answer these questions in the current study, titled Batch Enrollment for an AI-Guided Intervention to Lower Neurologic Events in Patients with Undiagnosed Atrial Fibrillation (BEAGLE), Noseworthy and colleagues prospectively recruited 1,003 patients (mean age 74 ; 38.2% women) who had risk factors for stroke but no known AF and who had a routine 12-lead ECG.

The AI ​​algorithm divided the participants, who came from 40 US states, into high- and low-risk groups based on ECGs. All received a continuous ambulatory heart rate monitor, which was worn for up to 30 days (mean 22.3 days).

Surveillance showed that patients identified as having a high risk of AF were indeed more likely to have the arrhythmia than those classified as having a low risk. The high-risk group also had a significantly higher AF burden (mean 20.32% vs. 4.97%; P = 0.016), with no difference between risk groups in terms of longest AF episode or time to AF diagnosis. About three-quarters of patients with a new diagnosis of AF and available clinical follow-up data started anticoagulation.

In a secondary analysis, investigators assessed whether the AI-ECG strategy increased detection of AF compared to usual care through a median follow-up of 9.9 months. And it does, but only for the high-risk group (10.6% versus 3.6%; P < 0.0001).

“To our knowledge, our study is. . . the first to assess the effectiveness of the AI-guided targeted screening program compared to usual care, which may inform the design and implementation of large-scale atrial fibrillation screening programs,” write Noseworthy and para.

Some limitations, clinical perspectives

Commenting for the TCTMD, Jagmeet Singh, MD, PhD (Massachusetts General Hospital, Boston), said that as one of the first prospective studies evaluating risk stratification with AI-ECG, “it adds a lot” to the literature on AF screening.

The algorithm “helps stratify the risk of a population at relatively high risk for stroke, and then helps quantify whether or not that risk stratification was appropriate,” Singh said. “It also gives an idea of ​​which patients to really focus on. Because A-fib is such a common arrhythmia that is increasing in prevalence, in order to have a scalable risk stratification strategy, you need something that is relatively easy to do and is done fairly regularly.

Singh, however, had some reservations about interpreting the study. It is important to note that there is a question of generalizability, as the AI-ECG algorithm was derived from a Mayo Clinic population that includes a very high proportion of white individuals; in this study, less than 4% of participants were non-Caucasian. There is also a question about the impact of digital literacy on the interpretation of the results, he said, pointing to the fact that the study included 1,003 people out of more than 15,000 initially invited.

“I think AI-based algorithms should be consumed with some caution because of the dataset they come from and the population they’re ultimately studied in,” Singh said.

The researchers conducted an “elegant and pragmatic” study, but a randomized study with a larger cohort and longer follow-up would help address some of these issues, he said, adding that studies evaluating surveillance strategies longer term beyond 30 days would be useful. to make sure the AF is not missed. “You really want to know if they can end up getting A-fib over the next few months or years,” Singh said.

As the next step before clinical implementation of the AI-ECG approach and approval from the US Food and Drug Administration, Noseworthy said there is a need to conduct a large, multi-center trial showing that patients diagnosed with AF with screening can be anticoagulated or otherwise treated to reduce their risk of stroke. “Ideally, we’d like to get this algorithm into the hands of as many people as possible to run these kinds of screening programs once we’ve demonstrated that it makes a difference to patients’ lives and their risk of stroke.” , did he declare.

Conducting this type of trial will be financially challenging, Singh said, but he suggested that a strategy like the one studied here, perhaps combined with others, could have a clinical role in the future. coming.

“We should adopt machine learning approaches to examine clinical covariates in more elegant ways to predict the risk of A-fib. Beyond that, the addition of the Mayo Clinic AI-ECG strategy can further add value to the algorithm to stratify patient risk even better,” Singh said.

The AI-ECG algorithm “actually creates an opportunity to provide remote monitoring to these patients, using not just wearable devices or patch monitors, but even implants,” he suggested, indicating the importance of “monitoring them for longer periods of time so you can pick up A-fib and in turn also show the impact on clinical outcomes.