In 14,665 adults, an AI electrocardiogram flagged slightly more atrial findings in 90 days, and one swapped event would erase the result

What the research found

Circulation published a single-center open-label randomized trial on 7 November 2026. Chiao-Chin Lee, Wen-Yu Lin, and colleagues enrolled 14,665 adults ages 40 to 85 with no prior atrial fibrillation or flutter. 7,337 were assigned to an AI electrocardiogram pathway and 7,328 to usual care. High risk was called in 84 people (1.1 percent) versus 78 people (1.1 percent). Of the 84 high-risk people in the AI arm, 48 (57.1 percent) were monitored, and 48 people matched on age and sex from the usual-care arm were monitored the same way. The primary outcome, new atrial fibrillation, atrial flutter, or other atrial findings that can precede atrial fibrillation, within 90 days, occurred in 54 people (0.74 percent) versus 35 people (0.48 percent). The odds ratio was 1.54 (95 percent CI 1.01 to 2.37, P = 0.046). Moving one AI-arm event into the non-event column makes P = 0.057. The result differed by baseline risk (interaction P < 0.001): 32.1 percent versus 2.6 percent in the high-risk group, and 0.4 percent versus 0.5 percent in the low-risk group. Atrial fibrillation or flutter itself was found in 0.3 percent versus 0.2 percent. Starts of antiarrhythmic therapy at 180 days were 2.3 percent versus 2.4 percent. Among people who were monitored, skin or discomfort events were 22.9 percent versus 18.8 percent. Registration NCT06847932. The positive result is fragile. DOI 10.1161/CIRCULATIONAHA.126.083540. Not medical advice.

Evidence level: Clinical trial. This is among the stronger kinds of evidence, but one study rarely settles a question.

Read the original at Circulation