Study protocol for Rule-Out Acute Myocardial Infarction Using Artificial Intelligence Electrocardiogram 2 (ROMIAE-2) in patients with chest pain: a multicenter randomized noninferiority trial.
文献摘要
Objective Acute myocardial infarction (AMI) requires rapid identification in the emergency department (ED). Although artificial intelligence (AI)-based electrocardiogram (ECG) interpretation systems have shown strong diagnostic accuracy, no multicenter randomized controlled trial (RCT) has evaluated their clinical impact. This protocol describes a RCT for assessing whether real-time AI-ECG can be safely integrated into routine emergency care for undifferentiated chest pain. Methods ROMIAE-2 is a prospective, multicenter, randomized, open-label trial with blinded endpoint adjudication conducted at 12 academic tertiary EDs in South Korea. A total of 4,670 adults with acute chest pain will be randomized 1:1 to AI-ECG-guided or standard care. In the intervention arm, a deep learning algorithm (AiTiAMI Ver 2.00.00) will analyze the initial 12- lead ECG and generate an AMI probability score, which is combined with the HEART score to classify patients into four integrated risk categories for clinical decision-making. The primary outcome is 30-day major adverse cardiovascular events (MACE), assessed using a noninferiority framework with a 3.0% margin. The trial will be powered at 80% with a onesided significance level of 0.025. Secondary outcomes include missed AMI at the index visit, ED length of stay, early discharge, 7-day ED revisit, hospital admission, hospital length of stay, AI-ECG diagnostic performance, additional testing, and direct medical costs. Clinical endpoints, including MACE and AMI, will be adjudicated by two emergency medicine specialists blinded to the group allocation. Discussion ROMIAE-2 will provide randomized evidence on whether integrating AI-ECG into routine emergency care can preserve patient safety while improving efficiency in patients with undifferentiated chest pain.