AI-Assisted Screening May Enhance Detection of Transthyretin Myloid Cardiomyopathy: JAMA
USA: AI-augmented screening shows promise in improving the identification of patients with transthyretin amyloid cardiomyopathy (ATTR-CM), potentially detecting cases missed by standard clinical care. However, prospective randomized trials are needed to confirm whether this leads to better patient outcomes.
In a new study published in JAMA Cardiology, researchers led by Dr. Sneha S. Jain from the Division of Cardiovascular Medicine and the Cardiovascular Institute, Stanford University, Palo Alto, California, and colleagues, evaluated an artificial intelligence (AI)-based clinical program designed to enhance the detection of ATTR-CM—a condition that remains significantly underdiagnosed despite advances in treatment.
The team developed and tested a machine learning model named ATTRACTnet, which utilized electrocardiogram (ECG) waveforms, echocardiographic data, demographic details, and diagnostic codes linked to orthopedic manifestations of amyloidosis. The study was a nonrandomized, multisite clinical trial conducted within a large healthcare system to assess how well this AI model could identify potential cases of ATTR-CM in a real-world setting.
Eligible patients included those with a left ventricular wall thickness of 12 mm or more and an ATTRACTnet score of 0.5 or higher. Exclusion criteria involved patients with prior ATTR-CM testing, hypertrophic cardiomyopathy, limited life expectancy (under one year), or advanced dementia. Once identified, eligible participants were offered nuclear scintigraphy and monoclonal protein testing following physician approval.
Key Findings:
- The AI model identified 1,471 patients as potentially positive for ATTR-CM.
- Of these, 256 met the inclusion criteria, and 50 underwent further amyloidosis testing.
- Among those tested, 24 patients (48%) were confirmed to have ATTR-CM.
- Treatment was initiated in 21 of these diagnosed patients (88%) within three months.
- The diagnostic positivity rate was 2.8 times higher than historical control groups, showing a marked improvement in detection.
- The program also achieved an 18% relative increase in new diagnoses compared to the previous year.
- The AI model, ATTRACTnet, demonstrated strong diagnostic performance with an area under the ROC curve of 0.85 in internal testing and 0.82 in external validation.
- Its accuracy was consistent across racial and ethnic groups, including Hispanic, non-Hispanic Black, and non-Hispanic White patients.
However, the authors acknowledged certain limitations. ATTR-CM’s underdiagnosis makes it difficult to create large, diverse training datasets, though this study used one of the largest so far. Its single-system, single-arm design may restrict generalizability, and prioritizing specificity over sensitivity (0.56) meant that some potential cases were likely missed.
Despite these limitations, the study highlights the potential of AI-driven clinical programs to address diagnostic gaps and identify patients who might otherwise remain undiagnosed. According to the authors, integrating AI models like ATTRACTnet into clinical workflows could transform early detection strategies and facilitate timely intervention for cardiac amyloidosis.
"The findings suggest that AI-assisted screening may enhance the detection of ATTR-CM across diverse patient populations, paving the way for broader adoption of machine learning tools in cardiovascular diagnostics. Future large-scale, randomized trials are warranted to confirm these promising results and assess their impact on long-term patient outcomes," the authors concluded.
Reference:
Jain SS, Sun T, Pierson E, et al. Detecting Transthyretin Cardiac Amyloidosis With Artificial Intelligence: A Nonrandomized Clinical Trial. JAMA Cardiol. Published online November 10, 2025. doi:10.1001/jamacardio.2025.4591
Next Story