Detecting Transthyretin Cardiac Amyloidosis With Artificial Intelligence: A Nonrandomized Clinical Trial

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Diagnostics

Background:

• Transthyretin amyloid cardiomyopathy (ATTR-CM) is frequently underdiagnosed, limiting timely treatment initiation.
• Expanding therapeutic options necessitate improved diagnostic strategies for ATTR-CM.

Purpose of the Study:

• To develop and evaluate an artificial intelligence (AI)-augmented clinical program for enhanced ATTR-CM detection.
• Assess the real-world performance of the AI model, ATTRACTnet, in identifying eligible patients for ATTR-CM testing.

Main Methods:

• An AI model, ATTRACTnet, was constructed using ECG, echocardiography, demographics, and diagnosis codes.
• A single-system, multisite, open-label trial evaluated ATTRACTnet's performance in identifying patients with left ventricular wall thickness ≥12 mm and ATTRACTnet score ≥0.5.
• Eligible patients were offered nuclear scintigraphy and monoclonal protein testing.

Main Results:

• ATTRACTnet demonstrated good discrimination for ATTR-CM detection with an area under the receiver operator characteristic curve of 0.85 in the internal set and 0.82 in the external set.
• Of 256 eligible patients, 50 underwent testing, leading to 24 (48%) diagnoses of ATTR-CM, with 21 (88%) initiating treatment within 3 months.
• The ATTR-CM positivity rate was 2.8 times higher than historical controls, representing an 18% relative increase in new diagnoses.

Conclusions:

• AI-augmented screening can significantly improve ATTR-CM detection rates compared to usual care.
• This approach may identify patients with ATTR-CM who are currently missed, facilitating earlier intervention.
• Further prospective randomized trials are warranted to confirm improved patient outcomes.