BridgeBio (BBIO) Corporate presentation summary

Study overview and objectives

• Nationwide, multicenter deployment of AI for multimodal screening of ATTR cardiomyopathy across the US, leveraging federated learning and format-agnostic inference software run locally at multiple health systems.

• Objectives included validating diagnostic performance of AI algorithms, examining association with adverse cardiovascular events, and estimating prevalence of probable ATTR cardiomyopathy.

Screening methods and AI model performance

• AI-ECG and AI-Echo models detect cardiac amyloidosis patterns imperceptible to clinicians, with AI predictions diverging up to 3 years before clinical diagnosis.

• Diagnostic performance measured by AUC: pooled cohort AUC for AI-ECG was 0.769 and for AI-Echo was 0.882, indicating strong predictive ability.

• AI-ECG is less specific, likely identifying a broader range of cardiovascular disease.

Study cohort and deployment

• Over 400,000 individuals aged 50–95 from diverse US health systems participated, with ECG and echo data from the last 5 years.

• Only de-identified phenotypic information was shared, ensuring HIPAA compliance.