Key Takeaways: AI-Guided Decision for LAAO vs. DOAC in AFib Patients

Key Takeaways: AI-Guided Decision for LAAO vs. DOAC in AFib Patients

(JACC: Clinical Electrophysiology, Feb 12, 2025)

1- AI Identifies Optimal Candidates for LAAO

• Researchers at Mayo Clinic developed an artificial intelligence (AI) algorithm to identify atrial fibrillation (AFib) patients who may benefit from left atrial appendage occlusion (LAAO).

• The study was published in JACC: Clinical Electrophysiology.

2- Challenge: Lifelong Anticoagulation Therapy

• Direct oral anticoagulants (DOACs) are preferred over warfarin for stroke prevention in AFib.

• However, lifelong anticoagulation increases bleeding risk and has poor adherence.

• LAAO offers an alternative, but selecting the right candidates remains unclear.

3- Study Design & AI Model Development

• Researchers analyzed data from 744,000 AFib patients (2015–2019), focusing on the 1.9% treated with LAAO due to high stroke risk.

• Propensity score matching was used to compare 14,000 LAAO patients with a similar DOAC-treated cohort.

• AI models, based on causal forest (CF) techniques, were developed to predict whether LAAO would be beneficial, neutral, or harmful.

4- External Validation Confirms AI Accuracy

• The model was validated using data from 377 LAAO patients and 26,000 DOAC patients (2016–2021).

• Propensity score matching resulted in 371 comparable patient pairs.

• The model confirmed that patients identified as likely to benefit from LAAO indeed had higher stroke risks.

5- Key Patient Characteristics Favoring LAAO

• Older patients and those with comorbidities—dementia, pneumonia, respiratory failure—benefited more from LAAO over DOAC therapy.

6- Clinical Implications

• This is the first study to apply causal machine learning to a national database for predicting treatment effects in AFib patients.

• The AI algorithm can assist general practitioners and non-specialists in identifying patients for further evaluation for LAAO.

7- Limitations & Considerations

• The AI model’s prediction is one factor in decision-making.

• Other elements—patient preferences, clinical outcomes, cost, and timeframe—should also be considered.