Adverse Event Landscape of Pulsed Field Ablation Catheters: A Computational Review of FDA MAUDE Post-Market Surveillance Data.

Abstract

Background: Pulsed field ablation (PFA) is Food and Drug Administration (FDA)-cleared for pulmonary vein isolation in atrial fibrillation. Using a keyword-based natural language processing pipeline applied to free-text narratives in the FDA MAUDE database, we aimed to characterize and compare adverse event and device malfunction profiles across the four cleared PFA systems.

Methods: We retrieved 6930 MAUDE reports for FDA product code QZI (pulsed field ablation catheter) spanning December 13, 2023 through April 30, 2026, covering all four FDA-cleared PFA systems. Reports were deduplicated and assigned to a specific device system using manufacturer and brand name fields. Concatenated free-text narratives were classified into 11 predefined adverse event categories using multi-label keyword-based natural language processing. Device malfunction reports were additionally subclassified into one of 11 hardware-specific subtypes. Descriptive statistics were used throughout.

Results: Of 6930 reports, 4185 (60.4%) were classified into at least one adverse event category. The three most common categories were device/catheter malfunction (2068; 29.8%), cardiac tamponade/pericardial effusion (915; 13.2%), and stroke/TIA/cerebral embolism (600; 8.7%). PFA-specific physiologic adverse events-ST-elevation myocardial infarction (STEMI), hemolysis/acute kidney injury, and phrenic nerve palsy-accounted for 615 combined reports (8.9% of all reports). Fatality proportions were highest for vascular access complications (6.8%), esophageal/atrioesophageal injury (6.0%), and coronary artery spasm/STEMI (5.3%). Among malfunction reports, retraction/retrieval failure predominated for Farapulse (48.9%), while electrical/energy delivery failure was the leading subtype for PulseSelect (29.4%).

Conclusions: Natural language processing (NLP)-assisted keyword classification enables systematic, scalable characterization of adverse event patterns across nearly 7000 real-world PFA reports. PFA-specific physiologic adverse events are identifiable through post-market surveillance data and constitute a distinct safety phenotype separate from device malfunction. This approach provides a reproducible framework for near-real-time surveillance of novel cardiac ablation systems.