Artificial Intelligence-Assisted Signal Detection of Medtronic Onyx Coronary Stent Complications: A Post-Market Analysis of the Manufacturer and User Facility Device Experience (MAUDE) Database

Read full text
Authors: Giuseppe Biondi-Zoccai, Marco Failla Mulone, Mattia Gallone, Edoardo Roberto Ginghina, Francesca Murri, Francesco Pezzullo, Ludovica Ruggiu, Giulia Testa, Sibilla Marie Tirabosco, Laura Vacciano, Anna Sirignano, Marco Bernardi
+3 more Luigi Spadafora, Salvatore Giordano, Arturo Giordano
Cited by: 1

Abstract

Background: Medtronic Onyx zotarolimus-eluting stents (ZES) are widely used in complex percutaneous coronary interventions (PCI), yet real-world safety signals remain underexplored outside of clinical trials. We conducted a focused case study leveraging the Food and Drug Administration (FDA) Manufacturer and User Facility Device Experience (MAUDE) database to characterize complications and technical failure modes specifically associated with Onyx platforms. Methods: We applied a ChatGPT 5-assisted extraction and curation workflow to the 500 most recent MAUDE reports retrieved using the query Manufacturer = “Medtronic” AND Product = “Onyx” (as of September 30, 2025). The model accurately parsed free-text narratives to identify device problems, clinical outcomes, and event severity, guided by a standardized taxonomy, which were then summarized providing row counts and prevalence rates, with exact 95% confidence intervals. Notably, out of 30 randomly selected reports manually verified by two independent reviewers, no inaccuracies were evident for death (0 [0-11.6%]) and two for non-fatal events (6.7% [0.8%-22.1%]). Results: We identified 499 reports (one was excluded because focusing on a non-coronary device), with death occurring in 21 cases (4.2% [2.8%-6.3%]). Devices were predominantly Onyx Frontier (n=419, 84.0% [80.5%-86.9%]), with Trucor (n=68, 13.6% [10.9%-16.9%]) and Trustar (n=12, 2.4% [1.4%-4.2%]). Leading issues included stent dislodgement/displacement (n=400, 80.2% [76.4%-83.4%]) and positioning failure (n=183, 36.7% [32.6%-41.0%]); balloon breakage/burst and balloon malfunction occurred in 58 (11.6% [9.1%-14.7%]) and 42 (8.4% [6.3%-11.2%]), respectively. Other reported outcomes included myocardial infarction in 19 (3.8% [2.5%-5.9%]), bleeding in 10 (2.0% [1.1%-3.6%]), and stroke in 3 (0.6% [0.2%-1.8%]). Conclusions: This case study demonstrates the feasibility and accuracy of ChatGPT 5 in extracting structured safety signals from unstructured MAUDE narratives. Complications linked to Medtronic Onyx stents were predominantly procedural, centering on delivery and positioning challenges, with case fatality of 4.2%. These findings are promising and insightful, albeit hypothesis-generating. They underscore the need for procedural optimization and enriched linkage with clinical registries to contextualize risk.

Fields of Study & Subjects

Coronary Interventions and Diagnostics Artificial Intelligence in Healthcare and Education Explainable Artificial Intelligence (XAI)

Publication details

Journal / source
Top Italian Scientists Journal
Publisher
Top Italian Scientists
Volume
3
Issue
1
Publication date
12 January 2026