Post-Market Adverse Event Patterns in Venous and Pulmonary Mechanical Thrombectomy Systems: FDA MAUDE Analysis, 2015–2025

Purpose: To characterize adverse events, temporal trends, and failure-mode patterns in venous/pulmonary mechanical thrombectomy using the FDA Manufacturer and User Facility Device Experience (MAUDE) database.

Materials and Methods: MAUDE was queried 01/2015–09/2025 using FDA product codes QEW and QEZ, augmented with brand keywords (FlowTriever/Triever and ClotTriever, Inari Medical, Irvine, CA; Indigo/Lightning, Penumbra, Alameda, CA; AlphaVac, AngioDynamics, Latham, NY). Reports were de-duplicated by report number and a narrative-similarity rule. Event type (malfunction, injury, death) was extracted. Narratives were coded into failure modes: device separation/fracture (SEP), embolization/migration (EMB), vessel injury/perforation (VSI), access/bleeding (ACS), console/suction/power (CON), other (OTH). Two raters performed coding with inter-rater agreement assessed by κ on a random 100-report subset. Yearly injury-proportion trend was tested with the Cochran–Armitage test (across years with non-zero injury counts, 2019–2025). Disproportionality of injury by failure mode was estimated with proportional reporting ratios (PRR) and 95% CIs. A sensitivity analysis repeated key results excluding reports tagged as purely operational/user-error. Public, de-identified dataset.

Results: N=507 unique reports after de-duplication: malfunction 85.2%, injury 8.9%, death 4.3%. Injury-labeled reports increased 2019–2025 (1→10; z=7.25, p< 0.0001). Most frequent failure modes were OTH 264 (52.1%), SEP 160 (31.6%), CON 46 (9.1%). ACS showed the highest disproportionate association with injury (PRR 7.17, 95% CI 3.31–15.52). Sensitivity excluding operational/user-error labels yielded similar directionality.

Conclusion: In this venous/pulmonary MAUDE review (QEW/QEZ), injury-labeled reports increased; access bleeding (ACS) showed the strongest injury link. Training/checklists for access, hemostasis, and anticoagulation may mitigate risk. Because MAUDE is voluntary/narrative—often incomplete or duplicated and lacking denominators—true rates or brand comparisons aren’t inferable. Compared with prior reviews, we add trend testing, PRR-based signal flagging, de-duplication, and rater agreement.

During the preparation of this work, the author used OpenAI ChatGPT in order to improve grammar, clarity, and organization. The author reviewed and edited the content and take full responsibility for it.