Skip main navigation

More Versus Better: Learning From the Medtronic Valiant Navion Recall

Originally published Cardiovascular Interventions. 2022;15

Endoleaks occur following ≥10% of the 8000 thoracic endovascular aortic repairs done yearly in the United States and can lead to aortic expansion, reinterventions, and rupture.1 In February 2021, Medtronic’s Valiant Navion endograft was recalled partially because of type IIIb endoleaks (T3bE). In a letter dated May 21, 2021, Medtronic reported 8 total T3bE identified among 404 clinical trial participants and commercial patients.2 T3bE are of interest because, as leaks through the graft fabric, they may be related to endograft manufacturing characteristics. However, given the infrequency of T3bE, which affects <2% of patients, and the lack of T3bE-specific diagnosis or procedure codes for T3bE, there are few data sources large enough to permit comparison of T3bE between endograft types.1 We hypothesized that a signal of association between endograft type and T3bE would be observable in the Food and Drug Administration’s (FDA’s) Manufacturer and User Facility Device Experience database (MAUDE).

The MAUDE database is maintained by the FDA and compiles medical device event reports from mandated reporters (manufacturers, importers, and device user facilities) as well as voluntary reports from health care professionals, patients, or consumers.3 We reviewed 11 years of MAUDE data (January 1, 2010—December 31, 2020) for reports pertaining to thoracic endografts (Bolton/Terumo RELAY, Cook Zenith TX2, Cook Zenith Alpha, Gore TAG, Gore CTAG, Medtronic Valiant, and Medtronic Talent). Reports pertaining to T3bE were identified through review of all reports containing the words “leak” or “endoleak.” These were tallied by device by year and a signal of association between device and T3bE was assessed using proportional reporting ratios (PRRs).4 Because MAUDE lacks denominator information, the PRR compares the proportion of all T3bE reports for a specific device to that same proportion for all other devices in the database. If the proportions are equal, the PRR will be 1. The PRR and the accompanying statistical detection threshold (≥3 cases, PRR≥2, χ2≥4) were developed for identifying potential signals in medication databases like the FDA Adverse Event Reporting System, and have been used extensively for this purpose.4 We also chronologically analyzed T3bE-related reports by device to assess whether any other devices showed trends towards fulfilling PRR-related criteria; only results for devices meeting criteria for a signal of association overall are reported. This study was exempted from institutional review board review given its use of publicly available, deidentified data. The data that support the findings of this study are available from the corresponding author upon reasonable request.

There were 7328 MAUDE reports for analysis (Table). The Medtronic Valiant device had 64 T3bE-related reports, a PRR of 2.07 (95% CI, 1.72–2.42), and a χ2 of 17.3. This signal of association was ascertainable as early as 2013, in which there were 10 T3bE-related reports for the Medtronic Valiant device with a PRR of 2.64 and a chi-squared of 5.60. In no year between 2013 and 2020 was the cumulative PRR <2.64 or chi-squared <5.60.

Table. Thoracic Aortic Endograft Types and Signals of Association With Type IIIb Endoleaks

Device typeYear of first reportTotal reportsType IIIb endoleaksProportion of reports related to type IIIb endoleaksPRR (95% CI)χ2χ2P valueFisher exact P value*
Medtronic Talent2010991180.01821.10 (0.60–1.59)0.130.720.69
Gore TAG2010132890.00680.36 (0–1.03)9.840.002<0.001
Medtronic Valiant20112520640.02542.07 (1.72–2.42)17.3<0.001<0.001
Cook TX2201138980.02051.24 (0.53–1.95)0.360.550.54
Gore CTAG20121479110.00740.39 (0–1.01)9.810.002<0.001
Bolton RELAY201231980.02511.53 (0.82–2.24)1.390.240.26
Cook Alpha201630250.01660.99 (0.10–1.87)<0.010.97>0.99

PRR indicates proportional reporting ratio.

* Displayed in addition to χ2 test and P value due to the presence of expected frequencies of approximately 5 for some devices.

The Medtronic Valiant device was recalled based on 8 T3bE identified exclusively in trial and commercial patients.2 We identified (at least) 56 additional T3bE related to Medtronic Valiant using MAUDE and furthermore found that a signal between Medtronic Valiant and T3bE was identifiable as early as 2013. Although PRRs are not perfect and MAUDE is known to have significant shortcomings, the gap between what was identifiable in MAUDE and was reported during the Valiant recall proceedings suggests that available reports are not used to proactively identify safety signals, or even to quantify the scope of the problem after signals have been identified.3 Most efforts to improve the FDA’s postmarket surveillance capabilities have focused on data collection (largely in relation to possible under-reporting), avoiding false-positive safety signals, and enforcing corrective actions. Much less attention has been paid to the efficient and appropriate use of gathered data. The National Evaluation System for Health Technology Coordinating Center, designed to leverage real-world data in medical device surveillance, is one promising vector for improving surveillance. Although National Evaluation System for Health Technology Coordinating Center was formed around issues of data provenance, it has expanded to include active surveillance. However, as laid out in the recent National Evaluation System for Health Technology Coordinating Center draft active surveillance roadmap, initial efforts will focus on hypothesis-driven analyses (signal refinement), rather than monitoring for new signals (signal detection).5 Although the proposed improved data availability and signal refinement analytic approaches will certainly improve surveillance capabilities, there remains a risk of missing rare but meaningful unknown unknown signals if signal detection continues to be a secondary concern. Therefore, it is the authors’ hope that the FDA and National Evaluation System for Health Technology Coordinating Center will additionally consider incorporating proactive surveillance into their future plans, whether that be through using PRRs to identify signals for further investigation or unsupervised learning approaches. Regardless of the approach taken, our analysis suggests that regularly repeating surveillance, such as yearly, may help identify concerning signals early, before undue risk of patient harm occurs.

Article Information

Disclosures None.


This manuscript was sent to Frederick G. Welt, MD, Senior Guest Editor, for review by expert referees, editorial decision, and final disposition.

For Sources of Funding and Disclosures, see page 631.

Correspondence to: G. Chad Hughes, MD, Division of Cardiovascular and Thoracic Surgery, Duke University School of Medicine, Box 3051 DUMC, Durham, NC 27710. Email