Medical device surveillance is only as strong as the reports behind it
Medical devices move from trial to bedside through a regulatory pathway that often relies on smaller and shorter premarket studies than the evidence base for drugs. Once a device is in widespread use, performance can shift in ways that were never visible during early testing: operator experience changes outcomes, device settings vary, and complications may emerge only after thousands of implantations. In that setting, postmarketing surveillance becomes the main defense against delayed recognition of harm.
The BMJ editorial centers on a basic but consequential problem: adverse event surveillance can fail when the reporting system is incomplete, delayed, or too imprecise to support meaningful interpretation. The linked cross-sectional study examined summary reports submitted to the FDA’s manufacturer and user facility device experience, or MAUDE, database between 2019 and 2025. More than 6000 reports represented about 160 000 adverse events, mostly involving high-risk cardiovascular devices. More than half of the reports lacked meaningful information about the device problem, 423 reports incorrectly listed a single event despite undercounting 11 902 adverse events, and 55% missed reporting deadlines.
One sentence captures the clinical problem plainly: a reporting system meant to expose danger can itself become an unreliable filter.
Why spontaneous reports can signal harm but cannot measure it
Spontaneous reporting systems occupy an important place in device safety because they are public, broad, and sensitive to unusual events that may not appear in premarket testing. A cluster of reports can alert regulators, clinicians, and manufacturers to a previously unrecognized defect. Individual reports also preserve public visibility, which matters when a device is used across many sites by many operators and the first sign of trouble may come from scattered bedside observations.
Even so, spontaneous reports have structural limits that are hard to avoid. They suffer from underreporting, incomplete detail, uncertain causality, and no reliable denominator showing how many patients actually received the device. Without that denominator, incidence cannot be estimated cleanly and competing devices cannot be compared fairly. The result is a system that can generate hypotheses quickly but struggles to quantify risk or distinguish a real signal from noise.
That limitation is not theoretical. In the MAUDE analysis summarized here, 57% of reports lacked meaningful information about device problems, while reporting delays were common. When a report misclassifies thousands of adverse events as one event, the burden of harm is pushed out of view. For clinicians, that means the public record may look cleaner than the bedside reality.
Registries supply the denominator MAUDE lacks
Clinical registries emerged in part because spontaneous reporting alone could not answer the questions clinicians actually ask: how often does a complication occur, which patients are at higher risk, and does one device outperform another over time? A registry collects standardized information on a defined cohort, links outcomes to the number of exposed patients, and permits longitudinal follow-up. In cardiovascular medicine, that model has proven especially useful because device performance often depends on anatomy, implantation technique, and long-term mechanical stress.
The editorial points to several examples. INTERMACS identified rising pump thrombosis rates among recipients of the HeartMate II left ventricular assist device, and National Cardiovascular Data Registry programs have shown that prospective surveillance of high-risk cardiovascular devices is feasible. Linked registry analyses have also allowed long-term assessment of prosthetic heart valve performance. These examples matter because they show how registry data can move beyond signal detection into quantification, trend analysis, and comparative evaluation.
Registry surveillance has also extended beyond cardiology. The National Joint Registry in England and Wales identified substantially higher revision rates among recipients of metal-on-metal hip prostheses. In contrast, a large MAUDE analysis of more than 83 000 reports related to the same implant class found little ability to detect meaningful warning signals for most manifestations of suspected cobalt toxicity. The contrast is stark: registry data captured a measurable safety problem, while spontaneous reports offered limited traction for risk estimation.
Public reporting still matters when defects move fast
MAUDE should not be dismissed as a relic. Public adverse event systems can identify dangerous patterns before a purpose-built registry exists and before longitudinal follow-up is available. A recent analysis of a recalled pacemaker found hundreds of safety mode failures and serious adverse events in MAUDE reports before the manufacturer recall occurred. That kind of early visibility can matter in real time, especially when a defect is rare, catastrophic, or concentrated in a specific device batch.
Still, the editorial does not treat early signal detection as enough. The analysis did not show that MAUDE itself prompted earlier regulatory action, and the authors argued for more sensitive and specific postmarketing surveillance methods. The practical issue is timing: a delayed corrective action can leave larger patient populations exposed before the pattern is recognized.
Transparency and trust rise or fall together
The linked study evaluated the completeness, accuracy, and timeliness of publicly available summary reports, not the underlying registry infrastructure. That distinction is important because the editorial is arguing for reform of the reporting pathway, not abandonment of registry-based surveillance. The solution is not to choose one system and discard the other. It is to preserve the strengths of each while correcting the weakest points in the public reporting chain.
Regulators should require accurate and timely summary reporting, standardized device identification, and transparent linkage to underlying registries when appropriate. Researchers should integrate spontaneous reports, registries, and linked healthcare data rather than treating any single source as definitive. In practice, that means a signal from MAUDE should trigger deeper evaluation, while registry data should provide the denominator and clinical context needed to judge whether the signal represents a true increase in risk.
Public confidence depends on that architecture. When safety information is incomplete, delayed, or difficult to interpret, trust in device oversight erodes. A cross-national survey cited in the editorial found that acceptance of medical technologies depends not only on perceived benefit and harm but also on confidence that regulators and healthcare professionals will identify and communicate emerging safety concerns appropriately. In other words, transparency is not a public relations detail; it is part of the safety system itself.
What this means for device oversight now
The editorial’s central argument is straightforward: MAUDE and clinical registries serve different, complementary purposes. Public reports provide openness and early warning. Registries provide denominators, clinical detail, and the ability to estimate risk. Cardiovascular devices illustrate both sides of that equation especially well, because small changes in design or implantation can lead to large differences in complication rates over time.
The FDA is currently transitioning to a new adverse event monitoring system intended to consolidate existing reporting systems and publish reports in real time. That shift raises a clear test for regulators: can the new platform improve timeliness without sacrificing accuracy, device specificity, and interpretability? If the answer is no, the same problem will persist under a different interface.
This editorial does not argue for more data in the abstract. It argues for better data governance, because postmarketing surveillance fails when completeness, transparency, and clinical context are treated as optional. The unresolved issue is whether future reporting systems will close that gap before the next high-risk device reaches widespread use.