Medical device adverse events: what 5 years of FDA MAUDE data says about 16 of the largest OEMs

Sixteen of the largest medical device makers are named in 4.8 million FDA adverse event reports from the last five years, a third of everything MAUDE received. Read the raw counts and you learn almost nothing. Read the failure modes, the product concentration and the recall records underneath them, and each company's file resolves into a distinct fingerprint: Abbott's reports are dominated by readings that were wrong, Medtronic's by breaks and lost signal, Intuitive's by fraying instruments, Olympus's by scopes reported as inadequately cleaned, and Philips', overwhelmingly, by one word in the FDA's taxonomy: Degraded.

We pulled every MAUDE report, every enforcement recall, and every 510(k), PMA and active product listing for 16 of the biggest OEMs, September 2021 through August 2026, matched by corporate family. Covidien counts under Medtronic, Abiomed under J&J, Respironics under Philips, Bard under BD. Then we cut it by the FDA's own device problem codes, the structured taxonomy behind every report, with terms like Material Deformation, Premature Discharge of Battery and Failure to Clean Adequately. That taxonomy is where this gets interesting.

The instinct to read MAUDE like a scoreboard is understandable, and wrong. FDA itself cautions that MAUDE cannot establish event rates, comparative safety or causation: reports come from manufacturers, importers, hospitals, clinicians and patients, they are unverified, and there is no exposure denominator behind them. A company selling five million glucose sensors a year will out-report a company selling ten thousand anesthesia machines at any conceivable quality level. Volume mostly measures exposure. The composition of the reports is what carries the information.

An adverse event profile is a fingerprint of product architecture. Show me how a company's devices are reported to fail and I will tell you what it builds.

The numbers

The raw numbers, company by company

Start with the shape of each company's file before the totals. The donuts below show how much of each OEM's adverse event volume rides on its single biggest product line, in each company's own brand color.

Widen to each company's top five product lines and the texture shows. Some files are one giant product plus rounding error. Others are a spread of near-equal franchises. Bubbles carry each company's brand palette.

The full picture, one row per company:

Signature failure mode is the most-cited informative FDA device problem code; generic codes (Adverse Event Without Identified Device or Use Problem, Insufficient Device Problem Information, Defective Device, term-not-available) are excluded, and a single report can cite multiple problem codes. Recall events are distinct FDA enforcement event IDs initiated in the window. Listed product codes come from current FDA registration and listing. Corporate families include acquisitions for the full window.

Volume itself is a diabetes story. Medtronic's three diabetes codes account for 920,665 reports and Abbott's two Libre codes add 456,908. Two companies, five product codes, roughly 1.4 million reports, close to 10 percent of every device adverse event FDA received from any manufacturer in five years. That is what continuous patient attachment does to a reporting system in which a malfunction on a device people dose insulin against routinely clears the bar for a reportable event.

These 16 companies book roughly $270 billion in combined medtech revenue against a global market in the $570 to 600 billion range, which puts their revenue share somewhere between 45 and 50 percent. Treat the two numbers as context, not a safety rate: revenue is not an exposure denominator, and FDA warns against inferring relative event rates from MAUDE counts. The other two thirds of MAUDE comes from thousands of smaller manufacturers.

This number needs context, because it is not spread across Abbott's portfolio. It sits almost entirely on one franchise.

• 324,321 Abbott reports cite incorrect, inadequate or imprecise readings, and another 109,366 cite low readings.
• Nearly all of it lands on three FreeStyle Libre sensor codes, the two current iCGM generations plus the original 14-day sensor.
• Nothing breaks and nothing leaks. The number on the screen is simply wrong, and for a device people dose insulin against, that is its own hazard class.

Five failure families cover the whole cohort

Sort every company's most-cited problem codes and the 16 OEMs collapse into a handful of families. The groupings track product physics, not corporate quality culture, and companies that compete fiercely in the market show nearly identical report profiles.

Philips deserves its own line outside the families. A single problem code, Degraded, appears on 299,475 Philips reports, 96 percent of its CPAP/BiPAP code's volume. That is the polyester-based polyurethane foam of the Respironics recall, still the dominant fact of the company's adverse event file more than five years after the June 2021 recall began, with Contamination (14,680) trailing behind it.

Concentration is extreme: a few products carry most of each company's reports

How tightly do reports cluster on each company's biggest products? The bars below show the share of each OEM's adverse events sitting on just its top three product codes.

The spread maps to business model. Intuitive puts 95.2 percent of its reports on one product code, the da Vinci platform, and 98.5 percent on its top three. GE's top three are 95 percent of its file, all anesthesia and ventilation. Platform companies carry platform risk: one product line's bad quarter moves the whole safety profile, and one foam formulation can define a company's regulatory decade.

Stryker sits at the other pole. Its largest code is 7.7 percent of its reports and its top three sum to 21 percent, spread across ankle implants, surgical drills and hospital stretchers. J&J's top three reach only 26 percent. Diversified surgical conglomerates diffuse MDR exposure the way they diffuse revenue, and their problem-code profiles are correspondingly flat: a little break, a little rupture, a little of everything.

Normalizing for portfolio breadth makes the same point. Reports per actively listed product code run from Intuitive at 4,350, Medtronic at 2,829 and Abbott at 2,110 down to Stryker at 123, Terumo at 79 and Siemens at 17. Stryker lists 442 active product codes, nearly Medtronic's 575, on 3.3 percent of Medtronic's volume. What a company sells, and how long it stays attached to a patient, sets its reporting gravity.

Adverse events vs revenue, company by company

Reports are not a business metric, and putting them next to revenue makes that concrete. Flip through the companies below: revenue in the brand color on the left axis, MAUDE reports on the right.

The interesting cases are the ones where the lines detach. Intuitive grew revenue about 70 percent over the window while reports grew tenfold. Abbott's device revenue climbed 44 percent while its final-year reports tripled. Stryker grew revenue 43 percent with a report line that barely moved, and Siemens added three billion dollars of revenue while its report count drifted down. When the dashed line outruns the solid one, something changed in the field or in the reporting, and either answer is worth knowing.

Five-year trends: who is moving and why

Market-wide MAUDE reporting grew about 20 percent over the five years, from 2.93 million to 3.51 million. The cohort grew much faster, from about 814,000 reports in 2021-22 to 1.4 million in 2025-26, up 72 percent. Two companies detached from even that baseline.

Intuitive grew reports 974 percent across the window, essentially all da Vinci instruments and almost all in the materials family: split, torn, frayed. Olympus grew 617 percent, concentrated in GI endoscopes and led by the reprocessing and imaging codes. Abbott's fifth year tripled to 407,269 on the Libre franchise's accuracy codes. Some of this is genuine field signal, some is procedure volume, and some is reporting practice, since summary reporting and software-logged malfunctions can step-change counts without the installed base changing. The honest read: the growth curves mark where FDA scrutiny goes next, not proof that a device got three times worse.

The quieter moves matter too. Philips' file peaked in 2021-22 at 95,077, dipped, and climbed back over 100,000 in 2024-25 as Trilogy ventilator issues layered onto the foam legacy. Baxter, Zimmer Biomet, Smith+Nephew and Siemens all ended the window below where they started, the closest thing the cohort has to quiet files. Nobody's line is flat.

Recalls: frequency and severity are different leagues

Recall behavior barely correlates with adverse event volume, and frequency does not equal severity. Medtronic and Philips tie for the most distinct recall events at 198 each, but Siemens ran 150 recalls on 7,390 adverse events, the most recall-intensive posture in the cohort relative to its file, and nearly all of it Class II corrections on diagnostics. That pattern reads as disciplined field correction, not crisis. Across the cohort, imaging equipment and software corrections generate more recall events than any other category, which is why the imaging-heavy companies recall so often and so quietly. Implants and catheters follow well behind.

Class I, where device use can kill, is the column that separates. Medtronic leads with 29 Class I events, then Baxter 26, Philips 23, J&J 22. As a share of their own recalls, J&J at 26 percent and Baxter at 24 percent run hottest: when they recall, it is disproportionately serious. And the Class I file is current, not historical. Abbott issued a Class I correction for certain FreeStyle Libre 3 sensors in November 2025 over carbon buildup causing false low readings, its accuracy fingerprint escalating into its recall record. Medtronic recalled the Harmony valve delivery catheter in May 2026, J&J recalled Impella CP introducer kits in June 2026, and Philips was still recalling Trilogy Evo ventilators in March 2026.

Then the clean sheets. Stryker initiated 84 recall events, every one Class II. Zimmer Biomet went 44 for 44, Smith+Nephew 25 for 25, Terumo 6 for 6. Four companies, 159 recalls, zero Class I in five years. Edwards barely recalled at all: 8 events on a portfolio where every product is life-critical. The ortho and cardiovascular-surgical world runs a fundamentally different recall economy than the electronics-heavy, patient-attached world.

Size the recalls in physical units and the skew is startling. The cohort's roughly 1,400 recall events covered about 700 million individual devices. The median event involved about 1,500 units. The average was over 500,000, because a handful of enormous disposable recalls carry the total, and almost none of the giants were the dangerous ones.

Units are as reported in the FDA enforcement quantity field, which counts whatever the firm distributes: devices, kits, consumables or components. Roughly 96 percent of the cohort's enforcement records carry a quantity.

The pattern in that table is the recall economy in one screen: the biggest recalls by unit count are cheap, high-volume disposables pulled at Class II or III, while the recalls that can kill someone, the Class I file, mostly involve thousands of units, not millions. Recall size and recall severity are almost unrelated.

Deaths

A note on what follows: the figures in this section are reproduced directly from FDA's public MAUDE database. A death report means a death was reported to FDA in connection with a device under mandatory medical device reporting rules. It is not a finding of causation, fault or responsibility by FDA, by any manufacturer, or by XO Medtech, and it should be read as an objective summary of what the database contains, nothing more.

Death reports deserve exactly one table and a caveat. They concentrate where device acuity is highest, not where quality is worst: five categories, HeartMate LVADs, Impella, CPAP, TAVR and Watchman, account for more than half of the cohort's 45,588 death reports, and every one is a device placed in or on critically ill patients. Edwards' rate is the cohort's highest for precisely that reason: structural heart is nearly its entire portfolio.

Rates track portfolio acuity: the ortho and imaging companies in the cohort run 0.5 to 3 death reports per thousand MDRs.

The XO take

We run this data for a living, and the failure families are the commercially useful layer. Problem codes describe reported symptoms, not validated root causes, so read what follows as commercial hypotheses. Held that way, every fingerprint in that table maps to a supplier capability that is, right now, under pressure somewhere in a sourcing organization.

Material split, torn and frayed at Intuitive is an instrument life-cycle problem, which means reinforcement, coatings and cable assemblies. Fluid path leaks at BD and Baxter are joining, welding and seal validation problems at extreme unit volumes. Fracture at the ortho trio is metallurgy, fatigue testing and design margin. Wireless and power failures at Medtronic and Boston Scientific are battery chemistry, connectors and RF engineering. Reprocessing failures at Olympus are surface engineering and cleaning-compatible design, arguably the most open capability gap in the whole dataset. A supplier who reads MAUDE this way knows which engineering conversations each OEM cannot avoid this year. And a Class I event, 29 of them at Medtronic in five years, 26 at Baxter, puts quality, regulatory and often sourcing teams on the clock with FDA watching, whether the fix ends up being a redesign, a software patch or new labeling.

Five years, 4.8 million reports, 1,400 recalls, 700 million recalled units, one public taxonomy that sorts the biggest companies in medtech into five ways of failing. Most of the industry still reads it as a compliance chore.

It is a map. Read it like one.

Methodology: the cohort is 16 of the largest device OEMs, selected for the scale and recognizability of their device businesses rather than a strict top-16 revenue ranking (players like Medline, B. Braun and Alcon are not included). MAUDE reports were matched on manufacturer-name variants per corporate family, acquisitions included for the full window (Covidien, C.R. Bard, St. Jude, Thoratec, Abiomed, Respironics, Hillrom, Wright Medical, Varian and others), and the five-year window is defined on the MAUDE date-received field, September 1, 2021 through August 31, 2026. Reports reach FDA from manufacturers, importers, user facilities, clinicians and patients; each MAUDE record is counted once, and some summary records cover multiple events. Manufacturer-name variation affects any MAUDE search, so figures are best read to the first digit, not the last. Recall events are distinct FDA enforcement event IDs; unit counts are the enforcement quantity field as reported. Revenue figures are approximate medtech-segment values from company filings; market size from published industry estimates. Data via openFDA.