Why Sterile Medical Devices Get Recalled: A 2015–2026 FDA Evidence Map and Supplier-Control Playbook

In this guide

• Rare by event, enormous by volume
• The 2015–2026 evidence map
• How bad do sterility recalls actually get?
• The supply-chain finding
• Your supplier's last recall predicts the next one
• Packaging: mildest in the record, riskiest at the point of use
• The sterilization-capacity squeeze
• What the 2026 rules change for supplier evidence
• The evidence packet, keyed to observed failure modes
• Two different worlds: recalls versus infection reports
• Methodology and limitations
• Frequently asked questions
• Conclusion

Rare by event, enormous by volume

Takeaway: sterility failures are uncommon as recall events and enormous as recall volume, because one upstream failure propagates through kits, trays and catalog variants. Two independent datasets — Sedgwick's recall index and our own FDA cohort — show the same shape.

Start with the current quarter. Sedgwick's U.S. Product Safety and Recall Index counted 146.63 million recalled medical-device units in Q1 2026 across 219 recall events — devices were the only one of five tracked industries where both events and units fell quarter-over-quarter ¹⁴⁵. By event count, the leading causes were device failure (41 events), software (24) and mislabeling (24). Sterility was not among them. By units, sterility ranked first — 86.35 million, "primarily driven by a single recall of 86.07 million convenience kits" ¹.

One event. Eighty-six million units. That asymmetry is not a quirk of one quarter. Across full-year 2025, Sedgwick counted 978 FDA device recalls (down 7.6% from 1,059 in 2024) with causes led by device failure (141), software (137) and mislabeling (93) — sterility again absent from the top three by event count — while total recalled units reached 490.2 million, the second-highest figure in a decade ⁶.

Our own cohort shows why. In the FDA recall database, one corrective action generates one product-level row per affected catalog number, size, kit or configuration. Collapse the 5,543 sterility-related rows since 2015 by recalling firm, initiation date and exact reason text, and 998 underlying episodes remain — 5.55 rows per episode, against 2.80 for the average device recall run through the identical collapse ². Sterility accounts for 8.6% of recall episodes but 17.1% of recall rows. The record inflation the trade press occasionally notices in FDA data is not generic — it is specifically severe for sterility, because sterility failures happen to processes and components that many products share. (The largest single episode in our cohort produced 183 product rows.)

The government's own numbers land in the same place. A December 2025 report by the U.S. Government Accountability Office examined 3,934 medical-device recalls across FY2020–FY2024 — all voluntary — and found 88% classified Class II, roughly 9% Class I and 3% Class III ⁷. Our independently computed all-device episode split for 2015–2026 is 89.5% / 6.9% / 3.6% — a different window and a different counting unit (GAO counts recalls; we count collapsed episodes), but the same shape. GAO also found that 74% of recalls exceeded FDA's three-month termination goal: a recall, once opened, stays open far longer than most commercial plans assume ⁷.

One more piece of context a buyer should hold: FDA's device-inspectorate leadership said in April 2026 that Class I recalls increased 232% between 2020 and 2025 — "that is a concerning number for us" ⁸. Whatever the driver (better detection, more kitted products, more complex supply chains), the severe tier of the recall system is getting busier. The rest of this report is about where, in the sterility record specifically, that severity actually comes from.

The 2015–2026 evidence map

Takeaway: sterility recall volume has been flat for eleven straight years — no improvement, no deterioration. Regulation and industry practice have not moved the number, which is precisely why buyer-side control matters.

The 998 episodes distribute across the years 2015–2025 in a band from 72 to 106, averaging 86.8 a year, with partial-2026 at 43 episodes through early July ². There is no trend. Sterility's share of all device-recall episodes is equally flat: between 7.3% and 10.1% every single year. Nothing in eleven years of guidance documents, standards revisions and enforcement cycles has visibly changed the rate at which sterility-related recalls occur.

Read as a control chart rather than a scoreboard, that stability is the finding: the industry's sterility-failure generating process is stationary. A buyer cannot rely on the system improving around them.

What the reasons say, at episode level ²:

• Packaging / sterile-barrier integrity — 534 episodes (53.5%); 3,285 records (59.3%)
• Sterilization process / dose / validation — 199 episodes (19.9%)
• Bioburden / microbial / endotoxin — 125 episodes (12.5%)
• Other sterility-related — 89 episodes (8.9%)
• Labeling / expiry / sterile-status mismatch — 51 episodes (5.1%)

Two honesty notes travel with that table. First, these are categories of language in a regulatory text field, assigned by a documented first-match rule — a map of how failures are described, not a validated root-cause distribution. Second, the FDA's own controlled root-cause field says the causes are less settled than any taxonomy implies: "Under Investigation by firm" is the single largest label at 21.1% of episodes, and with "Unknown/Undetermined" added, 23.6% of sterility episodes have no settled root cause on the public record ²⁹. A buyer should treat any confident claim about "the" cause of sterile-device recalls — including this report's — as a probability statement, not a verdict.

How bad do sterility recalls actually get?

Takeaway: sterility recalls are about half as likely to reach the most serious recall tier as device recalls generally — but the Class I events that do occur cluster in time, concentrate in bioburden and purchased components, and arrive as multi-firm cascades.

The previous version of the public conversation — where it exists at all — assumes sterility recalls are inherently catastrophic. The record says otherwise. Joining the cohort to FDA's enforcement classifications (99.4% of episodes matched), 3.3% of sterility episodes are recall-Class I — FDA's "reasonable probability of serious adverse health consequences or death" tier — against 6.9% across all device recalls in the same window. Ninety-five percent of sterility episodes are Class II ³.

So the realistic planning assumption for a distributor is not "patient-harm headline." It is: you will be pulling stock off shelves at Class II, with product-level rows fanning out across your catalog. That is an inventory, traceability and cash-flow event more often than a safety catastrophe — which is an argument for taking traceability seriously, not for relaxing.

But the severity that does exist is not where the volume is. Class I share by failure category ³:

The failure mode you will see most often is the one that hurts least, and the failure mode you will see least often is the one that hurts most. Fourteen of the 33 Class I sterility episodes — 42% — are bioburden, microbial or endotoxin events, off one-eighth of the cohort's volume. The counts are small, and we show them beside every rate for that reason; but a 14× spread between 0.8% and 11.3% does not survive on noise alone.

Those 33 Class I episodes in 11.5 years are also episodic, not steady: 13 arrived in 2021 and 9 in 2023, with several years contributing one or none ³. The 2021 cluster is a single contaminated topical consumable recurring as thirteen separate firm-level recall episodes — one upstream product, thirteen downstream companies forced into the most serious recall tier. That mechanism is the subject of the next section.

Two verbatim fragments from the FDA recall-reason field, quoted without naming any firm, summarize what the severe tier actually looks like ²:

"Sterilization certificates could not be validated by the supplier." (2024)

"Convenience kits labeled as sterile have not gone through the sterilization process." (2025)

The first is the single most useful sentence in the dataset for a buyer: a certificate that cannot be validated is precisely the artifact most qualification processes accept as proof. The second is the reason lot-release evidence — not paperwork volume — is the thing to demand.

One boundary to keep sharp: recall classification measures potential health consequence, assessed by FDA. A Class I recall is not evidence anyone was harmed, and Class II is not evidence the problem was trivial.

The supply-chain finding

Takeaway: episodes described as involving a supplier or purchased component are several times more likely to be Class I — about 4× excluding the largest single contamination cluster, about 10× including it. Two independent FDA fields point the same way. For a private-label owner, upstream control is not a compliance nicety; it is where the severe risk lives.

This is the number the report exists to publish. Split the 992 classified episodes by whether the FDA recall reason names a supplier, vendor, third party, component or raw material ²³:

Roughly ten times the Class I rate. And because a finding this strong deserves its own stress test: the 2021 contamination cluster described above contributes many of those Class I events. Remove it entirely, and the split is 7.0% (5 of 71) versus 1.8% (16 of 902) — still roughly 4×. The direction is robust; the magnitude depends on whether you count the one giant cascade. We publish both numbers, always, and we suggest reading the cluster not as an outlier to be discarded but as the mechanism at full scale: one upstream supplier's contamination forced thirteen separate downstream firms into Class I recalls in a single year.

Three caveats are load-bearing. The flag is language in a free-text field — it measures how firms describe recalls as much as what caused them. Firms may be likelier to name an upstream source when an event is serious enough to require explanation, which would inflate the association. And n = 82 episodes with 16 Class I events is small. So: "episodes described as involving a supplier or purchased component," never "episodes caused by suppliers."

But the free-text flag is not the only witness. FDA's controlled root-cause vocabulary — a separate field, assigned per recall — points upstream in 13.8% of all episodes ²⁹:

Note the sixth row. "Release of material or component prior to receiving test results" is a named FDA root cause — shipping before the test came back happens often enough that the agency has a controlled label for it. It is also one of the most directly checkable controls a buyer can put in an evidence packet.

The cascade shows up in product structure too. Episodes whose recall reason names a kit, tray, pack or set fan out across 9.29 product rows on average, against 4.67 for everything else — assembling other people's sterile components roughly doubles the SKU blast radius of any single failure ². And the single most-recalled product code in the entire cohort, by both episodes (46) and records (415), is "General Surgery Tray" ²¹⁰ — a product that is, by definition, an assembly of other manufacturers' components. The thesis of this report is sitting at the top of the product-code table.

Where does all this concentrate? Five FDA review panels carry 75.2% of episodes ²¹⁰:

General & Plastic Surgery (27.9% of episodes, and the widest fan-out at 7.76 rows per episode), Orthopedic (13.3%), General Hospital (12.1%), Cardiovascular (11.8%, fanning at 7.81) and Gastroenterology & Urology (10.1%). These are the single-use, kitted, high-SKU categories — which is not a coincidence; it is the same mechanism viewed by specialty. (FDA's "medical specialty" field is a review-panel label, not a clinical setting; we use it as such.)

One more result that should change procurement behavior: 24.2% of sterility recall episodes involve products that are device-Class I — the lowest-risk regulatory tier, largely exempt from premarket review: drapes, gowns, trays, basic instruments ²¹⁰. To be explicit about the two scales, because they collide here: device class (I/II/III) describes a product's regulatory risk tier; recall classification (I/II/III) describes the severity of a specific recall — and in this paragraph we mean device class. A buyer who scales supplier scrutiny to device class will systematically under-audit the products behind a quarter of sterility episodes. Sterility assurance is a process property. The process does not get simpler because the device is simple.

Your supplier's last recall predicts the next one

Takeaway: repeat firms produce seven in ten sterility episodes; when a second episode comes it usually comes within months, and nine in ten repeat firms fail in the same mode as before. A recall history is not a disqualifier — it is a diagnostic that tells you which control to audit.

Across the cohort's 455 normalized recalling-firm strings, 303 (66.6%) appear exactly once. The other 152 (33.4%) produce 695 episodes — 69.6% of the total ². Concentration runs deep: the 9.7% of firm strings with five or more episodes account for 42.7% of all episodes, and the 3.3% with ten or more account for a full quarter. Records concentrate harder still — the top ten firm strings hold 39.0% of all 5,543 product rows.

Say the caveat in the same breath: firm strings are not deduplicated legal entities (subsidiaries and name variants split), so concentration is understated — and none of this is portfolio-adjusted. A firm shipping thousands of sterile SKUs has more opportunities to recall than one shipping ten. This is a concentration measurement, not a quality ranking, every firm identity in our working data is hashed, and no reader should try to reverse-engineer a league table from it. The operational point survives all of that: recall history is unevenly distributed, which is exactly why asking about it is informative.

Timing, across 543 consecutive-episode gaps at the 152 repeat firms ²:

The median gap between a firm's consecutive sterility episodes is 159 days. A quarter arrive within 36 days — largely one underlying problem being written up as several actions while scope expands, which is itself worth knowing: the first recall notice is often not the whole problem. And 69.1% of next episodes arrive within twelve months. When a supplier tells you last year's recall is a closed chapter, the base rate disagrees for the first year after the event.

The most actionable finding in the entire dataset is what repeats. For each repeat firm, take the share of its episodes falling in its most common failure category: the mean is 72.1%; 91.4% of repeat firms concentrate at least half of their episodes in one category; 34.9% never leave one category at all ².

Firms do not fail randomly; they fail in their signature mode. That converts a recall history from a red flag into a targeting instruction: a history of seal failures says audit sealing validation and packaging change control; a history of bioburden says audit incoming material, environmental monitoring and dose audits; a history of status-mismatch says audit lot release and sterilizer logistics. The right response to a supplier's recall history is a specific question, not a reflexive rejection — and a supplier who can show you the CAPA, and clean operating evidence for the twelve months after it, has given you better assurance than one with no history and no evidence.

Packaging: mildest in the record, riskiest at the point of use

Takeaway: packaging leads recall volume but almost never reaches Class I — yet peer-reviewed simulation shows sterile-barrier configurations failing aseptic presentation up to 85% of the time in trained hands. The packaging risk doesn't disappear; it moves to the operating table, where no recall database can see it.

Half the recall record is packaging language, and only 0.8% of those episodes reach recall-Class I ³. Read carelessly, that says packaging barely matters. Read correctly, it says the recall record is the wrong instrument for measuring packaging risk — because a package that opens badly does not generate a recall. It generates a contaminated field, a discarded device, a delayed procedure, sometimes an infection nobody attributes — and none of that appears in a recall row.

The best evidence for that lives in a 2023 peer-reviewed study in AAMI's Biomedical Instrumentation & Technology: 40 experienced clinical nurses, 14 devices, five sterile-barrier configurations, 280 aseptic-presentation trials. Vent-bag configurations failed aseptic presentation 85% of the time. Header bags failed 72%. And 37 of the 40 nurses could not distinguish the protective packaging from the sterile-barrier system on a coronary stent package ¹¹. Packaging that passes every seal-strength and integrity test in the validation file can still fail at the point of use, in trained hands, most of the time.

This is why the standards framework treats usability as a design requirement, and it is worth being precise about which edition says so, because the industry frequently gets it wrong: the usability-evaluation requirement for aseptic presentation entered ISO 11607-1 in the 2019 second edition ¹¹¹². The 2023 amendments to both parts — ISO 11607-1:2019/Amd 1:2023 and ISO 11607-2:2019/Amd 1:2023 — are titled "Application of risk management", not usability ¹²¹³. FDA recognizes both amended parts as consensus standards ¹⁴¹⁵.

For the buyer, the packaging ask therefore has three layers, all of which the evidence-packet table below carries: Part 1 evidence that the designed barrier system protects the device through sterilization, distribution and aging; Part 2 evidence that the forming, sealing and assembly processes are validated and controlled — a supplier can hold a suitable pouch and an uncontrolled sealer, or a validated sealer and an unsuitable pouch; and a documented aseptic-presentation usability evaluation for the exact configuration you are buying, which after 2019 is not a nice-to-have but a requirement of the recognized standard.

The sterilization-capacity squeeze

Takeaway: half of America's sterile devices depend on roughly ninety buildings run by about fifty companies, the rule governing them has changed three times in two years, and the industry says the facilities are already running at capacity. Capacity contingency is now a supplier-qualification question, not a compliance footnote.

Sterilization is the one step in the sterile supply chain that most buyers never see and cannot substitute on short notice. The numbers describe a genuinely concentrated system: about 50% of sterile medical devices in the U.S. — more than 20 billion units a year — are sterilized with ethylene oxide ¹⁶¹⁷, in approximately 90 commercial facilities owned by roughly 50 companies ¹⁸²⁰. The National Academies put the full modality split at roughly 50% EtO, 40% cobalt-60 gamma and 10% e-beam ²².

The regulatory ground under those facilities has moved three times in two years, and an article citing only the first move would be materially out of date:

• March 2024 — EPA finalized amended air-toxics standards requiring more than 90% EtO emission reductions, with tiered compliance deadlines through 2026–2027 and an estimated $313 million capital cost across the industry ¹⁸.
• July 2025 — a Presidential proclamation exempted the facilities listed in its annex for two years beyond the rule's deadlines, pushing their technology-standard compliance to April 5, 2029 ¹⁹.
• March 2026 — EPA proposed rescinding the rule's risk-based standards outright and relaxing monitoring requirements, stating that commercial sterilizers process "50 percent of all medical devices in the U.S. every year — or 20 billion medical devices — with no viable alternative on the market" ¹⁹²⁰. By EPA's own impact analysis the revision would emit 7.8 tons per year more EtO than the 2024 rule ²¹.

Whichever way the final rule lands, the structural fact for buyers is the one AdvaMed put on the record: "EtO sterilization facilities have been operating at capacity, and taking a few facilities offline could affect timely access to medical equipment" ²¹. And the pipeline of alternatives is thin in the one place it can be measured: FDA's three sterilization master-file pilot programs — created precisely to make sterilization changes less disruptive — have attracted a combined seven participants in seven years ¹⁶²⁶. FDA has meanwhile issued enforcement discretion allowing Class III devices to ship from new sterilization sites ahead of the usual approval supplement, explicitly to "prevent or mitigate… supply chain disruptions or medical device shortages" ²⁷ — a policy that only exists because the risk is real.

The radiation side is concentrated differently but no less tightly. Cobalt-60 — the isotope behind the gamma share — reaches the market through a supply chain in which one company's two locations serve roughly forty customers worldwide ²⁵. The United States consumes about 50% of global cobalt-60 supply while relying entirely on foreign sources, and the first domestic production program targets deployment in the mid-2030s ²³²⁴. A buyer's ten-year sourcing plan for gamma-sterilized product runs straight through that dependency.

What to actually ask a supplier — four questions that take ten minutes and reveal more than most audits:

1. Which facility, which modality, which validated family? Site name, chamber or cell, and whether your configuration sits inside the currently validated product/load family.
2. What is the qualified backup? A second validated site or modality, with the change-control and regulatory pathway already mapped — or an honest "none," which is an answer you can plan around.
3. What happens if your sterilizer's compliance date moves? Suppliers running product through facilities on the exemption list have a 2029 date today and litigation risk tomorrow.
4. Has any product ever shipped before sterilization or before test results? "Release of material/component prior to receiving test results" is a named FDA root cause; the Class I record includes kits that skipped the sterilizer entirely ²⁹.

What the 2026 rules change for supplier evidence

Takeaway: QMSR makes supplier-audit and management-review records inspectable in the U.S.; EU MDR names the exact validation tests a sterile-device file must contain; and notified-body capacity makes the certificate itself a scheduling risk. The paperwork era of supplier qualification is ending on both sides of the Atlantic.

In the United States, FDA's Quality Management System Regulation took effect on February 2, 2026, incorporating ISO 13485:2016 by reference into 21 CFR Part 820 ²⁸. The change buyers should care about is inspectional: FDA states that management-review, internal-audit and supplier-audit reports — previously shielded from routine review — are now within an investigator's reach, and that records created before the effective date can be reviewed ²⁹. A supplier's supplier-control file is no longer a private artifact. If it will not survive an FDA investigator, it should not survive your qualification either.

In the EU, the MDR already writes the sterile-device evidence list into law. For any device placed on the market sterile, the technical documentation must contain validation reports for packaging, sterilization and maintenance of sterility — and the regulation names the tests: "bioburden testing, pyrogen testing and, if applicable, testing for sterilant residues" (Annex II §6.2(e)) ³⁰. A sterile Class I device — the tier that covers a quarter of our cohort's episodes — cannot self-certify: the sterility aspects go through a notified body (Article 52(7)(a)), and the packaging must keep its integrity "clearly evident to the final user" (Annex I §11.4) ³⁰. For a distributor, Annex II §6.2(e) is a ready-made, regulation-backed document request. Use the regulation's own words in your RFQ.

Then comes the queue. Team-NB's 2025 survey — 41 of the 52 designated notified bodies responding — reports 13,953 MDR certificates issued in 2025 against 25,978 applications, with only 31% of new certificates issued within 12 months and notified-body technical staffing falling for the first time in a decade ³¹. A Commission-commissioned survey adds that at 42% of notified bodies, fewer than a quarter of legacy-directive clients have completed the MDR transition ³². Translation for sourcing: a supplier "in the process" of MDR certification for your sterile product line is quoting you a timeline they do not control, with a median well past a year. Ask for the notified-body agreement, the submission date and the review stage — not the intention.

The evidence packet, keyed to observed failure modes

Takeaway: ask for ten linked evidence items, each tied to a failure mode the FDA record actually shows — and weight your scrutiny by observed severity, not by paperwork convenience.

Everything above converts into one table. This replaces the generic supplier checklist: each gate names the evidence, the observed failure category it defends against, and that category's measured share of episodes and Class I rate from this cohort ²³. (For the full supplier-qualification workflow — roles, entity/site verification, transfer baselines, quality agreements and lot-release mechanics — see our companion buyer's guide to medical-device contract manufacturing; this table is the sterility-specific core.)

Three knockout conditions, regardless of score: no applicable sterilization validation for the purchased configuration; no lot-to-load traceability; any indication product can ship before quality release. And one weighting rule: gates 2, 3, 6 and 7 defend against everything the record marks severe — if audit time is scarce, spend it there, not on re-reading certificates.

A note on scoring: we deliberately publish no points system. The record shows failure modes are firm-specific and repeat in kind — so the productive audit is targeted by the supplier's own history and product structure, not averaged into a number that hides the one gate that matters.

Two different worlds: recalls versus infection reports

Takeaway: manufacturing-sterility recalls and device-associated infection reports describe different device populations and different problems. Reading either dataset as the other produces confident, wrong conclusions — in both directions.

It is tempting to "complete" a recall analysis with adverse-event data: surely infection reports tell you which sterility failures hurt people? They do not, and the mismatch is itself a finding worth publishing.

We screened all 4,623,139 MAUDE reports from 2025 through mid-2026 for infection-type patient outcomes. After excluding a single dental-implant reporting surge that alone accounted for 67.6% of raw matches — a disclosure without which the numbers mislead — 88,752 non-dental infection-flagged reports remain, 1.92% of the window ³⁶. Their structure is the reverse of the general record: 94.1% are coded Injury and 2.2% Death, against a baseline of 38.0% Injury and 0.45% Death across all reports. Most device reports are about devices failing; infection-flagged reports are about patients harmed. In the hard systemic-infection subset (sepsis, bacteraemia, endocarditis, meningitis, peritonitis, osteomyelitis), 10.1% of 9,482 reports are coded Death ³⁶.

But look at which devices: pacemaker electrodes, implantable leads, defibrillators, implanted ports, peritoneal-dialysis sets, ventricular-assist pumps — long-dwell implanted and indwelling devices, clustering at the opposite end of the product universe from the single-use kits, trays and instruments where sterility recalls concentrate ³⁶. These are device-associated infections arising from implantation and dwell time. A central line infected after three weeks in a patient says nothing about whether it left the plant sterile.

So the two rules, stated once and bluntly: do not read MAUDE infection volume as a supplier-quality signal, and do not read a clean MAUDE record as evidence of sterility assurance. The recall record measures manufacturing-control failures caught by the system; the adverse-event record measures (mostly) clinical device-associated infection in implants. This report is built on the first. Nothing in the second either indicts or exonerates a sterile-device supplier.

Methodology and limitations

Takeaway: every number above is reproducible from public FDA data with the rules stated here — and lands where the government's own cohort lands.

Cohort. FDA Medical Device Recall data, snapshot of July 24, 2026 ². Inclusion: recall initiation date (posting-date fallback) on or after January 1, 2015, and a word-boundary match in the reason-for-recall field on sterility vocabulary (sterile/sterility/sterilization variants, non-sterile, aseptic, bioburden, microbial, bacterial, fungal, endotoxin, pyrogen). Result: 5,543 product records. Episodes collapse records sharing normalized recalling firm + initiation date + exact normalized reason text: 998. The same collapse applied to all 32,483 device-recall records in the window yields 11,609 episodes — the 2.80 comparator. The collapse can split one corporate action written up under two reason strings and merge same-day/same-reason actions; publishing both units keeps the transformation visible.

Severity. Recall classifications joined from FDA Enforcement Reports (99.7% of records, 99.4% of episodes matched); episode severity = the most serious class among its records ³. Category and panel joins from the FDA Product Classification Database (99.4% of records resolved) ¹⁰. Repeat-firm analysis hashes firm identities before aggregation; no names are retained in working data.

Triangulation. GAO's independent FY2020–24 cohort (Recall Enterprise System, non-device records excluded) reports Class II 88% / Class I 9% / Class III 3% ⁷; our all-device 2015–2026 episode split is 89.5% / 6.9% / 3.6%. Different windows, different units — GAO counts recalls, we count collapsed episodes — same shape.

Limitations, all load-bearing. There is no denominator: no units-shipped, procedure-volume or installed-base figure exists for this cohort, so nothing here is a rate, a probability of failure, or a company quality score. The text screen captures recalls described in sterility language, not every sterile device recalled. Root causes are unsettled in 23.6% of episodes. Firm strings are not legal entities. 2026 is partial (through July 7) and never treated as a full year. The taxonomy is a documented first-match text rule, not FDA's judgment. Recall classification measures potential consequence, not observed harm. U.S. data only. Sedgwick figures are cited from the publisher's press materials and trade reporting because the underlying report is gated ¹⁴⁶. Chart data files published with this report preserve every displayed value and caveat.

Frequently asked questions

How many sterile medical device recalls happen each year? In the FDA record since 2015, sterility-related recalls have run between 72 and 106 episodes a year (mean 86.8), generating roughly 300–700 product-level records annually — with no upward or downward trend across eleven years ². Sterility consistently represents 7–10% of all U.S. device-recall episodes. Counts are records and episodes, not rates: no public denominator of sterile units shipped exists.

What is the most common reason sterile devices are recalled? By language in the FDA reason field, packaging and sterile-barrier integrity leads — 53.5% of episodes since 2015 ². But the most common reason is also the least severe: only 0.8% of packaging episodes reach recall-Class I, against 11.3% for bioburden/microbial/endotoxin episodes ³. Frequency and severity point at different controls.

Are packaging-related recalls dangerous? Rarely, as recalls: 4 of 532 packaging episodes since 2015 were Class I ³. The larger packaging risk sits outside the recall record: in peer-reviewed simulated-use trials, some sterile-barrier configurations failed aseptic presentation 72–85% of the time in experienced hands ¹¹. The recall database cannot see a package that compromises the sterile field at the point of use.

What is the difference between a Class I recall and a Class I device? Two unrelated scales that share a name. Device class I/II/III is a product's regulatory risk tier (Class I = lowest, e.g. drapes and trays). Recall classification I/II/III is the severity of a specific recall (Class I = most serious). Nearly a quarter of sterility recall episodes involve device-Class I products ²¹⁰ — low device class does not mean low sterility risk.

What evidence should I request from a sterile-device supplier? The linked chain, not a certificate: product/claim boundary; incoming-material and bioburden program; supplier/component controls with change notification; sterile-barrier design file including the aseptic-presentation usability evaluation (ISO 11607-1:2019); sealing-process validation (ISO 11607-2:2019); sterilization validation plus routine release for your exact family; contract-sterilizer interface and capacity contingency; distribution and shelf-life evidence; label/status reconciliation; and recall traceability with a tested mock recall. EU MDR Annex II §6.2(e) names the core tests in law: bioburden, pyrogen and, where applicable, sterilant residues ³⁰.

Does a past recall disqualify a supplier? No — but it redirects the audit. Nine in ten repeat firms concentrate their episodes in one failure category, and 69.1% of next episodes arrive within twelve months ². Ask what the CAPA was and for operating evidence from the year after it. A supplier with a recall and a demonstrable fix can be a better-understood risk than one with neither history nor evidence.

Is there an ethylene oxide shortage? No realized shortage is documented as of mid-2026, but the system runs tight: ~50% of U.S. sterile devices flow through ~90 EtO facilities that the industry describes as operating at capacity, compliance deadlines have shifted three times since 2024 (some now to 2029), and FDA maintains explicit policies to prevent sterilization-driven shortages ^16192127. Buyers should ask suppliers for their sterilization-capacity contingency rather than assume continuity.

Conclusion: audit the chain, not the certificate

Four takeaways carry this report.

Sterility recalls are rare per event and enormous per event. One action averages 5.55 FDA rows — twice the all-device average — and the current quarter's record was set by one 86-million-unit kit recall. For anyone holding inventory or their name on a label, the planning unit is the cascade, not the recall notice ¹².

The record's volume and its severity point at different places. Packaging fills the record and almost never reaches Class I; bioburden and purchased components rarely appear and dominate the severe tier. Episodes described as involving a supplier or component are several times — up to an order of magnitude — more likely to be Class I ²³. Audit attention should follow severity, not frequency.

History is a diagnostic. Repeat firms fail in their signature mode on a five-month median clock. The productive question is never "have you had a recall?" but "show me the CAPA and the twelve months after it" ².

The chain now has a capacity dimension. Half of U.S. sterile devices depend on ninety-odd buildings whose regulatory ground moved three times in two years, and the gamma alternative runs on an imported isotope until the mid-2030s ¹⁶¹⁹²³. "Who sterilizes this, and what is the backup?" belongs in every RFQ.

VEMERIX manufactures sterile, single-use minimally invasive surgical and wound-care devices as the brand of Weihai Medison. We publish this analysis because we sit inside the evidence chain it describes — and we are prepared to be audited against it. If you are qualifying a sterile-device supply chain, ask us for the evidence packet for the exact SKU and market you are evaluating: registration status, packaging and sterilization validation summaries, lot-release records and traceability evidence. The useful conversation starts with the chain, not the certificate.