FDA Enforcement — Recalls & Warning Letters
FDA enforcement by the numbers — and the firms that keep coming back
Of 56,777 FDA recall reports in the database, 86% are Class II (probable but not certain serious harm) and 9.2% are the most-serious Class I. Medical devices account for 68.9% of recalls, drugs for 31.1%. 99.4% of recalls are firm-initiated — only 0.5% are FDA-mandated. Of 4,204 distinct recalled firms, the 109 with 100 or more recalls account for 46.8% of all recall records. Aggregate counts only; no individual named.
Key findings
recall reports in the FDA enforcement database — plus 3,574 warning letters, bringing the total enforcement action count to 89,553. These are public FDA regulatory records, each tied to a specific product, firm, and reason.
of all recalls are Class II — the middle severity tier, meaning probable but not certain serious harm. Class I (the most serious, reasonably likely to cause serious adverse health consequences) accounts for 9.2%; Class III (unlikely to cause harm) for 4.8%.
of recalls involve medical devices (39,095 recalls), with drugs accounting for the remaining 31.1% (17,682 recalls). Sterility and cGMP manufacturing failures dominate drug recall reasons; component and packaging defects lead for devices.
of recalls are firm-initiated — the company pulls the product without being ordered to. Only 0.5% (302 recalls) are FDA-mandated. The FDA rarely orders a recall; the enforcement system depends almost entirely on firms acting before a mandate.
each appear in 100 or more FDA recall records, and together account for 46.8% of all 56,777 recalls. The 233 firms with 50 or more recalls account for 62.2%. Recall volume is highly concentrated: most of the recall record traces to a small set of large-scale regulated manufacturers.
recalls are listed under ZIMMER BIOMET INC — the most-recalled firm in the database, appearing across 9 distinct years from 2012 to 2020. The top 10 firms by recall count all operate over multi-year spans, the longest running 16 years continuously. Firm names here are public FDA records; no individual is named.
At a glance
Recall volume by year (2015–2025)
Calendar-year recall counts, with the Class I / II / III breakdown. The dataset runs through 2026-06; 2026 is a partial year. Volume peaked in 2018–2019 (above 4,600 per year) and dropped in 2020–2021 before stabilizing near 3,600–3,800. The 2023 Class I spike (18.0% of that year's recalls) was the highest Class I share in the ten-year window.
| Year | Total recalls | Class I | Class II | Class III |
|---|---|---|---|---|
| 2015 | 4,640 | 202 | 4,178 | 260 |
| 2016 | 4,228 | 175 | 3,774 | 279 |
| 2017 | 4,454 | 203 | 4,033 | 218 |
| 2018 | 4,694 | 652 | 3,831 | 211 |
| 2019 | 4,798 | 418 | 4,133 | 247 |
| 2020 | 3,532 | 287 | 3,099 | 146 |
| 2021 | 3,412 | 295 | 2,973 | 144 |
| 2022 | 3,667 | 275 | 3,246 | 146 |
| 2023 | 3,723 | 670 | 2,932 | 121 |
| 2024 | 3,832 | 384 | 3,343 | 105 |
| 2025 | 3,608 | 422 | 3,019 | 167 |
By recall classification
The FDA's three-tier severity classification across all 56,777 recall records. Class I is the highest risk tier; Class III the lowest.
| Classification | Recalls | Share |
|---|---|---|
| Class II | 48,820 | 86% |
| Class I | 5,237 | 9.2% |
| Class III | 2,720 | 4.8% |
Class I: reasonably likely to cause serious adverse health consequences or death. Class II: may cause temporary or medically reversible harm. Class III: unlikely to cause harm.
Class I recall share by year
The share of each year's recalls classified as Class I — the most clinically significant signal in the year-over-year trend. The rate jumped in 2018 (13.9%) and again in 2023 (18.0%), the highest in the ten-year window. Pre-2018 years held a narrower 4–5% band.
| Year | Class I recalls | Total recalls | Class I share |
|---|---|---|---|
| 2015 | 202 | 4,640 | 4.4% |
| 2016 | 175 | 4,228 | 4.1% |
| 2017 | 203 | 4,454 | 4.6% |
| 2018 | 652 | 4,694 | 13.9% |
| 2019 | 418 | 4,798 | 8.7% |
| 2020 | 287 | 3,532 | 8.1% |
| 2021 | 295 | 3,412 | 8.6% |
| 2022 | 275 | 3,667 | 7.5% |
| 2023 | 670 | 3,723 | 18% |
| 2024 | 384 | 3,832 | 10% |
| 2025 | 422 | 3,608 | 11.7% |
By product type
Medical devices represent more than two-thirds of all recall records. This reflects the sheer volume and variety of Class II device categories (custom kits, diagnostic reagents, surgical instruments) rather than a higher per-product defect rate.
| Product type | Recalls | Share |
|---|---|---|
| Medical device | 39,095 | 68.9% |
| Drug | 17,682 | 31.1% |
What triggers a drug recall
The most common drug recall trigger is a sterility or contamination concern — the FDA and pharma regulators treat any doubt about a sterile product's sterility assurance as sufficient cause for recall, even before a contamination event is confirmed. cGMP manufacturing deviations (the catch-all for process-control failures) are the second largest category.
| Reason category | Drug recalls | Share |
|---|---|---|
| Sterility / contamination concerns | 6,669 | 37.7% |
| cGMP / manufacturing deviations | 2,982 | 16.9% |
| Contamination (non-sterility) | 2,219 | 12.5% |
| Labeling errors | 1,611 | 9.1% |
| Potency / out-of-specification | 1,100 | 6.2% |
| Temperature / storage excursion | 64 | 0.4% |
| Other | 3,036 | 17.2% |
Reason categories are keyword-derived from the free-text “reason for recall” field; categories are mutually exclusive in the order shown.
What triggers a device recall
Device recall reasons are more varied — more than half fall into the residual “design / component defect” category, reflecting the breadth of device types and failure modes. Sterility and packaging integrity failures are the second-largest cluster, driven by recalls of custom surgical kits and procedure trays.
| Reason category | Device recalls | Share |
|---|---|---|
| Other / design / component defect | 21,156 | 54.2% |
| Sterility / packaging integrity | 6,575 | 16.8% |
| Labeling / IFU errors | 4,190 | 10.7% |
| Leak / seal failure | 2,746 | 7% |
| Software / firmware defect | 1,969 | 5% |
| Component fracture / breakage | 1,390 | 3.6% |
| Contamination | 591 | 1.5% |
| Quality system / validation failure | 429 | 1.1% |
Reason categories are keyword-derived from the free-text “reason for recall” field; categories are mutually exclusive in the order shown.
Voluntary vs FDA-mandated
The FDA rarely mandates a recall (0.5%, 302records). The agency's enforcement model relies on firms acting first — often after an inspection, a field complaint cluster, or an internal quality audit triggers a voluntary pull. A firm-initiated label does not mean the FDA was not involved: in many cases a warning letter or inspection observation preceded the voluntary action.
The firms that keep coming back
Recall volume is highly concentrated. Of 4,204 distinct firms with at least one recall record, 109 (2.6% of firms) each appear in 100 or more records — and together account for 46.8% of all recalls. The 233 firms with 50 or more recalls account for 62.2%.
High recall counts at a single firm reflect primarily two factors: a large regulated product catalog (more products = more surface area for any single quality event) and the FDA's lot-level reporting convention (one recall event can generate hundreds of records if it covers hundreds of product variants). A high record count is an administrative fact about the published database, not an assessment of relative risk.
Top 10 firms by recall count
The ten most-recalled firms in the database. Firm names are public FDA regulatory records published at recalls.fda.gov; no individual is named. Recall counts are administrative record counts, not a severity or defect-rate ranking.
| Firm (FDA public record) | Recalls | Class I | Class I % | Year span | Distinct years |
|---|---|---|---|---|---|
| ZIMMER BIOMET INC | 875 | 3 | 0.3% | 2012–2020 | 9 |
| MEDLINE INDUSTRIES LP - NORTHFIELD | 822 | 72 | 8.8% | 2021–2025 | 5 |
| SIEMENS MEDICAL SOLUTIONS USA INC | 790 | 16 | 2% | 2011–2026 | 16 |
| CUSTOMED INC | 682 | 412 | 60.4% | 2014–2016 | 3 |
| SIEMENS HEALTHCARE DIAGNOSTICS INC | 660 | 13 | 2% | 2010–2026 | 16 |
| ICU MEDICAL INC | 651 | 28 | 4.3% | 2012–2026 | 15 |
| BOSTON SCIENTIFIC CORPORATION | 605 | 138 | 22.8% | 2012–2026 | 15 |
| ARROW INTERNATIONAL INC | 478 | 57 | 11.9% | 2012–2024 | 13 |
| GE HEALTHCARE LLC | 475 | 41 | 8.6% | 2011–2026 | 12 |
| BAXTER HEALTHCARE CORPORATION | 405 | 67 | 16.5% | 2011–2026 | 14 |
Firm names are normalized from the FDA's published recalling firm field. High recall counts reflect large product catalogs and lot-level reporting, not a per-product defect rate. Confirm any product's current recall status at recalls.fda.gov.
Reproduce it
Re-derive every figure on this page from the published artifacts:
- Reproducible SQL — the exact aggregation queries, with expected-result comments.
- Download JSON · Download CSV — the committed aggregate snapshot.
Re-check the source snapshots
Every figure here traces to signed source snapshots, not our word for it. Each daily pull is content-hashed and chained; you can re-hash the published bytes against the attestation yourself.
Re-check a snapshot → — re-hash any Fonteum snapshot and confirm the bytes match the chained attestation.
Methodology
The study aggregates two tables in the fonteum-platform warehouse. Recall counts come from public.fda_enforcement_reports (56,777 rows) keyed on FDA recall number and recall_initiation_date. Warning-letter and total enforcement-action counts come from public.fda_enforcement_actions (89,553 rows). Year breakdowns use the recall_initiation_date field; the 2015–2025 window covers complete calendar years with stable coverage. Firm names use the normalized firm_name_normalized field; each distinct value is one firm. Reason categories are keyword-derived from the free-text reason_for_recall field using case-insensitive pattern matching; categories are evaluated in priority order (first match wins). Classification values (Class I / II / III) are the FDA's own published field. Every published figure is re-derivable from the SQL linked above, whose expected-result comments match the committed JSON snapshot exactly.
How to cite this
Fonteum (2026). FDA Enforcement Trends 2026: Recalls and Warning Letters by the Numbers. Derived from FDA Enforcement Reports and Enforcement Actions databases (snapshot 2026-06-14). https://fonteum.com/gov/research/fda-enforcement-trends-2026
Canonical URL: https://fonteum.com/gov/research/fda-enforcement-trends-2026 · License: U.S. Government Works (public domain; 17 U.S.C. §105)
Related federal enforcement evidence
- The anatomy of the federal healthcare blacklist — 68,055 OIG exclusions, by statutory authority →
- Federal Suspension & Debarment Scorecard — the US exclusion registry, ranked →
- The US + EU Sanctions Universe — composition & growth →
- Federal records questions, answered →
- Government records evidence — all studies →
Limitations
- Recall counts are administrative record counts — one recall event covering many lot numbers generates multiple rows. The firm-level table reflects database records, not unique product-safety events.
- Reason categories are keyword-derived from free-text fields; the FDA does not publish a controlled vocabulary for recall reasons. Category assignments are a best-fit classification, not a formal regulatory designation.
- The year-by-year table covers 2015–2025 complete calendar years. Earlier records exist in the database but coverage is sparser; pre-2015 figures are not included in the trend tables.
- Firm names are normalized from the FDA's published recalling firm field. Entities operating under multiple names or entity structures may appear under more than one normalized name.
- Recall status changes over time. A record counted here reflects the FDA's published database as of the 2026-06-14snapshot. Confirm any specific product's current status at recalls.fda.gov.
Sources
Two FDA enforcement databases, both U.S. Government Works (public domain). The recall reports database covers product recalls reported to the FDA across drugs and medical devices; the enforcement actions database covers recalls plus warning letters.
Source: U.S. Food & Drug Administration — Enforcement Reports and Enforcement Actions (recalls.fda.gov), snapshot 2026-06-14. Public domain (U.S. Government Works). Confirm current recall status at recalls.fda.gov. Confirm current status at SAM.gov →
Frequently asked questions
What is an FDA recall classification, and what does each class mean?
The FDA classifies recalls into three tiers based on health risk. Class I is the most serious: the FDA has determined that use of or exposure to the product will cause serious adverse health consequences or death. Class II covers products where use may cause temporary or medically reversible adverse health consequences, or where the probability of serious harm is remote. Class III applies to products unlikely to cause adverse health consequences. As of this snapshot, 9.2% of the 56,777 records are Class I, 86% are Class II, and 4.8% are Class III.
Why do firms initiate almost all FDA recalls themselves?
The FDA has authority to mandate recalls under the Food Safety Modernization Act (FSMA) and similar statutes, but uses that authority rarely. The FDA's enforcement model relies heavily on voluntary compliance: the agency inspects, issues warning letters, and signals concerns, and most firms respond by pulling products themselves before a mandate is issued. In this database, 99.4% of recalls are labeled firm-initiated, with only 0.5% as FDA-mandated. This structure means recall data reflects both firm-level quality-control failures and the FDA's inspection and warning activity upstream of any formal mandate.
Why do some firms appear in hundreds of recall records?
Each recall record corresponds to a specific product lot, NDC, or device model — not a single event. A large manufacturer that initiates a recall of multiple product variants or lot numbers generates one record per line item. Volume reflects both the scale of a firm's regulated product portfolio and the breadth of any single quality event. The 109 firms with 100 or more recall records are predominantly large device manufacturers and pharmaceutical distributors with extensive product catalogs. Recall volume is an administrative count, not a severity judgment.
What years does this study cover?
The recall database (fda_enforcement_reports) spans back to the 1930s, but the year-by-year breakdown in this study focuses on 2015–2025 — the ten complete calendar years with stable coverage. The snapshot was ingested on 2026-06-14. The full table (56,777 rows) includes earlier records, but pre-2015 coverage is sparser and less uniformly coded.
What is the difference between a recall and a warning letter?
A recall (85,979 of the 89,553 actions in this database) is an action to remove or correct a product that violates FDA regulations. A warning letter (3,574 actions) is a regulatory notice to a firm that the FDA has found significant violations and is requesting corrective action — it precedes but does not require a recall. Warning letters are stored in fda_enforcement_actions; recall records come primarily from fda_enforcement_reports.
Does this study name any individual?
No. Every figure is an aggregate count — by year, classification, product type, reason category, or firm recall volume. Firm names in the top-N table are public FDA regulatory records published at recalls.fda.gov. No individual consumer, patient, employee, or officer is named, and the study produces no per-person report, dossier, or risk score. To confirm a specific product's recall status, search the official FDA recall database at recalls.fda.gov.