FDA Cleared 3,052 Devices Via Shortcut in 2024 — and 31 via PMA

The problem: When a US medical device makes the news for being recalled, killing a patient, or generating a class-action lawsuit, almost no coverage explains how the device actually got onto the market. Most readers — and many reporters — assume "FDA-approved" means the device went through clinical trials. For 95-98% of US-marketed devices, it didn't. It went through a 1976-vintage shortcut called 510(k), which only requires the manufacturer to show the device is "substantially equivalent" to one already on the market.

What this post is. A documentary audit of FDA 2024 data. Three numbers: how many devices entered the market via 510(k), how many entered via the rigorous PMA pathway, and what share of 2024 recalls involved devices that had not undergone PMA review. Why it matters: the asymmetry between the two pathways is large, the recall distribution is skewed, and the data is freely available via openFDA but rarely pulled together in one place. Use it when: reporting on a specific device recall, FDA reform debate, medical-device safety policy, or any story that turns on how devices reach the US market.

Key findings

• 3,052 medical devices cleared via 510(k) in 2024, against ~31 approved through the full PMA pathway.
• 98-to-1 ratio of fast-track clearances to clinical-trial-required approvals (originals only).
• Non-PMA pathways accounted for 96.8% of 2024 recalls — 3,169 of 3,275 total recalls.
• 2.6 million device adverse event reports logged in MAUDE in 2024 — roughly one every 12 seconds.
• Average 510(k) review: 168.9 days. Average PMA review: 363.2 days (down from 760.8 in 2023).
• Industry baseline: 95-98% of all US-marketed medical devices reach market through 510(k), per longstanding HHS OIG and trade-press estimates.
• Headline ratio applies to originals only — openFDA returns 2,230 PMA "decisions" for 2024, but most are supplements amending existing PMAs, not new devices.

In this article: The asymmetry · What 510(k) actually is · Recalls · Adverse events · Methodology · Caveats · Sources · FAQ

Quick answer

• What this is: A 2024 audit of three openFDA endpoints — device/510k, device/pma, device/recall — plus the device/event (MAUDE) endpoint.
• The headline: In 2024 the FDA cleared 3,052 medical devices via 510(k) and approved 31 via the PMA pathway. Devices that did not go through PMA accounted for 97% of 2024 recalls.
• What "substantial equivalence" means: A new device can be cleared by demonstrating it is similar to a device already on the market — no clinical trial required.
• What this post does not claim: That 510(k) is the cause of the recall asymmetry. The recall counts reflect device-population scale as well as pathway risk; the per-device rate would require additional analysis we did not do.
• Main caveat: "Non-PMA pathway" includes 510(k), De Novo, exempt, HDE, and pre-amendment devices. The 96.8% figure is "non-PMA," not "510(k)-only."

The asymmetry: 2024 FDA clearances vs approvals

The two main pathways for getting a medical device onto the US market are 510(k) clearance and Premarket Approval (PMA). They are not interchangeable. PMA requires clinical trials, manufacturing inspections, and a panel review; 510(k) requires demonstrating substantial equivalence to an existing device. The 2024 totals show how dominant the shortcut has become.

Sources: openFDA device APIs, MD+DI 2024 medtech FDA approval volume analysis. Figures based on data publicly available as of May 2026 and may be revised when CDRH publishes its full-year totals.

In other words: of the roughly 3,083 unique device authorisations in 2024, 98.7% went through 510(k) and about 1.0% went through full PMA. The remaining ~1.5% were De Novo classifications, a hybrid pathway for novel low-to-moderate-risk devices that have no predicate.

The headline asymmetry — 98 to 1 — applies to originals. If you include PMA supplements (amendments to existing PMAs, not new devices), the PMA "decision" total rises to 2,230. That is the more flattering number for the rigorous pathway, but it is not what readers usually mean by "FDA-approved": most of those decisions modify devices that were already on the market.

Five-year trend — the asymmetry is structural, not a 2024 anomaly

Querying the same openFDA endpoints across each calendar year from 2020 to 2024:

Source: openFDA device/510k, device/recall, device/event endpoints, queried by decision_date / event_date_initiated / date_received per calendar year, captured 8 May 2026.

Three patterns:

1. 510(k) clearance volume is roughly stable at 3,000-3,400 per year. The 2024 figure (3,129) is in line with the 5-year range, not an outlier.
2. Recall volume jumped 26% in 2024 — from a 2020-2023 average of about 2,435 per year to 3,275 in 2024. Whether this is FDA enforcement scaling up or genuine product-quality regression is a separate analysis we do not perform here.
3. Adverse-event reporting grew 12% from 2023 to 2024 (2.34M → 2.63M). MAUDE reporting is voluntary; the rise reflects a combination of larger active-device populations and easier reporting tools (the FDA's eMDR system has been progressively simplified).

The 5-year trend reinforces, rather than challenges, the headline. The 510(k) shortcut handles essentially the entire device market in every year of the window — the dominant pathway by orders of magnitude — and recall and adverse-event volumes scale with the device population it produces.

What 510(k) actually is — the substantial-equivalence pathway

Section 510(k) of the Federal Food, Drug, and Cosmetic Act, enacted in 1976, requires a manufacturer to notify the FDA at least 90 days before marketing a new device. The submission must demonstrate the device is "substantially equivalent" to a legally marketed predicate device — meaning it has the same intended use and either the same technological characteristics or different characteristics that do not raise new questions of safety and effectiveness. If the FDA agrees, it issues a clearance letter. The device can then be marketed without any clinical trial data on the new device itself.

PMA is the alternative for higher-risk devices (Class III). It requires the manufacturer to submit clinical evidence demonstrating safety and effectiveness, and the FDA's review typically includes manufacturing facility inspections and an advisory-panel meeting. PMA is what most lay readers picture when they hear "FDA-approved." The official FDA page on the difference is here.

The structural critique of 510(k) is the predicate chain. A device cleared in 2024 is usually substantially equivalent to a predicate cleared in, say, 2018. That predicate was substantially equivalent to one cleared in 2010. That one to a 1998 predicate. And so on. Some chains trace back to pre-1976 devices that were grandfathered into the market and never independently reviewed. The International Consortium of Investigative Journalists' Implant Files investigation in 2018 documented this dynamic across multiple device categories. The HHS Office of Inspector General has flagged similar concerns going back to its 2013 report on FDA's 510(k) process.

The FDA's defence of the pathway is that it works in practice: most cleared devices generate few problems, the 90-day review cycle is appropriate for incremental design changes, and full PMA would be impractical for the volume of moderate-risk devices entering the market each year. The 2024 numbers in this post are not, in themselves, a verdict on that argument. They are the input to it.

Story A — the shortcut handles essentially the entire device market

Industry estimates place 95-98% of US-marketed medical devices on the 510(k) pathway. The 2024 originals counts are consistent with that range: 3,052 of approximately 3,131 originals (510(k) + PMA + De Novo) is 97.5%.

The pathway dominates not because PMA was abolished, but because PMA reviews scaled down. In the early 2000s the FDA approved roughly 40-60 PMA originals per year. The 2024 figure of ~31 is in line with the lower end of the recent range. PMA review times have improved sharply — 363.2 days in 2024 vs 760.8 days in 2023 — but the volume has not bounced back.

The FDA itself acknowledges the 510(k) share. The agency's own 510(k) clearances database publishes the running total. Trade press tracking (NAMSA, MD+DI, the FDA Group) reports the same range. This post does not surface a contested number — it surfaces a widely-cited number against the underlying openFDA totals so the asymmetry is in one place.

Story B — devices that skipped PMA review account for 97% of recalls

In 2024, openFDA's device/recall endpoint returned 3,275 recalls. Of those, 106 (3.2%) involved a PMA-pathway device, identified via the _exists_:pma_numbers filter. The remaining 3,169 (96.8%) did not have a PMA flag — meaning the recalled device went through 510(k), De Novo, exempt classification, HDE, or pre-amendment grandfathering.

Source: openFDA device/recall endpoint, 2024 calendar year, queried 8 May 2026.

Two things about that 96.8% figure:

1. It is not an indictment of 510(k) on its own. The "non-PMA" bucket includes Class I exempt devices (sutures, examination gloves, tongue depressors), which generate routine quality-control recalls in large volumes. Without a per-pathway breakdown of recalls — which openFDA's recall schema does not directly provide — the 96.8% should be read as "devices that did not undergo PMA review" rather than "510(k)-cleared devices."
2. It does not, by itself, prove pathway-causation. Non-PMA devices represent ~98% of authorisations; the recall share is roughly proportional to the authorisation share. The recall rate per pathway, normalised by cumulative clearance population since each pathway's 1976 inception, tells a different story.

Recall rate per 1,000 cumulative cleared devices

Cumulative clearance counts retrieved 8 May 2026 via the fda-510k-clearances and fda-pma-approvals actors in aggregateBy: decision_date.year and aggregateBy: supplement_type modes respectively.

The PMA-pathway per-device recall rate is 4× higher than non-PMA. That sounds backwards: PMA is the rigorous pathway. The interpretation is that PMA covers only Class III devices — the highest-risk category. Per device, those devices generate more recalls; in absolute terms, the 510(k) pathway dominates because its volume is two orders of magnitude larger.

The headline (96.8% of recalls are non-PMA) holds. The rate analysis surfaces the nuance a regulatory reporter will ask about: pathway-rate is not pathway-volume.

A second caveat on this rate analysis: cumulative-clearance-since-inception over-counts the active device population (some devices retire, get superseded, or were never marketed). A more precise rate calculation would back out actively-marketed devices using FDA establishment-registration data, which this post does not do. The 18.1 / 72.7 figures should be read as order-of-magnitude per-pathway-rate comparisons, not 3-significant-digit precision.

What it does establish is that the population of devices generating US recall coverage week after week — saline bag contamination, surgical-staple misfires, hip-implant fractures, infusion-pump dosing errors — is overwhelmingly the population that did not go through clinical-trial-grade review. That fact is the input to the policy question; it is not the answer.

Top recalling firms in 2024

The named-entity breakdown of the recalls dataset is its own news. The eight firms below initiated 1,142 of the 3,275 total 2024 recalls — 35% of the year's total, distributed across companies whose names regularly appear in mainstream medical-device coverage.

Source: openFDA device/recall aggregated by recalling_firm.exact, 2024 calendar year. Single-firm recall counts include all recalls initiated by the firm regardless of severity classification.

Medline Industries — the largest single recalling firm in 2024 by a wide margin — is a global medical-supply distributor and manufacturer whose product range covers exam gloves, surgical gowns, wound-care products, and patient-handling equipment. Its 382 recalls in one year reflects the volume of low-risk Class I and Class II products it manufactures, not necessarily a higher per-product failure rate; without normalising by units shipped, the ranking is by raw recall count, not by quality. Boston Scientific (#3) and Philips North America (#4) are large device manufacturers with diversified high-risk product lines, and their recall counts include several Class I designations — the most serious.

Mini case study — the Philips Mobile Cardiac Telemetry recall, 2024

A concrete example is useful to make the abstract numbers tangible. The 2024 FDA-classified Class I recall of Philips' Mobile Cardiac Outpatient Telemetry (MCOT) monitoring service sits in the dataset as one row of the 3,275 total 2024 recalls. It is also the kind of recall that most readers would assume came from a device cleared via clinical trial.

The device: Philips' BTPS-1000 wearable sensor patch, paired with the company's MCOT monitoring service application — software that processes streaming ECG data from the patch and routes critical events to cardiology technicians. About 130,000 BTPS-1000 patches had been distributed in the US to more than 41,200 healthcare-provider customers. The recall ultimately affected approximately 21,000 devices.

The clearance pathway: Philips' family of cardiac telemetry monitoring devices is regulated under 21 CFR 870.1025 — a Class II classification. The relevant openFDA recall records for the Philips IntelliVue Information Center family (the predecessor product line in the same regulation) carry no PMA numbers, indicating the devices entered the market via 510(k) substantial-equivalence clearance rather than the PMA pathway. The MCOT software changes followed the same regulatory regime.

What went wrong: Between July 2022 and July 2024 — a two-year period — the MCOT software failed to flag certain critical ECG events for cardiology-technician review. Patient data was being collected as expected; the analysis-and-routing step was not surfacing every dangerous arrhythmia. Philips alerted healthcare providers in November 2024 and the FDA classified the recall as Class I — the most serious designation, indicating reasonable probability that use of the device will cause serious adverse health consequences or death.

The reported harm: 109 patient injuries and 2 deaths have been associated with the routing failure to date. Both deaths were tied to the device's failure to flag critical ECG alerts that, had they been surfaced and reviewed, would have triggered clinical intervention.

Why this case is the right shape for the data story: The MCOT failure is not a manufacturing defect, a battery issue, or a packaging error. It is a software-routing failure in a device whose entire clinical value proposition is reliable software-routing of cardiac signals — and that device entered the US market under a substantial-equivalence pathway that does not require manufacturers to demonstrate, with clinical evidence on the new device itself, that the routing software performs as claimed. The post above does not argue that PMA review would have caught this specific failure (the failure mode emerged years after market entry, in production-scale data). It does establish that this is the regulatory environment 510(k)-cleared cardiac monitoring devices operate in.

The Philips recall is not anomalous in the 2024 dataset — it is one Class I recall out of several hundred Class I designations the FDA issued during 2024. It is selected here because it has the unusual combination of a named global manufacturer, a software-routing failure mode, a public Class I designation, and confirmed patient deaths — making it citable in a way most recall entries are not.

Story C — 2.6 million adverse event reports, one every 12 seconds

The FDA's Manufacturer and User Facility Device Experience (MAUDE) database collects voluntary reports of medical-device-associated injuries, deaths, and malfunctions. In 2024 the database received 2,627,150 reports.

That works out to:

• One report every 12 seconds, around the clock
• Roughly 7,200 reports per day
• About 217,000 reports per month

The MAUDE figure is a floor, not a ceiling. Adverse-event reporting is voluntary for healthcare professionals and patients, mandatory only for manufacturers and device user facilities, and well-documented as under-reported in the medical-device safety literature. The FDA itself notes the limitations on the MAUDE landing page.

The 2.6 million figure also covers everything from sutures and glucose-monitor strips to surgical robots and pacemakers. The size of the number reflects the size of the active US device population (hundreds of millions of devices in use) at least as much as it reflects per-device safety. Treat it as scale context, not a per-device hazard rate.

Methodology

All figures in this post were pulled from openFDA's public REST APIs. openFDA is a public-data initiative operated by the FDA itself; no authentication is required and no commercial terms apply. The exact endpoints queried for the 2024 figures are listed below.

The queries were issued via four ApifyForge actors that wrap the openFDA endpoints with pagination, retry, and schema validation: fda-510k-clearances, fda-pma-approvals, fda-device-events, and fda-device-recalls. The wrappers do not transform the data; they call the same endpoints anyone can hit directly. The headline figures were re-verified with raw curl requests against openFDA before publication.

Endpoints queried:

• https://api.fda.gov/device/510k.json?search=decision_date:[2024-01-01+TO+2024-12-31]&limit=1 — 2024 510(k) decisions
• https://api.fda.gov/device/pma.json?search=decision_date:[2024-01-01+TO+2024-12-31]&limit=1 — 2024 PMA decisions (note: includes supplements)
• https://api.fda.gov/device/recall.json?search=event_date_initiated:[2024-01-01+TO+2024-12-31]+AND+_exists_:pma_numbers&limit=1 — PMA-flagged recall count
• https://api.fda.gov/device/recall.json?search=event_date_initiated:[2024-01-01+TO+2024-12-31]+AND+_missing_:pma_numbers&limit=1 — non-PMA recall count
• https://api.fda.gov/device/event.json?search=date_received:[2024-01-01+TO+2024-12-31]&limit=1 — MAUDE adverse-event report total

For each endpoint, the value used in this post is the meta.results.total field returned by the API, queried 8 May 2026.

The 510(k)-vs-PMA originals split is taken from the MD+DI 2024 medtech FDA approval volume analysis, which extrapolates full-year originals from the FDA's reported H1 2024 figures (1,507 510(k); 13 PMA originals; 24 De Novo). openFDA does not directly distinguish originals from supplements in the PMA decisions endpoint — that distinction is the single largest source of ambiguity in the 2024 figures and is the reason the post leads with the originals split rather than the raw device/pma decisions count.

Industry baseline (95-98% via 510(k)) is the figure cited consistently across HHS OIG reporting, FDA-watcher trade press (MD+DI, NAMSA, the FDA Group), and academic medical-device policy literature. The figure has been broadly stable over the past two decades.

The CDRH 2024 Annual Report — the FDA's own published full-year totals — was not yet available in PDF form at the time of writing. When it publishes, the 2024 originals counts in this post should be reconciled against it. The relevant index page is here.

Caveats

The data here is solid enough to ground reporting on, but five caveats matter for any journalist who plans to use it.

Non-PMA ≠ 510(k). openFDA's recall schema flags whether a device went through PMA. Everything else lands in a single "non-PMA" bucket, which includes 510(k)-cleared, De Novo, exempt (Class I), HDE, and pre-amendment devices. The 96.8% headline is "non-PMA," not "510(k) only." Do not write "510(k)-cleared devices accounted for 97% of recalls" — that is a stronger claim than the data supports.

Recall counts are not distinct-device counts. A single product type can generate multiple recalls across different lots and distribution channels. Recall counts therefore overstate the count of distinct problematic devices. For per-device rate analysis, normalise by lot-recall vs product-recall.

Adverse-event reports are voluntary. The 2.6 million MAUDE figure is what was reported. The actual incidence of device-associated injuries and malfunctions is meaningfully higher; under-reporting is well-documented in the medical-device safety literature.

PMA supplements vs PMA originals. The openFDA device/pma endpoint returns 2,230 decisions for 2024. The ~31 figure used in the headline refers to original PMAs — new devices entering the market — and is taken from MD+DI's full-year extrapolation. Both numbers are correct for their definitions; readers care about originals.

Causation is not in the data. This post documents the asymmetry between pathways and the recall distribution. It does not claim that 510(k) clearance causes harm or that PMA prevents it. The recall counts reflect device-population scale at least as much as pathway risk. The per-pathway recall rate (recalls per 1,000 active devices, normalised over time) is a separate computation we did not do here.

Data gaps a follow-up could close

Three things would tighten the analysis:

1. A 2020-2024 trend. Five years of (510(k), PMA originals, recalls, MAUDE reports) totals would show whether the asymmetry is widening, stable, or narrowing. The openFDA queries are trivial to repeat with adjusted date filters.
2. Top manufacturers and product codes by 2024 recall count. openFDA supports count=recalling_firm.exact and count=product_code.exact aggregations. Named-entity rankings convert abstract numbers into reporting leads.
3. Active-device-population reconciliation for the per-pathway rate. The 18.1 / 72.7 per-1,000 rates above use cumulative-cleared-since-1976 as the denominator — an over-count of the actively-marketed population. A more precise rate would back out devices that have been retired, superseded, or never marketed using FDA establishment-registration data. Available follow-up.

When this matters for reporting

• A specific recall is in the news and you want to put it in context — name the pathway the device cleared through, and use the 2024 totals as the comparator.
• An FDA reform bill or hearing is on the agenda — the 510(k) vs PMA split is the structural fact most reform debates turn on.
• A manufacturer is contesting a recall or class-action by pointing to its FDA "approval" — check whether the device was cleared via 510(k) or approved via PMA. They are not the same thing, and the distinction matters legally.
• An academic or policy paper cites the "95-98% of devices via 510(k)" range — this post grounds that range against the most recent year of openFDA totals.

Frequently asked questions

What is the difference between FDA 510(k) clearance and FDA PMA approval?

510(k) clearance and PMA approval are two different routes to the US medical-device market. 510(k) requires the manufacturer to demonstrate the device is "substantially equivalent" to a legally marketed predicate device — no clinical trial on the new device is required. PMA requires clinical evidence of the device's own safety and effectiveness, plus typically a manufacturing inspection and an advisory-panel review. Most lay use of "FDA-approved" describes PMA; most US devices reach the market via 510(k).

What percentage of US medical devices go through clinical trials?

Industry estimates and HHS OIG reporting place 95-98% of US-marketed medical devices on the 510(k) pathway, which does not require clinical trials on the new device. The remaining few percent go through PMA (which requires clinical evidence) or De Novo (a hybrid pathway). In 2024, ~31 of approximately 3,131 originals went through PMA — about 1%.

Why does the openFDA PMA endpoint show 2,230 decisions in 2024 if there were only 31 PMA approvals?

Because the openFDA device/pma endpoint counts every PMA-related decision — most of which are supplements amending existing PMAs, not new devices. The ~31 figure used in this post refers to original PMAs (new devices entering the market) and is taken from MD+DI's full-year extrapolation of FDA's H1 2024 figures. Both numbers are correct for their definitions; the originals figure is the relevant one for the "how many new devices got rigorous review" question.

Did 510(k)-cleared devices cause 97% of 2024 recalls?

The data does not support that phrasing. What the openFDA recall endpoint shows is that 96.8% of 2024 recalls involved devices that did not undergo PMA review — a bucket that includes 510(k), De Novo, exempt (Class I), HDE, and pre-amendment devices. Devices that skipped PMA also represent ~98% of authorisations, so the recall share is roughly proportional to the authorisation share. A causal claim about 510(k) specifically would require per-pathway recall rates normalised by the actively-marketed device population. The cumulative-clearance-based rates in Story B (non-PMA 18.1 per 1,000 vs PMA 72.7 per 1,000) are a useful approximation but over-count active devices.

Where does the "one adverse event every 12 seconds" figure come from?

The FDA's MAUDE database received 2,627,150 device adverse-event reports in 2024. Divided across a calendar year (31.6 million seconds), that's one report every ~12 seconds. The figure is a floor, not a ceiling — adverse-event reporting is voluntary for healthcare professionals and patients, and under-reporting is well-documented in the medical-device safety literature. The MAUDE database is queryable via openFDA's device/event endpoint.

Has the 510(k)-vs-PMA balance always looked like this?

The 510(k) pathway has been dominant since shortly after its 1976 creation, and the share of devices going through it has been broadly stable in the 95-98% range for two decades. PMA originals have ranged roughly 30-60 per year over the past decade. What has changed is review time: PMA average review fell from 760.8 days in 2023 to 363.2 days in 2024, a sharp improvement. The volume balance has not shifted.

What is the predicate chain critique of 510(k)?

A 510(k) clearance is granted on the basis of substantial equivalence to a predicate device that was itself usually cleared via 510(k). Trace the predicates of predicates back far enough and many chains end at pre-1976 devices that were grandfathered into the market without independent FDA review. The ICIJ Implant Files investigation in 2018 documented this dynamic across hip implants, surgical mesh, and other categories. Whether the chain matters in practice is contested; it is, however, the canonical structural critique of the pathway.

Sources

• FDA — 510(k) Clearances
• FDA — PMA Approvals
• FDA — Medical Device Safety and the 510(k) Clearance Process
• FDA — CDRH Annual Reports
• openFDA — Device APIs
• openFDA — MAUDE adverse event endpoint
• openFDA — device recall endpoint
• HHS OIG — FDA's Clearance of Medical Devices Through the 510(k) Process (2013)
• MD+DI — 2024 Medtech FDA Approval Volume Trends Down
• ICIJ — The Implant Files

Related ApifyForge analyses of public-data corpora: WHO GHO API, UN Comtrade trade data, World Bank project data.

For journalists — press lift-out paragraph

A May 2026 ApifyForge analysis of FDA openFDA data found that 3,052 medical devices reached the US market via the 510(k) substantial-equivalence shortcut in 2024, against just 31 devices approved through the rigorous PMA pathway with clinical trials — a 98-to-1 ratio. In the same year, devices that did not undergo PMA review accounted for 3,169 of 3,275 medical device recalls (96.8%), and the FDA logged 2.6 million device adverse-event reports — about one every 12 seconds.

Reuse without permission. Suggested attribution: "ApifyForge analysis of openFDA data, May 2026." Source endpoints and methodology in the Methodology section above. For data questions, replication, or follow-up cuts (per-manufacturer recall counts, per-product-code aggregates, multi-year trends), email [email protected].

Ryan Clinton publishes Apify actors and MCP servers as ryanclinton and writes about open-data audits and regulatory monitoring at ApifyForge.

Last updated: May 2026.

This post focuses on US FDA data, but the pattern of "documented dual-pathway regulation, asymmetric recall distribution" applies to other jurisdictions and other regulated product categories where parallel approval routes exist.