Where can I get a list of FDA-cleared medical devices?

The record that answers it is the openFDA Device 510(k) Clearances API: one row per premarket notification a manufacturer filed before marketing a device of that type in the United States. It held 175,814 records at the 2026-08-10 data load, with documented coverage from 1976 to present - the full history of American 510(k) market entry in one queryable table rather than a stack of PDF letters.

Quality scoring puts it at the top of the pool: Datadory rates it 10 out of 10, one of only 145 of the 1,744 datasets Datadory catalogs to reach that mark against a catalog-wide average of 7.81.

One scope note keeps the question honest. A 510(k) is a premarket notification demonstrating substantial equivalence to a predicate, so this list enumerates cleared devices; class III devices that reached market through the Premarket Approval (PMA) pathway sit in separate databases indexed by the FDA Medical Devices Portal. If your definition of "cleared" includes approved class III implants and diagnostics, say so when you request a sample and the delivery gets widened to match.

Each record answers who, what and when in a single flat structure:

Because date_received and decision_date ship on all 175,814 rows, the list doubles as a turnaround dataset without enrichment: K142820 was received 2014-09-30 and decided 2015-04-30, roughly seven months. Sample rows arrive with every delivery so the shape is verifiable before anything ongoing starts.

Every cut below runs on fields already in the table - no enrichment step required:

1. Cut by category. Filter product_code to isolate one device type (LPL for soft contact lenses), or filter advisory_committee for an entire review panel such as Ophthalmic.
2. Cut by maker. Filter applicant and you have that firm's complete clearance history - Abb Optical Group's BIOLENS Sphere contact lens, K142820, decided 2015-04-30 out of Alameda CA, sits in the sample rows.
3. Bound it in time. Pair the filters with decision_date ranges to enumerate entries per year or since a competitor's last launch.
4. Split by pathway. clearance_type plus expedited_review_flag and third_party_flag separate traditional filings from expedited and third-party-reviewed ones.
5. Derive turnaround. Subtract date_received from decision_date per row and the list becomes a review-latency series across five decades.

That five-cut pattern produces a defensible enumerated list per category, per applicant or per year. Working examples of the same cuts are collected on the developers builders use cases page.

A cleared device is not yet a sellable unit, and this is where the list needs a second corpus. The openFDA Unique Device Identifier (GUDID) API is the industry's master-data backbone: 5,083,948 device identifier records, one per model version, phased in from the 2013 UDI rule and refreshed weekly. Records carry brand and catalog numbers, version or model numbers, GMDN terms with definitions, product codes, sterilization attributes and implantable flags.

GUDID's nested premarket_submissions field links each model back to its 510(k), PMA or De Novo number, which makes it the cleanest join key from a row on the cleared-device list to actual catalog units - and commercial_distribution_status distinguishes models still in commercial distribution from those with a commercial_distribution_end_date. One caveat worth planning around: FDA announced in December 2024 that Preferred Term (PT) codes will be removed from GUDID in favor of freely accessible GMDN codes, so build item masters against GMDN now.

A procurement team can therefore answer both halves of the question at once: who was cleared, and exactly which models exist under that clearance today. Both corpora ship keyed for that join out of the box.

Clearance is a point-in-time event, so a static list ages the moment it is exported. The companion screen is the openFDA Device Recalls API: 58,999 recall records at the 2026-08-20 data load, classified from November 1, 2002 to present, and scored a perfect 10 like the clearances corpus. Each event names the recalling firm and FEI number, describes the product, gives the reason and root cause, states the classification, and records a distribution pattern reaching down to state level including international shipments.

For adverse-event context behind the same devices, the related openFDA Device Adverse Events (MAUDE) API adds 25,711,469 reports, pairing safety signals with corrective actions across the post-market picture.

A list of cleared devices says nothing about how many firms could have filed one, which is the gap between the regulatory view and the market view. Three named datasets close it inside the same industry slice:

• Supply base - U.S. Census Bureau Data Portal (quality score 8). County Business Patterns publishes establishment counts, employment and payroll by NAICS industry and geography annually since 1964, with API datasets from 1986; CBP 2023 exposes 28 variables. Attribution strings are mandatory in products built on it.
• Demand side - World Bank Health Topic DataBank (score 9). 658 health indicators across 295 economies; hospital beds per 1,000 people is the standard proxy for device-absorbing care infrastructure.
• Evidence pipeline - ClinicalTrials.gov API v2 - Studies Database (score 10). Roughly 600,000 registered studies overall, 89,921 of them matching a device-intervention query at research time.

Together they take a market-entry analyst from "here is every clearance since 1976" to "here is the manufacturing base, the hospital infrastructure absorbing these devices, and the trials feeding the next round of filings." Requested as one combined delivery, they land as separate typed tables sharing geography keys.

As rows, not exports you babysit. Datadory normalizes the clearance, GUDID, recall and MAUDE corpora into typed tables keyed on the K-number and product code, with the field dictionary attached, so the join from clearance to catalog unit to recall history runs inside one schema instead of four bespoke pipelines.

Upstream loads differ per corpus - monthly on clearances, weekly on identifiers and recalls - but your delivery schedule should follow the decision the data feeds, not the load calendar. A competitive-intelligence desk watching new market entry wants checks as filings surface; a quarterly audit wants one clean pull. Those are separate settings, and neither requires you to track when any particular upstream refresh lands. Watermarking comes standard: each delivery carries its load date, because any stored total count is already a floor.

Competitive-intelligence and market-entry analysts enumerate new clearances by applicant and product code. A new K-number in your category is the earliest structured signal of competitor market entry, and decision-date filtering makes "since their last launch" a one-line cut.

Procurement and catalog managers start from GUDID instead: 5,083,948 model-level records joined to internal SKUs, with the nested premarket_submissions link proving each unit traces back to a lawful marketing pathway.

Quality and sourcing-risk teams run the recall screen continuously rather than quarterly, flagging any clearance whose K-number appears in a new recall event before contracts renew.

Investors and quant researchers treat clearance velocity and recall events as factors - the how investors quants use health care supplies data post reframes the same feeds as R&D-output measures and event studies.

Three limits matter before anyone treats the list as more than it is.

Cleared does not mean active. Clearance proves a ruling happened; commercial_distribution_status in GUDID is what says whether a specific model still ships, and even that is labeler-reported.

Cleared does not mean safe. Recall status freezes after classification beyond Enforcement Report corrections, and MAUDE reports are submitted rather than audited - causality cannot be established from a single report, and under-reporting means incidence cannot be calculated from counts.

The 510(k) list is not the whole regulated market. Class III devices approved through the PMA pathway live outside it, and devices exempt from premarket notification under 21 CFR never appear at all - GUDID marks those with an is_pm_exempt flag. Scope the definition first; the delivery follows the definition.

Start with the health care supplies data guide, the pillar overview of all 20 pooled datasets in this industry - 8 primary plus 12 related - with coverage, field dictionaries and quality scores side by side.