Constat · Stage 1 · Premarket evidence

Find 510(k) predicates—and the evidence FDA actually accepted.

Name: Constat Precedent — FDA 510(k) predicate-device and evidence corpus
Creator: Health AI

Search FDA AI/ML 510(k) and De Novo summaries by intended use, then compare predicate chains, validation design, endpoints, PCCP disclosure, and post-clearance context. Every extracted field stays attached to its FDA source quote.

Presence rates, never pooled.Every figure below is “reported in X of Y audited devices.” Performance values are never averaged across devices — each stays with its analysis unit, task, and a verbatim FDA quote. A device we haven’t parsed yet is queue, never “FDA accepted thin evidence.” This is comparative public-record research, not a recommendation that a device is an appropriate predicate for a submission.

Research method: how to find and compare 510(k) predicate devices · how to compare validation evidence

Search FDA 510(k) predicate devices by intended use

Keyword-based retrieval over the parsed corpus (names, algorithm descriptions, endpoints, source quotes) — not semantic search. Try: · ·

Prefer a directory? Browse every indexed FDA AI/ML clearance record.

Device evidence examples

Review source-quoted FDA clearance evidence

Start with these representative records, then follow their cited predicates and related-device links.

K210187

Overjet Dental Assist

bench and clinical validation evidence

K230144

Denti.AI Detect

three validation studies and two cited predicates

K241681

Overjet Image Enhancement Assist

dental-image enhancement testing

K212519

Overjet Caries Assist

clinical and predicate evidence

K200921

qER

radiology AI validation evidence

Browse recent FDA AI/ML clearances

Latest source-quoted 510(k) and De Novo evidence

newest decision dates first

Alturion Series Diagnostic Ultrasound System

K260667

Philips Ultrasound, LLC

IYN · Radiology · 2026-06-29

ADAS 3D

K253969

Adas3D Medical S.L.

QIH · Radiology · 2026-06-29

Auto-Seg (SO-0012), Spine Auto-Seg (SO-0012)

K253628

Agada Medical, Ltd.

QIH · Radiology · 2026-06-29

SOMNUM (SOMNUM)

K253390

Honeynaps Co., Ltd.

OLZ · Neurology · 2026-06-27

MammoScreen® (5)

K260714

Therapixel

QDQ · Radiology · 2026-06-26

syngo.CT Coronary Cockpit

K253786

Siemens Medical Solutions USA, Inc.

QIH · Radiology · 2026-06-25

Enhanced Boundary for PCCT

K253560

Ge Medical Systems, LLC

JAK · Radiology · 2026-06-24

CMN Capillary Function with Oncology Virtual Expert

K260316

Cercare Medical A/S

QIH · Radiology · 2026-06-22

RX-1 Sleep

K253148

Vivaquant, Inc. Dba Rhythm Express

MNR · Anesthesiology · 2026-06-22

Synapse PACS (7.6.0)

K261713

FUJIFILM Healthcare Americas Corporation

QIH · Radiology · 2026-06-18

Vascular Assist Occlusion Triage (VAOT) (v1.0)

K260729

GuideAI Health

QAS · Radiology · 2026-06-18

ARTHUR

K260190

ROPCA ApS

IYO · Radiology · 2026-06-18

RadOncAI

K253050

Informai

MUJ · Radiology · 2026-06-18

ACUSON Sequoia Diagnostic Ultrasound System;ACUSON Sequoia Select Diagnostic Ultrasound System;ACUSON Origin Diagnostic Ultrasound System;ACUSON Origin ICE Diagnostic Ultrasound System

K260844

Siemens Medical Solutions USA, Inc.

IYN · Radiology · 2026-06-16

MediAI-OA

K253172

Crescom Co., Ltd.

QIH · Radiology · 2026-06-15

AiORTA - Plan v2.0

K254207

ViTAA Medical Solutions, Inc.

QIH · Radiology · 2026-03-30

ECG-AI Pulmonary Hypertension (PH) 12-Lead algorithm (1020)

K252360

Anumana, Inc.

SAT · Cardiovascular · 2026-03-28

S-scan Open (100001800)

K260746

Esaote, S.p.A.

LNH · Radiology · 2026-03-27

Automated Aortic Stenosis Software (AutoAS)

K254161

GE Medical Systems Ultrasound & Primary Care Diagnostics, LLC

POK · Radiology · 2026-03-27

Spectral CT Verida Family

K253649

Philips Medical Systems Technologies , Ltd.

JAK · Radiology · 2026-03-27

Vantage Fortian/Orian 1.5T, MRT-1550, V10.0 with AiCE Reconstruction Processing Unit for MR

K253625

Canon Medical Systems Corporation

LNH · Radiology · 2026-03-27

EPIQ Series Diagnostic Ultrasound System, Affiniti Series Diagnostic Ultrasound System

K253595

Philips Ultrasound, LLC

IYN · Radiology · 2026-03-27

Contour ProtégéAI+

K253270

Mim Software, Inc.

QKB · Radiology · 2026-03-27

Butterfly Gestational Age Tool

K252148

Butterfly Network, Inc.

IYN · Radiology · 2026-03-27

Corpus coverage — all panels

1,506 of 1,466clearances parsed, most-recent-first · 0 in parse queue · latest decision 2026-06-29

The hatched segment is our parse queue, not device data. Queued devices never count in any rate below — every denominator is audited devices only.

Original findings from the corpus

129 days

Median FDA review time, received → decision

Across all 1,146 matched AI/ML 510(k)s; the slowest decile takes 8.5+ months (p90 ≈ 260 days). Computed from openFDA decision dates — not published anywhere else for this cohort.

10.3%

PCCP adoption in 2025 AI clearances

Up from 1.3% in 2023 (8.2% in 2024; 11.2% across 2026 so far). Roughly nine in ten AI/ML clearances still file without a Predetermined Change Control Plan — a live gap as FDA pushes the mechanism. Across all 1,490 parsed 510(k) and De Novo devices, 4.9% carry a PCCP.

What the audited corpus reports

Evidence reporting

Any sensitivity metric432 of 1,506

Canonicalized — includes per-finding sensitivities, not just the summary's top-line slot.

Sensitivity and specificity357 of 1,506

Any performance metric790 of 1,506

Clinical (not bench-only) data1,067 of 1,506

Subgroup / demographic reporting839 of 1,506

Predetermined Change Control Plan74 of 1,506

“Not reported” is a real finding about what FDA accepted — the summary was audited and does not state it.

Metric presence (canonicalized)

sensitivity432 of 1506
specificity361 of 1506
auroc194 of 1506
ppv141 of 1506
npv82 of 1506
accuracy171 of 1506
f127 of 1506
dice244 of 1506
iou10 of 1506
detection_rate9 of 1506

Presence only — values are never pooled or averaged across devices. Each measurement lives on its device's page with its analysis unit and verbatim quote.

Median predicate age: 1.9y across 1682 of 2577 predicate edges (65% datable, dates backfilled from openFDA) — how old each cited predicate was when the child device cleared.

Study design & generalizability

Where FDA review is tightening — and what almost no premarket dataset reports. Each rate is presence over audited devices; a device that describes no clinical study is honestly “not reported” here.

Prospective or reader (MRMC) study254 of 1,506

Primary validation is prospective-clinical or a multi-reader design — not a retrospective bench run.

Multi-site validation647 of 1,506

External / out-of-distribution testing744 of 1,506

Validation data explicitly independent of the training/development set.

Scanner / device diversity700 of 1,506

Tested across more than one scanner or device manufacturer.

Subgroup performance broken out741 of 1,506

Performance reported per subgroup, not just an overall number.

Extraction quality. Each reported field includes its FDA source quote. Across all 1,490 parsed 510(k) and De Novo devices we check every one of ~25,600 quotes against its source summary: 98.3% grounded (99% mean token recall; a reproducible, published method). Two records are also hand-audited against the full documents at 100% field accuracy, including a deliberate no-new-testing device to catch invented metrics. We measure fabrication, not interpretation — and publish the method.

One corpus, whole lifecycle

This page

Clearance evidence

1506 devices parsed with source-quoted fields

Stage 2

Postmarket signals

1,519 devices snapshotted · 1192 of 1519 carry a drift signal

Stage 3

Reimbursement pathways

40 pathways across 37 devices (seed corpus, growing)

Every device page connects all four stages: clearance → evidence → postmarket → payment. Stages without data render as “not yet tracked”, never as empty or fabricated.

Machine access

The same retrieval over MCP — 13 FDA/CMS tools across the lifecycle

curl -s https://constat.dev/api/mcp \
  -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call",
       "params":{"name":"evidence_search",
                 "arguments":{"panel":"Radiology","reports_any_sensitivity_metric":true}}}'

device_evidence_lookup, evidence_search, predicate_chain, device_postmarket_lookup, reimbursement_lookup, and more. Every extracted field carries a source quote and page — descriptive only, never a compliance judgment.

Connect the MCP endpoint

Predicate-device search questions

How to use FDA clearance evidence without losing the source

What is a 510(k) predicate device?

A predicate is a legally marketed device used in a 510(k) submission to support a substantial-equivalence comparison. FDA recommends starting from the device classification and recently cleared devices. Predicate suitability still requires a submission-specific regulatory assessment.

How do I find possible predicate devices?

Start with the intended use, FDA product code, classification, and recent 510(k) clearances. Then compare indications, technological characteristics, validation design, and the predicates cited by each candidate. Constat supports this public-record comparison; it does not select a legally appropriate predicate for you.

How is Constat different from the FDA 510(k) database?

FDA is the authoritative clearance source. Constat adds source-quoted structured fields from public summaries, intended-use search, predicate chains, validation-study details, postmarket context, and reimbursement links while preserving citations back to the source record.

Authoritative starting point: FDA guidance on finding and using predicate devices.

Need a source-linked 510(k) evidence comparison?

Share the intended use, product code, or candidate device. Health AI will scope a bounded comparison of relevant clearances, predicates, validation evidence, and disclosed limitations.