FDA MRI Safety Labeling and GUDID Status: Data Analysis of 5.15M Device Records
Data analysis of 5.15 million FDA GUDID records evaluating MRI safety status distribution, ASTM F2503 labeling rules, implantable device coverage gaps, and 2023 guidance compliance.
When a patient with a metallic, conductive, or electronic medical device undergoes magnetic resonance imaging (MRI), the device can be subject to radiofrequency (RF)-induced heating, magnetically induced displacement force and torque, gradient-induced voltages, and image artifact. Two separate instruments address this in the United States: a binding regulation — 21 CFR 830.310(b)(8), which requires labelers to report to GUDID whether a patient may be safely exposed to MRI while using the device or while it is implanted — and a non-binding guidance — FDA's October 2023 Testing and Labeling Medical Devices for Safety in the Magnetic Resonance (MR) Environment, which recommends how to test and how to word the labeling.
That distinction is the whole story of this article. The obligation to report a status is mandatory. The obligation to have done the testing is not, in the general case. And an empirical scan of the full GUDID database shows exactly what that gap produces: more than 90% of published device identifiers in GUDID — and nearly 80% of implantable device records — report that labeling does not contain MRI safety information.
Direct Answer: What Do 5.15 Million GUDID Records Reveal?
Under 21 CFR 830.310(b)(8), labelers must report to GUDID "whether a patient may be safely exposed to magnetic resonance imaging, nuclear magnetic resonance imaging, or magnetic resonance tomography while using the device, or while the device is implanted in patient." GUDID accepts four controlled values for the MRI Safety Status element: MR Safe, MR Conditional, MR Unsafe, and Labeling does not contain MRI Safety Information.
MedDeviceGuide streamed the complete AccessGUDID full database release (207 XML parts, July 2026 release) and tallied every published Primary Device Identifier record — 5,152,799 in total:
- Labeling does not contain MRI Safety Information: 4,650,874 records (90.26%)
- MR Conditional: 260,550 records (5.06%)
- MR Unsafe: 162,998 records (3.16%)
- MR Safe: 78,377 records (1.52%)
Filtering to records where the GUDID implantable flag is true (1,135,724 records), 902,914 (79.50%) carry no MRI safety information, 178,441 (15.71%) are MR Conditional, 30,947 (2.72%) MR Safe, and 23,422 (2.06%) MR Unsafe. Restricting further to implantable records in active commercial distribution (1,014,251 records) barely moves the picture: 78.73% (798,484) carry no MRI safety data, 16.57% (168,013) are MR Conditional, 2.65% (26,832) MR Safe, and 2.06% (20,922) MR Unsafe. Discontinued legacy records are not what is driving the gap.
[!NOTE] Which release, and why the total differs from our earlier GUDID analysis. This scan uses the July 2026 AccessGUDID full release. Our earlier analysis of GUDID device identifier volume, classes and labeler concentration used the release available in June 2026 and reported 5,083,948 records. The 68,851-record difference is normal database growth between releases, not a discrepancy. GUDID totals should always be quoted with the release they came from.
┌─────────────────────────────────────────────────────────────────────────────┐
│ OVERALL GUDID MRI SAFETY STATUS (5.15M RECORDS) │
├─────────────────────────────────────────────────────────────────────────────┤
│ Labeling does not contain MRI Info ████████████████████████████████ 90.26% │
│ MR Conditional ███ 5.06% │
│ MR Unsafe ██ 3.16% │
│ MR Safe █ 1.52% │
└─────────────────────────────────────────────────────────────────────────────┘
┌─────────────────────────────────────────────────────────────────────────────┐
│ COMMERCIAL IMPLANTABLE DEVICE MRI STATUS (1.01M RECORDS) │
├─────────────────────────────────────────────────────────────────────────────┤
│ Labeling does not contain MRI Info ██████████████████████████████ 78.73% │
│ MR Conditional ███████ 16.57% │
│ MR Safe █ 2.65% │
│ MR Unsafe █ 2.06% │
└─────────────────────────────────────────────────────────────────────────────┘
What Is the FDA Guidance on MRI Safety Testing and Labeling?
FDA’s final guidance Testing and Labeling Medical Devices for Safety in the Magnetic Resonance (MR) Environment was issued on October 10, 2023 under docket FDA-2019-D-2837. It supersedes the December 11, 2014 guidance Establishing Safety and Compatibility of Passive Implants in the Magnetic Resonance (MR) Environment — an important scope change, because the 2014 document covered passive implants only, while the 2023 guidance covers all medical devices that might be used in the MR environment.
Scope of the Guidance
The guidance applies to any medical device that might enter an MR environment: items permanently implanted in tissue (orthopedic plates, vascular stents, pacemakers), devices fastened to or worn by a patient (insulin pumps, wearable monitors), and equipment brought into the scanner room by staff (IV poles, patient monitors, surgical instruments). Being non-implantable is not, by itself, a reason to skip the assessment.
Core Terminology and ASTM F2503 Definitions
FDA adopts the terminology and iconography of ASTM F2503 (Standard Practice for Marking Medical Devices and Other Items for Safety in the Magnetic Resonance Environment). Note the edition drift: the October 2023 guidance cites both ASTM F2503-20 and ASTM F2503-23, so confirm the edition currently listed in FDA's Recognized Consensus Standards database before writing a declaration of conformity.
- MR Safe: An item that poses no known hazards resulting from exposure to any MR environment. MR Safe items are composed entirely of materials that are electrically non-conductive, non-metallic, and non-magnetic (for example pure plastic, ceramic, or silicone).
- MR Conditional: An item with demonstrated safety in the MR environment within defined conditions. Those conditions must specify at minimum the static field strength, spatial field gradient, time-varying gradient field limits, RF excitation mode, maximum whole-body or local SAR and/or B1+rms, coil type, and any scan-duration or anatomical restriction. A bare "MR Conditional" claim with no stated conditions is not a usable claim.
- MR Unsafe: An item that poses unacceptable risk to the patient, medical staff, or other persons within the MR environment (ferromagnetic surgical tools, external pacemakers, unshielded electronic monitors).
- Not a fourth ASTM term — an FDA labeling option. For devices that have historically provided no MRI safety information, FDA describes labeling stating that the device "has not been evaluated for safety in the MR environment," that it has not been tested for heating or unwanted movement, and that performing an MR exam may result in injury or device malfunction. FDA says you should NOT use that option if (a) there are any known adverse effects or adverse events due to MR exposure for the device or device type, or (b) the device or device type has typically been labeled as MR Conditional or MR Unsafe. FDA also asks for a rationale in the premarket submission explaining why the option is appropriate.
[!NOTE] "MR Compatible" is retired. The term appeared in a 1997 FDA draft guidance and remains in circulation in older IFUs and marketing material. It is not a current ASTM F2503 designation and should not be used in new labeling. If your device master record still carries it, that is a good proxy for labeling that has not been reviewed since the UDI era began.
What Do 5.15 Million GUDID Records Reveal About MRI Safety Status?
To understand how labelers handle MRI safety declarations in practice, MedDeviceGuide scanned the full GUDID release. The dataset reflects device identifier submissions across thousands of medical device manufacturers.
Overall GUDID Distribution
The complete database distribution across all 5,152,799 device identifier records is summarized in the table below:
| MRI Safety Status Value | Record Count | Percentage of Total GUDID | Clinical & Regulatory Context |
|---|---|---|---|
| Labeling does not contain MRI Safety Information | 4,650,874 | 90.26% | Default submission for non-evaluated devices, legacy records, or non-implanted instruments. |
| MR Conditional | 260,550 | 5.06% | Evaluated devices safe under specific magnetic field and operating limits. |
| MR Unsafe | 162,998 | 3.16% | Devices known to pose hazards (ferromagnetic, active electronics, wireless aids). |
| MR Safe | 78,377 | 1.52% | Entirely non-metallic, non-conductive devices (silicone drains, plastic catheters). |
| Total Database Records | 5,152,799 | 100.00% | Full AccessGUDID database release, July 2026. |
A useful way to read that table: only 501,925 records (9.74%) say anything explicit at all, and 48.1% of those (241,375) say MR Safe or MR Unsafe — two statuses that can often be assigned from materials and construction with little or no scanner testing. Just 260,550 records carry MR Conditional, the status that requires an actual test program and a stated set of scan conditions. That is about one device identifier in twenty — and MR Conditional is precisely the status a radiologist needs in order to scan a patient with an implant rather than defer the exam.
Historical Trend: Are Labelers Adding MRI Data Over Time?
Tallying every record by devicePublishDate — the year the UDI record was published in GUDID — gives the full annual series rather than a selected sample. "Explicit" means the record carries MR Safe, MR Conditional, or MR Unsafe rather than the no-information default.
| Publish year | Records published | With explicit MRI status | Share explicit |
|---|---|---|---|
| 2014 | 35,275 | 8,255 | 23.40% |
| 2015 | 450,150 | 69,711 | 15.49% |
| 2016 | 762,198 | 59,603 | 7.82% |
| 2017 | 311,456 | 29,640 | 9.52% |
| 2018 | 416,645 | 33,197 | 7.97% |
| 2019 | 421,876 | 32,999 | 7.82% |
| 2020 | 392,339 | 35,131 | 8.95% |
| 2021 | 406,882 | 30,179 | 7.42% |
| 2022 | 613,330 | 42,103 | 6.86% (series low) |
| 2023 | 430,768 | 36,689 | 8.52% |
| 2024 | 397,977 | 38,327 | 9.63% |
| 2025 | 360,756 | 55,477 | 15.38% |
| 2026 (partial year) | 153,144 | 30,611 | 19.99% |
The shape of this curve tracks UDI compliance policy almost exactly, and reading it correctly matters:
- 2014–2015 (23.4% → 15.5%): The first cohort was Class III devices, whose September 24, 2014 compliance date pulled in exactly the implants most likely to have been MR-tested already.
- 2016–2022 (falling to 6.86%): The Class II compliance date of September 24, 2016 and the Class I / unclassified date of December 8, 2022 flooded GUDID with high-volume commodity items — gauze, tubing, general instruments — for which no MR evaluation had ever been performed. The denominator grew far faster than the numerator. This decline is a composition effect, not evidence that labelers got worse.
- 2023–2026 (6.86% → 19.99%): Publication rates rise sharply after FDA's October 10, 2023 final guidance, at the same time the Class I backlog clears from the denominator. The 2026 figure is a partial year and could move in either direction.
[!CAUTION] This series measures newly published records per year, not the installed base. Even at 20% for new 2026 records, the cumulative database remains 90.26% blank, because 4.65 million legacy records are never revisited. Improving new-record quality does not fix what is already published — that requires deliberate remediation of existing device identifiers.
Why Do Nearly 80% of Implantable Device Records Lack MRI Safety Data?
The most critical finding from the GUDID analysis centers on implantable devices. In clinical settings, when a patient with a spinal fusion construct, total joint replacement, or dental implant requires an urgent MRI, radiologists and technologists rely on GUDID and product labeling to verify scanning parameters.
Cross-tabulating the GUDID implantable flag against FDA product code reveals where the documentation voids concentrate — and, more usefully, where they do not.
The Twelve Largest Implantable Product Codes, Fully Broken Out
The table below is a complete status breakdown restricted to records flagged implantable = true, so every row sums to 100% of that product code's implantable records. It covers the twelve product codes with the largest implantable record populations in GUDID:
| Product Code | Device Description | Implantable Records | No MRI Info | MR Conditional | MR Safe | MR Unsafe |
|---|---|---|---|---|---|---|
| MAX | Intervertebral fusion device with bone graft, lumbar | 108,970 | 102,490 (94.1%) | 4,479 (4.1%) | 202 (0.2%) | 1,799 (1.7%) |
| NKB | Thoracolumbosacral pedicle screw system | 105,866 | 95,947 (90.6%) | 8,532 (8.1%) | 1 (0.0%) | 1,386 (1.3%) |
| NHA | Abutment, implant, dental, endosseous | 96,537 | 56,193 (58.2%) | 35,298 (36.6%) | 1,293 (1.3%) | 3,753 (3.9%) |
| HWC | Screw, fixation, bone | 63,498 | 41,777 (65.8%) | 18,957 (29.9%) | 1,326 (2.1%) | 1,438 (2.3%) |
| KWQ | Appliance, fixation, spinal intervertebral body | 49,675 | 46,690 (94.0%) | 1,907 (3.8%) | 0 (0.0%) | 1,078 (2.2%) |
| HRS | Plate, fixation, bone | 48,320 | 34,704 (71.8%) | 12,945 (26.8%) | 235 (0.5%) | 436 (0.9%) |
| KWP | Appliance, fixation, spinal interlaminal | 42,559 | 39,193 (92.1%) | 3,168 (7.4%) | 0 (0.0%) | 198 (0.5%) |
| JWH | Knee prosthesis, semi-constrained, cemented | 41,971 | 34,067 (81.2%) | 7,895 (18.8%) | 0 (0.0%) | 9 (0.0%) |
| DZE | Implant, endosseous, root-form | 40,782 | 28,646 (70.2%) | 9,330 (22.9%) | 507 (1.2%) | 2,299 (5.6%) |
| MNH | Orthosis, spondylolisthesis spinal fixation | 29,123 | 27,961 (96.0%) | 534 (1.8%) | 0 (0.0%) | 628 (2.2%) |
| MNI | Orthosis, spinal pedicle fixation | 27,715 | 26,635 (96.1%) | 501 (1.8%) | 0 (0.0%) | 579 (2.1%) |
| ODP | Intervertebral fusion device with bone graft, cervical | 24,114 | 21,978 (91.1%) | 1,033 (4.3%) | 0 (0.0%) | 1,103 (4.6%) |
MedDeviceGuide analysis of the AccessGUDID full database release. Counts are of published Primary Device Identifier records flagged implantable, matched to the first FDA product code on each record. 816 implantable records carry no product code and are excluded from this table.
Intra-Code Divergence: Why It Is a Labeling Issue, Not a Physics Issue
The most important number in that table is not the highest percentage — it is the spread between product codes that face identical physics.
Endosseous dental implant abutments (NHA) and bone fixation screws (HWC) are small, passive, mostly titanium implants. Their labelers populate an explicit MRI status 41.8% and 34.2% of the time respectively. Bone fixation plates (HRS) reach 28.2%. Now compare spinal constructs made from the same material families: pedicle screw systems (NKB) reach 9.4%, spondylolisthesis fixation orthoses (MNH) 4.0%, and lumbar interbody fusion devices (MAX) 5.9%.
Titanium pedicle screws and titanium dental abutments do not present fundamentally different MR test problems. A four- to ten-fold difference in disclosure rate between them cannot be explained by physics.
The divergence is also visible within a single code. Among NKB pedicle screw systems, 95,947 records carry no MRI safety information while 8,532 records — from competing manufacturers, in the same product code — are registered MR Conditional. Among HWC bone screws, 18,957 records are MR Conditional against 41,777 blank. Whatever testing burden exists, some labelers in every one of these codes have already met it and recorded the result.
The reasonable inference is administrative rather than technical: many manufacturers tested and labeled the physical implant as MR Conditional in their Instructions for Use, then left the MRI Safety Status field at its default when submitting or updating the UDI record. The IFU says one thing; GUDID says nothing.
Top MR Unsafe Product Codes
Across the whole database, the 162,998 MR Unsafe records cluster heavily in active electronic and ferromagnetic head/neck products:
| Product Code | Device Description | MR Unsafe Records | Why |
|---|---|---|---|
| OSM | Hearing aid, air-conduction with wireless technology, prescription | 36,709 | RF receivers, internal coils, batteries |
| ECM | Band, preformed, orthodontic | 9,265 | Ferromagnetic alloys, displacement and artifact |
| EJF | Bracket, metal, orthodontic | 6,856 | Displacement force and severe image distortion |
| KLW | Masker, tinnitus | 5,982 | Active electronic sound generators |
| DSA | Cable, transducer and electrode, patient | 4,148 | Conductive leads acting as RF antennas |
| NHA | Abutment, implant, dental, endosseous | 3,753 | Ferromagnetic alloy variants within the code |
| DZC | Wire, orthodontic | 3,174 | Ferromagnetic archwire alloys |
| ESD | Hearing aid, air-conduction, prescription | 3,128 | Electronic components and batteries |
Note that OSM alone accounts for 22.5% of every MR Unsafe record in GUDID. Hearing instruments overall (OSM + ESD = 39,837) and orthodontic hardware (ECM, EJF, DZC and orthodontic tubes DZD = 22,001) together make up 37.9% of all MR Unsafe declarations in the database — which is worth knowing if you write pre-scan screening questionnaires, because those are exactly the two device categories patients most often fail to mention when asked whether they have "an implant."
Differences Between 1.5 T, 3.0 T, and 7.0 T High-Field MRI Scanner Physics
Evaluating medical devices across different static magnetic field strengths requires understanding physical field interactions. A common misconception among medical device engineers is that passing testing at 3.0 Tesla automatically qualifies a device for 1.5 Tesla or 7.0 Tesla MRI scanners.
┌─────────────────────────────────────────────────────────────────────────────┐
│ STATIC MAGNETIC FIELD STRENGTH COMPARISON │
├───────────────────┬───────────────────┬───────────────────┬─────────────────┤
│ Parameter │ 1.5 Tesla (1.5 T) │ 3.0 Tesla (3.0 T) │ 7.0 Tesla (7.0 T│
├───────────────────┼───────────────────┼───────────────────┼─────────────────┤
│ Larmor Frequency │ ~64 MHz │ ~128 MHz │ ~298 MHz │
│ RF Wavelength in │ ~52 cm │ ~26 cm │ ~11 cm │
│ Tissue │ │ │ │
│ SAR Scaling │ Baseline ($B_0^2$)│ 4x higher than │ ~22x higher │
│ │ │ 1.5 T │ than 1.5 T │
│ Displacement │ Governed by the product B0 x (dB0/dz), NOT by B0 │
│ force │ alone. Scanner-specific; must be measured, not scaled.│
└───────────────────┴──────────────────────────────────────────────────────┘
Key Physical Phenomena
- Wavelength Standing Wave Nodes: At 1.5 T (~64 MHz), the RF wavelength in biological tissue is approximately 52 cm. At 3.0 T (~128 MHz), the wavelength drops to 26 cm, which is comparable to anatomical torso dimensions. This creates constructive wave interference (hotspots), altering where maximum RF heating occurs along metallic implants.
- RF power scaling with $B_0^2$: For an equivalent flip angle, RF power deposition scales roughly with the square of the static field strength, so 3.0 T deposits about four times the power of 1.5 T and 7.0 T about twenty-two times. Note the caveat: this is a scaling relationship for a fixed sequence, not a guarantee that a given scan runs at four times the SAR — clinical protocols are constrained by the IEC 60601-2-33 operating mode limits regardless of field strength.
- Displacement force does not scale with $B_0$ alone: A common error is to assume the projectile risk at 3.0 T is exactly twice that at 1.5 T. Magnetically induced displacement force depends on the product of the field and its spatial gradient, and the spatial gradient depends on the specific magnet's bore geometry and shielding. A shielded 1.5 T magnet can present a steeper spatial gradient than an unshielded system at higher field. This is why ASTM F2052 specifies measurement at the location of maximum spatial gradient rather than a calculation from field strength.
- Ultra-High Field 7.0 T Scanners: 7.0 T clinical scanners operate at ~298 MHz with a tissue RF wavelength of roughly 11 cm. Dielectric resonance and severe wave interference dominate, so device evaluation generally relies on electromagnetic simulation rather than extrapolation from lower fields. Note that the 2 cm / 3 cm RF heating test exclusion described below applies only at 3.0 T or less and gives no relief at 7.0 T.
The governing principle is explicit in FDA's guidance, and it is stricter than most engineering teams assume: the safety and conditions of use of a medical device should be assessed — or an adequate scientific rationale provided — for each magnetic field strength (e.g., 0.25 T, 1.2 T, 1.5 T, 3.0 T, 7.0 T), for each RF transmit coil type, and for each set of scan conditions. There is no field strength at which a claim is presumed to carry over from another. A device labeled MR Conditional at 1.5 T is, absent a rationale, simply unevaluated at 3.0 T — not implicitly safe and not implicitly unsafe.
Active Implantable Medical Devices (AIMD): Testing under ISO/TS 10974:2018
For active implantable medical devices (AIMDs)—such as pacemakers, implantable cardioverter-defibrillators (ICDs), deep brain stimulators (DBS), spinal cord stimulators (SCS), and cochlear implants—evaluating MRI safety requires specialized testing far beyond passive implant standards. AIMDs contain electronic circuits, battery power sources, telemetry antennas, and long conductive leads that act as antenna structures in the RF field.
FDA recognizes ISO/TS 10974:2018 (Assessment of the safety of magnetic resonance imaging for patients with an active implantable medical device, Edition 2) as the consensus framework for AIMD MR Conditional evaluation. Each hazard has its own numbered clause, and getting the mapping right matters because test labs quote and accredit against clause numbers:
| Field source | Hazard | ISO/TS 10974:2018 clause |
|---|---|---|
| RF | RF field-induced heating of the AIMD (including lead tip heating) | Clause 8 |
| Gradient | Gradient field-induced device heating | Clause 9 |
| Gradient | Gradient field-induced vibration | Clause 10 |
| $B_0$ | $B_0$-induced force | Clause 11 |
| $B_0$ | $B_0$-induced torque | Clause 12 |
| Gradient | Gradient field-induced lead voltage (extrinsic electric potential) — unintended stimulation | Clause 13 |
| $B_0$ | $B_0$ field-induced device malfunction | Clause 14 |
| RF | RF field-induced device malfunction and RF rectification | Clause 15 |
| Gradient | Gradient field-induced device malfunction | Clause 16 |
| Combined | Combined fields test | Clause 17 |
[!IMPORTANT] ISO/TS 10974:2018 is scoped to 1.5 T only. The Technical Specification applies to AIMDs used in 1.5 T cylindrical-bore whole-body scanners operating at approximately 64 MHz with whole-body coil excitation. It does not, on its own, support a 3.0 T MR Conditional claim — that requires additional justification, and it is a routine source of deficiency letters. ISO also lists the 2018 edition as "to be revised," with ISO/DIS 10974 in draft, so labs and sponsors planning multi-year programs should track the revision.
The RF Lead Tip Heating Hazard
The single highest clinical risk for AIMD patients in an MRI scanner is RF-induced lead tip heating. When the scanner's RF transmit coil generates high-frequency electromagnetic fields (~64 MHz at 1.5 T; ~128 MHz at 3.0 T), long metallic cardiac or neurostimulation leads absorb RF energy along their length. The lead behaves as a linear antenna, conducting current toward the uninsulated electrode tip, where high local energy density can cause thermal tissue injury.
ISO/TS 10974 Clause 8 offers tiered approaches of increasing complexity. The most rigorous — the Tier 3 or "domain decomposition" method — characterises the lead by its transfer function, a position-independent RF response measured on the bench. The incident tangential electric field ($E_{tan}(z)$) produced by the RF transmit coil is computed for clinically relevant exposures, then propagated along the modelled lead path using that transfer function to predict local SAR and temperature rise ($\Delta T$) at the tip. In practice this is combined with numerical human body models across many lead trajectories and body positions, which is why AIMD MR evaluation programs are measured in quarters rather than weeks.
Failure Modes When Un-Evaluated Devices Enter the MRI Suite
The hazard categories FDA's guidance asks manufacturers to address are not theoretical — each corresponds to a documented clinical failure mode. These are the mechanisms a regulatory or quality team should be able to argue against for its own device:
- RF thermal burns from conductive leads and loops. Monitoring cables — ECG leads, pulse oximeter probes — left on a patient during scanning couple with the transmit field and can produce contact burns where a lead touches skin or forms a conductive loop. This is why ISO/TS 10974 treats lead geometry, not just lead material, as the controlling variable.
- Displacement of ferromagnetic implants. The best-known case is the 1992 death of a patient with a ferromagnetic cerebral aneurysm clip who underwent a 1.5 T scan, reported by Klucznik and colleagues in Radiology in 1993. The clip had been assumed non-ferromagnetic based on the model designation; it was not. Displacement force is why ASTM F2052 uses the 45° deflection criterion, and why "same model number" is not the same as "same alloy lot."
- Malfunction of un-evaluated active implants. Legacy non-MR Conditional pacemakers and ICDs exposed to MR fields have shown reed switch closure, asynchronous pacing, inappropriate sensing inhibition, and telemetry corruption — the hazards mapped to ISO/TS 10974 Clauses 14 through 17.
- Projectile events from equipment brought into Zone IV. Non-implanted items — oxygen cylinders, IV poles, wheelchairs, external infusion pumps clipped to a patient's belt — remain among the most consequential MR incidents precisely because they fall outside the mental model of "implant screening." This is the category the 2023 guidance widened its scope to capture.
[!NOTE] These failure modes are drawn from the published MR safety literature and the hazard categories enumerated in FDA's guidance and ISO/TS 10974. They are presented to illustrate mechanisms, not as an incidence estimate. FDA's Manufacturer and User Facility Device Experience (MAUDE) database is a passive, voluntary-plus-mandatory reporting system: report counts reflect reporting behaviour, not event rates, and cannot be used to compute the frequency of MR-related harm.
How Do ASTM F2503 Symbols and Test Standards Apply to Labeling?
To achieve FDA compliance for MR safety claims, manufacturers must adhere to standardized testing protocols and labeling presentation rules.
Standardized ASTM F2503 Marking Icons
FDA recommends labeling devices using the standard ASTM F2503 color-coded icons on physical packaging, labeling, and electronic IFUs:
┌─────────────────────────────────────────────────────────────────────────────┐
│ ASTM F2503 SYMBOL SUMMARY │
├──────────────┬────────────────────────┬─────────────────────────────────────┤
│ Symbol Term │ Visual Design │ Meaning / Operational Requirement │
├──────────────┼────────────────────────┼─────────────────────────────────────┤
│ MR Safe │ Green Square with │ Safe in all MR environments without │
│ │ white "MR" letters │ restriction. Non-conductive/metallic│
├──────────────┼────────────────────────┼─────────────────────────────────────┤
│ MR │ Yellow Equilateral │ Safe only under tested conditions. │
│ Conditional │ Triangle with "MR" │ IFU must state field & SAR limits. │
├──────────────┼────────────────────────┼─────────────────────────────────────┤
│ MR Unsafe │ Red Circle with slash │ Poses known safety hazard in MR │
│ │ over white "MR" │ scanner room. Do not enter. │
└──────────────┴────────────────────────┴─────────────────────────────────────┘
Consensus Testing Standards Cited by FDA
When evaluating a medical device for MR Conditional labeling, FDA expects testing performed according to recognized consensus standards:
| Hazard | Standard cited by FDA | What it measures |
|---|---|---|
| Magnetically induced displacement force | ASTM F2052-21 | Translational attraction from the static field gradient, via a deflection-angle test. A deflection greater than 45° means the magnetic force exceeds the device's weight. |
| Magnetically induced torque | ASTM F2213-17 | Rotational torque tending to align the device's long axis with $B_0$. |
| RF-induced heating | ASTM F2182-19e2 | Temperature rise ($\Delta T$) on or near a conductive passive implant in a tissue-simulating gel phantom under $B_1$ exposure. |
| Image artifact | ASTM F2119-07(2013) | Size, shape, and signal-void extent of the artifact in spin-echo and gradient-echo sequences. |
| Active implant assessment | ISO/TS 10974:2018 | Full hazard framework for pacemakers, ICDs, neurostimulators, cochlear implants (1.5 T scope — see above). |
| Scanner operating modes | IEC 60601-2-33 (FDA guidance cites the 2010+A1:2013+A2:2015 consolidated version) | Defines Normal Operating Mode (whole-body SAR $\le 2.0\text{ W/kg}$) and First Level Controlled Operating Mode (whole-body SAR $\le 4.0\text{ W/kg}$). |
Passive Implant Testing Exemption (The 2 cm / 3 cm Rule)
FDA's October 2023 guidance contains a testing exclusion for small passive implants that can save a manufacturer a full RF heating study. It is narrower than it is usually quoted:
[!TIP] RF heating test exclusion (quoting the guidance): "A passive implant with dimensions of less than 2 cm in all directions and at least 3 cm away from another passive implant does not need to be tested with respect to RF induced heating at 3.0 T or less, as it is expected to generate a temperature increase of less than 2 °C in Normal Operating Mode, over the course of 1 hour of exposure."
The exclusion is not valid when:
- multiple replicas of the implant (for example multiple metallic anchors, closely spaced bone screws, or dental posts) are implanted within 3 cm of each other — the 3 cm separation exists to avoid RF coupling with neighbouring implants; or
- part of the implant is outside the patient (percutaneous or transcutaneous components).
Where multiple replicas can be joined, the guidance directs you to assess the worst-case joined condition. Condition (2) is the one most often missed, and it removes the exclusion for a large share of external fixation and percutaneous hardware.
Hospital EHR / RIS Integration & Automated UDI Lookup
Modern hospital radiology departments are increasingly automating MR safety verification workflows. Rather than manually hunting for paper IFUs or searching third-party websites during pre-scan patient intake, healthcare networks are integrating automated GUDID API calls into Electronic Health Record (EHR) systems (e.g., Epic, Cerner) and Radiology Information Systems (RIS).
┌─────────────────────────────────────────────────────────────────────────────┐
│ HOSPITAL AUTOMATED UDI MRI VERIFICATION WORKFLOW │
├─────────────────────────────────────────────────────────────────────────────┤
│ 1. Patient Intake: Scan Device UDI Barcode on Implant Card or EHR Record │
│ 2. Automated API Query: Ingest GUDID `MRI Safety Status` & IFU Attributes │
│ 3. RIS Pre-Scan Checklist: Flag MR Conditional / MR Unsafe Constraints │
│ 4. Technologist Alert: Display Scanner $B_0$ & SAR Limits on MRI Console │
└─────────────────────────────────────────────────────────────────────────────┘
Where such a lookup is implemented, a record returning "Labeling does not contain MRI Safety Information" cannot resolve the safety question — it is indistinguishable, to an automated check, from a device that was never assessed. The workflow then falls back to manual IFU retrieval or radiologist adjudication, which is precisely the delay the automation was meant to remove.
This is the commercial argument for remediating GUDID records that the compliance argument misses. A manufacturer whose MR Conditional IFU is not reflected in GUDID has, in effect, made its device look unevaluated to every automated screening tool that queries the database — while a competitor in the same product code, with the same testing, resolves cleanly.
[!NOTE] Degrees of automation vary widely between health systems, and many still run MR screening entirely on paper or on institution-maintained reference tables. The point here is directional: as screening automates, the value of a populated
MRI Safety Statusfield rises, and the cost of a blank one shifts from the manufacturer to the patient's scan schedule.
Manufacturer Audit Checklist for GUDID MRI Safety Records
A regulatory or quality team can close this gap without new laboratory work in most cases. The five steps:
- Extract your full published catalog. Pull every device identifier published under your labeler DUNS numbers. AccessGUDID provides a public download and a search API; your own GUDID account holds the authoritative submitted view.
- Flag the blanks. Filter for records currently carrying
Labeling does not contain MRI Safety Information, and split them by implantable flag and commercial distribution status — implantable and in-distribution is where remediation pays first. - Reconcile against what your IFU already says. For each flagged record, check the current approved IFU and the premarket file (510(k), De Novo, or PMA) for ASTM F2052 / F2213 / F2182 / F2119 reports, an ISO/TS 10974 assessment, or a documented 2 cm / 3 cm exclusion rationale. In most cases where an MR Conditional IFU exists, the record is simply out of sync and no new testing is needed.
- Correct the GUDID record. Update the MRI Safety Status attribute through the GUDID Web Interface for low volumes, or by HL7 SPL submission through the FDA Electronic Submissions Gateway for bulk correction. Under 21 CFR 830.330(b), an update to required information must be submitted no later than the date the device is first labeled with the changed information — or, where the information does not appear on the label, within 10 business days of the change.
- Confirm publication. Verify the corrected value appears in the public AccessGUDID record before closing the action. Do not assume same-day publication; check rather than schedule.
[!TIP] Where the labeling genuinely has not been evaluated, the honest answer is not to leave the field blank and hope. FDA's guidance describes explicit "not evaluated" labeling language, and asks for a rationale in the premarket submission. But it also bars that option where the device type is typically labeled MR Conditional or MR Unsafe — which, given the intra-code data above, rules it out for most spinal, orthopedic, and dental implant codes.
Frequently Asked Questions (FAQ)
Is a "No Information" GUDID result the same as MR Unsafe?
No. In GUDID, "Labeling does not contain MRI Safety Information" simply means the labeler did not submit an explicit ASTM F2503 safety status code to the database. It does not mean the device is proven dangerous. However, in clinical hospital settings, if a technologist cannot verify MR Conditional parameters in GUDID or the IFU, they will treat the device as potentially hazardous and defer the scan.
Can a device tested at 3.0 T automatically be labeled MR Conditional at 1.5 T?
No — and the reverse is equally untrue. FDA's guidance asks that safety and conditions of use be assessed, or an adequate scientific rationale provided, for each static field strength and RF transmit coil type. The physics is why: RF wavelength in tissue is roughly 52 cm at 1.5 T (~64 MHz) and roughly 26 cm at 3.0 T (~128 MHz), so standing-wave interference patterns and heating hotspots occur at different anatomical locations. Higher field strength is also not uniformly worse for every hazard — displacement force scales with $B_0$ and its spatial gradient, while RF power deposition scales with $B_0^2$, so the binding constraint can change between fields.
Does ISO/TS 10974:2018 cover 3.0 T active implant testing?
Not on its own. The Technical Specification is scoped to 1.5 T cylindrical-bore whole-body scanners operating at approximately 64 MHz with whole-body coil excitation. A 3.0 T MR Conditional claim for an active implant needs justification beyond a straight ISO/TS 10974 test report — a point that regularly surfaces in FDA deficiency letters.
Where must MRI safety information appear in product labeling?
FDA recommends — this is guidance, not a binding requirement — that MRI safety information appear in a separate labeling section titled "MRI Safety Information," and that this section be included in the labeling's table of contents where one exists. The guidance also asks that the information remain readily accessible for the life of the device even when original labeling is not to hand, for instance on the manufacturer's website or by telephone, with the device clearly and unambiguously identified.
Does 21 CFR 830.310(b)(8) apply to Class I devices?
Yes. The reporting requirement applies to every version or model required to bear a UDI on its label, regardless of class. Class I and unclassified devices reached their GUDID submission compliance date on December 8, 2022 (Class III was September 24, 2014; Class II September 24, 2016) — which is why the 2022 publication cohort is both the largest and the least likely to carry an explicit MRI status.
Our IFU says MR Conditional but GUDID says "no information." Are we non-compliant?
The record is inaccurate, and 21 CFR 830.310 requires the submitted information to reflect the device. Under 21 CFR 830.350, FDA may notify a labeler and delete or correct information in GUDID that appears incorrect or potentially misleading, and the labeler must provide corrected information — or explain why the information is correct — within 30 days of that notice. The practical exposure is less about enforcement than about a clinician being unable to confirm a scan condition that your own IFU already supports.
Strategic Recommendations for Regulatory & Quality Teams
Three actions, in priority order:
- Audit before you test. Cross-reference your published GUDID records against your approved IFUs. Given that competing manufacturers within nearly every implantable product code have already registered MR Conditional status, the most likely finding is a records gap rather than a testing gap — and records gaps are cheap to close.
- Remediate implantable, in-distribution records first. Of the 902,914 implantable records with no MRI safety information, 798,484 are in active commercial distribution. Those are the ones a radiology department will actually query.
- Apply the 2 cm / 3 cm exclusion where it genuinely applies — and document why. The exclusion covers small, fully internal passive implants at 3.0 T or less. It does not cover closely spaced replicas or any implant with a component outside the patient. Where you rely on it, record the rationale in the premarket file rather than leaving the GUDID field blank.
Related Knowledge Base Guides & Data Analyses
Explore our related regulatory knowledge base guides and medical device dataset studies:
- What 5.08 million GUDID records show about UDI compliance — Comprehensive analysis of GUDID record volume, device classes, and labeler market concentration.
- ISO 15223-1 device symbols guide — How ASTM F2503 MRI icons interface with ISO 15223-1 labeling symbols on medical device packaging.
- How to verify a device UDI in GUDID — Step-by-step hospital workflow for querying GUDID database APIs and interpreting UDI attributes.
- MRI and CT scanner recall patterns — MAUDE safety teardown analyzing imaging equipment component failures and patient safety incidents.