Passive Implants and Retained Foreign Bodies in MRI — Aneurysm Clips, Stents, Orthopaedic Hardware, Orbital Metal and Retained Fragments: What the Evidence Shows and Where It Stops

Passive Implants and Retained Foreign Bodies in MRI — Aneurysm Clips, Stents, Orthopaedic Hardware, Orbital Metal and Retained Fragments: What the Evidence Shows and Where It Stops

MRIninja Knowledge Base | Child Page — MRI Safety (deep dive S8) Version 1.0 — October 2026

Parent page: MRI Safety — Master Page (hazard table in Section 3). Builds on: Implant and Device MR Safety Framework (test methods, label content, matching a label to the scanner) · MR Site Zones and Access Control (screening and access control) · Projectile and Torque Hazards in the Static Field (force and torque in the static field)

1. Scope, Status and Boundaries

This page gathers the evidence on passive implants (implants without a power source or electronics) and on retained metallic foreign bodies: intracranial aneurysm clips, coronary and other stents, orthopaedic hardware, metal in the orbit, and retained bullets and fragments. It states what the sources read report about harm and about the conditions under which scanning has been done, and where those sources stop. Neighbouring topics are held elsewhere and are not repeated: how a label is read and matched to the scanner, and the ASTM test standards behind it (Implant and Device MR Safety Framework); the screening workflow and access control (MR Site Zones and Access Control and Patient Preparation); the physics of force and torque (SAR, Bioeffects, and MRI Safety Physics Section 5.2, Projectile and Torque Hazards in the Static Field); RF heating of conductors and burns (RF Heating and Burns); and active implants, including cardiac devices (Cardiac Implantable Electronic Devices in MRI, Active Implants Other Than Cardiac Devices in MRI).

Reading status. The page rests on the fifth edition (July 2026) of the UK MHRA guidelines (general sections read verbatim; Section 5.5 on implanted devices was not available in the text read), two documents by a single expert author (May 2025), one narrative review with an incident-database search and national survey for aneurysm clips, a handful of case reports, retrospective reviews and a cost-utility model, and two news reports of laboratory work on bullets. Almost all sources were read at abstract level or through a summarising tool, as stated for each in the bibliography. The ACR Manual on MR Safety, the ASTM standards, current manufacturer labels and databases of implant labels were not read. No systematic review of any topic in this page was read.

What this page does not do. It gives no scanning conditions to be applied to a patient: the label of the exact implant, where there is one, is the authority (Implant and Device MR Safety Framework), and the conditions quoted from a single-author document are that author’s proposals, not a manufacturer’s label. Statements that are the page’s own inference are marked Reasoning; where a source needed to complete a statement was not found they are marked (to be sourced).

2. Sources Read and What They State

Table 1. Documents used and how they were read.

3. Mechanisms and the Role of the Label

The MHRA guidelines state that the effects of the MR environment can adversely affect medical equipment, including passive and active implantable medical devices; that implantable medical devices or foreign bodies could move within the body, potentially causing tissue damage; and that time-varying gradients induce eddy currents within implants [1]. They define MR Safe, MR Conditional and MR Unsafe and add that an item without a label is termed MR Unlabelled, and they state that for MR Conditional implants exposure risks are minimised by adhering to the manufacturer’s instructions [1]. The tests and the content of a label are in Implant and Device MR Safety Framework; the mechanism of force and torque is in SAR, Bioeffects, and MRI Safety Physics and Projectile and Torque Hazards in the Static Field.

Reasoning. For a passive implant the three concerns are movement (force and torque), heating of the implant and the tissue around it, and image artefact; there is no device function to be disturbed. A foreign body without a label is the same problem without the documentation: its composition, size, position and age are what has to be estimated, which is why the sections on clips, the orbit and fragments below turn on identification and risk-benefit reasoning and not on a label.

4. Intracranial Aneurysm Clips

4.1 The Fatal Case and the Incident Databases

The first reported fatal case, in 1993, concerned a clip that had been recorded as nonferromagnetic: the patient developed an acute intracerebral haemorrhage in the unit, imaging suggested that the clip had torqued, and autopsy showed a torn middle cerebral artery from clip movement and a ferromagnetic clip [5]. A narrative review that combined a literature search, a search of incident databases and a Dutch national survey found records of 10 incidents: two deaths probably caused by the clip (1992 and 2001), two other serious events probably clip-related (1993 and 1997), four reports of MRI without adverse events in patients with unknown or expected ferromagnetic clips, and two post-MRI deaths judged unlikely to be clip-related [4]. Figure 1 shows the distribution. The authors note that incident databases are likely to under-report and that some attributions were uncertain [4].

Figure 1 — Records of aneurysm clip incidents in MRI found in incident databases (10 records)

4.2 Prevalence, Implantation Year and Heating

The review reports a retrospective study in which 0.03 per cent of referred outpatients had a ferromagnetic aneurysm clip; a survey of 11 neurosurgical departments in the Netherlands (all responded) in which no ferromagnetic clips had been implanted from 2000 onward, with hospital-specific cut-off years ranging from 1980 to 2000; a catalogue of 349 clip models in a public database (173 MR Safe, 157 MR Conditional and 18 MR Unsafe, a total that does not add up to 349, and the paper gives 154 and 157 MR Conditional models in different places); the authors’ estimate that in-vivo heating stays below 1 °C at 3 T and below, so that no additional SAR restriction is needed; and larger artefacts with cobalt-chromium and other alloy clips than with titanium clips [4]. The authors’ conclusion is that patients with clips implanted in the Netherlands from 2000 onward can generally be scanned at 1.5 T or 3 T, and that for older or unknown clips a stepwise risk-benefit assessment is recommended: implantation records, prior MRI history, clip model, importance of the diagnosis, consent, and precautions such as head fixation and centred positioning [4].

The same review states that the probabilities of damage in its tables are expert estimates tied to implantation year and hospital, that the literature search was run in 2016 and the database searches in 2021, that only papers with English abstracts were included, and that the survey relied on the recall of department heads and treated all clips before the cut-off years as ferromagnetic [4]. The cut-off years are those of one country and cannot be applied elsewhere without local data (to be sourced).

Reasoning. A previous uneventful scan does not prove that a clip is safe: the 1993 case shows that a recorded composition can be wrong, and the sources read give no way of checking composition without the model. The recommendation to identify the model, or to estimate risk from the year and place of implantation when it is unknown, is the review’s own and rests on expert estimates.

5. Coronary Stents and Other Vascular Implants

A single-author guideline of May 2025 proposes that patients with coronary artery stents, drug-eluting or bare metal, single or multiple, including stents with unknown labelling, may be scanned at 3 T or less, with no restriction on static field direction or spatial gradient, a SAR of 2 W/kg (Normal Operating Mode) for a stent inside the transmitted RF area and up to 4 W/kg (First Level Controlled Operating Mode) for a stent entirely outside it, and up to 15 minutes per pulse sequence [3]. It states that no waiting period is needed, that the earlier belief in a wait of six weeks or more has been refuted because no coronary stents are known to be ferromagnetic, and that there has never been an adverse event reported in association with MRI in patients with these implants; it rests on about 25 references and states that it must be adopted as a written policy after review by the supervising radiologist, that deviations need prior approval, and that it must be reviewed annually [3]. The studies of early scanning that it cites were not read.

Reasoning. “No adverse event reported” is an absence of reports; it does not give an event rate and cannot exclude events that were not reported. The guideline does not give manufacturer-specific data, and stents other than coronary stents are not covered by it. The sources read contain no guidance for peripheral, neurovascular, aortic or other stents and grafts (to be sourced).

6. Orthopaedic and Other Passive Implants

The same author’s guideline for passive internal orthopaedic implants proposes conditions of 3 T or less, no static field direction restriction, no spatial gradient value listed, SAR of 2 W/kg in Normal Operating Mode for an implant inside the transmitted RF area and up to 4 W/kg in First Level Controlled Operating Mode for one entirely outside it, and up to 15 minutes per pulse sequence; for the post-operative patient it states that a nonferromagnetic passive implant can be imaged immediately after implantation at 3 T or less, that very few passive implants state a wait period in the instructions for use, that weakly ferromagnetic implants that are rigidly fixed may also be imaged immediately, and that RF heating must still be considered; it excludes external fixation, programmable or magnetically controlled implants, bone fusion stimulators and prostheses with microprocessors, states that heating is a theoretical risk for some spinal fixation systems without substantial heating or burns documented, and says that if the implant cannot be identified or tissue integrity is in doubt the patient should not be exposed [2]. An expert review of 2025 states in its abstract that modern implants are mostly non-ferromagnetic, which lowers the risk of displacement, that no confirmed cases of tissue damage from MRI-related heating near orthopaedic devices appear in the literature, and that MRI is safe and diagnostically valuable for monitoring patients with orthopaedic hardware [6].

Reasoning. The two sources agree that confirmed harm is not documented, and both rest on expert opinion and absence of reports. They do not replace the instructions for use of a given implant, and the claim that a wait period is generally unnecessary comes from one author; it conflicts with nothing read but was not checked against labels (to be sourced). Implants with a label stating a wait period or a SAR limit follow the label (Implant and Device MR Safety Framework). The sources read give no data for spinal fixation heating, joint prostheses at 3 T or implants in children (to be sourced).

7. Orbital and Intraocular Foreign Bodies

An accident list in a 2001 case series, quoted from the authors’ references and not read at source, includes unilateral blindness from an intraocular metallic foreign body [10]. On how to screen, a cost-utility model compared clinical screening with radiographic screening before MRI: with base-case estimates and no discounting the then-current guideline of radiographs cost 328,580 US dollars per quality-adjusted life-year saved; screening cost was the most critical variable; and the authors concluded that clinical screening before radiography improves cost-effectiveness about tenfold, that asking the patient whether a doctor removed a foreign body is a sufficient clinical screen, that occupational history alone does not justify radiographic screening, and that practice guidelines should be altered [7]. A conference review (35 articles found, 28 accepted) states that radiography is the best method to identify intraocular foreign bodies, notes that no research addresses the lowest dose that gives adequate orbital image quality, and reports no sensitivity figures in its abstract [9]. In a retrospective series of 204 trauma cases and suspected cases, plain films without clinically evident penetration (177) showed no foreign body and CT in patients without evident penetration (9) showed none, while with evident penetration a foreign body was clinically visible in 19 of 27 and all 12 CT scans were positive [8].

Reasoning. The trauma series concerns patients with suspected penetrating injury and does not measure how well radiographs detect an occult foreign body in a person without symptoms who is about to be scanned. The model gives its recommendation under assumptions of 2000 and cannot say which screening policy a department should adopt. The sources read do not report the prevalence of orbital metal in screened populations, the sensitivity of radiographs or CT for fragments of the size that matter for MRI, or the proportion of questionnaires that miss a foreign body (to be sourced). Orbital screening is a screening-workflow question and the workflow itself is in MR Site Zones and Access Control and Patient Preparation.

8. Retained Bullets, Shrapnel and Other Fragments

A retrospective review of 10,322 screening questionnaires at one centre found 24 patients who reported fragments; after exclusion of 7 without confirming evidence, 17 patients (18 examinations at 1.5 T) remained, with injuries 2 to 39 years earlier and fragments of 1 to 10 mm; one patient (5.9 per cent) reported superficial migration of a fragment of about 1 cm after a knee examination, and none reported heating [11]. The authors describe the work as preliminary and advise weighing each case by fragment proximity to vital organs, size, field strength and technical safeguards [11]. Bench data are known only from news reports: a conference presentation reported that, among 32 commercial bullets and 7 shotgun pellets in phantoms at 1.5, 3 and 7 T, non-steel items showed no magnetic effects, steel-containing items deflected, an armor-piercing steel-core bullet had torque and force above safe levels, and maximum heating of five bullets at 3 T was 1.7 °C above background [12]; and a laboratory study reported that CT and X-ray could distinguish ferromagnetic from nonferromagnetic fired bullets in gelatin, that a triage algorithm was proposed, that ferromagnetic unfired bullets moved or rotated, and that none heated above 2 °C [13].

Reasoning. The clinical series is small and self-reported, and the bench data are secondary reports on phantoms and gelatin; neither gives a rate of harm for a fragment near a critical structure. What both indicate is that composition is the determinant, and composition is rarely known for fragments from combat, terrorism or industry. The sources do not cover fragments in or near the eye or the spinal canal, nor what to do when imaging cannot tell steel from lead (to be sourced).

9. Where the Documents Differ and Open Questions

Table 2. Differences and unresolved questions.

10. Limits of the Evidence and What Was Not Read

The evidence is incident databases, case reports, small retrospective series, a cost model, bench data reported in the news and the opinions of two expert sources; no systematic review was read and no study measured an event rate. MHRA Section 5.5, the ACR Manual on MR Safety, the ASTM standards and the implant labels themselves were not read. Absence of reported events is not evidence of safety. The statements on conditions are those of a single author and must not be applied without the label of the exact implant or, where there is none, a local policy adopted by the responsible radiologist and MR safety expert.

11. Topics Not Covered

Label content and test standards (Implant and Device MR Safety Framework); active implants and cardiac devices (Active Implants Other Than Cardiac Devices in MRI, Cardiac Implantable Electronic Devices in MRI); heating mechanisms and burns (RF Heating and Burns); the physics of force and torque (Projectile and Torque Hazards in the Static Field); the screening questionnaire and access control (MR Site Zones and Access Control, Patient Preparation); the unconscious or non-communicating patient (screening: Patient Preparation, Section 3.3; sedation and monitoring: Claustrophobia, Sedation, Monitoring, Equipment and Emergencies in the MR Environment); implants in pregnant patients and pregnant staff (pregnancy in general: MRI in Pregnancy and the Pregnant Worker; no implant-specific data were read); implants at 7 T and in ultra-high-field research; dental, penile and breast implants, tissue expanders, tattoos and body piercings; and the rules of individual jurisdictions (Incident Learning, Governance, Quality and Jurisdictions in MR Safety).

12. Evidence-Based References

No systematic review and no study of event rates was read for any topic of this page; the sections below follow the hierarchy of the sources actually read, and the two expert guidelines are single-author documents.

A. Guidelines, Consensus, Standards and Regulation

B. Reviews

C. Case Reports, Retrospective Series and Models

D. Bench Data Known Only from News Reports

End of document — Passive Implants and Retained Foreign Bodies (S8) — MRIninja v1.0 — October 2026. Parent: MRI Safety master.