The Danish company received FDA Breakthrough Device Designation in November 2024, which allowed the agency to provide more intensive guidance during the review process. The 510(k) clearance itself rests on substantial equivalence to a legally marketed predicate; it is a market authorization, not a premarket approval that requires new clinical trial data demonstrating improved outcomes. Whether the extended EEG record the EpiSight produces will improve seizure outcomes or time to surgical candidacy determination will require prospective clinical evidence.

The gap the EpiSight is designed to fill is well-documented. Standard ambulatory EEG captures 24 to 72 hours of brain activity — a narrow window in a disease where most patients experience seizures days, weeks, or months apart. Repeated short-term recordings are expensive, logistically burdensome, and frequently capture no diagnostic events. Clinicians are often left adjusting antiseizure medications or making escalation decisions based on limited or incomplete electrical data.

The EpiSight implant is placed subcutaneously — beneath the skin but above the skull, without penetrating brain tissue. It transmits recordings wirelessly to an external receiver, allowing neurologists to accumulate a longitudinal EEG record between clinic visits. The device carries MRI conditional labeling at 1.5 and 3 tesla, the field strengths used by the large majority of U.S. scanners, so patients can undergo neuroimaging without removing the implant.

Epilepsy affects approximately 3.4 million Americans. About one-third of patients develop drug-resistant epilepsy — defined as failure to achieve sustained seizure freedom after at least two adequately trialed antiseizure medications. For that group, accurate long-term seizure characterization informs decisions about whether to escalate pharmacotherapy, evaluate for vagal nerve stimulation, or pursue resective surgery candidacy.

Correction, 2026-06-22: An earlier version of this article stated that “about one-third of patients do not achieve seizure control with the first antiseizure medication prescribed.” That framing misapplied a standard epidemiological benchmark: the one-third figure refers to patients who develop drug-resistant epilepsy — those who fail to respond to at least two appropriately chosen and adequately trialed antiseizure medications — not to first-drug failure specifically. First-drug failure rates are substantially higher (~50%). The sentence has been corrected.

UNEEG has not disclosed a U.S. commercial pricing or specific availability timeline. The EpiSight system has been marketed in Europe since receiving CE mark.

A market withdrawal differs from a Class I or Class II recall in a precise regulatory sense: the product was not distributed through Dexcom’s own authorized channels, and the company is acting voluntarily to remove it rather than responding to a formal FDA-mandated recall order. That distinction does not diminish the clinical stakes.

The two affected lots carry distinct hazard profiles. Lot 1725204004 was pulled because the sensors were never properly sterilized before being scrapped; patients who insert them face an elevated risk of skin or subcutaneous infection at the application site. Lot 1725069002 was condemned internally for an elevated testing-failure rate; sensors from this lot risk delivering no glucose readings whatsoever — a silent failure with no alarm, no alert, and no data.

For insulin-dependent patients, particularly those with type 1 diabetes, a silent CGM failure is not a minor inconvenience. Without real-time glucose data, hypoglycemia can go undetected for hours. The G7 is marketed specifically to replace fingerstick monitoring during routine daily management, meaning users may have no secondary alerting system in place.

Dexcom traced the stolen product to Pharmsource LLC, an unauthorized distributor that supplies a network of independent pharmacies and durable medical equipment distributors across the United States. Pharmsource is not an approved Dexcom distribution partner. The company has not disclosed how many sensors from the two lots reached patients.

As of the announcement date, Dexcom reported no severe adverse events associated with either lot. The company has established a verification portal for patients and providers to check lot numbers, and it is offering free replacements to anyone who received an affected sensor.

Patients and pharmacists can verify devices by checking the lot number printed on the G7 sensor packaging. Anyone holding either lot should stop use immediately and contact Dexcom customer support.

The distinction between mass units and molar units is not bureaucratic. Lp(a) particles vary substantially in size because of a polymorphic protein called apolipoprotein(a), whose isoform diversity can cause measured Lp(a) mass to differ by up to 200 percent between two patients with identical molar — and therefore identical cardiovascular — risk. The European Society of Cardiology has endorsed nmol/L as the preferred reporting unit precisely to neutralize that variability. Diazyme’s assay uses an isoform-independent method, so results should hold across the genetic diversity labs actually encounter.

The clinical stakes are rising fast. An estimated one in five people globally carries elevated Lp(a), a genetically determined risk factor for atherosclerotic cardiovascular disease that no approved drug yet specifically targets. That is about to change: Novartis’s pelacarsen is in the Lp(a)HORIZON outcomes trial, and Amgen’s olpasiran anchors the OCEAN(a) Phase 3 study, which enrolled patients at a threshold of ≥200 nmol/L. Both trials use molar endpoints, creating a near-term need for labs to run companion diagnostics in compatible units.

The 510(k) pathway authorizes Diazyme’s device through substantial equivalence to a legally marketed predicate — it is a clearance, not a PMA approval. For hospital and reference labs still running older mass-unit platforms, the Diazyme clearance opens a second vendor option as they upgrade cardiovascular risk panels ahead of what could be a busy regulatory period for Lp(a)-lowering therapies.

The Modius Spero applies transcutaneous electrical stimulation to the mastoid process — the bony prominence behind the ear — at parameters the manufacturer, Neurovalens, says modulate brainstem circuits connected to the autonomic nervous system and limbic system. The company’s rationale is that attenuating hyperarousal through this pathway can reduce PTSD symptom burden.

The pivotal evidence supporting the authorization has not been published in a peer-reviewed journal. The FDA’s decision was based on data submitted in the De Novo request, including a sham-controlled trial; the full trial report, protocol, and results are not available in the public scientific literature as of this edition. Neurovalens has indicated its intention to submit the data for publication, but no manuscript has appeared.

The absence of a published trial record is unusual for a first-in-class device authorization in a high-prevalence psychiatric indication. PTSD affects approximately 13 million Americans in any given year, according to the National Center for PTSD. Existing pharmacotherapy and psychotherapy options leave a substantial proportion of patients with inadequate symptom control — the unmet need is real. What is not yet independently verifiable is the magnitude of benefit the Modius Spero provides.

The FDA’s authorization letter specifies adjunctive use — not as a standalone treatment — and includes a post-market requirement for performance data within 36 months. Clinicians considering recommending the device should note that the primary efficacy evidence remains in regulatory files rather than the peer-reviewed record, and apply appropriate clinical judgment in that context.

FDA De Novo DEN240012, Neurovalens Modius Spero, June 19, 2026. National Center for PTSD, U.S. Dept. of Veterans Affairs.

The model was developed by a team led by lead author Darui Jin, with senior co-corresponding authors Moritz Gerstung at the German Cancer Research Center (DKFZ) and Felix Sahm at Heidelberg University Medical Faculty and Heidelberg University Hospital. Validation drew on more than 11,000 digitized tissue sections from 9,606 patients across 11 medical centers on four continents, making it one of the largest multi-institutional assessments of an AI pathology tool for CNS tumors to date.

Head-to-head performance. In a histology-only comparison using 210 cases spanning the full spectrum of CNS tumor types, Hetairos achieved a top-1 accuracy of 0.68 — the proportion of cases in which the model’s single best prediction matched the molecular ground truth. The five neuropathologists, working from the same slides without molecular data, reached a mean top-1 accuracy of 0.30. At top-3 (whether the correct subtype appeared among the three most likely predictions), Hetairos scored 0.84 versus 0.50 for the specialists. For cases in which the model assigned its highest confidence scores, accuracy reached 0.87.

The study also included a prospective component: Hetairos ran in parallel with routine clinical diagnostics on 111 consecutive tumors without influencing treatment decisions, confirming the 12-minute turnaround.

What it does — and does not — do. The model predicts molecular subtype from morphology alone; it does not perform or replace molecular testing. Sahm stated in a release from DKFZ that the tool is “not intended to replace molecular analyses, but rather to specifically complement and accelerate them.” Rare subtypes underrepresented in training data remain a challenge, and the authors note that the validation centers were predominantly high-income academic institutions — a limitation with direct implications for generalizability.

Equity dimension. Because Hetairos runs on digitized versions of H&E sections — the same stain used in virtually every pathology laboratory worldwide — proponents argue it could extend sophisticated subtype classification to settings where DNA methylation arrays are unavailable or unaffordable. Gerstung described it as demonstrating the tool’s “enormous potential of AI-supported digital pathology to provide rapid and widely available diagnostic methods that were previously only possible with considerable technical effort.” Whether that potential translates to low-resource settings will depend on digital-pathology infrastructure — slide scanners, connectivity, and local implementation capacity — none of which is addressed in the current validation.

The model’s code is publicly available via the Gerstung lab’s GitHub repository. No confidence intervals for the accuracy figures were available from sources accessible to this desk; the primary paper is paywalled.

Correction — June 18, 2026: An earlier version of this article stated that Hetairos classifies 102 methylation-based CNS tumor subtypes. The Gerstung lab’s official GitHub repository (Tumor_label_mapping.yaml) contains 108 distinct class entries. Corrected to 108. Identified by The Vital Record’s post-publication fact-check.

CEREPAK coils are used in endovascular embolization of intracranial aneurysms: a neurosurgeon or interventional radiologist threads a microcatheter through a blood vessel to the site of the aneurysm and deploys the coil, which fills the sac and promotes clotting to prevent rupture. When a coil fails to detach from its delivery wire as intended, it cannot be positioned correctly — creating risks of hemorrhagic stroke (from aneurysm rupture or vessel injury) or ischemic stroke (from thrombus formation), and potentially necessitating open surgical intervention.

J&J MedTech and its subsidiary Cerenovus initiated a voluntary recall on October 2, 2025, after an internal evaluation identified that the detachment failure rate may be higher than anticipated. As of October 14, 2025 — the data cut-off date cited in FDA records — the company had received reports of four serious injuries and one death associated with the defect. The FDA formally announced the Class I classification on February 5, 2026.

Affected Devices

The recall covers approximately 12,000 units across eleven product lines in the Uniform, Heliform, and Freeform families, including the CEREPAK Uniform, Uniform XL, Uniform 3D, Heliform Soft, Heliform XtraSoft, Heliform XL, Heliform XtraSoft XL, Freeform, Freeform Mini, Freeform XtraSoft, and the CEREPAK Detacher Handle.

A note on classification terminology: The CEREPAK is a Class II medical device cleared via FDA’s 510(k) pathway (clearance K220040) — not a Class III PMA-required device as originally stated in this article. This is distinct from the recall’s Class I designation, which describes the severity of the health risk posed by the defective units, not the regulatory authorization pathway.

What Clinicians Should Do Now

Facilities holding affected inventory should immediately stop using all affected units; identify and physically segregate affected stock; notify all relevant clinical staff and any external facilities that received affected products; and contact J&J MedTech sales representatives for help identifying suitable alternatives. No date for return to market has been announced. Clinicians with concerns about patients who received procedures with affected coils should follow institutional adverse event protocols and may report to the FDA’s MedWatch program.

Correction — 2026-06-14: An earlier version of this article incorrectly described the CEREPAK Detachable Coil System as a Class III medical device requiring premarket approval (PMA). FDA records confirm it is a Class II device cleared via 510(k) pathway (clearance K220040). The terminology note in the article above has been corrected. Desk: Devices & Diagnostics.

The clearance arrives in the wake of ISO 80601-2-61:2026, a newly published revision to the global pulse oximeter performance standard that, for the first time, requires device makers to demonstrate accuracy across defined skin-tone groups rather than a predominantly light-skinned validation cohort. According to Medtronic, Nell-EQ is the first commercially cleared pulse oximeter to meet that updated standard.

The FDA had previously granted the technology its Safer Technologies Program (STeP) designation — a pathway reserved for devices reasonably expected to meaningfully improve on the safety of currently marketed alternatives, and one that affords closer FDA collaboration during development and review.

The urgency behind both the new standard and the STeP designation traces back to a well-documented clinical problem. Studies conducted during the COVID-19 pandemic found that conventional pulse oximeters overestimate blood-oxygen saturation in patients with darker skin, with SpO₂ readings running 0.6 to 1.5 percentage points higher than measured arterial saturation. That gap translated into false-negative rates for hypoxemia that were up to 35 percentage points worse in darker-skinned patients — delays that likely contributed to higher COVID-19 mortality among Black and Hispanic patients.

Nell-EQ addresses the problem through advanced signal processing that incorporates patient-specific and sensor-specific optical signal characteristics, reducing the distorting influence of melanin on light-absorption measurements.

Clinical validation spanned 146 adult subjects and 319 pediatric cardiac patients across the full spectrum of skin tones. Medtronic reported that accuracy met or exceeded the threshold set in FDA’s draft guidance for diverse-population performance.

The Nellcor platform is one of the most widely deployed pulse oximetry systems in hospital critical care, making the Nell-EQ upgrade potentially significant in scale as well as design.

In analytical performance studies submitted to the FDA, the BACTEC FXI demonstrated a mean time-to-detection (TTD) for common bloodstream pathogens — including Staphylococcus aureus, Escherichia coli, and Candida albicans — approximately three hours shorter than the predicate BACTEC FX system under equivalent clinical loading conditions. (Note: Candida albicans is a fungal, not bacterial, pathogen; blood culture systems detect both bacteria and fungi, making “bloodstream pathogen” the accurate collective term.)

Blood cultures are the diagnostic backbone of sepsis workup. Sepsis mortality rises with each hour of delay before effective antibiotic administration; reducing the laboratory phase of that delay even by a few hours has downstream clinical implications. A three-hour reduction in median TTD would, if replicated in real-world hospital settings, compress the window from blood draw to positive culture and enable faster escalation from empirical broad-spectrum coverage to targeted therapy.

The BACTEC FXI is cleared for aerobic and anaerobic bacterial and fungal detection in adult and pediatric blood culture bottles. The 510(k) pathway indicates the FDA found the device substantially equivalent to a legally marketed predicate; it does not involve the full premarket approval (PMA) process used for higher-risk, novel-mechanism devices. The 510(k) clearance is distinct from PMA approval.

Waters has not disclosed a commercial availability date. The BACTEC FX is currently installed across many reference laboratories and hospital microbiology departments; the FXI is expected to be offered as an upgrade pathway for existing installations.

Correction, June 10, 2026: Three errors were identified and corrected in post-publication fact-check. (1) Company attribution: the original article attributed the clearance announcement to Becton, Dickinson and Company and cited a BD press release URL. The clearance was announced by Waters Corporation, which acquired BD’s Biosciences and Diagnostic Solutions business (including the BACTEC franchise) via a Reverse Morris Trust transaction in 2025; the source citation has been updated to the Waters Corporation press release on PRNewswire. (2) Statistical measure: the original article stated “median time-to-detection”; every primary source, including the Waters press release, reports this improvement as mean time-to-detection (approximately 3 hours, from ~20 hours to ~17 hours). Corrected to “mean.” (3) Organism classification: Candida albicans is a fungus, not a bacterium; grouping it under “bacteremic pathogens” is a scientific classification error. The collective term has been changed to “bloodstream pathogens,” which correctly covers both bacteria and fungi.

The alert covers Giraffe and Panda Warmers fitted with the Integrated Resuscitation System (iRes) and the Giraffe Stand-alone Infant Resuscitation System, specifically units containing the M1091607-R blender. Per the FDA notice, the air-oxygen blender knob shaft on certain systems can loosen, affecting the delivered oxygen concentration so that it does not match the user’s set concentration — potentially resulting in hypoxia or hyperoxia in the patient.

That failure mode is unforgiving in the population it serves. Titrated oxygen is central to neonatal resuscitation: too little risks hypoxic injury, while excess oxygen exposure in newborns — particularly preterm infants — is linked to oxidative harm. A blender that silently drifts off its set point undermines the one control clinicians rely on most in the first minutes of life.

What the alert says, and what it doesn’t

GE HealthCare sent affected customers a letter dated May 8, 2026. The FDA flagged the issue publicly on June 8, 2026 — about a month later — through an Early Alert, which the agency describes as a notice about a potentially high-risk device issue.

As of May 20, 2026, GE HealthCare has not reported any serious injuries or deaths associated with this issue.

The FDA’s recommended actions: inspect Giraffe and Panda iRes and stand-alone resuscitation systems to determine whether a unit contains an affected blender, and if so, remove it from patient use and contact a GE HealthCare service representative. U.S. customers can reach GE HealthCare Service at 1-800-437-1171.

For units that must remain in service while awaiting correction, the FDA describes an interim mitigation: perform the system’s pre-use checkout and Multipoint Oxygen Concentration Check as outlined in the user manual, and if those pass, GE HealthCare recommends verifying oxygen concentration with an independent oxygen analyzer on a frequent basis until the unit is corrected. This step does not substitute for removing an affected unit, but it guards against undetected drift in the interim.

The number of affected units and the distribution timeline were not disclosed in the alert.

For now, the practical signal for neonatal and NICU teams is concrete: confirm whether your resuscitation systems carry the M1091607-R blender, run the pre-use checkout and Multipoint Oxygen Concentration Check, verify delivered oxygen with an independent analyzer frequently on any unit still in use, and remove and report affected units per the FDA’s guidance.

The clearance is a Special 510(k), the pathway used when a manufacturer modifies a device it already markets, with the agency finding the new module substantially equivalent to its predicate. The submission was received April 27 and cleared in roughly five weeks. It sits under product code QNP — the gastrointestinal-lesion software category that the original GI Genius created when the FDA granted it a De Novo authorization on April 9, 2021. The device is a Class II computer-aided detection (CADe) tool under regulation 21 CFR 876.1520, reviewed by the Gastroenterology-Urology advisory panel.

What this clears, and what it doesn’t

The 510(k) record establishes equivalence and market authorization; it does not, on its own, publish new clinical effectiveness data. The widely cited performance figure for GI Genius — what Medtronic’s 2021 De Novo announcement described as “a 14% absolute increase in ADR compared to colonoscopy alone for both flat (42% increase) and polypoid (36% increase) lesions” — comes from that release for the original module, not from this filing. The release attributes the figure to a single study (Repici et al., Gastroenterology 2020) and does not disclose a sample size, confidence interval, or p-value; it also does not state whether the 42% and 36% lesion-type figures are absolute or relative.

A Special 510(k) clears a hardware-and-software upgrade as substantially equivalent — it is a regulatory equivalence finding, not a fresh efficacy trial.

Medtronic markets GI Genius as the exclusive worldwide distributor of the Cosmo-developed hardware. The companies have not yet posted summary documentation detailing what changed in the Module 300 generation.

The alert follows a voluntary Medical Device Correction that Insulet announced on May 26 for specific lots of three product families — the Omnipod 5 Automated Insulin Delivery System, Omnipod DASH, and the original Omnipod (Eros) system — in the U.S. and affected international markets. By the company’s own count, roughly 7 million Pods fall within the scope of the action, about 60% of which have already been consumed or expired. That represents about 8.5% of 2025 global Pod production.

A leak you may not feel

The hazard here is under-delivery, not overdose. According to Insulet’s release, the tear sits between the Pod and the point where the cannula enters the body, so insulin can escape outside the Pod. Patients might notice wetness, sticky adhesive, or the smell of insulin — but the FDA is blunt that detection can fail: “this issue may happen without triggering any alert, meaning you could receive less insulin than needed without realizing it. Do not rely only on Pod alerts to know if something is wrong.”

“In the most severe cases, prolonged and persistent high blood glucose levels can lead to diabetic ketoacidosis (DKA) … and can be life-threatening if not treated.” — FDA Early Alert

As of May 20, Insulet had logged 24 serious injuries associated with the issue and no deaths, the FDA reports; the company’s release characterizes the same 24 as serious adverse events tied to high blood glucose, including hospitalization and DKA. Both the FDA and Insulet confirm no deaths.

What the FDA tells affected patients to do

Critically, the agency does not tell patients to stop insulin therapy — the opposite. The Early Alert’s core protective instruction is to swap the suspect Pod for a working one so dosing continues: “If you are currently using a Pod from an affected lot, immediately change your Pod to resume insulin delivery.” If changing the Pod then leaves you short, the FDA says to “talk to your healthcare provider about other methods of insulin delivery.” The action, in other words, is to replace — not abandon — insulin delivery, either with an unaffected Pod or, if none is available, an alternate method arranged with your clinician. No reader should infer they should simply stop using their pump without a substitute in place.

The FDA stresses that this is a separate, new issue from Insulet’s March 12 removal of certain Omnipod 5 Pods — itself a Class I recall, the agency’s most serious category, covering 49 lots and 1,240,115 units, in which a manufacturing defect caused an internal cannula tear that leaked insulin into the Pod. Patients can check whether their lot is affected using Insulet’s lot-lookup tool and request free replacements; the company says it has sufficient supply and does not anticipate a shortage. This is reporting, not medical advice — patients with affected Pods should follow the FDA’s instruction to change the Pod immediately and continue insulin delivery, and should contact Insulet and their clinician with questions.

This is a software correction, not a device removal. No one is being asked to explant a working pacemaker. Boston Scientific is distributing a software update, Brady Software Maintenance Release 6 (SMR6), to be applied during routine in-office device-check appointments across the Accolade, Proponent, Essentio, Altrua 2, Visionist, and Valitude lines.

The FDA’s posted guidance is that prophylactic replacement before high battery impedance is confirmed is not recommended. Patients should keep their scheduled device checks and talk to their care team — not seek pre-emptive replacement surgery.

What can go wrong

The root issue is a battery problem, not a coding bug. In late 2024, Boston Scientific identified a subpopulation of Accolade pacemakers and cardiac resynchronization therapy pacemakers (CRT-Ps) with an increased potential to enter Safety Mode during telemetry because of high internal battery impedance. Safety Mode limits pacing functionality. An earlier release (SMR5) disabled wandless ZIP telemetry to keep ambulatory devices out of Safety Mode, but the manufacturer and the Heart Rhythm Society notice describe residual behaviors that SMR6 is meant to resolve — including incomplete ZIP disablement, magnet-induced false-positive battery-impedance testing, and a “prolonged voltage recovery state” that can leave battery diagnostics stuck. The advisory population has now expanded to include all Accolade CRT-P and dual-chamber extended-life (DR-EL) devices, which face a risk of early replacement.

According to the FDA correction notice as reported by trade press, Boston Scientific had reported four deaths and 2,557 serious injuries associated with this issue as of March 18, 2026. The Heart Rhythm Society’s clinician notice independently describes the four deaths: three patients whose devices entered Safety Mode from high battery impedance had syncope requiring hospitalization and later died, and one patient died after a replacement procedure. (See the verification note below regarding the serious-injury figure.)

The two underlying recall records are classified and reachable through the openFDA enforcement database. The lead record (Z-1770-2026), covering non-CRT pacemakers, lists 718,456 units in commerce and is marked “Open, Classified”; the companion CRT-P record (Z-1771-2026) covers another 685,776 units, for a combined 1,404,232 devices. Both were initiated March 19, 2026, posted April 24, 2026, and distributed worldwide.

The FDA’s posted guidance is measured. Prophylactic replacement before confirming high battery impedance is not recommended. Patients at risk who have not yet received the prior SMR5 update should, per the notice, schedule prompt in-person follow-up if four or fewer years of device longevity remain; everyone else should keep to the standard follow-up cadence. Patients can check whether a device is affected by model and serial number through Boston Scientific’s lookup tool. This is reporting on a regulatory action, not medical advice — clinicians and patients should follow the manufacturer notice and their own care team.

Verification note: On review, the cited CDRH recall front-end record (res.cfm?id=219239) and the FDA correction page returned errors to automated retrieval, so the device counts, dates, classification, and correction mechanism here are taken directly from the openFDA enforcement records (Z-1770-2026, Z-1771-2026). The four deaths and the “no prophylactic replacement” guidance are confirmed by the Heart Rhythm Society’s primary clinician notice. The “2,557 serious injuries as of March 18, 2026” figure could not be reconfirmed against a directly reachable FDA primary page and is attributed to the FDA correction notice as reported by trade media (MassDevice, Becker’s, Cardiovascular Business).

According to the FDA recall database, CareFusion 303, the BD subsidiary that makes the line, issued an Urgent Medical Device Recall on April 30, 2026. The recall reason the agency recorded is blunt: “Fingerprint scanner may overheat to a temperature to cause 1st degree burn.” The agency lists the root cause as “Component design/selection.”

Two records cover the action. The larger, Z-2298-2026, applies to the BD Pyxis MedStation ES and covers 15,933 units in commerce. A companion record, Z-2297-2026, covers the older Pyxis MedStation 4000 and lists a single unit. Both carry product code BRY and name all serial numbers of the affected configurations. (The cabinets themselves are regulated as Class I devices — a designation that describes the product’s risk category, not the severity of this recall.)

A familiar name on the recall list

Rather than pull cabinets from service, CareFusion told facilities to keep dispensing as normal but to stop using the Bio-ID scanner if it feels hot, update local procedures for when the scanner is unavailable, and return a Customer Response Form. Distribution was US nationwide plus roughly twenty other countries, including Spain, the UK, Canada, Australia and Brazil.

The hardware most clinicians never think about — the scanner you tap to pull a dose — is now the part the FDA flagged.

The recall lands against a difficult regulatory backdrop for the Pyxis franchise. In November 2024, the FDA issued BD a warning letter over quality-system violations at the same San Diego operation, according to MedTech Dive’s reporting on the inspection, which cited 111 open software defect tickets classified as catastrophic or severe patient harm and more than 92,000 complaints about system freezes, shutdowns or malfunctions that delayed medication delivery.

The current recall does not involve software or dispensing logic; it is a thermal hardware fault in the biometric reader. But it adds a physical-injury risk to a device line the FDA was already scrutinizing.

Correction (6 June 2026): An earlier version of this story described the recall as Class I, the FDA’s most serious tier. The FDA recall database lists both records (Z-2298-2026 and Z-2297-2026) as Class II; the Class I label applies to the cabinet’s device regulatory category, not to this recall. The headline, standfirst, chart and text have been corrected, and the November 2024 warning-letter figures are now attributed to MedTech Dive’s reporting. Flagged by The Vital Record’s independent verification pass.

This is journalism, not medical or investment advice.

The action is split across two device recall records, both initiated May 6, 2026 and classified by the FDA on June 3 as Class 2, the agency’s tier for products that may cause temporary or medically reversible harm. The larger record, Z-2276-2026, covers 582,030 units across seven Pencan spinal-tray variants in 24-, 25-, and 27-gauge configurations, two of them packaged with DuraPrep skin prep. A companion record, Z-2278-2026, covers 104,720 units of B. Braun’s customizable Design Options Spinal Tray, spanning 35 hospital-specific material variants. Both carry the same FDA product code, OFU, and both were distributed nationwide.

A drug recall in a device’s clothing

The FDA filings are explicit that these are “sub-recalls”: B. Braun is pulling the trays not because of a fault in the kit itself, but because the bundled bupivacaine is implicated in Huons’ upstream action.

“A sub-recall of a product recall from Huons Co, Ltd., the manufacturer of Bupivacaine Hydrochloride in Dextrose Injection, USP.”

The root cause remains listed as “under investigation by firm,” and the records do not specify the nature of the defect. Bupivacaine is the standard local anesthetic for spinal blocks in cesarean delivery and lower-body surgery, making the affected trays a routine stock item in many hospital anesthesia carts.

The failure mode leaves little margin. According to the FDA’s notice, the device’s proximal clasp can disconnect from the outer control tube, leaving the graft unable to unclasp from the delivery system. A surgeon may notice “a lack of resistance felt when sliding the apex holder back accompanied by a failure to release the proximal stent.” Once that happens the implant cannot be recaptured, and, the FDA adds, “this failure mode is not able to be recognized until it occurs during the procedure.”

There is no device-level rescue. The agency states plainly that “no device-based bailout method has been identified for this specific scenario.” When the techniques in the instructions for use fail, clinicians are told that “clinical judgment should guide timely decision-making, including consideration of conversion to open surgery” — a high-risk fallback in patients undergoing what is meant to be minimally invasive thoracic aortic repair.

What the deaths involved

As of April 23, 2026, Bolton Medical, a subsidiary of Terumo Aortic, had reported three deaths associated with the issue. One patient died of an aortic perforation; two others died of stroke after the procedure was converted to open surgery.

“This failure mode can occur without prior warning and no device-based bailout method has been identified for this specific scenario.” — FDA recall notice

The recall (FDA number Z-2160-2026) covers all lots of the RelayPro N4 non-bare-stent configuration, 32 mm and above, spanning 49 reference numbers. The devices were distributed worldwide, including Puerto Rico, Argentina, Chile, Canada, Colombia, Mexico and the United Kingdom. The FDA logs the action as a labeling correction rather than a physical removal, and Bolton is directing customers to consider alternative stent-grafts until mitigation measures are in place.

The company initiated the action on April 22, sent customer letters on April 23, and the FDA issued an Early Alert on April 28 before classifying the recall as Class I on May 20. The root cause remains “under investigation by firm.” The RelayPro is used to repair aneurysms, dissections and tears of the thoracic aorta.

The pressure in question is pulmonary capillary wedge pressure (PCWP), a measure of how congested a failing heart is. Today it is captured either by right heart catheterization or by an implanted pulmonary-artery sensor — both invasive. The software instead reads signals from a wearable chest sensor and infers the value.

The authorized indication is narrow: adults with heart failure with reduced ejection fraction (HFrEF), defined as a left ventricular ejection fraction of 40% or less, who have New York Heart Association (NYHA) functional class II, III, or IV symptoms.

What the evidence shows

The authorization rests on SEISMIC-HF I, a multicenter prospective study published in JACC: Heart Failure. Investigators enrolled 310 patients with HFrEF (EF ≤40%) across inpatient and outpatient settings (DOI). The wearable — Cardiosense’s CardioTag — records electrocardiography, seismocardiography, and photoplethysmography, and a machine-learning model maps those signals to an estimated PCWP.

The reference standard was right heart catheterization, with a blinded core laboratory adjudicating the tracings to set the criterion-standard labels. The model was then tested on a held-out set unseen until final evaluation.

In that test set, estimated PCWP differed from catheter-measured PCWP by 1.04 ± 5.57 mm Hg, with limits of agreement of −9.9 to 11.9 mm Hg.

The cohort’s mean measured PCWP was 18.1 ± 9.45 mm Hg. Patients averaged 61 ± 13 years, were 38% female, and were 44% White and 39% African American; the authors described performance as consistent across sex, race, ethnicity, and body-mass index (DOI).

Two caveats temper the result. The wide limits of agreement — roughly ±11 mm Hg — mean an individual reading can land well off the catheter value, and the authorized use is confined to HFrEF (EF ≤40%) patients with NYHA class II–IV symptoms; the larger preserved-EF population was not studied here. Whether a non-invasive estimate translates into the hospitalization reductions seen with implanted sensors is a separate, unanswered question. The De Novo establishes a device class; it is not evidence that monitoring changes outcomes.

The company positions DeepView as a clinical decision-support adjunct: a tool that “provides clinicians with” a healing-potential assessment, intended to inform rather than replace the clinician’s own burn-wound assessment and any decision to excise or graft. The device pairs multispectral imaging with a proprietary algorithm to assess healing potential, flagging areas judged unlikely to heal within 21 days — wounds that, in the company’s words, “may require significant medical intervention.” The openFDA record lists product code SHY under a Direct De Novo pathway, the route the agency uses to classify novel low-to-moderate-risk devices that have no existing predicate.

A new device class, at the bedside

Because it was granted through the De Novo pathway, the authorization establishes a new device classification rather than matching an existing one. Under the FDA’s general De Novo framework, a granted device can subsequently serve as a predicate for future 510(k) submissions — though the openFDA record for DeepView itself makes no statement about future predicate use.

The DeepView System is intended for use in burn care in various settings, including burn centers, trauma centers, and emergency departments.

Per the company, image capture takes about 0.2 seconds and AI classification 20 to 25 seconds, with the model trained on a database it describes as more than 340 billion pixels of burn-wound imagery.

One caveat for readers: the openFDA record does not yet publish the device’s decision summary, and the company’s announcement reports no sensitivity, specificity, or comparator-against-clinician figures. As a result, the device’s standalone accuracy in flagging non-healing burn areas cannot be independently assessed from the primary record. Until the FDA’s summary posts, the pivotal-performance numbers behind this authorization remain unstated — and current clinical practice should not change on the basis of this authorization alone.