Every OEM quality lead I talk to after a public recall asks a version of the same question: could this happen to us? After a decade of sitting across the table from device companies, I can tell you the honest answer is almost always yes, unless you can point to the specific gate that would have caught it. Recall notices tell you what failed. They rarely tell you which manufacturing control was supposed to stop it. That gap is where the useful conversation lives.
The Abiomed Impella cluster is a good teaching case because the failures are so specific, and because the FDA notices stop exactly where an OEM's own root-cause work has to start.
A cluster of recalls with one recurring theme
Between 2024 and 2025, Abiomed (J&J MedTech) worked through a string of FDA Class I recalls across the Impella heart-pump platform. FDA classified the removal of certain Automated Impella Controllers as Class I after a pump driver circuit assembly issue and possible capacitor failure, associated with one reported death (FDA Safety Communication, Catheter Controller Recall). J&J voluntarily removed 69 of those controllers, and trade coverage noted the action followed a prior 2023 FDA warning letter to Abiomed (MedTech Dive).
That was not the only vector. FDA also issued a Class I recall of Impella CP Sets with SmartAssist that were out of specification and could cause low purge pressure from the onset of a case, with potential loss of mechanical circulatory support; the action was linked to pump exchanges and a death (FDA Recall Notice, Impella CP Sets). A later 2025 controller recall traced purge retainer failures such as cracks to the force applied when inserting and removing purge discs, and was linked to five serious injuries (MedTech Dive).
Two more actions round out the picture but sit in a different category. FDA issued a Class I correction to the Instructions for Use for Impella left-sided blood pumps over left-ventricle perforation risk, with 129 serious injuries and 49 deaths reported (FDA Recall Notice, IFU correction), and a separate Class I recall addressed cybersecurity vulnerabilities in network and physical access, handled as a correction rather than a physical removal (AHA News).
Here is the pattern I want OEM teams to notice. The mechanical and assembly failures, the capacitor, the out-of-spec purge set, the crack from an assembly-force interface, are exactly the failure modes a mature build-gate program is designed to intercept. They are not exotic. They are the kind of thing that shows up when component qualification, in-process inspection, force validation, or lot-release acceptance has a gap.
The gates that catch these failures before they ship
Under the QMSR, the FDA quality-system framework now incorporates ISO 13485:2016 by reference. That matters here because each Impella vector maps to a clause an auditor can name. I am not the person who runs the incoming inspection or writes the validation protocol, but I have watched enough programs succeed and fail to know which gates the good ones treat as non-negotiable.
Component and supplier qualification. A capacitor that fails, or a circuit assembly that does not meet current specifications, is a purchasing-control question before it is anything else. ISO 13485:2016 Clause 7.4 requires manufacturers to set criteria for evaluating and selecting suppliers and to verify purchased product against specified requirements, proportionate to the component's effect on device quality. A common gap in supplier control programs is treating qualification as a one-time event rather than an ongoing obligation—ISO 13485:2016 Clause 7.4 requires that supplier evaluation criteria and purchased-product verification remain proportionate to each component's effect on device quality, which means requalification should be triggered whenever a component or its specifications change.
Assembly-force validation. The purge retainer cracking traced to disc insertion and removal force is the clearest example of a step where you cannot inspect quality in after the fact. ISO 13485:2016 Clause 7.5.6 requires validation of production processes where the output cannot be fully verified by later monitoring, with defined acceptance criteria and revalidation as appropriate. A controlled joining or force-dependent operation is precisely that kind of step. If the force window is not validated and monitored, a crack that forms during assembly can pass a visual check and still fail in the cath lab.
Set-level acceptance and lot release. An out-of-specification CP set that shipped is, at root, a final-acceptance question. ISO 13485:2016 Clause 8.2.6 requires documented acceptance activities across receiving, in-process, and finished-device stages, and prohibits release until those activities are satisfactorily completed and authorized. Clause 8.3 governs control of nonconforming product so that out-of-spec assemblies get segregated rather than distributed. The uncomfortable question every OEM should ask after reading that CP set recall is simple: how did non-conforming product clear final inspection and reach distribution? That is the gate to pressure-test in your own operation.
Cleanroom and contamination control. None of the acceptance gates hold up if the build environment introduces defects. ISO 13485:2016 Clause 6.4 requires documented arrangements for the work environment where conditions can adversely affect product quality, including contamination control for product and personnel. The classification itself runs through ISO 14644-1:2015, which defines air cleanliness by particle concentration. For a blood-contacting circulatory support device, the cleanroom class is not a formality, it is part of the acceptance chain.
Why a labeling correction does not close a build problem
The IFU correction is worth a separate note, because I see OEMs draw the wrong lesson from labeling recalls. An IFU update addresses how the device is used. It does not touch how the device was built. When the remediation for a Class I risk is a labeling change, the upstream design-transfer and assembly controls that determine whether the device performs as intended are left where they were.
I am not second-guessing Abiomed's specific corrective actions; I do not have their root-cause files and neither does anyone reading this. The point for your own program is narrower. When you read a recall that was resolved with an instruction change, ask whether the same failure could have been addressed upstream, at a build gate, rather than downstream, at the bedside. In my experience the customers who treat every recall (their own or a peer's) as a chance to test their own gates are the ones who do not become the next notice.
The checklist I would hand an OEM team
If you want to convert the Impella pattern into something actionable, here is where I would start. This is the operational lens, not a substitute for your quality team's analysis:
- Re-verify supplier qualification on every safety-critical component, especially electronic parts like capacitors, and confirm the qualification is current, not a one-time record (Clause 7.4).
- Identify every force-dependent or non-verifiable assembly step and confirm it is validated with a defined acceptance window and revalidation triggers (Clause 7.5.6).
- Trace one recent lot through your acceptance gates end to end and confirm no path exists for out-of-spec product to reach release without authorization (Clauses 8.2.6 and 8.3).
- Confirm your cleanroom classification matches the device risk and that contamination-control arrangements are documented, not assumed (Clause 6.4, ISO 14644-1:2015).
- Check that finished-device history records actually show the inspection was performed, not just that a step exists on paper.
A recall cluster like this one is a warning label for the whole industry. The vectors are specific, they map to named clauses, and they are catchable. The OEMs that come out of a review like this in good shape are the ones who can point to the gate, show the record, and prove the authorization happened before release.
For OEM teams evaluating contract manufacturing partners, the controls discussed here — component qualification, in-process inspection, and finished-device acceptance — are worth confirming are built into the production line by design, not added reactively. LSO's cleanroom assembly and kitting capabilities are structured around these same gates.