Philips Respironics: how known foam risk became a 5.5M-device recall

Anyone who has closed an audit knows this shape. A signal appears in complaint data. It gets logged. A risk assessment is commissioned, maybe even filed. Then nothing structural changes, and the product keeps shipping. The foam in a Philips CPAP machine did not cost the company billions. What cost billions was the distance between knowing and doing — the gap FDA inspectors walked into when they looked for the CAPA that should have closed the loop and found it absent or ineffective.

The situation

Philips Respironics was a dominant supplier of positive airway pressure therapy devices — the machines that keep airways open for sleep-apnoea patients, and in some cases ventilators for people with severe respiratory conditions. Inside each device, a block of polyester-based polyurethane foam did a straightforward job: it dampened the blower motor so patients could sleep. That foam sat directly in the air path delivering therapy to the patient's lungs, mouth and nose.

Patient complaints about black debris, particles, and chemical odour in the device airway had been documented well before the recall. The foam was degrading. Not in a fringe failure scenario — under the normal, continuous airflow and environmental conditions the device was designed for.

How it unfolded

The timeline is the familiar shape of a slow-moving quality collapse. Complaints arrived and were registered. What happened between registration and action is where the case turns. In a functioning system, a pattern of foam-degradation complaints triggers a material review, an engineering investigation, accelerated life testing, and a CAPA with verified effectiveness. The signal sat in the system instead.

The recall came in 2021, and it was not the quality system self-correcting. The gap between what was known internally and what had been communicated to regulators and patients had grown unsustainable. FDA inspectors examined whether Philips had identified the risk and whether its CAPA process had responded with adequate effectiveness. The consent decree that followed — halting US production and requiring remediation under external oversight — is the regulatory endpoint for a quality system that has demonstrated it cannot close its own loops.

Root-cause anatomy

The technical root cause is not mysterious to anyone who has run a PFMEA on a polymer in continuous service. Polyester-based polyurethane foams degrade through hydrolysis and oxidation. Heat, humidity, and constant airflow accelerate the breakdown. The foam crumbles, sheds particulate, and releases volatile organic compounds. This is materials science, not an ambush.

An FMEA line item for "sound-attenuation material degrades in air path, generating inhalable particulate" is not speculative. It is the kind of failure mode a competent cross-functional team should identify before volume production — not after five and a half million units are in the field.

The organisational root cause is where the real damage lives. The degradation was not a hidden defect that escaped detection. Complaints existed. Awareness existed. What did not exist was the machinery to convert that awareness into an effective corrective action — a change to the material, the design, the barrier, or the labelling — before external forces forced the issue.

Where the quality system failed

The failure touched every quality discipline that should have caught this, and it failed at each one in sequence. The PFMEA either did not identify foam degradation as a failure mode with a severity rating reflecting direct patient inhalation exposure, or it did and the rating did not drive design action. The risk management file under ISO 14971 should have connected the complaint trend to the residual-risk assessment and forced a formal benefit-risk re-evaluation. The CAPA system — where FDA inspectors focused — is the precise point where a registered signal must convert into verified action. The change-control process had an opportunity to flag material substitution or barrier design changes and did not drive the fix to completion.

A risk you have identified but not closed is not risk management — it is deferred liability with a serial number.

Each gate had a job. Each gate had a record. The records either documented inaction or documented action that was ineffective. Both are system failures, not foam failures.

What would have caught it

Accelerated life testing would have caught it first. Foam samples run through thermal, humidity, and airflow cycles that match years of patient use would have shown the degradation before volume production. If the test method did not replicate real-world conditions, the test method itself was the gap — and test-method adequacy is an FMEA input, not an afterthought.

Second, the system needed hard thresholds, not judgement calls. A rule that when particulate or odour complaints cross a defined rate per shipped unit, the CAPA gate opens automatically — no discretion, no monthly review queue. I have audited too many quality systems where the trigger depends on someone deciding the signal is serious enough this week.

Third, CAPA effectiveness verification against field data. If any corrective action was taken before the recall, continued complaint rates should have reopened it. Persisting complaints mean the CAPA failed verification. No exceptions, no sign-offs, no closing a finding that the field has not confirmed.

The PFMEA itself should have been revisited. The standard asks you to re-score when field failure modes emerge. If the degradation pattern was not fed back into the document, the feedback loop was broken at the file level — and a broken file is worse than no file, because it creates the illusion of compliance.

My take

I have spent two decades building and fixing quality systems in automotive and aerospace, and the pattern never changes. The costly failure is never the defect. It is the open loop. I have sat in plants where the CAPA log was a graveyard of good intentions, where 8D reports had action items with no owners and no closure dates, where the audit finding was always the same: you identified the risk, you just did not finish the job.

The 50% reduction in audit findings I delivered in one cycle at Airbus did not come from cleverer inspectors. It came from closing CAPAs properly and verifying they worked. The 70% defect-cost reduction I led at SNOP came from forcing the quality system to treat every signal as something that had to be resolved, not parked. In aerospace, an unaddressed nonconformance on a flight-critical part can ground an aircraft. In medical devices, the same organisational failure lands in a patient's lungs. The chemistry differs. The system failure does not.

What I look for in any quality system I audit or lead is simple. How fast does a signal become an action, and how do you prove the action worked? If the answer to either is vague, you are Philips before the recall — just with different math.

The Philips recall will be remembered as a foam problem. It was a loop-closure problem — the most expensive kind, because the system had every signal it needed and failed to act. The foam was always going to degrade under continuous airflow. That was knowable, known, and fixable. The quality system's job was to close the loop before five and a half million patients were breathing the evidence.