The number is 105,000.
Between April 2021 and the middle of 2023, the U.S. Food and Drug Administration logged more than 105,000 medical device reports referencing degraded sound-abatement foam in Philips Respironics sleep apnea machines, including 385 reports of death. Later FDA tallies pushed the death figure past 500. That number anchors nearly every account of the recall, and it is the ambient justification for why a question about corporate whistleblowing is being argued this month in a federal courthouse in Austin, Texas, on behalf of a lawsuit that will be tried in Amsterdam.
It is also a number that does not mean what most people who cite it think it means. Worth understanding, then, before we get to the venue problem — which is where the actual strategy lives.
What the 105,000 actually counts
A medical device report is a filing, not a finding.
Under 21 CFR Part 803, a manufacturer that becomes aware that one of its devices may have caused or contributed to a death or serious injury has 30 days to tell the FDA. Importers and hospitals file too. Clinicians and patients can file voluntarily through MedWatch. All of it lands in the same database. Nobody adjudicates it on the way in. The FDA's own boilerplate says the reports are unverified, that causality is not established, and that the data cannot be used to calculate incidence rates.
So the intake mechanism, in order: a patient reads about a recall, connects a symptom to a device, calls a lawyer or a clinician, a report is filed, a manufacturer is now on notice and files its own report on the same event, and the counter moves — sometimes twice for one incident. Recall publicity drives reporting volume. That effect is well-established in pharmacovigilance; it has a name, stimulated reporting, and it is the reason report counts routinely spike by an order of magnitude in the months after a product makes the news. How much of the 105,000 is that spike is not measurable from the outside. Anyone who tells you they've isolated it is telling you about their priors.
Here's the trap. Philips's own device count rose past five million units. Divide 105,000 by five million and you get roughly two percent, a figure that looks reassuring and is meaningless — the numerator is a self-selected pile of unverified filings and the denominator is registered hardware, not patient-years. The arithmetic is available. It just doesn't compute anything.
What the number can't do
It can't establish that foam particulate caused any individual cancer. It can't separate mechanical particulate exposure from volatile organic compound off-gassing, which are different hazards with different toxicology. It can't tell you whether a given death would have occurred anyway in a population that is, by definition, older and sicker than average.
And — this is the part that matters commercially — it can't tell you anything about what Philips knew, or when.
The Dutch case doesn't need it to. Investor claims are not patient claims. A shareholder suing over a product recall is not required to prove the product hurt anyone; the claim is about the interval between internal knowledge and public disclosure, and about who bought stock inside that interval. ProPublica and the Pittsburgh Post-Gazette reported in September 2022 that Philips had fielded complaints about foam degradation for years before the June 2021 recall. If that reporting is right, the operative number in the case is not a body count. It's a date on an internal document.
The financial context is not in dispute. Philips resolved U.S. consumer economic-loss claims for $479 million in September 2023 and personal-injury claims for roughly $1.1 billion in April 2024, both without any admission of liability. A January 2024 consent decree with the Justice Department barred new sleep apnea device sales in the United States pending compliance. Shares fell on the order of 70 percent from their 2021 peak — tens of billions in market value, most of it belonging to pension funds and asset managers who now want to know when the clock started.
Can a Dutch lawsuit force testimony in Texas?
Yes — through 28 U.S.C. § 1782, a statute that lets a U.S. district court order a person "found" in its district to give documents or testimony for use in a foreign proceeding. The requesting party doesn't need to be American. The foreign case doesn't need to be filed yet, only within reasonable contemplation. Applications are frequently made ex parte, which means a subpoena can issue before the target has said a word. What the target gets is a motion to quash, argued after the fact.
This is not a loophole. It is a statute Congress wrote on purpose, expanded in 1964, and it has become one of the most consequential asymmetries in transnational litigation.
How a claim in Amsterdam reaches a courthouse in Austin
Walk it in order.
First, the U.S. securities forum closes. Morrison v. National Australia Bank (2010) held that federal securities fraud claims reach only transactions in securities listed on U.S. exchanges or executed in the United States. A Dutch pension fund that bought Philips shares in Amsterdam is out. Not weakened — out.
Second, a Dutch forum opens. The Netherlands' WAMCA regime, effective January 1, 2020, allows a representative foundation to seek collective damages, not just the declaratory relief and court-blessed settlements the older WCAM framework produced. Amsterdam had already become a settlement venue for global investor claims; WAMCA made it a litigation venue.
Third, the foundation runs into Dutch procedure. Civil-law systems don't do American discovery. There is no general document production, no deposition practice, no compelled pretrial examination of an adverse party's former executives. A claimant who needs to prove when they knew has, in Amsterdam, very few tools for getting it.
Fourth — and this is the whole strategy — the claimant looks for a witness standing on U.S. soil. A former Philips director living in Austin is, for statutory purposes, a person found in the Western District of Texas. Reporting on the filings identifies him as Dr. Mohamed Elzayat, whose account of internal knowledge is the reason anyone is fighting over his calendar. In March, a magistrate approved the application. The subpoena issued. Philips moved to quash. A hearing is set for August 24.
Fifth, if the deposition happens, a transcript taken under U.S. rules crosses the Atlantic and enters a Dutch file that could never have generated it.
The four questions the judge is actually asking
Intel Corp. v. Advanced Micro Devices (2004) gave district courts four discretionary factors once the statutory requirements are met. They are the entire argument.
| Factor | What it asks | Why it's contested here |
|---|---|---|
| Participation | Is the witness a party to the foreign case? | He isn't — which cuts toward discovery, since a Dutch court can't reach him |
| Receptivity | Would the foreign tribunal accept this evidence? | Dutch courts generally admit § 1782 material; the dispute is over how readily |
| Circumvention | Is this an end run around foreign proof-gathering limits? | The defense's strongest ground — and the hardest to distinguish from "using a statute as written" |
| Burden | Is the request unduly intrusive? | Usually resolved by narrowing scope, not by quashing outright |
Note what's absent from that list: whether the underlying allegations are true. A § 1782 ruling decides who gets to ask, not what the answer was.
An honest rule of thumb
When you see a recall story quoting a report count, ask what the litigation actually has to prove — and then go find the number that bears on that. In investor cases it is almost never the harm tally. It is a date, and a document, and whoever was in the room.
The honest uncertainty here runs in both directions. That degraded foam shed particulate is well-established; the company recalled on that basis. That it caused a specific quantum of disease across the exposed population is plausible but thin — the epidemiology has never been done at the scale the report count implies. That Philips knew years earlier is documented by serious reporting and contested by the company's own procedural posture, which is a different thing from being settled fact.
This week, try one thing. Open CourtListener's RECAP archive, search the Western District of Texas for § 1782 applications, and read a single motion to quash — skip straight to the section arguing circumvention. Twenty minutes. You will come away able to spot, in any future recall, the difference between a company defending a record and a company defending a calendar.
The report count tells you how loud the aftermath was; only the docket tells you when the silence started.
