By Aeroz Editorial September 2026
Fresenius Medical Care AG describes itself as the world's leading provider of products and services for individuals with renal diseases. According to the FDA's warning letter, its Ogden, Utah plant manufactures large volume parenteral (LVP) bag drug products, including Delflex Peritoneal Dialysis Solution. The FDA conducted an inspection from March 2 to 6, 2026. The FDA's warning letter cites two CGMP violations: failure to thoroughly investigate complaints about leaking bags (21 CFR 211.192) and inadequate production and process control procedures, including the visual inspection program (21 CFR 211.100(a)).
The letter describes the leaking-bag complaints as a “complaint trend.” According to the letter, in August 2025 the firm initiated an investigation due to a complaint trend for leaking bags of Delflex Peritoneal Dialysis Solution, and that investigation noted the firm eventually received 35 complaints covering approximately 156 bags from multiple batches. The investigation that followed was, according to the warning letter, not thorough and did not appropriately evaluate the risk to product quality. The letter states the firm failed to identify and implement adequate and timely corrective and preventive actions (CAPAs), and that once it knew of the trend “there should have been a commensurate increase in urgency to contain the issue.” According to the FDA, the firm chose not to recall after its August 2025 investigation; it issued voluntary recalls of Delflex Dextrose Peritoneal Dialysis Solution for lack of assurance of sterility on April 6, 2026, roughly eight months after the investigation opened and one month after the inspection ended. The FDA's warning letter is dated August 25, 2026.
“The moment that deserves the most scrutiny is the gap between the complaint trend and any visible corrective escalation.” Moe Alsumidaie, Clinical Trial Vanguard, September 8, 2026
The complaint file said property damage
The detail that carries the most weight in the warning letter is not the leak rate. It is the severity classification.
Fresenius maintained a Risk Management Matrix for this product. That matrix identified peritonitis as a potential harm associated with a compromised peritoneal dialysis bag. Peritonitis in a dialysis patient is not a quality inconvenience. The FDA’s own language in the letter describes exposure to potentially non-sterile drug products as risking peritonitis with serious and potentially life-threatening complications.
According to the FDA, although the matrix recommended the highest severity level, the firm assigned the lowest severity level, with the potential harm recorded as “damage of property.”
So the company's own risk framework named the clinical hazard, while its investigation recorded the potential harm as property damage. In its response, according to the FDA, the firm acknowledged the risk assessment should have been based on peritonitis risk, reassessed the risk, recalled the impacted lots and revised its procedure; the FDA found the response inadequate. Everything downstream followed from the lower number. In a typical quality system, a property-damage classification does not trigger batch extension or field action.
Assumptions doing the work of evidence
Fresenius did not decline to recall in August 2025 without reasoning. According to the FDA's warning letter, its justifications included that the leaks should be readily detected before treatment, and:
- No confirmed cases of peritonitis had been attributed to the affected bags.
- Product labeling instructs users to inspect for leaks before use.
- Escaping fluid would collect in the outer overwrap, making a compromised bag visible.
Each of these sounds defensible in a conference room. None of them was a measurement.
Following the inspection, Fresenius stated that it reexamined part of one batch and “found presence of perforations in the primary container without substantial fluid present in the overwrap,” which the FDA said demonstrated the error in assuming users can consistently detect leaks. That single finding undercuts the overwrap assumption directly, and the label-inspection and detection assumptions with it. If fluid does not pool where the assumption says it will pool, then the inspection the label asks for is looking for a sign that is not reliably there. The user is being asked to detect a perforation by observing its absence of evidence.
The first proposition is the weakest and the most common in complaint files across the industry. Peritonitis in a dialysis patient has many plausible causes, presents days after exposure, is treated clinically rather than forensically, and is almost never traced back to a specific bag from a specific batch. A batch signature is only found when a system is built to produce one. The absence of attributed cases does not establish the absence of cases. It is the absence of attribution, and those two things are only the same thing if you have a mechanism that would have caught the difference.
There is one more detail worth sitting with. The investigation attributed the leaks to holes caused by printing. The FDA's letter asks Fresenius to provide an independent review of its printing controls and an analysis of alternative printing technologies. Printing is typically how human-readable information is applied to a flexible container, and here the printing step was named as the cause of holes in the container. Anyone whose product identity layer is physically imposed on the primary package should read that sentence twice.
The second violation is about human detection
The warning letter cites a second CGMP violation under 21 CFR 211.100(a), covering the visual inspection program for sterile injectables. The letter states that the firm failed to appropriately qualify personnel who performed visual inspections, that its qualification kit did not include adequately representative visible particulates (e.g., size), and that its qualification records lacked sufficient detail to determine the qualifications of visual inspection staff.
The agency's framing of the underlying problem is worth quoting: “Visual detection of particulates is a probabilistic process,” and qualification kits “should typically contain particles in the range of 100-150 microns to adequately challenge inspector detection capabilities.”
Read the two violations together and a single shape emerges. Both findings involve relying on people to see a defect: users expected to spot leaking bags, which the FDA said “should not be relied upon,” and plant inspectors whose qualification records, per the FDA, lacked sufficient detail to establish their qualifications.
When verification depends on a person noticing something, verification fails quietly and the failure leaves no data behind.
What the FDA is asking for now, and what it costs
The remediation list in the warning letter is broad. The letter gave Fresenius 15 working days to respond and asked it to provide, among other items:
- A comprehensive independent assessment of root causes, including failure mode analysis for the leaking bags
- An independent review of printing controls and alternative printing technologies
- An assessment and remediation plan confirming the quality unit has the authority and resources to function
- A comprehensive assessment of the deviation investigation system
- An independent review of in-process criteria for defects, with a five year retrospective evaluation
- A process capability assessment with a statistical monitoring program
- An independent review of equipment qualification and validation
- Engagement of a CGMP consultant under 21 CFR 211.34 and a six system audit of the entire operation
Two of those items tell you how the agency is reading this. The request for a plan to ensure the quality unit “is given the authority and resources to effectively function” can be read as asking whether quality could have escalated even if someone had wanted to. And the five year retrospective evaluation of defect criteria is the expensive one. The FDA is not asking what went wrong with one product. It asks for at least five years of batch performance summaries covering the period of recurring leaking incidents, including every batch with atypically high defect rates.
That is the general cost of a complaint system that cannot prove which units were affected. You cannot bound the question yourself, so the regulator bounds it for you, and the boundary it chooses is everything you have made.
This is a data architecture failure before it is a diligence failure
The failure pattern documented in the Fresenius warning letter is the same one the FDA has flagged in enforcement actions across the pharmaceutical and medical device industries for years. In that broader pattern, complaints are received, logged, and closed without thorough investigation, without extension to other potentially affected batches, and without timely corrective action that prevents recurrence.
The warning letter is the last signal in a chain that should have triggered serious response much earlier. But it is worth being precise about why the chain broke, because the obvious reading is the wrong one.
The obvious reading is that someone was careless. The more useful question is whether a complaint management system connects individual complaint events to a unit-level manufacturing record that can identify which batches, production windows, and production conditions are associated with the failures; the warning letter does not address this directly. Without that connection, every complaint is an isolated event rather than a signal in a pattern. There are three questions a complaint file like this needs to answer:
- Which specific units are affected, beyond the ones a customer happened to complain about
- What was different about the line, the shift, the printing run, or the environment when those units were made
- Where did the rest of that population go, and can it be reached
A complaint system that cannot answer those is not a system that detects trends. It is a system that stores anecdotes and waits. The pattern only becomes visible in retrospect, when the FDA arrives with a Form 483.
We should be honest about the limit of this argument. Unit-level traceability does not make the severity call. The firm assigned the lowest severity level despite its own risk matrix, and better data does not automatically overrule that decision. What it changes is how long that call survives. A severity downgrade is easy to sustain when the only counter-evidence is 35 complaints with no shared structure. It would be much harder to sustain if the record showed, for example, those complaints clustering on a few production windows on one line, along with how many more units from those windows were still in clinics.
The Aeroz Angle
Aeroz creates a real-time, unit-level chain of custody that connects every individual unit to its exact production environment, handling history, and downstream distribution path. When a complaint arrives, the affected unit is traceable to a specific lot, line, and time window in under 60 seconds.
Concretely, in a file like this one, that changes four things.
The first complaint carries a manufacturing fingerprint. Not a lot number printed on the bag, which points to a population of thousands of bags, but the actual production conditions behind that unit. A single complaint becomes a query rather than a record.
The trend is bounded in August, not after the inspection. Thirty-five complaints across multiple batches can be recognized as a trend and still be underweighted when the only shared field is the product name. The same 35 complaints looks like a cause when they resolve onto a narrow set of production windows. Fresenius identified the trend in August 2025; what a unit-level record adds is its scope.
The affected population is bounded, not estimated. The recall question stops being a judgment call about exposure and becomes an inventory question with an answer. That is the difference between an eight-month gap from investigation to recall and a defensible field action in days, and it is also the difference between a recall scoped to what you can prove and a recall scoped to everything you cannot rule out.
The assumption gets tested instead of asserted. Fresenius assumed fluid would collect in the overwrap. Its own post-inspection reexamination of part of one batch found perforations without substantial fluid in the overwrap. In a unit-level system, that assumption is checkable against real field returns tied to real production conditions, long before a regulator checks it for you.
The FDA found the Fresenius investigation was not thorough and did not appropriately evaluate risk. Connecting a complaint trend to specific units and production conditions supports that kind of investigation, and that connection is what Aeroz makes possible from the moment of manufacture.
None of this is on a compliance calendar. The FDA did not cite Fresenius for failing to adopt traceability technology, and no current CGMP requirement names it. The citations were for failing to investigate thoroughly and evaluate risk appropriately, and for an inadequate visual inspection program. The point is that thorough investigation has a data prerequisite, and most complaint systems in this industry do not have it. The gap between a complaint trend and a corrective escalation is not a gap in intent. It is a gap in what the record can say.
Aeroz, Making Authenticity Checkable. Visit aeroz.io to learn more & get in contact with our team via [email protected].
2026 Aeroz all rights reserved.
Sources
- U.S. FDA — Warning Letter 320-26-119, Fresenius Medical Care AG & Co. KGaA (25 Aug 2026)
- Fierce Pharma — Fresenius unit dinged in FDA warning letter flagging 'complaint trend' from customers (3 Sep 2026)
- Clinical Trial Vanguard — Fresenius Got a Warning Letter Six Months After the Complaints Started. That Gap Is the Real Problem. (8 Sep 2026)
- Fresenius Medical Care — Company profile
Updated 11 October 2026: headline revised, claims checked against primary sources and corrected, and sources added.