Users Are Not a Detection Control

Users Are Not a Detection Control

There is a point where a risk assessment stops managing risk and starts negotiating with it.

One common tell:

The control is the user catching the problem.

That sounds practical in a meeting. It can even sound reasonable on paper.

The labeling says inspect before use.
The defect should be visible.
The user should notice.
The issue should be contained by normal handling.

Maybe.

But "the user should catch it" is not the same thing as control.

What FDA Said

FDA's August 25, 2026 warning letter to Fresenius Medical Care AG & Co. KGaA describes a complaint trend involving leaking bags of Delflex Peritoneal Dialysis Solution.

According to FDA, Fresenius initiated an investigation in August 2025 after complaints about leaking bags. FDA said that investigation included 35 complaints covering approximately 156 bags from multiple batches, and that the firm attributed the leaks to holes caused by printing.

FDA also said the firm's risk assessment identified peritonitis as a potential harm, but the severity assigned in the assessment did not reflect that patient risk. FDA noted that the firm initially chose not to recall product after the investigation, relying in part on assumptions that users would inspect the bags, that leaks would be readily detected before treatment, and that leaking fluid would collect in the overwrap.

FDA's position was blunt: users should not be relied upon to detect leaking bags.

FDA also said that once the firm knew of a trend involving leaking bags in marketed product, the urgency to contain the issue should have increased. The warning letter further criticized the firm's visual inspection program and said the response did not include sufficient CAPA, including improvements to detect leaking units during manufacturing before release and distribution.

Those are the source facts.

Here is the operator lesson:

If your risk assessment depends on the user detecting the defect, your control strategy is already too far downstream.

Detection at the End Is Not Prevention

There is a difference between an instruction and a control.

An instruction tells someone what to do.

A control reduces the chance the failure reaches them in the first place.

That distinction matters.

Especially when the potential harm is not administrative, cosmetic, or theoretical. In the Fresenius letter, FDA focused on the patient risk tied to leaking peritoneal dialysis solution bags, including peritonitis.

That is not a paperwork inconvenience.

That is the kind of risk that should pull leadership, quality, operations, medical, regulatory, and supply chain into the same room with a real decision to make.

Not a slide to polish.

A decision.

Complaint Trends Are Signals, Not Background Noise

One complaint can be noise.

A complaint trend is different.

When complaints show the same failure mode across multiple batches, the question changes from "Can we explain this one event?" to "Do we still understand our process?"

That is where teams can get into trouble.

They treat the trend like a documentation problem.

Open the investigation.
Add the rationale.
Discuss detectability.
Write the conclusion.
Move on.

But the system is already talking.

Complaints, field alerts, rejects, inspection findings, deviations, supplier issues, and customer reports are not separate piles of paper. They are signals. If the signals point in the same direction, the quality system has to connect them fast enough to matter.

Risk Assessments Have to Name the Real Harm

A risk assessment is only useful if it is honest about the harm.

If the real concern is patient infection, product contamination, loss of sterility, treatment interruption, or exposure to compromised product, the risk assessment has to say that plainly.

Not in a footnote.

Not as a theoretical note that gets diluted by a lower operational category.

Plainly.

Because the harm drives the urgency. The urgency drives containment. Containment drives the decision.

If the severity is softened, everything downstream gets softer too.

The recall discussion.
The CAPA scope.
The inspection response.
The management attention.
The timeline.

That is how a risk assessment can look complete and still fail the job.

Visual Inspection Has to Match the Real Defect

Visual inspection is not magic.

It is a process.

That means it has to be designed, qualified, challenged, trended, and improved like a process.

If the defect is a small leak, the inspection method has to be capable of detecting that defect under realistic conditions. The qualification kit has to represent the actual failure mode. The records have to show what was inspected, what was found, what was missed, and what changed.

Otherwise, "we inspect for it" becomes another version of "someone should catch it."

That is not good enough.

A weak inspection process does not become strong because the risk assessment needs it to be strong.

Recall Decisions Need Urgency Before Certainty Feels Comfortable

Teams often want complete certainty before making a containment decision.

That is understandable.

Recalls are disruptive. They create noise. They cost money. They pull people away from already overloaded work.

But quality decisions do not always arrive with perfect certainty.

Sometimes the question is not "Do we know everything?"

Sometimes the question is:

Do we know enough to stop exposing users while we figure out the rest?

That is a different leadership posture.

It requires the organization to act before the narrative is fully comfortable.

Most teams say they want a strong quality culture until that culture asks them to make an expensive decision with incomplete information.

That is where the real test shows up.

Build the Control Upstream

When a complaint trend appears, the practical questions are not complicated.

They are uncomfortable, but they are not complicated.

What is the credible patient or user harm?

What evidence says the defect is contained?

What complaints, field alerts, rejects, visual inspection findings, deviations, or process signals point to a trend?

Can manufacturing detect the defect before release?

Is the inspection qualification kit representative of the actual defect?

What threshold triggers containment, recall discussion, escalation, or distribution hold?

Who has the authority to stop release or distribution?

What CAPA will prevent the failure from reaching the user again?

Those questions move the conversation from documentation to control.

That is where quality work belongs.

The User Can Be Careful and Still Not Be the Control

Users should follow instructions.

Patients and caregivers should inspect product when the labeling tells them to inspect it.

No argument there.

But that is not the same as making the user the detection system.

The quality system owns the defect before it leaves the facility.

Manufacturing owns detection before release.

Leadership owns escalation when the trend becomes visible.

CAPA owns prevention before the next batch becomes the next complaint.

If your control strategy needs the user to catch the problem, move the control upstream.

That is not consultant theater.

That is the job.

Website Publishing Notes

• Stage as Squarespace draft, not published.
• Author must be set to Two Pharma Guys.
• Add SEO title, meta description, URL slug, excerpt, source URL, and image alt text.
• Use the finished hero image at `outputs/users-are-not-detection-control-hero.png`.
• Confirm the draft appears in Blog Content before considering staging complete.