Interim Risk Control in a Device Recall: Was It Tested?

Engineer timing a manual double-check by comparing a monitor with a handheld reference meter at a test bench: testing an interim risk control in a device recall
In this post
  1. The two notices
  2. Two different devices, the same instruction
  3. Where that control sits in the hierarchy
  4. The uncomfortable part
  5. What testing an interim risk control would look like
  6. Information for safety is still a risk control
  7. A short checklist for the next one
  8. Where Nopmark fits
  9. FAQ
  10. Test the control, not just the code

FDA posted two Class 2 device recalls five days apart this month. Both name software design as the cause, and both ask a clinician to do something by hand until the fix ships. That instruction is an interim risk control, and it is usually the only control that has not been tested.

The two notices

The first of the two covers the Midmark Digital ECG. Its record is Z-3013-2026, posted 25 August 2026. The firm began the action on 20 July.

The manufacturer's stated reason runs to a single sentence. A communication anomoly may result in incorrect electrocardiogram waveforms of a patient and an inaccurate beats-per-minute calculation. FDA lists the determined cause as software design.

Then comes the part this article is about. Midmark recommends a manual clinical assessment of the patient's heart rate before starting therapy. That applies to any tracing that reads as a fast heart rate, until the corrected plugin software is installed.

The second of the two covers FUJIFILM Synapse PACS. Its record is Z-3030-2026, posted 26 August 2026. The firm began that action on 22 July.

Its stated reason is shorter still. A software issue may cause the ruler measurement to be inaccurate. FDA again lists the determined cause as software design.

The correction says the system may stay in use, provided the end user knows about the issue and takes precautions. One of those precautions is to perform the affected measurements on the ultrasound machine itself, before the study is transferred to the archive.

Two different devices, the same instruction

Read those two side by side and the scope is identical; the device produces a number but the number may be wrong. The clinician is asked to produce the number a second way, by hand, until a patch arrives.

This is a real control, and it may well be the right call as pulling every affected system out of service would carry its own harm. But it is worth being precise about what has just happened. A software defect has been converted into a human task. The safety of the device now rests on a busy clinician noticing a condition and doing extra work under time pressure.

Where that control sits in the hierarchy

Risk management practice puts risk controls in a priority order. You design the hazard out first. If you cannot, you add a protective measure, such as an alarm, a lockout, or an automatic check the device performs itself. Only when neither of those is available do you fall back on telling the user.

Telling the user is the least effective for a reason. Every control above it works whether or not the user is tired, busy, new, or covering two rooms. The last one only works if a person reads it, remembers it, spots the situation, and then acts. So an interim risk control of this kind is the weakest tool in the tool-box.

The uncomfortable part

Here is the question I would want answered if this were my device. Somebody wrote that instruction. Did anybody test whether it works?

For a design control, the answer would be obvious. You would not ship an alarm without evidence that it fires. You would also want to know it is audible in the real room, and that the user responds correctly. Risk controls have to be tested for effectiveness, not for presence. You show that the hazardous sequence no longer occurs, rather than showing that a control exists somewhere.

The interim control tends to get a different standard. Quality and regulatory staff write it, review it for accuracy and tone, approve it, and send it. It is a use-related requirement, and it may never go near a usability study. That gap decides whether the mitigation you reported to FDA actually reduces risk in a clinical setting.

What testing an interim risk control would look like

None of this needs a new discipline, because the methods already exist in usability work. The work is small when the instruction is small.

  • Detection: can the user tell when the condition applies? For the ECG case, the trigger is a tracing that reads as a fast heart rate, which is fairly easy to see. For the archive case, the trigger may be much harder, because a wrong ruler measurement looks exactly like a right one.
  • Availability: is the alternate method actually available? Performing the measurement on the ultrasound machine assumes the user can get back to it, and that the study has not already moved on.
  • Workload: does the extra step survive a busy shift? A control that adds a minute to every study is a control people may skip on the twentieth one.
  • Understanding: does the notice read the way the writer thinks it reads? Hand it to 5 clinicians who have not seen it. Ask them what they would now do differently, and compare that to what you intended. This is a short exercise and it tends to be revealing.

Detection is the one I would look at hardest in the FUJIFILM case. When a defect produces a plausible wrong number, asking the user to notice it is asking for something the defect is designed to prevent.

Information for safety is still a risk control

Usability standards treat information for safety as part of the device. If your control is a sentence, the sentence is subject to the same question as any other control. Does it work, and how do you know?

That understanding changes what a field correction letter is. The letter introduces a use-related requirement into a live clinical workflow, and it does so without the development cycle a requirement normally gets.

A short checklist for the next one

If you are drafting a correction notice, four questions cover most of it.

  1. What exactly triggers the user's action, and can the user see that trigger?
  2. Is the alternate method available at the moment they need it?
  3. What does this add to their workload, and for how long?
  4. Has anyone outside the writing team read it and told you what they would do?

None of that takes long, and it is far cheaper than finding out later that a mitigation you reported did not hold.

Where Nopmark fits

Two of the ideas above come from the same place. Testing a risk control for effect rather than presence sits in risk management practice, and evaluating whether an instruction works sits in usability engineering. Most teams keep those in separate departments, which may be why the interim control falls between them.

Nopmark's medical device courses cover both sides. They include how risk controls are verified, and how use-related risk is judged against real users.

FAQ

Is a warning ever an acceptable risk control? Yes, when the options above it have been looked at and ruled out. What makes it acceptable is the reasoning and the evidence, not the fact that a notice was sent.

Does an interim risk control need its own records? Treat it the way you would treat any control you rely on. If it is reducing risk in the field, the rationale belongs in the risk file rather than only in a customer letter.

We have very little time during a field action. Where do we start? Start with detection. If the user cannot reliably tell when the condition applies, the rest of the instruction does not matter. You may need a different mitigation.

Do these two recalls mean software recalls are rising? That is not a claim these records support. They are two entries, posted five days apart, and FDA posts recall records all the time.

Why does the posting date differ from the recall date? FDA records tend to appear weeks after a firm begins a field action. Both of these were initiated in July and posted in late August.

Test the control, not just the code

Nopmark's Medical Device Risk Management course teaches how to build a risk file that holds up. That includes verifying that a control reduces risk rather than simply existing. The course page is at nopmark.com/courses/medrisk.

Keep reading