Remediation of diagnostic devices always seems to be tricky, even when the object of the diagnostic device is a non-life-threatening disease. While it is certainly important that we make a good-faith effort to apply FDA's 510(k) decision-making guidance already discussed, I would also add that it is equally (if not more) important to remember that, at the end of the day, regardless of the thought-paradigms in the guidance documents, FDA ultimately looks at these scenarios through the lens of public health and welfare.
Though a device's false positives may indeed only be for a non-life-threatening liver disease, FDA will nonetheless require full consideration of the associated and reasonably foreseeable risks. Specifically, in the case of false positive liver diagnoses, it is important to be sure to understand the reasonably foreseeable severity of harm (if any) and probability of harm (if any) stemming from a patient's and physician's false beliefs that the patient has a liver disease, when in fact the patient does not actually have such disease.
A primary risk in false positive scenarios can be related to the unnecessary administration of medical treatment (e.g., certain drugs, medical procedures, etc.), and to the emotional stress for the patient. Disease treatment and emotional stress can be significant even, for a non-life-threatening disease. Consequently, design changes made to control such risk might be viewed by FDA as an intent to "significantly" improve the safety or effectiveness, or to "significantly" affect clinical functionality or performance specifications.
One way to get insights about the meaning of "significantly" in a given context is to do so in correlation with the ranking scales in the firm's risk management process, and also perhaps in correlation with FDA's Health Hazard Assessment (HHE) / Health Risk Assessment (HRA) tool.
Because the compliance risk of a wrong 510(k) decision can be significant, the Sponsor may find it beneficial to appropriately consult the agency on the issue to be sure a proper decision is made. Remember also that such transparency with FDA will inevitably lead to a discussion and consideration of the need (or lack thereof) for a recall (FDA uses the HRA for that). Whether or not the Sponsor will find such transparency acceptable depends, at least in part, on the Sponsor's philosophy regarding its public health obligations.
------------------------------
Kevin Randall, ASQ CQA, RAC (U.S., Europe, Canada)
Principal Consultant
ComplianceAcuity, Inc.
Ridgway, CO
United States
www.complianceacuity.com© Copyright 2020 by ComplianceAcuity, Inc. All rights reserved.
------------------------------
Original Message:
Sent: 25-Jan-2021 12:56
From: Anonymous Member
Subject: FDA - Changes to Software and Special 510(k)
This message was posted by a user wishing to remain anonymous
Hi RAC Community,
I have a question regarding change to SaMD and opinions on weather a Special 510(k) is needed.
The SaMD in question was recently 510(k) cleared and we have not yet started selling the software in the US. The device is used to analyze MRI images and aid in the diagnosis of a liver condition. The dieses is not life-threatening.
Recently we have realized that the device is producing a false positive (a very high value compared to the predicate's value) result for a certain image (only one image that was not part of the validation dataset used for 510(k) submission). The reason for this anomalous result was due to resizing of the image that is done before the SaMD does the analysis. To fix this anomaly and avoid a false positive result we are changing this resizing tool. As result of this modification there is a minor change in the sensitivity, specificity, negative and positive percent agreement values of the device. There is no change to the core software that does the analysis; no change to the intended or indication of use; no change in the technology of the device; and no new risk have been identified.
My question is do we require to submit a special 510(k) for this change or a letter to file will be sufficient?
Thanks