Regulatory Open Forum

 View Only

  • 1.  Substantial Equivalence Testing

    This message was posted by a user wishing to remain anonymous
    Posted 10-Jan-2020 08:59
    This message was posted by a user wishing to remain anonymous

    Hi. 

    When attempting to prove substantial equivalence, should the validation testing be blinded?

    We are developing a software tool for analysing the lungs and I was wondering if the regulations state that testing has to be blinded, e.g. you do not sit side by side with the predicate device when performing the validation testing. 

    To me, it is a given that you should be blinded but I am receiving some push back on this. 

    Thank-you for any comments


  • 2.  RE: Substantial Equivalence Testing

    Posted 10-Jan-2020 10:31
    First, there are no "regulations" on how to perform validation testing (or most any other testing). The regulations say you must verify and validate devices, and in some geographies there are standards that are generally followed.

    That said, there is an expectation that such testing is scientifically valid. If your intent is to "compare" the two devices, I can think of a couple ways you might approach it. One, as you point out, is to have "new" users and blind them to which device they are testing. Another would be to have all users test both versions, but randomize the order they are presented in. Ideally the tester also doesn't tell them which is the "new" vs "predicate" device, but it could be all users already are familiar with the existing device so that isn't possible.

    It might be worth your while to talk to a consultant experienced in user/validation testing to help you create the most robust design that takes into account the details of your situation.

    g-

    ------------------------------
    Ginger Glaser RAC
    Chief Technology Officer
    MN
    ------------------------------

    Original Message


  • 3.  RE: Substantial Equivalence Testing

    Posted 13-Jan-2020 11:13
    A couple of thoughts: I am not clear on who or what should be blinded. It sounds like you are analyzing lung images between a predicate device and your device. If so,  there should be some blinding.

    1. The case image set could be blinded, someone else beside the users of the software creates the blinded data set.
    2. The users, if possible could be blinded to the software they are using.

    You just want to prevent any bias in the data you collect.

    ------------------------------
    John DeLucia
    Consultant
    ------------------------------

    Original Message


Related Content