Regulatory Open Forum

 View Only
  • 1.  IVDR Clinical Performance Study Plan - Decision Algorithms

    Posted 01-Mar-2023 11:25

    Annex XIII, 2.3.2 of the IVDR mentions all the items that must be covered in the CPSP. One of them (point q) are ''decision algorithms''. What does the IVDR mean by this? And what would I be expected to write down here? When is something a software, and when is it a ''decision algorithm''? Can someone please elaborate on what manufacturers are expected to hand in here, and what the content of this section should be?

    Lisanne Karbaat

  • 2.  RE: IVDR Clinical Performance Study Plan - Decision Algorithms

    Posted 01-Mar-2023 13:27
    Edited by Rajeswari Devanathan 01-Mar-2023 13:27

    For IVDR - one good example of diagnostic decision-making Algorithm, is a pathology lab test algorithms - From tissue collection, staining, and scanning, down to the question of how the results should be displayed to the pathologist, to help him take an informed decision. it often relies on the analysis of large amounts of personal data to infer
    correlations or, more generally, to derive information deemed useful to make decisions.


     For your next question When is something a software?

    In a  software qualified as a device, reference databases and other sources of data used as the basis for its decision making. While the clinical algorithm can be implemented into software (e.g CDSS), the algorithm itself is detached from IT implementations and is often depicted via flowcharts, decision trees or protocols. It is a rather simple set of instructions, which can be straightforwardly executed by humans. 

     For Both updates and versions planned, proposed testing, monitoring has to be specified and data has to be stored as part of CPSP. 

    Raje Devanathan
    Amerisource Bergen
    TPIreg, Innomar Strategies
    Senior Manager - Regulatory Affairs, Medical Devices
    3470 Superior Court
    Oakville ON L6L0C4

  • 3.  RE: IVDR Clinical Performance Study Plan - Decision Algorithms

    Posted 02-Mar-2023 20:35

    Hi Lisanne,

    Agree with you that as worded in the Annex XIII for CPSP 2.3.2 point q is very vague - decision algorithm in a CPSP is very broad.  It could mean for example the algorithm to call a sample result based on positive and negative and internal control test results as part of the assay (perhaps these are not incorporated into software, or perhaps they are).  It could also mean as described by Rajeswari below discriminating between qualitative results by means of showing what constitutes 'positive' vs. 'negative' or describing it in terms of staining (i.e. number of minimal viable cells, staining and scoring interpretation, tumor proportion score, or %stain).  I believe this is left broad on purpose to cover the wide range of assay types within in vitro diagnostics.  

    Kind Regards,

    Lubna Syed RAC
    Global Director, Regulatory Affairs

  • 4.  RE: IVDR Clinical Performance Study Plan - Decision Algorithms

    Posted 22-Mar-2023 06:51

    Many thanks to you both!

    So is my understanding correct if I say that the decision algorithm can be literally anything and any action that contribute to the making of a decision. For example for a PCR test, sample taking, sample preparation, running the analysis, setting a threshold and then comparing the Ct value results to those of a control. All of these together form the ''decision algorithm''. Part of this can be in software, but it does not necessarily has to be software. Any other factors influencing the decision (positive or not), but outside the PCR test itself, (for example other clinical data from the patient) are also part of the decision algorithm, because they influence the final positive/negative result. 

    Correct? Or am I still missing something?

    Lisanne Karbaat
    Project manager medical devices

  • 5.  RE: IVDR Clinical Performance Study Plan - Decision Algorithms

    Posted 22-Mar-2023 08:30

    Hi Lisanne

    Decision algorithms as described in Annex XIII don't appear to be the algorithms implemented within the device. Rather, they are the decisions you plan to make within the context of running the study. For example, you might have some rules like

    If fewer than X patients are enrolled within 6 months, then ...

    If more than X% of test results are associated with adverse events, then ...

    Anne LeBlanc
    United States