Performance Evaluation Report (IVDD)

The Performance Evaluation Report contains the methods and results regarding scientific validity, analytical
performance and clinical performance.

There's a separate standard available for that: EN 13612:2002. It's very short and doesn't contain a whole
lot of information. Additionally, there are three IMDRF guidance documents:

• GHTF/SG5/N6:2012
• GHTF/SG5/N7:2012
• GHTF/SG5/N8:2012

Product

• Name: <product name>
• Version: <product version>
• Basic UDI-DI: <insert UDI-DI, if/when available>

Mapping of Requirements to Document Sections

1. List of Abbreviations

2. Product

• Name: <product name>
• Version: <product version>
• Basic UDI-DI: <insert UDI-DI, if/when available>
• UMDNS-Code:
• GMDN-Code:

3. Relevant Documents

• SOP Performance Evaluation
• Performance Evaluation Plan

4. Intended Use

Copy-paste the intended use of your device here.

5. Risk Analysis

Copy-paste the summary of your Risk Analysis Report here.

6. Medical Context and State of the Art

6.1 Medical Context

Summarize in which medical context your IVD is used. If it's an HIV test, it may be used for screening, or
maybe only for people who think they've recently gotten infected with HIV.

6.2 State of the Art

Describe how this is currently done. Continuing the example above: What happens currently to those patients
who are screened for HIV, or those who think they've gotten infected? Are there any specific tests out there
or is the state-of-the-art procedure another one, like (random example) doing a chest x-ray?

7. Scientific Validity

This is generally based on literature research. Whatever your IVD is measuring, the current scientific
knowledge has some sort of (valid) reason for this, because it is associated with some sort of condition. Can
you still follow?

Here's an example: You've developed an HIV test. Based on current scientific knowledge, it makes sense to do
HIV tests on people because it's established that HIV is a non-benign disease which will lead to AIDS some
time in the future. Early detection is useful because early treatment leads to favourable
outcomes. Therefore, it's scientifically valid to do HIV tests.

7.1 Scientific Validity: Literature Search Methods

Describe your methods for your literature research for scientific validity. You'll probably have a list of
keywords which you'll be entering in certain databases (or other literature sources).

Some example literature sources from guidance document GHTF/SG5/N7:2012:

• Scientific databases – bibliographic (e.g., MEDLINE, EMBASE);
• Specialized databases (e.g., MEDION);
• Systematic review databases (e.g., Cochrane Collaboration);
• Clinical trial registers (e.g., CENTRAL, NIH);
• Adverse event report databases (e.g., MAUDE, IRIS);
• Scientific databases – bibliographic (e.g., MEDLINE, EMBASE);
• Specialized databases (e.g., MEDION);
• Systematic review databases (e.g., Cochrane Collaboration);
• Clinical trial registers (e.g., CENTRAL, NIH);
• Adverse event report databases (e.g., MAUDE, IRIS);
• Reference texts

7.2 Scientific Validity: Literature Search Results

Describe your search results from your literature research

7.3 Scientific Validity: Literature Search Conclusion

This is a bit like the "discussion" section in a scientific paper. You reach some sort of conclusion, based
on your literature search. In the HIV test example, this would be something like "testing for HIV is useful
because HIV is the disease which subsequently leads to AIDS, and early detection of that is good".

8. Analytical Performance

Pretty simple. Describe the metrics by which your IVD detects whatever it should detect.

In the HIV test example, those could be sensitivity / specificity values, in other words: If I use this test
on 100 blood samples from different patients, what sort of analytical performance can I expect?

This will require you to run your test on some sort of test set and do some analysis of those results.

8.1 Analytical Performance: Methods

Describe your methods. If you have a Machine Learning model, you could describe your test set and why you
chose that specific dataset as a test set. You could also describe the metrics by which you evaluate the
performance of your ML model.

8.2 Analytical Performance: Results

Describe your results. Again, similar to a peer-reviewed paper.

9. Clinical Performance

Slightly harder to comprehend and a bit similar to Scientific Validity. These are the performance metrics of
your product in its intended patient population.

So, for the HIV test: You'll have some numbers for the analytical performance, but that's only on the
"reagent" level. What are the metrics when you actually use that test on real people? There's probably a
different sensitivity / specificity. Maybe certain comorbidities (like other viral diseases) may lead to
false-positive test results. So, this is like analytical performance, but in the real world, on real
patients.

You can also do a literature research for this, or do a clinical performance study.

9.1 Clinical Performance: Methods

Describe the methods of your clinical performance evaluation.

9.2 Clinical Performance: Results

Describe your results.

10. Conclusion

Conclude why your IVD is safe and effective to use. It makes sense to refer to your intended use, the risks
in your risk analysis, and the scientific validity, analytical performance and clinical performance.

11. Dates and Signatures

Date and sign the plan. If your document management system supports it, you can digitally sign by typing
e.g., your initials in the "Signature" field. Otherwise, you can still sign it the old-school way (print it
and sign the sheet of paper, ugh).

12. Qualification of the Responsible Evaluators

Attach CVs of the people who were involved in writing the Performance Evaluation. They must be "adequately
skilled and trained".

<!-- Links -->

Template Copyright openregulatory.com. See template
license.

Please don't remove this notice even if you've modified contents of this template.