Statistical justification of sampling size in V&V tests

[Ocean12]

Hi,
ISO 13485:2016 and the QSR both require a statistical based justification behind the selection of the sampling size in V&V testing.
For one of our V&V tests, we tested 3 samples because we followed an harmonized standard which is specific for this type of test and requires the use of 3 samples.
My question is: will the NB accept the reference to the harmonized standard to justify the selection of 3 samples without any further "statistical" justification? It makes sense to me that the regulatory body that accepts that standard should also accept the sample size of the standard. What do you think about it? What is your experience with you NB or FDA in relation to statistical justification? Thank you for your feedback!

 

[shimonv]

Hi,
It's difficult to say without knowing what the product is and to which harmonized standard you are referring to, but I can say this:
Sample size is risk-based; a high risk requires a statistical prediction with high confidence level and high reliability, which means a larger sample size.
You need to look carefully in the harmonized standard and see whether the recommended sample size of 3 matches the description of your V&V plan. If so, you can reference it in your V&V protocol and that should be sufficient.

Shimon

 

[Ronen E]

I would carefully look into the standard's wording. 3 datapoints is definitely not enough for any meaningful statistical prediction; however, 3 samples doesn't necessarily mean 3 datapoints as you could test each unit more than once (and there are statistical methods based on such an approach).

 

[shimonv]

Come to think of it, I have never seen a standard that is bold enough to prescribe a sample size for a specific test. Even if the test procedure is described in detail in the standard - the medical device can be used in a number of ways and in different clinical application. The risk levels might be different from one case to another, which means different levels statistical confidence might be required, which means different sample size...

 

[Highground]

Hi,
It's difficult to say without knowing what the product is and to which harmonized standard you are referring to, but I can say this:
Sample size is risk-based; a high risk requires a statistical prediction with high confidence level and high reliability, which means a larger sample size.
You need to look carefully in the harmonized standard and see whether the recommended sample size of 3 matches the description of your V&V plan. If so, you can reference it in your V&V protocol and that should be sufficient.

Shimon

This makes sense and helps me out...V&V plan actually has the sample size rationale.

 

[Ocean12]

Understood. Thank you Shimonv