Medical Device Design verification sample prototype

ringring

Registered
Hi,
I have been reading a lot of topics about this subject but can't really understand how it is done..
What should I do (step by step) to calculate the sample size for design verification. We need to do it on few samples but in the audit they told us to come up with a valid statistic argument to justify the sample size.
We are testing an active guidewire (DV tests: Biocompatibility, software, computer safety..)
Thanks for reading this
Ring
 

John Broomfield

Leader
Super Moderator
You verify the conformity of your prototype to your design before you then test to validate the design. That way you know exactly what you have tested to see if the prototype works as intended.

From your question it seems that you are seeking information on design validation not verification.
 

ringring

Registered
Thank you for your response.
No it is for design verification, validation will come after.
What I am looking for is how to justify (and chose) the sample size for the tests of DVerification
 

John Broomfield

Leader
Super Moderator
Sorry, my mistake.

Have you researched this site for Design of Experiments?

Your DoE with due regard for risk should determine your sample size.
 

John Predmore

Trusted Information Resource
When you do verification, there are four possibilities because of variation and measurement uncertainty. A hypothesis is true and your experiment supports the claim, or else you incorrectly conclude it is false. Or a hypothesis is really false and you prove it false, or else you are unable to show it false. Scenarios 1 and 3 are the outcomes you want. Scenario 2 is called a Type 1 error and scenario 4 is a Type II error. The statistical test you design (such as t-test for difference of means) has a factor in the calculation related to the sample size. As the experimental designer, you decide what level risk you can accept (balanced against the practical concern about excessive cost), and you choose a sample size appropriate for the risk you agree to accept.
 

Ronen E

Problem Solver
Moderator
When you do verification, there are four possibilities because of variation and measurement uncertainty. A hypothesis is true and your experiment supports the claim, or else you incorrectly conclude it is false. Or a hypothesis is really false and you prove it false, or else you are unable to show it false. Scenarios 1 and 3 are the outcomes you want. Scenario 2 is called a Type 1 error and scenario 4 is a Type II error. The statistical test you design (such as t-test for difference of means) has a factor in the calculation related to the sample size. As the experimental designer, you decide what level risk you can accept (balanced against the practical concern about excessive cost), and you choose a sample size appropriate for the risk you agree to accept.
This is roughly the traditional textbook spill about the topic; unfortunately when you go into it in greater detail and acknowledge all the assumptions and approximations made along the way, you might realise that there's a lot more to it. After >20 years in developing medical devices I'm still in search of the holy grail (caveat: I didn't directly deal with statistics and experiments the whole time... the other side of it is regulatory and standards, where things are even fuzzier!).
 

Ronen E

Problem Solver
Moderator
We are testing an active guidewire (DV tests: Biocompatibility, software, computer safety..)
Biocompatibilty testing sample sizes will be best advised by your accredited testing provider (3rd party lab / toxicologist), which in turn will rely on widely accepted test standards.
Software testing, AFAIK, doesn't require a "sample" because at least in theory all instances of the software are identical.
Not sure what you mean by "computer safety".
I do imagine that additional (bench) test types are involved in verifying a guidewire.
 

ringring

Registered
Thank you for your answers.
I do imagine that additional (bench) test types are involved in verifying a guidewire.
Yes from ISO 11070 mostly (see in link).
Sorry I'm just a rookie and I was told "we haven't a lot of samples for verification so we need to use a small amount and the auditer told us to have a "statiscal rational" to justify this small samples"
Not the easiest task :rolleyes:
 

Attachments

  • GUIDEWIRE TESTS.pdf
    100.1 KB · Views: 384

Ronen E

Problem Solver
Moderator
No, it's not an easy task, and looking at the list it's also not a small one... Actually it's a full-on mini-project for a statistician or another knowledgeable professional.
Just note that "a sample" in this case is not necessarily identical with a sample of physical device units. Unless it's a disposable device, you are generally allowed to run each unit multiple times to enlarge your sample size.
 

ringring

Registered
Thank you Ronen
Unless it's a disposable device, you are generally allowed to run each unit multiple times to enlarge your sample size.
Do you know how do we decide the number of times you can run 1 unit ?
 
Top Bottom