Statistics/sampling plan for medical device

[CCharlie]

Hi Gents,

I am looking for some help with regard to statistics/sampling plans etc. Traditionally we have use ANSI Z1.4 as our statistical basis for sampling (AQL sampling at the end of batch). I am looking for alternatives (with references) since the sampling of medical devices such as pen injectors is an expensive operation and these devices should already have under-gone design verification studies at the manufacturer. Any ideas are appreciated.

 

[Tidge]

Are you doing a study of attributes or variables? What is your tolerance for "false positives"?

 

[CCharlie]

Are you doing a study of attributes or variables? What is your tolerance for "false positives"?

Not sure what you mean by false positives...but we listed all the attributes and categorize them into minor/major/critical in our SOP and following AQL. But the problem is these devices are very expensive and we could lose a lot of products if strictly following AQL at the current batch size.

 

[Tidge]

Not sure what you mean by false positives..

Type I errors.

 

[Jim Wynne]

Not sure what you mean by false positives...but we listed all the attributes and categorize them into minor/major/critical in our SOP and following AQL. But the problem is these devices are very expensive and we could lose a lot of products if strictly following AQL at the current batch size.

It's not clear to me what the question is. Are you looking for a sampling plan that will make it easier for you to accept defective product? Please explain the types of defects you're looking for and why you think the current plan is inappropriate.

 

[CCharlie]

It's not clear to me what the question is. Are you looking for a sampling plan that will make it easier for you to accept defective product? Please explain the types of defects you're looking for and why you think the current plan is inappropriate.

Hi Jim, we currently use AQL level II which results in 500 products being sampled which are undergoing destructive testing. The goal is not to accept defective product but to find a reasonable way to reduce the total lost due to destructive testing.

 

[Jim Wynne]

Hi Jim, we currently use AQL level II which results in 500 products being sampled which are undergoing destructive testing. The goal is not to accept defective product but to find a reasonable way to reduce the total lost due to destructive testing.

This is the first mention that what you're doing is destructive. I asked what sort of defects you're looking for. Now I'll ask how many you typically find. Help us to help you.

 

[Bev D]

What is your AQL and your RQL?
Although since I’m not a ‘gent’ you may not want my answer

 

[CCharlie]

What is your AQL and your RQL?
Although since I’m not a ‘gent’ you may not want my answer

For critical defects is always 0/1 to reject the entire batch

 

[CCharlie]

This is the first mention that what you're doing is destructive. I asked what sort of defects you're looking for. Now I'll ask how many you typically find. Help us to help you.

Injectable pen product. Looking for volumes from each click, number of clicks/total volume, analytical and micro testing. The product is still in PV stage, not clear the amount of defect typically find. If I have that number, I probably can establish a new AQL through a risk assessment. But the issue is we do not want to lose so many products at this stage so potentially looking for an alternative approach.