[Jacky Cheung]

Dear Dr Taylor,

I have a few beginner questions regarding STAT-04 and STAT-09 from your book Statistical Procedures for Medical Devices. Suppose we have implemented 100% verification in production testing on a critical specification e.g. temperature measurement accuracy of a thermometer. Any production unit that does not meet the accuracy requirement will be rejected.

1) Is it acceptable to use production screening as a method to ensure requirements are met in production, and accept that some products that have all worst case tolerance stack up e.g. 2% will be rejected?

2) If production screening is allowed, I guess we cannot use worst-case testing in design verification as we know if we stack up all worst case tolerances the product will not meet the requirement. Does it mean we will need to use a sampling plan?

3) If we are to use a sampling plan, will a representative sample be after production screening, or before production screening? And will the design verification be to verify that <2% will be rejected, does it not fall into process validation?

When 100% production testing is designed to ensure production units meet the requirements, I am a bit confuse on what statistical technique and sampling is required
for design verification.

I hope my questions make sense and thanks in advance for your help.

[Author]

Posts