During a recent internal audit the auditor questioned the manner of re-testing following failure during QC testing of our finished product - the manufacture of our device has ben validated.

Currently this particular QC test - an appearance test, involves testing a fixed number of product withdrawn randomly from the batch. In the case that all of the criteria in this appearance test are not met, our procedure currently dictates that a further number of devices (the same number as sampled originally) should be removed for retesting. It was this sampling plan for which the auditor raised an observation - suggesting that we should withdraw a %, e.g. 5% of the remaining batch and evaluate a sample of - if not all of this product.

Would anyone be able to provide advice on how many samples should be removed when repeated QC testing is necessary.

Thanks in advance
Jampot

If your sample size is based on AQL levels and c=0 methodology you should be placing the lot on 'hold' and then making a determination based on risk.

Depending on your procedures you might consider choosing a sample size based on criticality of your defect. You would then take a statistically valid sample size based on confidence intervals (see any Stat book for the sample size charts) and then justify your sample size based on criticality. For example a defect that kills someone (Risk level Catastrophic) you may assign a 99% confidence interval and a cosmetic defect (risk level Negligible) a 90% confidence interval. In my example I would be testing 459 parts in the first case and 22 parts in the 2nd case.

Because you are inspecting for the defect someone must've come up with a requirement that you can track back to a FMEA or risk analysis. This is where you would start documenting your justification.

You want to justify your sample sizes based on risk, not on a flat percentage unless you can somehow back up a blanket 5% sample size. Just remember that FDA is now hammering on 'risk based approaches'.


:2cents: