Process Qualification of custom infusion sets

Hello,

I am new to the forum. I am also new to the industry and would really appreciate if you can answer few or all my questions below.

I work for a medical device company who contracts a CMO for custom assembly of infusion sets. These have tubings, luers and check valves which are bonded by a glue and cured by UV light. Then are packaged into tyvek pouches and gamma sterilized. The process I am helping our CMO to qualify is the gluing and UV curing process.

My questions:
1. Is the 'No failure sample size' of 59 is sufficient with 95/95 confidence and reliability? The risk associated for this process is defined as 'Tolerable' per pFMEA.
2. Can we make proxy units (just leur and check valves with out in-line filters and tubing) for this qualification or complete finish goods are required? We will be testing for bond strength (tensile and leak tests).
3. Should the units be sterilized for OQ/PQ? These are sterile devices. The test for sterility is planned for Manufacturing Validation with 3 lots after OQ/PQ is completed. But the tensile and leak tests are performed on AQL basis on sterilized units as part of batch release. If I test the bonds with out sterilization, does that mean it no longer represents routine production.?
4. Do we have to use validated equipment (leak tester and tensile tester) in verifying the OQ/PQ samples? If so, to have the validated equipment, shouldn't we perform the test method validation on units from qualified process?

I

Going to pass on the statistics question. Hopefully one of the experts will jump in.

Sterilization can affect characteristics so I would certainly run the devices through the sterilization cycle to ensure the gluing and curing is not affected after sterilization. Given that, I'd probably want to use finished goods.

If you choose other routes, you'll need a risk-based justification.

I think you should certainly use validated equipment for testing the samples! And yes, it's likely you should have test method(s) validation(s).

The 90/95 sample size is for lot to lot acceptance sampling for a continual stream of lots. We have discussed this many times so you might find a search valuable (although we don’t have the best search engine here)

You can use the 90/95 IF you ensure that you have parts from the extremes of the allowable variation (spec settings for input factors and stress conditions). If you use only extreme values you can reduce your sample size to a handful of units since if you pass a teh extremes you will pass at the nominal values.

You are probably better off using continuous data rather than categorical (pass/fail) data.