We are a start-up company with about 30 employees. One question that comes up frequently is what our sample size should be for packaging integrity testing (either at time zero or post-aging). For example, if testing sealed pouches by bubble leak testing, what is the “appropriate” test quantity to give an acceptable level of confidence and reliability? 10? 30? 100? Does it matter if the device is an implant vs. non-implant? Also, should the test quantity (and confidence and reliability) ratchet up as the product is used in greater numbers, in initial feasibility trials vs. pivotal trials vs. market release?
Regrettably, it is not possible to simply recommend an “appropriate” test quantity for attribute data whether in initial, pivotal or release stages. The subject of sample size determination is not a trivial one, and in the medical device area is directly related to managing risk. Risk varies significantly with the nature of the medical device, particularly for those devices designed for implantation. Also of significant concern is the reliability and
reproducibility of your measurement system(s), and the degree to which those systems have been validated. These and other considerations are generally considered toward the end of determining an appropriate AQL level, which can then be used to determine sample sizes for testing.
The following website offers an introduction to risk assessment, and may be useful for familiarizing yourself with sample size determination:
Additionally, we recommend the following as references: “The Handbook
of Applied Acceptance Sampling: Plans, Procedures and Principles,” by Kenneth S. Stephens, published by ASQ Quality Press, 1st Edition – 2001. The subject is also discussed in an article in MDDI, October 2006, “Sample Size Selection Using A Margin of Error Approach”, by Nick Fotis and Laura Bix.