In selecting a sampling plan, there are a number of considerations. I assume that the sampling plan is not prescribed by regulation, so you are free to pick the "best" plan.
In sampling plan in general a "representative sample" is a simple random sample. There is a product lot with a specified number of items. The sample is items from the lot drawn with equal probability, i.e., each item in the lot has the chance of being in the sample. The number of products to pick, the sample size, is determined by the sampling plan.
Sampling plans for lot acceptance fall into two broad categories – attribute and variable.
In an attribute sampling plan the items in the sample are classified by the attribute pass or fail. The sampling plan also has an accept number to determine whether to accept the lot. For example, you might have a sample size 20 and an accept number of 1. Take 20 items from the lot and classify the as pass/fail. If 0 or 1 fail accept the lot. If more than fail reject the lot. Many people don't like to accept a lot with 1 fail in the sample.
In a variables sampling plan, the items in the sample are measured and the sampling plan uses these results to estimate the statistical distribution of the process. The distribution (mean and standard deviation) are compared with the upper and lower specification limits to estimate the percentage falling outside the specification. Reject the lot if the percentage nonconforming is too large.
If, in an attribute sampling plan, you measure the sample item to classify it, then a variables sampling plan might be a better approach. The classification process discards information resulting in a larger sample size.
In cases where the lot size is small, there are alternate attribute sampling plans. One is called chain sampling in which the results of several lots are chained together to get better statistical performance.
The question of what to test depends on the production process. You may want to test the results of intermediate steps, such as media production. However, I think considering these as processes that require validation is a better approach. You control the process to provide a high degree of assurance that it produces only conforming product.
I generally recommend doing sample testing on the finished product only. This gives assurance that the product performs correctly for the patient.
------------------------------
Dan O'Leary CQA, CQE
Swanzey NH
United States
------------------------------
Original Message:
Sent: 22-May-2023 15:45
From: Colleen Flanagan
Subject: Mycoplasma sample size requirements
I don't know the answer to this question, but I would also love to get other member feedback on this query as I am struggling with the same question. I am working on a R&D/product development for combination product (HCTP's/device/drug) that has a low # of doses created per batch, and am struggling to find guidance on a sampling plan or sample number for finished product mycoplasma testing.
I also wondering what part of my finished product to test and would appreciate any insight or guidance from RAPS members. My product 'dose' consists of 2 hydrogel capsules that surround a human tissue allograft. These 2 capsules are delivered within 5mls of media and contained within a 50ml conical tube at ambient conditions .
Would it be preferred to test the media that the capsules are suspended within for mycoplasma or to take the capsules out (aseptically) and suspend them in media for mycoplasma assay?
Thanks for your advice and feedback.
------------------------------
Colleen Flanagan
Sr. Research Specialist
Ann Arbor MI
United States
Original Message:
Sent: 16-Mar-2022 04:52
From: Anonymous Member
Subject: Mycoplasma sample size requirements
This message was posted by a user wishing to remain anonymous
In biologics manufacturing, what constitutes a "representative sample" when you have limited sample available for mycoplasma testing (e.g., volume, cell density)?<o:p></o:p>