A Knapp test is carried out to ensure that the manual inspection is similar to the automated process of inspection. It is a very lengthy process, but this document is generally asked for if you qualify an automated machine for inspection by an auditor.
The Knapp test is as follows:
Firstly, the USP states “all parenterals will be 100% visually inspected.”
Secondly, below are my opinions only (experience) and if I fail to identify anything that a regulator may be looking for, or if someone with more experience can add to anything or correct anything, I will truly value the feedback.
When developing a method in which to qualify operators it is important to develop a process for qualifying that you can easily defend.
The following discussion is something we can call the ‘pseudo Knapp method’
The PDA presents papers on Julius Knapp’s methodology to use when validating a semi-automated inspection method (eg. siedenaders) or fully automated inspection machines (eg. EISAI) so these can be used for reference.
The best thing to do is collect particulate rejects from various batches (keeping the characteristics of the product the same – eg. viscosity, clear or yellowish, fill volume, etc…)
After collecting X numbers of particulates rejects from various batches, these vials can be blindly placed into Y number of uninspected vials from various batches.
Now with X + Y, have inspectors (let’s say three different inspectors) manually inspect these vials 10 times each (3 inspectors times 10, times each vial, is 30 inspections each vial) . Using Knapp’s probabilistic methodology, anything with a probability of 0.7 is classified as a reject.
Thus from the X + Y, any vials that were rejected 21 out of 30 times would be considered a reject. Using these vials classified as rejects and using vials that are classified as good vials (anything with a probability of 0.3 and less) a standard qualification set can be created.
Note: it is best to get reject vials that have a reject probability of 1.0 (30/30) to make things easier in the end.
Let’s say for example we create a set of 100 vials with 30 vials being rejects, and 70 being good vials classified per Knapp’s probabilistic model.
Other terms that need to be defined include:
RZE – Reject Zone efficiency = sum of reject probability of the 30 reject vials / 30 reject vials
RAG – Reject Accept/Grey = sum of probability of rejecting the 70 Accept and Grey vials / 70 vials
NOTE: I did not discuss grey vials, don’t see a need but this can be incorporated
Probabilities of Rejection
Now we should determine the RZE of the set. Adding up the probabilities of rejection from the 30 reject vials, (eg. let’s say its 28) Thus the RZE = 28/30 = 0.93. J. Knapp states that an RZE of 0.80 and 0.95 is what is normally seen in industry, so determine an RZE that suits your business, also your business would also want a low RAG (RAG should be less than 0.10, but is also based on what you sets RAG is) which is considered how your operators will react to good product.
Now you can test your operators using this set and calculate the RZE and RAG values to identify operators for inspection.
After writing this, I know that there are many things that have been left out, for example, you need to verify the visual accuracy of your inspectors (eye test done during hiring). In addition, your product set may degrade over time causing it to form particulates so you may need to store the set at the product requirements to impede degradation, or replace the set on an on-going basis.
I would be most grateful for any other comments on this subject.