10/11 March 2020
The topic 100 percent visual control of parenterals constantly raises new questions, especially as concerns the use of automatic inspection machines or if statistical issues emerge concerning AQL inspections. Following you will find some of these questions and answers by Dr. Tobias Posset (Head of Production Support at Roche Diagnostics).
Question: What are the differences between qualification, particle (Knapp test) and function test set?
Answer: The function test set serves for a sort of system suitability test, i.e. this test is used to test before and after each batch whether the cameras are functioning correctly. Usually, no challenging samples are used for this, but rather unities with particles having a detection rate of 100% such as particles with a size of 1000 µm, vials with missing stoppers....
Qualification test set: the qualification test set consists of product specific containers containing the product and having all known "static" defects (scratches, wrong flip-off, missing stopper,...). Usually, about 10-20% of the containers of the set have a defect. New failures or defects are added to the qualification test set.
Particle test set (Knapp-Test): Sets that contain only particles. These are particles having the size from 50µm to 1000µm and consisting of different materials (plastic, the material stoppers are made of, glass, metal). Hence, they are "non-static", i.e. the defect is in the container or in the drug solution. Particle test sets are part of the qualification test set at the same time.
Question: Should all these test sets be prepared artificially? Should this be done, for instance, by an external laboratory with defined failures?
Answer: We prepare the static defects in-house and a part of the non-static defects are produced externally. But it is also possible to have everything produced externally. But the static defects should be the same that are generated in the worst case by your production machines. Therefore, I advise to make these in-house and get the release from QA. In this way you would have representative bad samples from your process. In the best case you take bad samples from your process but it is not possible to do this for all samples.
Question: Requalification: If the regular requalification of the automatic inspection machine is carried out by means of the test of 5,000 does this mean that the 5,000 vials from one production batch must all be controlled manually and by the automatic inspection machine? I.e. is it necessary to repeat the man-machine-comparison that was carried out in PQ?
Answer: Almost correct. PQ consists of the particle (Knapp test) run, the run concerning the "static" defects and also of a run of 5,000 of GOOD vials. This run of 5,000 is carried out once a year for each product. The Knapp test and the "static" test are not repeated except when there have been changes at the particle detection stations. Without changes (something that occurs only very seldom) only the test of 5,000 is carried out, i.e. 5,000 vials from the automatic inspection machine are inspected by a person and by the machine.
Question: Is it also required to carry out the AQL test with 1,250 randomised vials for the requalification?
Answer: We carry out an evaluation of the AQL values instead of the test of 5,000. In the course of the AQL test the man-machine-comparison is already carried out so that the test of 5,000 can be skipped. Once a year a report is written for each product and the automatic inspection machine is considered to be requalified from AQL test to AQL test (= continuous requalification). This means that we are carrying out a man-machine-comparison for far more than 5,000 vials for our production of market products in one year. This means that we are getting a better statistic (depending on the product in some cases up to 30,000 vials). The automatic inspection machine has already decided "GOOD" and now this is verified by a person in the course of the AQL.
Question: We produce a lot of products with different formats. Until now we use a rejection rate of <2% as acceptance criterion and sets of samples from production to control the functionality of the machine.
Answer: An AQL test should be carried out for each batch. In the meantime this is expected by the authorities and inspectors. This is also described in the ECA Best Practice Paper on Visual Inspection. A control of the rejection rate is also expected but it rather serves for recognizing if a batch differs from the normal unobtrusive production.
Question: The problem is that sets of samples have to be remade regularly since they have expired. How do you prepare the function test set?
Answer: Usually, our sets keep several years. Single samples have to be replaced again and again. These single samples are then released by QA and we introduce them into the set. The sets should be controlled at least once a year and be released again afterwards.
Question: Can we have some representative test sets prepared by an external laboratory and use them in the long term?
Answer: On principal, test sets can also be prepared by external laboratories. But there should be an internal procedure to test the sets for their suitability and to release them. At least once a year it should be controlled if the set is still representative and then it should be released again. In the ECA Best Practice Paper we even recommend to carry out this practice twice a year.
Question: We have about 50 different aqueous solutions for a few hundred products (one solution for several formats). So far we have prepared one set of samples (function set, qualification set including Knapp set) from production rejects. This is very time-consuming. Do you think it might be useful to measure parameters such as viscosity, surface tension of the 50 solutions in order to group the products (bracketing)?
Answer: If I were in your place I would also carry out a bracketing. The rationale using viscosity is completely ok.
Question: Can an automatic inspection machine also inspect oily solutions?
Answer: There is no reason why an inspection of oily substances shouldn't be carried out. But in this case an internal qualification should be carried out since the viscosity changes the physical reaction of the particles.
Question: Should the set of samples for the qualification (training) of the staff be taken directly form the production rejects or should it be produced artificially based on typical defects?
Answer: Usually, the training set should be taken directly from the production rejects. Unfortunately, some failures occur only very rarely so that you have to produce some of them yourself. Finally, you have a set with failures that are typical for the production that has to be released. This guarantees that the failures are typical and that you can use them for training.