As a rule, in the PQ and even more so in the validation, the batch size typical of the future routine production are expected. But does it make sense to compare the detection rates of humans and machines based on complete batches with 20,000 vials, for example?
In the PQ/ validation of the system, the human-machine comparison is made by repeatedly testing the Qualification Test Sets. This is a product-specific test set that contains the approximately 20% defects that can occur in the real process. The ECA Best Practice Guide recommends a minimum of 10 repetitions of the test. For additional security, parts of a validation batch can be inspected as well, for example, 5000 units. It is also possible to combine samples from the three validation batches for a total of 5000 units. However, this mixes up the batches and afterwards, the inspectetd units may not be returned to their batches if they are intended for the market.
As part of revalidation, the 5000-unit test can be repeated regularly.