According to the requirements of pharmacopoeias (e.g. EU or US), batches of sterile injectables must be 100% visually inspected. This test, which is to be rated as an in-process control, must be validated; because of its probabilistic nature. However, it is not 100% free from errors. That means that the discovery of defects like particles, for example, is a question of probability. Visual inspection cannot rule out the presence of defects in a tested batch. The higher one raises the probability of detection, the higher the so-called False Reject Rate goes, meaning the rejection of actually error-free test pieces.
Now, how is one to handle the fact that the result of an officially required batch test might not be 100% correct?
For this, the US Pharmacopoeia intended an additional sampling inspection of containers which were found to be adequate in chapter <790>. After much delay (because of US-internal discussions), the additional chapter <1790> was published now, too. Intended are so-called AQL tests (acceptable quality limit). Depending on the number of containers in a batch, a dictated AQL table defines the size of the sample, which is then visually inspected and in which only a fixed number of defects may be found. The number thereby depends on the criticality of the defect. The number of permitted critical defects in an AQL test should be 0, numbers permitted for major and minor defects are also specified by the AQL limit. The USP set that limit at 0.65. This has the great advantage of making the pharmacopoeia requirement measurable that batches of parenterals must be practically or essentially free of particles.
The European Pharmacopoeia does not include a requirement about AQL testing, so far. Still, AQL testing is considered Good Industry Practice in Europe, as well, and is performed by many pharmaceutical companies. After all, GMP inspectors are also aware that a 100% (in-process) control is not absolutely free from errors.
On the other hand, the AQL testing may not be misused, either. Companies could release batches with known defects as long as they pass the AQL testing. Measures to investigate the defects and prevent them in the future wouldn't have to be taken, either. This does not correspond with the idea of quality in the pharmaceutical industry. This issue, however, is one of the reasons why the additional USP chapter <1790> has been published with so great a delay and only after the revision of some clarifying passages.