With the world moving towards automation and digitalisation, including the pharmaceutical industry, there is an increased emphasis on automation and digitisation of manufacturing equipment which yields many opportunities but is not without its challenges.
In this article we consider one such equipment-based process, Automatic Visual Inspection (AVI) and the digital solutions available, including an overview of the considerations necessary to employ such technology.
Visual inspection (VI) plays a key role in ensuring only safe products of the correct quality reach the pharmaceutical market. The guidance for the expectations for Visual Inspection methods, including relevant acceptance criteria for parenteral products, e.g., damaged primary packaging, are listed in the various pharmacopoeias: USP, EP and JP. All pharmacopoeias require that each container, where possible, is inspected i.e. 100% visual inspection. EudraLex Volume 4, Annex 1 also states that all filled containers of parenteral products should be inspected individually for extraneous contamination and other defects. Although this requirement seems straightforward to achieve, there are many challenges that may be encountered along the way.
The presence of particles in parenteral products, identified through visual inspection methods are a contributing factor to medicinal product shortages, which highlights the importance of ensuring you have a robust visual inspection programme in place. This is apparent when reviewing the FDA safety recall data, which attributes 33% of all recalls between 2017 and 2021 to the presence of visible particulates. Six recalls took place in 2020 and again a further six in 2021 due to visible particles alone.
On consideration of the type of Visual Inspection Programme to implement, it is important to understand one’s products, as well as the potential risks to the product and primary container that may impact the quality and safety of the product. Companies should consider the type of packaging being used and the preferred method of inspection, be that manual (by the human) or automatic (with equipment) means. For example, amber glass for light sensitive products does not lend itself well to visual inspection; if the product is opaque or a suspension, turbid products make it more difficult to identify true foreign particles and distinguish them from the characteristics of the product itself. If the product is sensitive to vibrations and excessive movement, the selected inspection method may cause it to shear. There are a multitude of characteristics that need to be considered within a Visual Inspection Programme commencing at laboratory level.
When moving towards an Automated Visual Inspection programme, it is important to ask the right questions when purchasing the equipment. Annex 1 is noticeably clear that the machine and automatic process must be as good as human inspection, if not better, so you need a solution that can detect new variances like a human would, without requiring months of machine learning.
The types of cameras and the resolution of the footage used with AVI equipment is of upmost importance. A variety of factors influence the probability of detecting a single particle, such as the size of the particle, its colour, opalescence, shape or even simply its position within the container.
It is a regulatory requirement that the inspection process should be validated to detect known defects – which may impact product quality or safety. The performance of the equipment should be challenged using representative defects prior to start up and at regular intervals throughout the batch. Therefore, building a robust defect library is a priority, as AVI can only ever be as good as the test pack used for training and the ability of the operators reviewing the inspection footage. Qualification of operators is still of upmost importance, even when switching to a digital or automated solution.
When considering Test Packs and Machine Learning, it is vital to define all the instances of what good may or may not look like, it is not only about particulates. Variables relating to the crimp consistency on the collar, the angle of the collar or even the glass tolerances on the vials must be considered. A human may not detect these minuscule differences; however, a well-trained computer will ensure they are identified and flagged. Could this increased sensitivity result in false rejects, and would you then need to put in place a program for humans to confirm all rejects?
Understanding what good acceptable vials look like and having a representative library of accepted vials, fit for human use is as necessary as the large defect library test pack needed to ensure the AVI system will not place defective product into the supply chain.
With the much-needed revision to Annex 11 in discussion with the drafting groups and the final release for publication set for the end of this year, further clarification regarding the inspectorates’ expectations in relation to digital transformation and the computerised systems used within the manufacturing process is on the way.
Although this article is focused on visual inspection of parental products, automatic visual inspection solutions can be used to support the release of many different dosage forms and packaging types.
USP <1> Injections and Implanted Drug Products (Parenterals) – Product Quality Tests
USP <790> Visible Particulates in Injections
USP <1790>Visual Inspection of Injections
EP 2.9.20 Particulate Contamination: Visible Particles
EP 5.17.12 Recommendations on Testing of Particulate Contamination : Visible Particles
JP 6.06 Foreign Insoluble Matter Test for Injections
JP 6.07 Insoluble Particulate Matter Test for Injections
Annex 1 Manufacture of Sterile Medicinal Products
Annex 11 Computerised Systems Concept Paper