Visual Inspection: Seeing Room for Improvement?

As a mandatory practice for injectable drug products by both FDA and the European Medicines Agency to ensure patient safety, visual inspection identifies and rejects parenteral containers and units with defects that have the potential to impact quality. According to engineer Davide Luisari, design manager with Bonfiglioli Engineering S.r.l., automatic visual inspection (AVI) is especially effective for the quality control of vials, glass bottles, pre-filled syringes, and blow-fill-seal containers. Specifically, visual inspection detects cosmetic defects—such as scratches, inclusions, and dents—in a container as well as evaluates the color, fill level, and identifies the presence of foreign particles within the drug product. The latter is the most critical benefit to ensure the safety of the end user/patient.

“[AVI machines use] a system of high-resolution cameras, illuminators, and computers that receive and process data to record a series of images of the product, photographed in 360° and in motion. Specialized software identifies any defects in the container and any particles or contaminants in the product,” says Luisari. “Products are classified as conforming or non-conforming; non-conforming packages are automatically rejected. The processing unit generally uses algorithms to determine compliance.”

The best visual inspection is non-destructive and inspects every manufactured product, Luisari adds. In addition, visual inspection can be either manual or automatic:

• Manual visual inspection: Performed by one or more trained operators following strict procedures.
• AVI: A machine performs the same tests done by manual operators. However, these machines often have equal or better performance in terms of detecting defects, false rejects, and speed.

Visual inspection installation

Compared to manual systems, AVIs are usually installed before labeling and packaging, according to Luisari.

"AVI systems are installed end-of-line, typically before labeling and packaging in order to have the products as close as possible to end-user conditions but with the greatest amount of uncovered surface to facilitate a complete inspection of the product,” says Luisari. “Infeed and outfeed for AVI machines can be completely in-line, either belt-to-belt or by means of infeed and/or outfeed trays. Products can be loaded and unloaded manually or automatically.”

For the visual inspection process to best be incorporated into manufacturing, John Shabushnig, principal consultant, Insight Pharma Consulting, LLC, shares that identifying particle types early is key. This identification aids in reducing the inherent particle load in the formulation.

Shabushnig says, “This carries into manufacturing with similar goals to identify and reduce sources of particle contamination. Visual inspection further supports stability studies during development and routine manufacture to assure the product and process function as intended.”

Visual inspection guidelines

According to Shabushnig, the guidelines for visual inspection in the United States, Europe, and Japan are as follows:

1. United States: United States Pharmacopeia (USP) General Chapter <790> Visible Particulates in Injections and the supporting information chapter USP <1790> Visual Inspection of Injections.
2. Europe: European Pharmacopeia (Ph. Eur.) 2.9.20 Particulate Contamination: Visible Particles and the information chapter 5.17.2 Recommendations on Testing of Particulate Contamination: Visible Particles.
3. Japan: Japanese Pharmacopoeia6.06 Foreign Insoluble Matter Test for Injections, which aligns with the manual inspection conditions found in the USP and EP.

“I would also add the Parenteral Drug Association (PDA) Technical Report 79: Particulate Matter Control in Difficult to Inspect Parenterals and PDA Survey: 2014 Visual Inspection as good references on supplemental inspection methods often useful for drug formulations encountered in the biopharmaceutical field as well as benchmarking of current inspection practices and results,” Shabushnig adds.

“USP <1790> sets out guidelines for the inspection of injectable materials for visible particles. The acceptance criteria described are based on a Knapp Test,” Luisari says. “This is a detailed test carried out to ensure that automated inspection is similar to manual inspection. The acceptance criteria for a successful Knapp Test requires the AVI to identify defective products at a rate equal to, or higher than, what human inspectors can identify in a fixed period of time in a controlled (light and background) environment.”

But how manufacturers define a “normal production line” significantly impacts the visual inspection process, according to Luisari—adding that there is a direct connection between the ability to detect a specific defect and the product features. As such, Luisari says it’s essential for manufacturers to first perform an adequate risk analysis when defining their product.

Best practices and artificial intelligence

When it comes to best practices, Shabushnig recommends manufacturers drive process improvement by using particle characterization and identification information. It would be ideal if particles never entered the filled containers (rather than requiring inspection after filling), as proactive prevention is more reliable and can reduce cost as compared to reactive prevention. Shabushnig adds that this applies to container and closure defects as well.

Moreover, advances in artificial intelligence (AI) and deep learning—according to Shabushnig—may allow AVI to better detect defects and have a reduction in false rejects from automated systems.

“The application of AI is increasing day by day, including visual inspection for pharmaceutical products. It is particularly beneficial to apply neural networks to visual inspection because the testing machine ‘learns’ from its mistakes, becoming more precise and reliable over time,” says Luisari. “Bonfiglioli Engineering utilizes neural networks and AI for the most challenging AVI applications, such as detecting defects in lyo cakes [lyophilization cakes] or finding cosmetic defects on a container. Over time, AI analysis refines the definition of ‘acceptable’ and ‘defective’ products, to generate an improved ‘defects directory’ and deliver better more refined results.”

The challenge, according to Luisari, is to continuously research advances in software and hardware, which are technologies that are constantly changing and improving.

“[I]t’s essential to be able to create the best conditions for product handling and lighting/optical setup, to fully emphasize the product features that facilitate the neural networks’ operation in defining a compliant or non-compliant product: there is no visual system in the world that can identify an unseen defect,” says Luisari.

Pain points and misconceptions

When asked about key problem areas, Luisari shares that many manufacturers have incorporated multiple testing methods—each of which provides different product integrity information—into their production processes. With multiple testing methods in place for a single drug product, the reliability of the results increases, which then leads to better quality assurance. However, the more testing methods that are in place means the more manufacturers must juggle, specifically additional overhead for installation, operation, and maintenance.

“While combining different test methods increases the precision and reliability of the results, it also presents challenges with regard to space, cost, and efficiency,” says Luisari. “These challenges can be met by combined machines that perform different types of tests in a single automated process. A combined testing machine for pharmaceutical products offers numerous advantages over traditional solutions, including space and cost savings and improved handling, as well as overall operational efficiencies.”

But when it comes to misconceptions for visual inspections, Shabushnig believes that one of the most common misconceptions is the belief that visual inspection assures one-hundred percent removal of all visible defects from a given batch.

“Visual inspection, human or machine-based, is probabilistic and many defects will not achieve 100% probability of detection. Defect size, shape, color, and location have a significant influence on detection probability,” says Shabushnig. “I would also add that many consider inspection to be primarily concerned with visible particles. A wholistic approach, including the primary container and closure, is also needed to address the full range of defects of greatest concern.”

What manufacturers should know

No process or practice in manufacturing is error-proof nor does it guarantee a particular result. However, meticulous upfront design and planning can significantly impact the visual inspection of injectable drug products.

“While the inspection process is not perfect, a well-designed and operated inspection program provides valuable information on the performance of the manufacturing process and contributes to the assurance of product quality,” says Shabushnig. “It is part of a larger quality system and relies on good process design (both inspection and manufacturing). It also provides valuable information on process and product stability and can be used to drive continuous process improvement to reduce or prevent future defects and product loss.”

Designing a detailed visual inspection program upfront that includes proactive prevention, keeping up to date with advances in software and hardware, incorporating AI technology, and shifting from a manual process to an automated one that combines machines together has the potential to significantly improve the visual inspection of parenterals—thereby, ensuring both safety and product efficacy.

About the Author

Meg Rivers is a senior editor for Pharmaceutical Technology, Pharmaceutical Technology Europe, and BioPharm International.

Article Details

BioPharm International
Vol. 45, No. 11
November 2021
Pages: 38-40

Citation

When referring to this article, please cite it as M. Rivers, “Visual Inspection: Seeing Room for Improvement?”, BioPharm International, 45 (11) 2021.