Visual Inspection of Parenteral Products Part 4: From Regula & Guidelines to Implementation and Production

We have discussed in previous Part3 the main regulations and requirements about visual inspection covered by USP, EP and JP pharmacopoeias.

It is crucial to point out that in most cases such regulations and requirements don't actually explain in detail which systems, procedures and technologies should be used in order to achieve the required results such as “Essentially free of visible particles" and “The packaging system should be closed or sealed in such a manner as to prevent contamination or loss of contents”. 

Furthermore, not only the choice and implementation of actual systems and technologies to be used is left to the manufacturer, but even more importantly and critically it is also manufacturer's duty to prepare and provide the necessary documents to prove that the whole manufacturing process is under control and able to fulfill all quality and safety requirements. 

This situation has always put a lot of pressure on drug manufacturers in the struggle to find the way to close this gap and be able to produce and sell their products to the market. 

Particle Inspection 

In order to fulfill the fundamental statement “There should be no visible particles” inside containers it is necessary to define first of all:

1) What we consider “particles”

2) How we define “visible” (and consequently non visible)

The term “particle” (or equivalent terms “particulates” or “particulate matter”) was initially defined as "mobile undissolved particles, other than gas bubbles, unintentionally present in the solutions” and we already discussed how they can be classified in Part 2. However, in this discussion we are more interested in how such "particles" can and should be detected during inspection and this leads to a first, very important factor:

<< The detection of particles is fundamentally based on their movement >> 

This is true because small particles present into containers are much easier to spot if they are moving.

Non-moving particles are much more difficult to detect because there is no easy way to distinguish them from small defects or spots on the container's walls.

When we consider how to define when a particle is visible or not things get even more complex. A big number of factors enter into play and affect how much a particle is visible such as: 

• size, shape, color, reflectivity and other optical characteristics of the particle(s)
• transparency of the container, transparency and movement of liquid and particle(s)
• quality of sight and analysis of inspector (or inspection system)
• speed and lighting conditions during the visual inspection proces

All those factors add up through a complicate cumulative function, too complex to be managed in day by day production and inspection process management. It is much more convenient to combine all of them into a single number: the Probability of Detection (PoD).

The PoD can range between 0% (the particle(s) is never detected) to 100% (always detected) and it's the most practical and useful measure of how much a particle is "visible".

In the past it was commonly considered that particles bigger than 50 or 70 µm are easily detectable and this was considered the standard minimum size to be detected during the visual inspection process.

However, more recent and precise studies have demonstrated that this is not true and that the actual size of particles that are consistently detected during visual inspection is significantly bigger. Today the size of particles reliably detected (POD≥70% Probability Of Detection) is considered 150 µm or larger.

The above chart, which combines the results of six different studies, shows another critical aspect of particle visual inspection: even when we try to measure the PoD of well-known particles we end up in a huge variability of results.

 If we consider for example 100 µm particles, the associated measured PoD ranges between 22% (by Ryan) to 70% (by Knapp) with another 4 studies falling in between. 

That's a 300% discrepancy and this is telling us that there are too many variables at play in this process to allow a single, precise result.

This kind of results has forced to change the whole approach to particle detection and to accept that:

<< Visual inspection process is intrinsically probabilistic >>

There is no way to know in advance if a particle in the range between 50 and 100 µm will be detected during the visual inspection process. We can only measure and evaluate its PoD and we need to base our entire quality control system on this statistical approach.

So according to all above information, how have the drug manufacturers managed to comply with the generic requirement “There should be no visible particles”? 

We can summarize the mostly used and accepted approach in two phases: 

• First, design and implement a Manual Visual Inspection (MVI) process able to provide the highest possible level of inspection and detection of particles.
• Second, use the statistical analysis of inspection data to check and validate every visual inspection system, including Automatic Visual Inspection (AVI) machines. 

It's important to recall once more how the development and update of official regulations and pharmacopeias has always been much slower than the production practices and the available technologies and since 70's the first automatic inspection machines have begun to be available on the market.

Even though at the beginning the automatic visual inspection machines were primitive and based on basic cameras, computers and lights, the potential and advantages of this technology was immediately clear and a lot of work was made in order to validate such systems and improve their performances.

The Knapp Test

Probably the most common procedure developed to assess and validate a new visual inspection system, specially an AVI, is commonly named "The Knapp Test". It derives from a paper published in 1980 by Knapp J., Kushner H.

It describes a method to collect a set of inspection data from two inspection systems, usually the MVI already used for Visual Inspection of products, and a new inspection system, usually an AVI Machine. 

These two sets of data are then compared on a statistical base to assess whether the new inspection system performances are equal or better than the original one.

We don't need to go into Knapp Test's details here, but it is important to highlight some of its key logical principles: 

• The analysis and comparison are based on a statistical base, many repeated inspections of many different containers. This is because it doesn't matter if one system is better on a specific container or defect, it is more important to find out which inspection system is overall, on average better.
• The original inspection system, usually the MVI, is considered to be good enough to fulfill requirements and warranty a good enough level of quality and inspection. Therefore, if the new system can prove to be equal or better, it can be considered acceptable too.
• Bad containers are identified only according to their PoD (Probability of Detection) by the "old" inspection system. It doesn't really matter if a container has a particle inside or how big it is. Only if its PoD is above 70% the container is considered bad and to be rejected.

Starting from 80's and 90's the Knapp Test has become more and more the standard procedure used worldwide to test and validate new visual inspection processes in pharmaceutical field, especially new Automatic Inspection machines.

AQL sampling

Gradually all international pharmacopeias and regulations have been evolving and adapting more and more to such new topics and technologies. More recently the concept of Acceptable Quality Level AQL has been introduced and added to the procedures recommended to further ensure the quality level of visual inspection. 

After 100% Visual Inspection a further check on several sampled containers is performed. The number of samples to be taken and the maximum number of acceptable bad among them is calculated according to ANSI/ASQ Z1.4-2003 or ISO 2859 tables.

This final test, once again purely statistical, is intended to confirm that the previous inspection process was correct and that the whole batch is "Essentially free of visible particles". If this last sampling and analysis complies with requirements, the whole batch can then be released.