There's something strange in the Annex!

Jeff Granger PhD of Pharmagraph talks about airborne particle counting for the pharmaceutical industry in 2004

Jeff Granger PhD of Pharmagraph talks about airborne particle counting for the pharmaceutical industry in 2004.

The amendment to the EU GMP Annex 1 published in September 2003 again focused the industry's attention on particle counting strategies for monitoring sterile production processes. The earlier introduction of the new cleanliness standard ISO14644-1 and ISO14644-2 heralded a period of review of the way particle counting is done and what, if anything, needs to be changed to meet both the new standards and guidelines and also to satisfy the MHRA/FDA inspector. This period of change has not yet finished, since the new FDA Guideline for Aseptic Processing is currently in draft format and will be with us shortly, and indeed, there is considerable pressure to thoroughly review Annex 1 itself.

From a particle counting standpoint, I have issues with Annex 1. On the one hand, it would be quite possible to design an excellent automated particle monitoring system that technically fails to meet the requirements of Annex 1. More worryingly, it would be quite possible to design a poor particle monitoring system that fully meets the requirements. As most people are looking for a well designed particle monitoring system that also meets Annex 1, this article considers some of the more commonly asked question and attempts to provide answers that fulfil both objectives.

Before attempting to describe best practice it is prudent to ask why we are doing particle counting in the first place. Clearly, it is important to demonstrate that we are meeting the various standards and guidelines, which may or may not in itself keep the Medicines Inspector happy. More importantly, however, the particle monitoring process should be an integral part of the overall quality assurance plan. Particle counting can contribute to this plan in three distinct areas:

    1. The Classification and Class Verification exercise demonstrates that the overall cleanroom is operating to the criteria originally specified when the facility was first commissioned.
    2. Continuous Monitoring is usually less rigorous spatially, but more informative temporally. It is rarely possible to design an automated particle monitoring system with sufficient sampling locations to classify a whole room, but the continuous, or semi-continuous monitoring of fewer carefully selected locations can provide extremely useful information about the likelihood of product contamination during manufacture.
    3. Continuous Monitoring is also a good indicator of how well the organisation is actually in control of their cleanroom.

Question 1: What is the difference between class verification and continuous monitoring?

The distinction between verification and continuous monitoring is often not well understood. When we refer to one of the cleanroom cleanliness standards, what is being defined invariably is class verification. There is usually a recommended time interval in which this process should be repeated, but this is not continuous monitoring. With class verification, the surface area of (usually) the floor is calculated, and a formula applied to determine how many sampling positions there should be. A particle counter is set up at each sampling position in turn and a number of replicate samples taken at each position. The minimum sample volume or sample period is usually set down in the standard. Typically, three or five, one minute (= 1 cu.ft or 28.3l) samples are taken at each location.

Continuous monitoring on the other hand is the process of taking repeated samples from one physical location, either indefinitely (24/7) or for the duration of a batch fill. ISO14644 makes little reference to continuous monitoring other than to concede its existence1, and to make some concession on the frequency of reverification for continuous monitoring. I am often asked to design an automated particle monitoring system that will meet the new ISO standard. Unfortunately, that would require designing a continuous monitoring system that meets the ISO requirements for verification, which would be impractical in most cases.

Question 2: How many sample locations do I need to classify my cleanroom?

The ISO14644-1 standard defines how many sample points are required to verify the cleanliness class in any defined area, and broadly, where they should be. ISO 14644 makes no recommendation about sampling positions for continuous monitoring.

Question 3: What is isokinetic sampling (and do I need to worry about it)?

It is well established that when taking samples from a unidirectional air flow, errors will creep in to the analysis if the velocity of the air-stream is different to the velocity of the air as it enters the sampling device. To understand why this should be, we first need to remember that real-world particles have a wide variety of sizes, shapes and densities. It is not too surprising then that different particles react differently in airstreams.

Let's take a simple example of two particles, both with the same density, and both perfect spheres. One measures 0.5µm diameter, the other is 5.0µm diameter. Ignoring sedimentation due to gravity for a moment, both travel at the same speed and in the same direction in unidirectional air flow. Now what happens if we force the airflow to change direction? The smaller particle has less inertia than the large one and so tends to follow the changed direction in air stream. The larger particle has more inertia and therefore a tendency to continue on its original path.

The velocity of air in the sample tube of a 1cfm particle counter is around 6.67 ms-1, which is nearly 15 times the velocity of the ambient air (0.45ms-1). If the air velocity in the sampling tube is the same as the air velocity of the unidirectional air being sampled, both small and large particles are sampled in the correct proportion.

If the air velocity in the sampling tube is greater that that of the air being sampled, as is the case with most particle counters, 'extra' air must be drawn into the sampling tube from the surrounding areas. Small particles tend to follow this air stream, but the larger particles will tend to pass by. The result is that the particle counter will under-count the larger particles.

For the air flow in the tube to be the same as the ambient velocity (0.45m/s) we need to increase the cross sectional area of the sample tube inlet until the velocities are equal. An isokinetic sampling probe is a funnel with an opening that is of the right diameter to ensure that the air velocity at the entrance point is equal to that of the air being sampled. Once the air sample is captured, its velocity increases as it passes down the sampling tube, but the concentration of both large and small particles accurately reflects the concentrations in the unidirectional air flow.

One other consideration is the cross sectional area that is actually being sampled. The fact is that in a perfect unidirectional air flow, and with a correctly sized isokinetic sampling probe, the only air being sampled is a cylindrical column of air the same diameter as the isokinetic sampling probe (36.5mm in the case of a 0.45 m/s air flow). This equates to a cross-sectional area of around 10cm2. It has been shown that under similar conditions a source of contamination such as a pinhole leak in a HEPA filter just 30cm to one side of a sampling location would not be detected.

One further thought; low flow rate particle counters operating at 0.1cfm would sample a cross sectional area of 1/10th of this, i.e. just 1cm2.

Question 4: If I want to do background sampling in turbulent air, do I need an isokinetic probe and which way should it be pointing?

In turbulent air flows it makes no difference whether the sampling tube is fitted with an isokinetic sampling probe or not, but as they are often fitted by default it is usually more convenient to sample with the probe attached, especially when supported on a tripod. Best practice is to point the probe towards the ceiling in turbulent air. This will err on the side of caution since the larger particles (5µm) will tend to sediment out of the air flows and will drop into the sampling probe. In uni-directional flow the probe should always be pointed into the air stream.

Question 5: Where should I locate my sampling points for continuous monitoring?

The idea is that we should attempt to sample the air that impinges on contamination susceptible areas of the process. On an aseptic filling machine, this would normally be the filling head itself, but should also be where open vials and stoppers are approaching the filling position, and where unstoppered, filled vials are exposed. Two or three sampling points on a filling machine is usually about right.

I would recommend locating the opening of the isokinetic probe 30cm upstream of the filling position and some 5-10cm away from it horizontally. Ideally, airflow studies should be done to determine how close it is possible to situate the sampling probe without disrupting the unidirectional airflow that is bathing the exposed product.

Question 6: Can I use a sequencing manifold to continuously monitor my filling process?

Annex 1 does not address this question specifically, though there are some interesting implications. Firstly, Annex 1 requires that Class A areas are "continuously monitored" and it recommends continuous monitoring is done on Class B areas. The first issue is to decide whether a sequencing manifold, which only samples the critical area at pre-defined intervals, meets the requirement for being "continuous". Clearly it is not continuous, but might be described as "semi-continuous". The second consideration is whether a sampling manifold can meet the requirement to sample a full 1m3 of air from each location. The point being that the sample taken from each location by such a device is typically 28.3 litres (1cu ft) at each visit.

Somewhat surprisingly, Paul Hargreaves of the MHRA stated at a Parenteral Society meeting last September, that provided that such a device sampled a total of 1m3 during a batch fill, this would probably meet the requirement. The suggestion was also made that sampling once every 10 minutes would probably be OK, but there might be a problem justifying sampling every 45 minutes. However, there is a much bigger problem in meeting Annex 1 using a manifold than sampling 1m3 of air. The transport efficiency of 5µm particles in low concentrations down long lengths of sample tubing is notoriously poor. How poor depends on the air velocity down the tube and the Reynolds Number, which indicates whether the flow is likely to be unidirectional or turbulent.

Additionally, it depends on the length of the sample tube, the number of bends in the tube and the severity of each bend. Typically, between 20% and 40% of the 5µm particles that enter one end of the sample tube do not emerge from the other end. If this was not bad enough, think about where the missing particles go: they stick to the walls of the tubing. Many of these particles will be held by weak forces and can be easily dislodged if the sample tube is knocked or shaken. Now, remember that under Annex 1, one 5µm particle/m3 is a pass, but two are a fail. So if you see two particles in a cubic metre of air how confident can we be that they originated in the cleanroom rather than the sample tubing?

Finally, validating the transport efficiency of a manifold sampling system with any degree of confidence is so fraught with difficulties as to be practically impossible.

To summarise, manifold-type particle sampling systems are in my opinion no longer considered to be suitable for critical point sampling. That is not to say that if you already have one installed you should immediately replace it with something else. If you have historical data to demonstrate that your system, as installed, has performed satisfactorily for a number of years then you can probably defend the case for keeping it with your medicines inspector. What I would say is that if you are planning a new facility then locally mounted point-of-use sensors like the Pharmagraph CPC-1 would be the product of choice.

Manifold type systems do still have a role in our overall particle monitoring strategy. They are, in fact, excellent trend monitors, and I would suggest that they provide a low cost solution to the background monitoring of the less critical areas within the facility. In a Class D area the loss of a few particles during the sampling process is a lesser evil than having no sampling at all.

Question 7: Can I use my portable particle counter to sample background points?

From an Annex 1 perspective there is no requirement to continuously monitor Class C and Class D areas, though there are class limits to be met for "in operation" conditions. One strategy for monitoring the background points is to utilise the portable particle counters that you may have on site for doing the class verification exercise.

An exciting new development by Pharmagraph that greatly increases the practicality of doing this is the revolutionary Smart-Socket technology. Using this approach, Smart-Sockets are installed adjacent to sampling locations. Portable counters can be simply plugged in to a socket which connects them directly to the Envigil software database. The software automatically sets the correct operating parameters on the particle counter, such as sample time, and then logs data from the unit for as long as it remains connected. The beauty of this procedure is that the user no longer has to rely on a paper printout of data from the portable counter and all data is securely logged onto the database where suitable alarm limits can be set for that particular location. This also gives compliance with 21 CFR Part 11 for electronic records

Question 8: If I take 36 minute (1m3) samples to be compliant with Annex 1, how can I respond to particle excursions rapidly?

The issue here is a real one. There are technical limitations to the sample flow that can be achieved in a particle counter, and the maximum flow available from manufacturers is 1cfm (28.3 l/min). If we literally take 1m3 samples then our system will only provide data every 36 minutes, since that is how long it will take to sample 1m3 of air.

The answer to this question lies in the facilities available in your data logging package. Our approach at Pharmagraph is to provide an "Annex 1" operation mode, where two complementary alarm limit test systems that effectively run in parallel. Particle counters take one minute samples. As 35.31 samples are required to obtain a cubic metre, we take 36, which is the nearest whole sample. Each minute, the data is tested against two parameters.

Firstly, does the value exceed the Annex 1 limits for 1m3? Clearly, if we record two particles at 5µm in one minute, we will have exceeded the one particle per m3 limit irrespective of how many of the remaining 36 samples record zero. Secondly, we test a rolling m3 of data i.e. the last 36 samples against the Annex 1 limits. In this way, we satisfy the Annex 1 requirement for analysing a full m3 of air, but at the same time the production manager will know within one minute if there is a serious problem.

Conclusion

The design of a particle counting strategy should use a range of different approaches to provide a comprehensive non-viable monitoring system. The change in recent years is from quantity to quality of monitoring. Ten years ago manifold systems were seen as an ideal solution and could provide more than 30 sampling locations at low cost. It is now realised that sampling issues and validation issues mean that these systems are probably not suitable for critical measurements, but do still have a role as trend monitors in non-critical parts of the operation.

More thought needs to be given to the exact positioning of sampling points in critical areas to ensure the data they provide is truly representative of the environment in the filling area.

Finally, with cost of ownership being a major consideration, we can look to increasing the value of the instruments we already have by using Smart-Socket technologies to integrate existing portable instruments into an automated particle monitoring system.

Companies