Revision of ISO 14644-1:1999 A progress report and explanation of some of the key issues and principles June 2012 Gordon J Farquharson Convenor ISO TC209 WG1, Critical Systems Ltd, 4 Greencroft, Guildford, Surrey, GU1 2SY Corresponding author: Gordon J Farquharson, Critical Systems Ltd, Guildford, Surrey, GU1 2SY, UK; email: gj.farquharson@gmail.com Introduction I have taken the unusual step of using some pages of this journal to share some very important issues relating to the revision of the fundamental cleanroom standard ISO 14644-1:1999 1 (ISO: International Standards Organisation). I decided to do this because the specifications of air cleanliness by particles form a fundamental part of our good manufacturing practices (GMPs) for sterile products in particular, and many other active pharmaceutical ingredients (APIs) and dosage forms. The pharmaceutical and life sciences industry uses cleanroom technology in many parts of its manufacturing and research operations, and, in these applications, the control of air cleanliness has a direct link to regulatory compliance and achievement of specified product quality. I have not explored the parallel process of revision of ISO 14644-2:2000 2 since, whilst an important standard, the issues at large are not as fundamental to the cleanroom community. The broader cleanroom community, beyond the pharmaceutical and life-sciences industry, is also directly influenced by the evolution and development ISO 14644-1:1999, whick is indeed a very important standard. The process of revision involves input by technical experts within an ISO committee environment. These experts should bring to the table current scientific and technical knowledge on the subject, and also table the opinions and requirements developed through national mirror organisations. This is not the first time that this journal has addressed the subject. This journal published a paper in January 2011 (Hartvig, et al.) 3, which was prepared to explain the basis of the sampling statistics in the ISO DIS 14644-1:2010 4 (DIS: Draft International Standard). Page 1 of 12
This report builds on the principles developed in that paper, and sets out to explain the areas of concern, technical argument, and in particular where the standards community reported difficulties with the relationship of this DIS to the pharmaceutical GMPs. Publication of ISO DIS 14644-1:2010 for review and national vote Close followers of the ISO standardisation process for cleanrooms under ISO TC209 (TC: Technical Committee) will recall that a draft international standard, ISO DIS 14644-1 was published at the end of 2010 for the statutory 6-month DIS review, comment and vote. As has been the case with the complete family of ISO TC209 cleanroom standards, the technical enquiry and vote was undertaken in parallel within the CEN standards community (CEN: Committee for European Normalisation). A convention, called the Vienna Convention, put in place to prevent unnecessary parallel work being carried out in CEN and ISO environments, has been applied to this work. The DIS enquiry and vote is a very important stage in the standardisation process because it is the first time that the draft documents are opened beyond ISO TC209 for broader public and industry scrutiny. Equivalent English, French and German language versions are published by national standards bodies, and formal comments are sought. The comments are then submitted back to the ISO central secretariat, together with a national vote. In the case of the recent cleanroom standards process, the national voting was in favour of approval of both ISO DIS 14644-1 and ISO DIS 14644-2 subject to comments being addressed. The ISO and CEN rules require that all comments are considered and adjudicated upon and reported for the record. Extent of comments I would now like to move on to explore the scale and nature of the comments made during the DIS enquiry process between December 2010 and May 2011. I am going to focus on ISO DIS 4644-1:2010 because this is the more important of the two standards, and one which received the majority of substantive comments. There were more than 100 substantive technical comments, and many more of an editorial nature. Every one of the comments has to be assessed and reported on. ISO TC209 WG1 (WG: Working Group) met on the 8th and 9th October 2011 to consider the comments and the main issues raised by industry (WG committee of experts nominated by national standards bodies to develop the content of particular standards). The Page 2 of 12
nature of the comments, and the likely changes to the DIS were so significant that a second DIS enquiry and vote will be required. This means a delay to the revision project. However, the topic of cleanroom classification is so significant that the concerns of industry have to be addressed. The majority of the comments came from the pharmaceutical and life sciences industry, but there were also some very important comments made by semiconductor and display screen manufacturing industry. One of the most interesting of these was a serious concern about how the standard should deal with super-large cleanrooms. To help experts understand scale, a Chinese presentation was made that illustrated single cleanroom facilities of up to 40,000m², with working zones of up to 10,000m². These are clearly outside the range of expectation envisaged during the evolution of cleanroom standards from the earliest US Federal Standard 209, up to the current ISO 14644-1:1999, and also the ISO DIS 14644-1:2010. I will address this issue later, when we consider the number of sample locations required for classifying a cleanroom or clean-zone. Sadly, I have to report that many of the national comments on the draft standard challenged or contradicted some of the important principles developed by their own experts during the 2-year evolution of the new draft. Whilst this is extremely disappointing I took the view, as convenor, that it was more important to work through the issues to ensure the broader acceptance of a revised standard in the future. When challenged, these experts explained that the DIS had been opened to broader scrutiny than had originally been possible in their national mirror committees. The main areas of concern identified in the vote and comments were as follows. The relevance of the particle concentration values in the classification table. The exclusion of particles 5 microns from the classification table for ISO Class 5. The increased number of sample points required for classification of a cleanroom or clean-zone compared to the current 1999 version. The locating of the sample points within the cleanroom or clean-zone (randomisation issues). Dealing with risk-based locations selected to specifically certify the cleanliness of critical locations within a cleanroom or clean-zone. The comments and issues in more detail The role and significance of the classification table Page 3 of 12
The subject I want to look at in more detail is Table 1 in both the current standard (Figure 1) and the proposals in the 2010 DIS (Figure 2). The first most important change between the existing standard and the DIS is that Table 1 becomes the foundation and basis of classification. The advantage of doing this compared with the use of formula (as in the 1999 version), is that the table can be used more effectively to constrain and guide readers and users into choosing appropriate particle sizes and cleanliness levels. Historians of cleanroom standardisation will remember that this was the basis of US Federal Standard 209E 5 and its US predecessors. Figure 1. Informative Table 1 from ISO 14644-1:1999 1 Page 4 of 12
Table 1 Classification table for particulate cleanliness classes ISO Classification number (N) Maximum allowable concentrations (particles per m 3 ) for particles equal to and greater than the considered sizes shown below a 0,1 m 0,2 m 0,3 m 0,5 m 1 m 5 m ISO Class 1 10 b d d d d e ISO Class 2 100 24 b 10 b d d e ISO Class 3 1 000 237 102 35 b d e ISO Class 4 10 000 2 370 1 020 352 83 b e ISO Class 5 100 23 700 10 200 3 520 832 e 000 ISO Class 6 1 000 237 000 102 35 200 8 320 293 000 000 ISO Class 7 c c c 352 000 83 200 2 930 ISO Class 8 c c c 3 520 832 000 29 300 000 ISO Class 9 c c c 35 200 000 8 320 000 293 000 Figure 2. Normative classification Table 1 from ISO DIS 14644-1:2010 4. Figure 3. Equivalent table found in US Federal Standard 209E 5. The following became apparent from review of the comments made on Figure 2 above. The issue equally applies to Figures 1, 2 and 3. Many readers were of the opinion that for a given ISO Class 1 9, the tabulated particle concentrations at different particle sizes actually represents typical particle distributions found in the cleanroom environment. This is not the case. A number of experts, including workers such as Whyte, Eaton, Ljungqvist and Page 5 of 12
Reinmuller have all carried out particle concentration measurement and assessment work that suggests very clearly endorses this position. During the ISO TC209 WG1 meeting in October 2011, the experts took account of these facts, and even considered proposing a new table that might better represent typical particle distribution at different sizes in cleanrooms. After much deliberation, it was decided that this was a step too far because it would mean a fundamental change to the current understanding of the ISO Classes 1 9. Following further discussion, it became clear that this only presents a problem when you try to use more than one particle size to classify a level of cleanliness. Within the existing 1999 standard, the last paragraph of clause 3.3 suggests it is legitimate to use more than one particle size, provided there is a reasonable separation in the sizes considered. One possible simple solution to this would be to simply remove this clause. At this juncture, it is important to remember that ISO 14644-1 is a standard specifically constructed for the purpose of carrying out a cleanroom classification. This is a formal process undertaken during the commissioning of a new cleanroom asset, and then periodically thereafter to prove continued compliance with the class. In addition to a formal classification, most practitioners, either based on good business practice, or as mandated by a regulatory authority, additionally monitor air cleanliness by particles is in real-time. Under these circumstances, selection of more than one particle size might be more appropriate because one should really be looking at deviation from a state of demonstrable control, rather than compliance with a relatively arbitrary value. The 5 micron particle problem The removal of the 5 micron particle concentration in ISO 5 compared to the 1999 version was a major concern for a number of reviewers (see Figure 2 above, and compare with Figure 1). In the 1999 version, the figure, of course, was the well-known 29 particles per cubic metre (see Figure 1). So why did experts feel the need to remove this 5 micron particle number concentration? The clear advice from the experts was that this low concentration of large particles is unreliable to count using our airborne particle counters, and, therefore, is considered an inappropriate size of particles to evaluate the ISO Class 5. Once again, there is a clear precedent for this in that this was exactly the situation in the discontinued US Federal Standard 209 E (see Figure 3). It is quite difficult to find a clear reason why the 29 particles limit was retained in 1999, but it is most likely to have occurred because ISO 14644-1:1999 included a significant European input. It is really important to realise that once the new standard is agreed as a classification based on Table 1, and the table includes no figure in the Page 6 of 12
5 micron column for ISO Class 5, then this means that you cannot use this particle size for classification in accordance with ISO 14644-1. This leaves us with an interesting problem about how to support the regulatory requirements set out in Annex 1 of the EU GMP. In the view of particle counting experts on WG1, the concentration of 5 micron particles in grades A and B of Annex 1 are just as inappropriate as 29 particle concentration in the ISO standard when considering classification. Monitoring may be a different matter, not considered in any version of ISO 14644-1. This is becoming an even more sensitive issue due to the recent replication of the EU GMP Annex 1 requirements in the PIC/S GMP, the WHO GMP, and most recently the new Chinese GMP regulations. It is because of this that many comments received on ISO DIS 14644-1:2010 requested that the 29 particle limit be replaced. Whilst this matter is not yet concluded, the expert opinion is still that Table 1 in the cleanroom standard should not include particle concentration sizes that are inappropriate. The experts are currently working on forms of words, to be included in the standard, which would allow pharmaceutical industry regulatory authorities to provide their own guidance outside the boundaries of the standard. The nature of the guidance being considered very seriously is to recommend complete separation of the activities of cleanroom or clean-zone classification from real-time monitoring. This is already clearly stated in Annex 1. Careful reading of the text in Annex 1 will indicate to the reader that the regulatory authorities are most interested in the operational state. The classification of cleanrooms or clean-zones in the at-rest state is more about defining a benchmark of effective performance of the environmental control system to give confidence that there is no external contamination contribution within the controlled environment. The real-time monitoring in operation, on the other hand, gives you much more important information about the sources of contamination immediately around the process in operation. The greatest concerns relate to the contamination sources from human beings, and failed technical systems. Under these circumstances, it would be much easier to consider 5 micron particle assessments for real-time monitoring alone. One of the greatest advantages of this approach would be that it would allow a better fit with the scientific approach to the revision of ISO 14644-1, and would remove some of the concerns being expressed about increased testing time due to the combination of a tight 5 micron particle limit and, in some cases, an increased number of sample locations proposed. Number of sample locations Page 7 of 12
The next important subject we have to consider is the basis of the new table of sample locations. The experts believe that a look-up table is a more appropriate approach because it allows the cleanroom or clean-zone to be divided into fundamental unit areas that have some tangible logic. Therefore, for very small clean-zones, fundamental unit areas of one square metre were considered appropriate, and for larger cleanrooms and clean-zones fundamental unit areas of 4m² were considered to be suitable. The DIS feedback clearly indicated that there was confusion between fundamental unit areas and sub-divided regions. The experts are working on improvement to the text to resolve this problem. It is a basic assumption within the DIS standard, and, indeed, a similar principle exists in the current standard, that within any single fundamental unit area the cleanliness level is assumed to be even throughout the area, and that a sample taken anywhere in a fundamental unit area will be representative of the whole area. The DIS feedback clearly demonstrated that the majority of the cleanroom community accepted that an increase in the number of sample locations to give at least 95% confidence that at least 90% of all locations do not exceed the class limits was an adequate level of assurance. It also allows the 95% upper confidence limit evaluation required for between 2 and 9 locations in ISO 14644-1:1999 to be discarded. This simplifies the classification process, one of the demands made by the user community at the outset of this revision. We now have to consider the issue of super-large cleanrooms. This was a problem brought to the table by Chinese experts, and the main concern was that the proposed new table of sample locations would lead to unnecessarily large numbers of sample locations for very large rooms. The Chinese experts reported that they felt comfortable with a combination of sample locations based on square root of the area of clean-zone, plus specific risk-based process-specific test locations. The problem is best illustrated by considering a single cleanroom area of 10,000m². In such a situation, if we use the square root rule in ISO 14644:1999, then we would require only 100 locations to carry out full classification. If you take the sample location requirement identified in ISO DIS 14644-1:2010, then you will require 520 locations because the draft stated that you have to divide the cleanroom into 500m² parts, and apply the sample location table to each part. Once again, this is still under consideration by the experts, but it is likely that the table of sample locations will be used for up to 900m², and then the square root rule would be applied for rooms or clean-zones greater than 900m². The logic of this can be seen on the graph below, where 900m² is the point where the confidence level and the square root rule lines cross (see Figure 4). Page 8 of 12
Square root rule ISO 14644-1:1999 Applies above 900m 2 Transition point at 900m 2 95/90 rule ISO DIS 14644-1:2010 Applies up to 900m 2 Figure 4. Sample point number comparison. From a statistical point of view, the confidence level would theoretically be much higher for cleanrooms greater than 900m² than for those less than 900m². The WG1 experts believe this is a reasonable pragmatic approach to give a reasonable number of locations. The other aspect of super-large cleanrooms that requires some additional consideration is those which have an extremely high ceiling. The experts believe the new DIS should contain an element that explains that you may need to define several test planes within the cleanroom, and carry out classification at multiple locations in each test plane. Random selection of sample locations This was the next bridge confronting the expert committee review. During the course of the development of the DIS, the experts had agonised over how to deal with the random selection of previously established fundamental unit areas. Initially, the concept developed involved dividing the cleanroom or clean-zone into equal areas based on the number of fundamental unit areas, and then randomly selecting from these the number of areas required to be sampled from the look-up table providing a number of sample locations. So, if we had 100m² of clean-zone, this would be divided into 25 fundamental unit areas, and from the look-up table, we would be required to sample from 16 of them. The 16 would be selected randomly. There is, of course, the possibility that the 16 could be grouped together, rather than be evenly distributed (Location Selection Option 1). To get over this, the DIS adopted a semi- Page 9 of 12
random approach that tried to ensure a more even distribution of sample locations (Location Selection Option 2). This received quite a lot of negative response during the DIS enquiry. The experts have further discussed these two options at length, and now the majority believe that the original plan should prevail. It is most likely that Location Selection Option 1 will be worked on in more detail in the next few months. Next we need to consider how to undertake the random selection. This becomes a great deal easier once we decide to choose Location Selection Option 1. A standard number generator without replacement can be easily found on the Internet for this kind of activity. Terminology used There is also some confusion about terminology used in the context of subdividing a cleanroom or clean-zone, with terms used such as sub-zone, region, area, sector and so on. These matters will be resolved by the experts, and will be coordinated with the other standards in the ISO TC209 family. The likely outcome of this, subject to further review and discussion, will be the following: cleanroom; clean-zone; fundamental unit area; number of sample points; sample point location. Conclusion and preview of the next draft ISO 14644-1 I want to try to summarise this complex story, and suggest that the following scope and principles are the most likely scenario for the second DIS 14644-1 to be published for comment and national vote sometime in 2012. 1. The classification will be by means of a table quite close to that in Figure 2, and will exclude the 5 micron particle number concentration for ISO 5. It should be noted that the table does not address the real-time monitoring, or special requirements for testing the presence of 5 micron particles as required by some regulatory authorities. 2. The sample size calculation will be based on exactly the same principle as in the current ISO 14644-1:1999. That is, you must sample sufficient volume that you would count 20 particles at the class limit for the largest considered particle size for the ISO Page 10 of 12
cleanliness class in question. This is the requirement within the existing standard and endorsed in the DIS vote, and was not questioned in the vote response. 3. There will be a look-up table that provides information for a cleanroom on the number of fundamental unit areas, and the number of these that needs to be selected for sampling. Figure 5 illustrates how this might look, and includes some sample values. Beneath the table will be a description on how to divide up the cleanroom or cleanzone, and how to use a random number generation process for selecting the ones to be tested. The standard will specify a procedure something like the following. Identify the number of fundamental unit areas from the table. Divide the total area of cleanroom or clean-zone into equal sized fundamental unit areas and identify them uniquely. Identify from the table the number of fundamental unit areas that need to be tested. Use a random number generator to define the ones to be tested. Locate the sample point anywhere within the defined fundamental unit areas for this test. (Since there is a basic assumption that each fundamental unit area is homogeneous in terms of cleanliness, then the absolute location of the sample point within the randomly selected fundamental unit areas can be determined by the tester.) Should additional risk-based sample points be required, then these are additional to those selected on a random basis, except when a risk-based sample point happens to be in one of the randomly selected fundamental unit areas. Cleanroom or clean- Fundamental Number of available Minimum number of zone available area m 2 unit area used fundamental unit areas fundamental unit areas required m 2 to be tested 6 2 3 3 36 4 9 9 52 4 13 10 104 4 26 16 116 4 29 18 900 4 225 29 >900 Use the square root of the area and the practice defined in ISO 14644-1:1999. (Note: the details of how this transition works are still to be developed by the experts of ISO TC209 WG1.) Page 11 of 12
Figure 5. Example sampling plan look-up table. 4. Take the minimum air sample at each point, and count the concentration of particles. This concentration would be normalised to particles per cubic metre. Should more than one sample be taken at any sample point, then it is acceptable to average the count prior to normalising the value. 5. The data collected shall then be evaluated on a point-by-point basis, and all points must comply with the class limit for a specific classification class. References 1. International Standards Organisation. ISO 14644-1:1999 Cleanrooms and associated controlled environments Part 1: Classification of air cleanliness. Geneva, Switzerland: ISO; 1999. 2. International Standards Organisation. ISO 14644-1:2000 Cleanrooms and associated controlled environments Part 2: Specifications for testing and monitoring to prove continued compliance with ISO 14644-1. Geneva, Switzerland: ISO; 2000. 3. Hartvig NV, Farquharson G, Mielke R, Foster M. Sampling plan for cleanroom classification with respect to airborne particles. EJPPS 2011; 16(1):5 12. 4. International Standards Organisation. ISO DIS 14644-1:2010 Cleanrooms and associated controlled environments Part 1: Air cleanliness classification by particle concentration (ACP). Geneva, Switzerland: ISO; December 2010. (Published by ISO for the DIS enquiry and national vote.) 5. US-FED-STD-209E Federal Standard Airborne particle cleanliness classes in cleanrooms and clean-zones. Published by US GSA June 16, 1988. Page 12 of 12