How many particles are needed to demonstrate statistical confidence for cleanroom control and how much sample volume is required to see them? There is increasing confusion over the requirements for compliance with EC GMP Annex 1 limits on 5.0 µm particles, especially when compared with the ISO 14644-1 calculations recommended to establish conditions for such testing. The problem arises when using the ISO 14644-1 calculations. The minimum sampling volume required by this formula for EC GMP compliance is huge. If we look at a Class A (ISO 5) environment and the size ≥0.5 µm, where C_{n.m} is 3500 nm^{-3} the equation becomes: Using a 1 cubic foot per minute (CFM) instrument this volume would take approximately 12 seconds to sample. A caveat in the rules requires that the sampling duration is a minimum of 1 minute, so that a full 28.3 litres would be drawn (28.3 l = 1 CFM). However, if we look at the 5.0 µm requirement for the same Class A environment we are now faced with a limit of 1 nm^{-3} and the equation now becomes: This would now require a sample at each location of 20 m^{3} and using a 1 CFM particle counter the sample would take approximately 706 minutes (11.8 hours). The question is how to gain statistical confidence while maintaining a reasonable sample period. ## History of Application The history of this application dates back to the Federal Standard on cleanroom cleanliness, FS209. When reviewing the original standard the minimum number of particles allowed in a sample was deemed to be 20. This number determined whether a sample was statistically significant either for class limits of the ‘U’ descriptor. If a sample did not yield sufficient particles a sequential sample technique could be employed. Sequential sampling is a technique that allows the total required volume to be divided into equal parts and essentially the proportion of required particles equally divided throughout each portion. Therefore, a large volume did not have to be sampled if the sub-sample showed sufficient evidence of cleanliness. The statistical confidence of a population of these random particles was determined to have a minimum cut off before special calculations were required to prove that lower numbers showed any statistical validity. When the revised ISO14644-1 was released it also contained this function of statistical confidence and also required a minimum of 20 counts per unit volume. There is evidence that for certification purposes, either FS209E or ISO14644-1 a minimum number of particles is required. As particle counters have a fixed flow rate (1 cfm, 0.1 cfm, 50 L/min), so too is a fixed sample period is determined by this calculation. ## Solutions for Certification Two routes are available when looking at room certification. Either one can adopt the recommended minimum sample volume written in the EC GMP Annex 1, or use the sequential sampling technique identified in both the FS209E and ISO 14644-1 standards. ## Minimum Sample Volume from EC GMP Annex 1 September 2003 In the notes section immediately below the EC GMP classification table, Note ‘a’ states: “For routine testing the total sample volume should not be less than 1 m^{3} for grade A and B areas and preferably also in grade C areas.” Therefore, if “routine testing” (i.e.: periodic room certification) the sample volume needs to be 1 m^{3} and not the calculated 20 m^{3} if using ISO alone. What is unclear and open to interpretation is whether the 1 m^{3} applies to the grade A/B area or to each sample point within each area. That is, if my room has five sample points do I need to sample 1 m^{3} at each location or is 1/5th 1 m^{3} at each location sufficient? Both approaches are used and both are accepted as long as a high degree of confidence can be shown over control of the environment. ## Sequential Sampling In sequential sampling the running total of the particles counted is compared with an expected count limit that is a function of the amount of sampling done. Sequential sampling typically requires less sampling than any single sampling plan having the same probability of false acceptance and false rejection. Figure 1 shows the boundaries of the sequential sampling plan that has been designed for use in this standard. The observed number of counts, C, is plotted against the expected number of counts, E, for air which is precisely at the class limit. A full single sample corresponds to E = 20. **Figure 1.** Sequential sampling boundaries Therefore, if the expected particle counts should be ten and only five have been measured, the sampling can stop, the location declared a pass, and sampling started at the next location. ## Solutions for Monitoring In accordance with GMP regulations cleanrooms need to be routinely monitored in operation. The monitoring locations must be based on formal risk analysis obtained during the initial classification of the room. For Grade A zones a continuous or frequent sampling particle monitoring system should be used. The Grade A zone should be monitored at such a frequency that all interventions and other transient events are captured and alarms triggered if excursions from defined operating norms occur. The sample sizes taken for monitoring purposes using automated systems is a function of the sampling rate of the system used. It is not necessary that the sample volume is the same as that used for formal classification of clean rooms and clean air devices. Sample periods of one minute are normal and can be as short as ten seconds for critical areas associated with high volume production. This will give much smaller sample volumes. However, statistical confidence over the process is now gained due to an increase in the number of individual samples taken. In Grade A and B zones, the monitoring of the 5.0 µm particle concentration count is significant as it is an important indicator of failure of sterility. An occasional indication of a single 5 µm particle count is typically a random count and has no foundation with an associated problem. However, consecutive or regular counting of low levels is an indicator of a possible contamination event and should be investigated. |