VOTE this article for Best JVT Article 2007

Establishing Alert Limits for Microbial Counts in Purified Water

B Y P R A M O T E C H O L AY U D T H

O
INTRODUCTION
Purified Water (PW) is widely used in pharmaceutical and other healthcare industries. Due to its property as an excellent microbial growth medium, the microbial level has become the most critical quality attribute in Purified Water and is the most common problem encountered in Purified Water systems. To control such a highly sensitive growth medium, understanding of the Purified Water system regarding microbiological contamination, proliferation, and survival (e.g., biofilm) as well as appropriate control knowledge, is essentially required. Although related guidelines including those of both the World Health Organization (WHO) and the U.S. Pharmacopoeia (USP) addressing microbial limits for Purified Water are available, establishing in-house limits is practically required to effectively control microbial contamination. This article is provided to introduce a scientific method for establishing microbial alert limits for Purified Water using statistical methods based on validation or historical data.

SPECIFICATIONS
The USP indicates the following specifications for Purified Water: Quality Attributes pH Conductivity Total Organic Carbon (TOC) Microbial Count Specification Limits 5.0 - 7.0 1.3 S/cm (25C) (USP 24 Specification) 500 ppb 100 CFU/mL

Microbial limits at various stages in water purification system suggested by WHO1: Sampling Points Suggested Maximum Limits (CFU/mL) Target Alert Action 500 500 500 100 100

300 200 Source (Raw) Water 300 50 Post Carbon Filter Feed to RO Unit 20 200 50 10 Feed to Polishing Unit * Purified Water 1 10 * May be EDI (Electrodeionization) Unit or Mixed Bed Column
44 Journal of Validation Technology

Pramote Cholayudth

UNDERSTANDING OF BINOMIAL DISTRIBUTION

Microbial data are countable (i.e., attribute or discrete) and not measurable (variable) data. Variable data may follow a Normal Distribution while the attribute data will follow a Binomial Distribution. Therefore, microbial count data will ideally follow the Binomial Distribution. If we repeatedly take samples of the same size n = 90 from a population containing only conforming (say 95% of the population) and non-conforming (5%) items, the numbers of non-conformity (success) will distribute according to the probability density function (pdf) governing the Binomial Distribution (Figure 1). This probability density function is expressed below as:

Figure 1
Binomial Distribution:

P=

n c

pq

c n c

n! pcq n c = c!(n c)!

CRITERIA FOR ESTABLISHING MICROBIAL ALERT LIMITS

From Figure 1 when the Control Chart limits 3 (approximately 99.73% of the normal curve area) are applied, the control limits for np Chart, one of the most common Attribute Control Charts, will be UCL = np + 3 npq = 10.7 CL = np = 4.5 LCL = np 3 npq = -1.7 = 0 Where: UCL = Upper Control Limit CL = Center Line LCL = Lower Control Limit

Where: P = Probability n = Sample size (n! = n factorial) c = Number of success (c! = c factorial) p = Non-conforming rate q = 1-p The relevant statistics for the curve are the following: Mean = np = 90x0.05 = 4.50 SD = npq = 90 x 0 . 05 x 0 . 95 = 2.07

Where: Mean = Mean of the distribution curve SD = Standard deviation of the curve n = Sample size p = Non-conforming rate q = 1-p In other words, Binomial Distribution is the sampling distribution of the non-conforming units (non-conformity) in samples taken from a population with known non-conforming rate (p). When sampling Purified Water from the system, distribution of the microbial counts will occur in the same manner.

Since the areas across (0, 10) and (0, 11) are 99.47 and 99.82% respectively, so the area across the control limit (0, 10.7) is approximately 99.5% of the binomial curve. To compute the area across (0, 10), for example, is to proceed as follows: Using MS Excel formula, P = BINOMDIST (10,90,0.05,TRUE) = 0.9947 Where: 10 = Number of success 90 = Sample size 0.05 = Proportion of non-conforming TRUE = Cumulative distribution
45

N o v e m b e r 2 0 0 6 Vo l u m e 1 3 , N u m b e r 1

Pramote Cholayudth

The alert limits for the microbial count in Purified Water will be also based on such limits ( 3). However, the more practical limits of 4 or even less stringent (e.g., 5) are preferably applied to compensate the possible water sampling and testing technique errors as shown below UAL = 4 = n p + 4 LAL = - 4 = n p - 4 Where: UAL = LAL = = = n = p = q n p q (1) n p q (2)

EXAMPLE CASES FOR ESTABLISHING MICROBIAL ALERT LIMITS

In a pharmaceutical plant where the author provides consulting service, the validation data Operational Qualification (OQ) for the Purified Water System is as follows:

Upper Alert Limit Lower Alert Limit Mean of distribution Standard distribution (SD) of distribution 100 (CFU/mL) = maximum count allowed Averaged proportion = averaged count per maximum count allowed = (1 p )

The value of n = 100 is based on the action limit (100 CFU/mL), which is the maximum number of colony forming unit (CFU) allowed per mL.

PROCEDURE FOR ESTABLISHING MICROBIAL ALERT LIMITS

All the data to be used must meet the action limit of 100 CFU/mL - i.e.: collected during the state-of-under-control period (validation or monitoring) of the PW system. The more important consideration is that these microbial test results are of the same sample size - e.g.: 1 or 100 mL. One should not mix up the data if two or more different sample sizes are taken. Therefore, the alert limits derived from the data for different sample volumes may be different. Procedure for preparation of data is described in the following steps: All individual results (from all points of use over an extended period) are averaged and then divided by 100 to be a proportion value ( p ) Find q which is equal to ( 1 - p) Find n which is equal to 100 Proceed to next step for calculating the Mean and SD and subsequently the alert limits using equation (1) and (2) above.

46

Journal of Validation Technology

Pramote Cholayudth

Figure 2
Microbial Count Data Microbial Count Results (CFU/mL) Recirculation Loop POU # Day # After After 4 POU # POU # POU # POU # RO EDI Return (+Hose) 1 2 3 4 <1 <1 <1 <1 <1 1 4 1 2 <1 <1 <1 <1 <1 <1 2 1 4 <1 <1 <1 <1 3 2 2 2 6 <1 <1 <1 <1 <1 4 1 3 5 <1 5 4 4 4 1 8 4 4 <1 <1 <1 <1 6 2 3 18 9 <1 <1 7 3 5 7 15 2 7 <1 8 1 2 8 6 6 1 4 <1 <1 9 4 1 6 15 23 10 <1 <1 <1 <1 10 8 5 1 2 Mean 2.06 7.50 3.90 5.30 POU: Point of Use, RO: Reverse Osmosis, EDI: Electrodeionization Note: Zero result is denoted as <1 and transformed to 1 when computed for mean value.

Figure 3
Purified Water System Diagram

Pretreated Water

RO Unit

EDI Vent Filter

Return Loop Purified Water System

Storage Tank

Pump UV

3 2 1 (Point of Use #)

N o v e m b e r 2 0 0 6 Vo l u m e 1 3 , N u m b e r 1

47

Pramote Cholayudth

Figure 4
Microbial Count Distribution in Purified Water

Calculation for the upper and lower alert limits (UAL and LAL) is simply carried out as follows: Count mean ( n p ) p q UAL = n p + 4 LAL = n p 4

= 2.06 = 2.06/100 = 0.0206 = 1-0.0206 = 0.9794

n p q = 7.7 = 8 n p q = -3.6 = 0

Area below the curve = BINOMDIST (8,100,0.0206,TRUE) = 99.98%

From a microbiological point of view, recording the zero result (0) is denoted as <1 in the Test Report. However, in computation of the mean value, it is more practical to use 1 (as in this article) or 0.5 instead of zero.

48

Journal of Validation Technology

Pramote Cholayudth

Figure 5
Microbial Count Distribution in Purified Water (via Hose Routinely Used)

Point of use number 4 is designed for washing a piece of process equipment, sampling the water is also carried out through the flexible hose routinely used while washing. Handling of the hose is Good Manufacturing Practice (GMP) compliant - i.e.: disconnected and hung with both ends down while not in use and also under a sanitizing program.

Count mean ( n p ) p q UAL = n p + 4 LAL = n p 4

= = =

7.5 7.5/100 = 0.075 1-0.075 = 0.925

n p q = 18.0 = 18 n p q = -3.0 = 0

Area below the curve = BINOMDIST (18,100,0.075,TRUE) = 99.99%

N o v e m b e r 2 0 0 6 Vo l u m e 1 3 , N u m b e r 1

49

Pramote Cholayudth

Figure 6
Microbial Count Distribution in Purified Water

Figure 7
Microbial Count Distribution in Purified Water

50

Journal of Validation Technology

Pramote Cholayudth

RO and EDI waters are also controlled for microbial level. Establishing their limits is undertaken in the same way as follows: Microbial Alert Limits for RO water Count mean ( n p ) p q = = = 3.9 3.9/100 = 0.039 1-0.039 = 0.961

quality in the range of 30-50 CFU/mL. This operational data would not justify establishing a less stringent specification of not more than 100 CFU/mL. In this case the method introduced here can be properly used. The pretreated water - e.g., feed to RO and RO water, is also required to have a controlled level of microbial contamination. Establishing their alert limits can be made using the same method. Based on data consistency, intended use, and related risks, the choice of 4 or 5

UAL = n p + 4 LAL = n p 4

n p q = 11.6 = 12 n p q = -3.8 = 0

Area below the curve = BINOMDIST (12,100,0.039,TRU E) = 9.9% Microbial Alert Limits for EDI water (PW ) Count mean ( n p ) p q = = = 5.3 5.3/100 = 0.053 1-0.053 = 0.947

UAL = n p + 4 LAL = n p 4

n p q = 14.3 = 14 n p q = -3.7 = 0

Area below the curve = BINOMDIST (14,100,0.053,TRUE) = 9.97%

DISCUSSION
Reproduced from the Human Drug cGMP Notes,5 firms should set and justify their own microbial limits for Purified Water (PW). The microbial limit for the water as a component should be more stringent than the limit set for the end product. For example, where a finished product has a microbial limit of not more than 100 CFU/g, the corresponding limit for the water should be less than 100 CFU/mL. According to the Notes, properly controlled and welldesigned PW systems should be capable of producing water
51

N o v e m b e r 2 0 0 6 Vo l u m e 1 3 , N u m b e r 1

Pramote Cholayudth

should be considered for its proper use. Trend data analysis for the microbial count data on periodic basis (e.g. monthly) is also required using the same method for establishing the trend limits. Construction of the Binomial curve will help us imagine how the microbial count will distribute and how justified the upper limit is. For readers who are interested in construction of the Binomial curve for the Purified Waters microbial data, MS Excel file used for forming the curve is obtainable from cpramote2000@yahoo.com.

Article Acronym Listing

EDI MS OQ PW RO SD TOC WHO Electrodeionization Microsoft Operational Qualification Purified Water Reverse Osmosis Sample Standard Deviation Total Organic Carbon World Health Organization Population Mean Population Standard Deviation

CONCLUSION
Alert limits should be established for Purified and other water systems. Furthermore such limits are specifically required for different sub-systems - e.g.: pretreatment, purification, and recirculation. When exceeded, it indicates that the system or sub-system may have drifted from its normal operating condition. Alert levels are to be statistically established based on historical, routine monitoring data and should reflect the seasonal variations of the source water. Therefore, comprehensive accumulated data up to one year is required. For new systems, alert limits may be temporarily established using the OQ data. The limits will be periodically reviewed and revised when necessary using the most updated accumulated data. Once the accumulated data can cover the annual seasonal variations (of source water) the limits can be finalized and used permanently. J

cepts and Practices of Pharmaceutical Process Validation. He is currently an Editorial Advisory Board member of the Journal of Validation Technology and can be contacted by fax at 662-740-9586, by e-mail at cpramote2000@yahoo.com, or by standard mail at the following address: Pramote Cholayudth, 6/756 Number One Complex Bangkok-Ram 2 Road Pravate District Bangkok 10250, Thailand

REFERENCES
1. WHO Supplementary Training Modules on Good Manufacturing Practice, Presentation File on Water for Pharmaceutical Use Part 2: Water Purification Engineering. 2. The United States Pharmacopeial Convention, Inc., General Information Chapter <1231> Water for Pharmaceutical Purposes, United States Pharmacopoeia 27th edition Philadelphia, PA: National Publishing. 3. WHO Technical Report Series # 929, Thirty-ninth Report, Annex 3: WHO Good Manufacturing Practices: Water for Pharmaceutical Use, World Health Organization, 2005. 4. European Medicines Evaluation Agency, Note for Guidance on Quality of Water for Pharmaceutical Use, London, CPMP/QWP/158-01, May 2002. 5. Human Drug CGMP Notes, Volume 5, Number 1, March 1997, What Should Firms Use as Microbial Limits for Purified Water? 6. Vincent, D. W., Qualification of Purified Water Systems, Journal of Validation Technology, Volume 10, Number 1, November 2003.

ABOUT THE AUTHOR

Pramote Cholayudth is Executive Director of Valitech Co., Ltd., a well-established validation and compliance consultant services company to the Pharmaceutical Industry in Thailand. He is a guest speaker on Process Validation to industrial pharmaceutical scientists organized by the local FDA. Pramote was a full-time lecturer in the School of Pharmacy of a private university for four years (1998-2001). Prior to entering the academic arena, he spent 23 years in the Pharmaceutical Industry with Bayer Laboratories (1974-1981) and OLIC (Thailand) Limited (1981-1997) - a leading and the largest pharmaceutical toll manufacturer for multinational companies. Pramote is the author of Con52 Journal of Validation Technology