The past year has seen a change in the way pharmaceutical manufacturers monitor their aseptic environments with the most recent revision to the guidance chapter <1116> in USP 35/NF 30, now entitled “Microbiological Control and Monitoring of Aseptic Processing Environments.”1
One of the major changes within this chapter is guidance for the assessment of contamination recovery rates (CRR) for environmental monitoring (EM) data, moving away from the Alert and Action level concept. The USP defines contamination recovery rate as “the rate at which environmental samples are found to contain any level of contamination.”2 Use of the CRR recognizes the limitations of traditional microbiological tests, creates a paradigm shift away from long-standing guidance in environmental monitoring, and places a focus on the evaluation of trends versus comparison against the standard colony forming unit count approach. The CRR concept also moves the industry closer to utilizing a strategy that does not rely on a single test methodology, and instead focuses on the incidence of contamination occurrences, making this USP revision highly conducive to implementation of more cost-effective and sensitive microbiological methods that speed the time-to-result and reduce sampling bias.
Traditional EM methodology and limitations
The method of bacterial detection typically used by the pharmaceutical microbiologist has been the Heterotrophic Plate Count method (HPC), which is defined by the detection of individual bacterial colonies from an environmental sample (air, water, surface, etc.) that may grow on a Petri dish, also known as colony forming units (CFUs). Interestingly, the CFU is a derivative quantity; the individual CFU from the HPC method represents merely a subset of the target measurement (microbes in the environment), rather than directly counting that quantity. There are two factors for this disparity: first, the HPC incubation conditions do not support recovery of all organisms (this concept has been extensively treated in literature; refer to citations 3, 4, 5, and 6 for a sampling of sources), and second, the probabilistic nature of a single finite sample, which cannot represent an entire environment in space or time.
It is tempting to believe that the CFU is an actual quantity that can be measured reproducibly. After all, the test yields a whole-number count, and tests over time in the same location using the same techniques typically give similar results. As volume and/or microbial concentration decrease, however, it becomes harder and harder to detect low levels of organisms using the HPC method. Broadly speaking, the number of samples needed to accurately quantify the microorganisms in an environment is inversely related to the concentration of microbes. Yet, these tests are valuable in that a detection of trend is possible, though the fidelity of that trend depends on the reproducibility of the sampling and test regime being employed.
Yielding to the temptation of the CFU
Trend detection is the rationale given for the use of this traditional approach, even in light of the limitations. In the USP 23/NF 18 revision of <1116>, it was deemed critical that “…an environmental monitoring program be capable of detecting an adverse drift in microbiological conditions in a timely manner that would allow for meaningful and effective corrective actions.” The problem arises when one tries to summarize the state of quality of an environment with a particular CFU count. Nevertheless, there still remains a temptation to assign a critical weight to the CFU values generated from these samples, and to ascribe a level of quality to the data generated. This represents one of the fallacies of microbiological monitoring: that the quality of the product being manufactured is represented by the measured environmental quality. An HPC result of “zero” does not represent an environment that is sterile or free of objectionable organisms. Rather, it indicates that trends otherwise detected at higher cleanroom grades or ISO levels cannot be detected at the low concentrations encountered in that environment.
Alert and Action levels, as indicated by the previous iteration of USP <1116>, reflect microbiological recoveries for which a certain activity (such as investigation) needs to occur. The temptation here is the belief that the lower the CFU counts are, the fewer remediation activities need to be performed. Taking this to its logical conclusion, CFU counts of zero would then be considered ideal. If one looks closer at the statistics involved with using the HPC method in these clean environments, however, the low counts begin to have a high rate of standard deviation, ranging from 20% at a mean CFU count of 25, to 100% at a mean CFU count of one.7,8 In accordance, counts generated in cleanrooms over long periods of time begin to resemble not a standard quantification of organisms, but rather a series of positive or negative incidence data, similar to tests performed using the most probable number (MPN) method. Probabilistic effects are apparent, and while collecting more samples or greater sample volumes may alleviate some of these statistical issues, it is very resource-intensive, to the point of impracticality.
Contamination recovery rates: Refocusing on science
With the USP 35 version of <1116>, a new paradigm of trend analysis has been proposed, called the Contamination Recovery Rate (CRR) approach. The paradigm is simple—in a given data set of samples over a monthly time period, a certain percentage of those samples can be expected to exhibit non-zero recoveries of contamination; the percentage of samples presenting contamination is then defined as the Contamination Recovery Rate. These percentages are listed in Figure 1.
For example, during a typical day of pharmaceutical manufacturing, a site within a cleanroom may be sampled once per shift. For that location, three data points each day are generated with an associated CFU result. An issue arises, however, when it is assumed that these three data points represent the totality of microbiological quality for that site throughout the day. This sampling regime only creates three individual ‘snapshots in time’ of the microbiological activity in that location. A month’s worth of data in this location would yield about 90 individual sample points from which a rate could be derived. Yet, because aggregation of data is required to determine a true rate, accurate conclusions can only be drawn about the localized trend after long periods of time. A key benefit of utilizing RMMs then becomes one of increasing this sampling frequency, so that the contamination rate can be determined more quickly and precisely. New RMMs are uniquely suited to this task, as they can sample at faster rates than traditional methods, often at lower cost-per-test, with some having the capability to sample remotely. Ultimately, by developing a way to assess a contamination rate daily, or even continuously using a rolling CRR analysis approach (enabled by continuous RMM data), manufacturers are empowered to maintain the manufacturing environment with a higher degree of control, detection, and responsiveness. Such an implementation allows the industry to meet the goals of an EM program guided by USP<1116>, “to demonstrate that that the aseptic monitoring environment is operating in an adequate state of control.” 2
Contamination recovery rates support process understanding
As pharmaceutical manufacturing processes continue to improve, the tools and statistics by which pharmaceutical manufacturers monitor and analyze those processes must also improve. Ready RMM tools exist with which to monitor these processes in an enhanced way, but their implementation has been limited by long-existing sampling and trending paradigms, which ultimately hamper process knowledge and control.
The revised USP <1116> guidance has refocused the industry on the goal of gaining knowledge and understanding of processes, to achieve a higher state of control, and ultimately a higher state of quality assurance.
References
1. United States Pharmacopeia. “<1116> Microbiological Control and Monitoring of Aseptic Processing Environment.” United States Pharmacopeia 35th edition; National Formulary 30th edition, 2012.
2. United States Pharmacopeia. “<1116> Microbiological Control and Monitoring of Aseptic Processing Environment.” United States Pharmacopeia 37th edition; National Formulary 32nd edition, 2013.
3. La Duc, M.T. et al. (2007) “Isolation and Characterization of Bacteria Capable of Tolerating the Extreme Conditions of Clean Room Environments.” Applied and Environmental Microbiology, 73, 8, 2600-2611.
4. Moldenhauer, J. (2011) “Use of Viability Methods—The Problem of Viable but Not Culturable (VBNCs).” Moldenhauer, J. (ed.) Rapid Sterility Testing 2011, 181-200.
5. Oliver, J. (2005) “The Viable but Nonculturable State in Bacteria.” Journal of Microbiology (Microbiological Society of Korea), February 2005, 93-100.
6. Roark, A.L. (1969) A Model for the Quantification of the Qualitative Microbial Sampling Problem. Report SC-RR-69-310, May 1969. Sandia Laboratories Publication, Albuquerque, NM. Bioscience Division, Office of Space Science Applications, NASA.
7. D. Hussong and R.E. Madsen. “Analysis of Environmental Microbiology Data From Cleanroom Samples,” Pharm. Technol. (Aseptic Processing Suppl.), 10–15 (2004).
8. C. Eisenhart and P.W. Wilson. “Statistical Methods and Control in Bacteriology.” Bact. Rev. 7(57), 137, 1943.
9. Parenteral Drug Association. “Interview with James Akers on Revised USP <1116>.” PDA Letter, March 2012, pp28-30.
Peter Noverini is a Field Applications Scientist at Azbil BioVigilant, Inc. (ABV), a company focused on delivering instantaneous microbial detection technologies to the pharmaceutical and life science industries. He joined the company in 2010 after working for more than ten years in the parenteral drug and medical device industries, specializing in the microbiological aspects of environmental monitoring, bioburden evaluation, and validation of sterilization processes. In his role at ABV, he is part of a team of scientists who support and guide customers throughout their RMM implementation and validation process.
This article appeared in the March 2014 issue of Controlled Environments.
