Developing a viable environmental monitoring program for nonsterile pharmaceutical operations.

Article Excerpt
No standards currently exist for establishing a viable environmental monitoring program for nonsterile pharmaceutical manufacturing operations. This article presents one approach to establishing such a program.

**********

In the pharmaceutical industry, current guidelines or standards do not exist for establishing a viable environmental monitoring program in a nonsterile drug manufacturing operation. As a result, inconsistent practices abound, and in some cases, monitoring is not being conducted at all.

Because of this lack of defined standards or guidelines, viable monitoring programs in nonsterile pharmaceutical manufacturing facilities range from comprehensive programs, paralleling those of aseptic operations, to nonexistent. From the number of warning letters about environmental monitoring posted on FDA's Web site, one can clearly recognize that the agency desires a formal, defined program. Therefore, the industry should establish voluntary guidelines before prescriptive FDA standards are mandated.

A viable environmental monitoring program should provide sufficient information to show that the manufacturing environment is operating at an adequate level of microbiological control. This requirement applies to nonsterile, terminally sterilized, and aseptically filled products. Although the required state of control will vary depending on the product being manufactured, the requirement is universal across all pharmaceutical manufacturing. In fact, the microbiological control of drug manufacturing areas is a statutory requirement of 21 CFR 211 as stated in Sections 211.46, 211.56, and 211.113 (1).

Surveillance of viable environmental microorganisms is one tool used to assess the integrated effect of the multifunctional controls of the manufacturing environment (e.g., HVAC, sanitization, gowning, personnel practices, etc.). Although guidance does exist for the environmental monitoring of sterile product manufacturing and controlled areas (2-4), regulatory requirements or compendial guidelines do not exist for the frequency of monitoring or the acceptable levels of microbial contamination, specifically in nonsterile drug manufacturing environments, which continues to be a subject of considerable discussion. This article describes one approach to establishing a viable environmental monitoring program for a nonsterile manufacturing operation.

Program rationale

One of the first challenges in establishing a viable monitoring program is educating and convincing senior management that such a program is required. Frequently, management does not understand why the company should monitor viable environmental microbial contamination because the company does not manufacture sterile products and the finished products do not have a microbial release specification.

The statutory requirements of 21 CFR 211 and the lack of viable environmental monitoring in nonsterile drug facilities mentioned in FDA warning letters should address the need for such a program (1). Furthermore, the results from a 1998 survey conducted by AAI International (Wilmington, NC) can be used to highlight current industrial practices (5). In this survey, 94% of respondents indicated that they conducted some viable monitoring. A total of 75% of respondents indicated that they monitored at least monthly, with 50% monitoring at least weekly. Of those respondents, 67% used both active air samplers and contact plates.

To supplement and update AAI's survey, PSI (Mundelein, IL) recently conducted an informal survey of 12 pharmaceutical manufacturers. PSI's findings confirmed those of AAI's survey and also highlighted the following industry practices:

* Monitoring is routinely conducted (i.e., weekly, monthly, or quarterly depending on the product).

* Monitoring sites are selected on the basis of product contamination risk assessment.

* Alert and action levels are established on the basis of system capabilities and are derived from historical data. These levels are evaluated on an ongoing basis.

The respondents' general consensus revealed that microbial excursions observed during environmental monitoring are anticipated and are not directly tied to any product release specification. However, the respondents concluded that excursion events should be investigated when they do occur.

Procedure

The protocol. An environmental monitoring program should be established and coordinated using a preapproved protocol. At a minimum, the program should address the following elements:

* sample site rationale

* site maps

* sampling procedure

* sampling frequency

* sample handling

* sample incubation

* data analysis

* investigative responses to exceeded action levels

* statistical data trending

* establishment of alert and action levels

* execution responsibilities

Because of the lack of regulatory requirements or guidelines for specific levels of viables in nonsterile pharmaceutical manufacturing areas, defined protocol acceptance criteria are not established for the viable microbial counts. The purpose of the protocol is to establish a base line for the indigenous flora in the manufacturing operation. Therefore, all data are of interest. The only stated acceptance criteria are that the microbiological media meet growth promotion requirements, that all equipment used is in a current state of calibration-validation, and that the study conducted is consistent with the protocol. The protocol should also state that preliminary alert and action levels will be established at the conclusion of the first 12 weekly samplings.

The program. Sample-site rationale. Sample-site selection should be based on an analysis of the risks associated with product contamination. A hazard analysis and critical control point (HACCP) evaluation should be used and documented for each selected sampling location. FDA's stated position is that any environmental monitoring program should assess the risk of possible product contamination when establishing sampling sites (6). Because atmospheric and personnel contamination are the primary sources of airborne and surface contamination, sampling sites should be selected in high-traffic areas where the potential exists for direct product contact or direct contamination of product contacting surfaces.

Site maps. All sample sites should be detailed on area maps for each room monitored to ensure that the sites are included during each sampling period. This procedure makes analysis and trending of the data more meaningful.

Sampling procedure. The initial monitoring frequency should be once a week for a total of 12 weeks, after which time preliminary alert and action levels can be established. Many procedures exist for the calculation of these levels. In most cases, environmental microbial data do not follow a normal distribution, although high-count data may approximate a normal distribution. In general, experimental data more closely resemble either a Poissen or negative exponential distribution. In these cases, one of the following procedures is used to estimate alert and action levels:

* Direct calculation of percentiles: An alert level is set at the 95th percentile, and the action level is established at the 99th percentile. This method is used most often and is more reliable when the number of data points is [greater than or equal to] 100. However, it is also used for smaller data sets.

* Negative exponential distribution (7): This model provides a reasonable approximation of the 95th and 99th percentiles, even though the data is not continuous. The 95th percentile is approximated by multiplying the mean of the data set by 3.0, and the 99th percentile is approximated by multiplying the mean of the data set by 4.0.

If the data set appears to approximate a normal distribution, the upper one-sided 95th and 99th percentiles can be approximated by the mean [+ or -] 1.645 standard deviation (SD) and the mean [+ or -] 2.326 SD. If the graphical trending of the data indicates that the weekly counts per area are consistent, the routine sampling frequency could be reduced to monthly for the next nine months (totaling one year from project initiation). After this time, the data are reevaluated to ascertain if further reduction in sampling frequency is warranted (i.e., it can be supported by the data). The minimum sampling frequency for a monitoring program is quarterly.

Outliers, data points or a cluster of points that lie far outside the range of other data, should not be used in any statistical analyses of historical data because they will skew the analyses, which can result in elevated calculated alert and action levels. An out-of-control state is not representative of the continuous process capability and, therefore, should not be used when calculating response levels.

Outliers may be a result of adverse conditions, an out-of-control state, a breach in aseptic technique, or other causes. Statistical methods exist for determining whether to reject outliers, many of which rely...