Sei sulla pagina 1di 2

Downloaded from journal.pda.

org on November 16, 2016

COMMENTARY

Setting Alert/Action Limits for Environmental Monitoring Programs


JAMES D. WILSON

Pharmaceutical Systems, Incorporated, Mundelein, Illinois

The principle purposes of an environmental monitoring or negative exponential distributions. Regardless of the
program are to provide an assessment of the general method employed, however, a special concern is how one
microbiological cleanliness of an operation and to assess the deals with outliers and clusters of unusual results in the data.
state of environmental systems control. Being cognizant of Including data taken from a period of unusually high counts,
the numbers and kinds of microbes present, a microbiologist where the process was out of control, will lead to inappropri-
can assist in designing an effective control system(s); ately high alert/action limits. Such clusters of obviously out
however, no environmental monitoring program, regardless of control data should be excluded before alert/action limits
of its design, can provide the level of confidence desired are calculated. For example, suppose a plot of the data
without subsystem management which is accomplished by versus date showed that counts from July and August were
adherence to current good manufacturing practices, facility about twice as high as typical. An investigation showed that
design control, effective supervision, sound corrective ac-
a new technician had been using poor technique and was
tion steps, and proper employee training.
retrained at the end of August. If this out of control period is
An environmental program is comprehensive. One impor- left in, the alert/action limits will be calculated inappropri-
tant component of an environmental surveillance system is
ately high. The out of control period is not part of the
the establishment of alert and action levels. It is dangerous to
underlying process capability and should not be used in
consider alert/action limits independently of other compo-
calculating limits.
nents of the system. The setting of absolute numerical limits
for microbiological monitoring, from either published re- Several ways to identify unusually high clusters of data
ports or guidelines, should be approached with caution due include (1) plots of the data over time; (2) histograms; and
to the diversity of operations and inherent variability of (3) statistical process control charts. The data should always
sampling/testing methodologies, notwithstanding statistical be plotted versus date for separate physical locations to help
implications. identify unusual data points.
There are several methods in use for establishment of Once the data have been cleaned up by removing obvious
alert/action levels. There is no clear consensus among out of control points, the underlying process capability can
pharmaceutical/device manufacturers on a preferred method. be determined. Several methods may then be used to
Challenges to microbiologists apparently occur when wide determine appropriate alert/action limits on the cleaned up
ranges in results are observed or when limited data are data:
available. The challenge for the microbiologist is exacer-
bated by the adoption of arbitrary limits imposed by some 1. Direct calculation of percentiles
Food and Drug Administration (FDA) investigators. If there are enough data points (perhaps 100 or more),
If historically based alert/action levels are desired, accept- the action level may be set to equal the 99th percentile
able microbial levels must be established based on sound of the data; the alert level may be set to equal the 95th
statistical surveys of historical data that have been collected percentile. This method is probably the most desirable
on a consistent basis using identical methods. Historically, because it does not make any assumption about the
environmental data have been treated as means ± 2 or 3 SD, shape of the distribution of the data.
cumulative frequencies, log transformations, etc., but these 2. Poisson
methods can ultimately lead to levels that may not represent Counts often follow the poisson distribution. If the
the actual process capability and/or an understanding of the data appear to be Poisson distributed, the estimated
same. 99th and 95th percentiles may be calculated with
Setting alert/action levels using standard statistical tools tables of the Poisson distribution or available software.
such as means ± 2 or 3 SD, while providing a method of 3. Negative Exponential Distribution
identifying out-of-limit trends, can be misleading because Although microbial counts are not continuous data, the
(1) these methods are based on a normal distribution and (2) exponential distribution may provide a reasonable ap-
these methods are based on two-tail probability. In practice, proximation. The estimated 99th and 95th percentiles
environmental data are usually not normally distributed; may be calculated by multiplying the mean by 4.6 (for
environmental data histograms generally resemble poisson the 99th percentile) or 3.0 (for the 95th percentile).
4. Other Distributions
If the data appear to be approximated by the normal
Received January 25, 1997. Accepted for publication April 10, 1997.
Correspondence address: Pharmaceutical Systems, Inc., 102 Terrace Dr., distribution, which does not often occur with microbial
Mundelein, IL 60060-3826. data, the upper one-sided 99th and 95th percentiles

Vol. 51, No. 4 / July-August 1997 161


Downloaded from journal.pda.org on November 16, 2016

may be approximated by the mean ± 2.326 SD (99th) This deficiency can happen when the method and distribu-
and by the mean ± 1.645 SD (95th). Sometimes taking tion used does not match the actual distribution of the data.
a transformation of the data, such as square root or This approach provides a reference point for adopting
logarithm, will normalize the data. Other distributions. official or corporate alert/action levels by using a sound
such as the Weibull, may also be used. scientific rationale.
It is advisable to compare the alert/action limits obtained
Acknowledgment
with a histogram of the data. If something is dramatically
wrong with the method used, it will be obvious when The valuable assistance of Mark Varney is gratefully
locating where the alert/action limits are on the histogram. acknowledged.

162 PDA Journal of Pharmaceutical Science & Technology

Potrebbero piacerti anche