Chemistry Reference
In-Depth Information
6.13.5 A
verAgIng
(r
eSAmPlIng
t
eStIng
d
AtA
)
Averaging of test data can be a valid approach, depending on the sample and its
purpose [20]. In some analytical techniques, several discrete measurements are
often averaged to report a test result. For example, an HPLC result may be deter-
mined by averaging the peak response from replicate injections of the same sample
preparation. In this instance, the average result is considered one test and one result.
Reliance on averages has the disadvantage of masking variability among individual
test results, however. For this reason, unless averaging is specified by the SOP, all
individual test results should be reported, along with a statistical treatment of the
variability. This is common, again, in content uniformity assays, where the standard
deviation is also reported.
6.13.6 o
utlIer
t
eStS
cGMP regulations require that statistically sound control criteria include acceptance
or rejection levels [21,22]. A result may qualify as a statistical outlier if it is markedly
different from the others in a series of results obtained by a validated method. The
use of an outlier test should be determined in advance, and again, documented in
the SOP, and it should specify the minimum number of results required to obtain a
statistically significant assessment. Because an outlier test is only a statistical analy-
sis, it cannot be used to invalidate the data, but is useful for the evaluation of the
significance of the result for batch evaluations. One note of caution: in cases where
the variability of the product is what is being measured (i.e., content uniformity), an
outlier test should not be applied, because a measurement thought to be an outlier
may in fact be an accurate result! Table 6.2 provides an example of outlier testing.
6.14 concludIng tHe oos InvestIgAtIon
Now it is decision time. To conclude the investigation, following the SOP, the results
should be evaluated, the batch quality determined, and a release decision should be
made. The goal of the investigation is to arrive at one of two conclusions: either the
batch fails and should be rejected (i.e., the OOS result is confirmed), or the OOS
result is invalidated and an assignable cause is revealed. The OOS result can only be
invalidated upon the observation and documentation of a test result that can reason-
ably be determined to have caused the OOS result. If the OOS result is confirmed,
the batch is rejected.
Of course, there is one other possible outcome. Despite all the controls in place,
assessing, identifying, and investigating results may still be inconclusive. In cases
where an investigation does not reveal a cause or confirm the OOS result, the OOS
result should be retained in the record and taken into account in the batch or lot
disposition decision.
Finally, for those products that are the subject of applications, regulations require
submitting a field alert report (within three working days) concerning any failure of
a batch to meet any of the specifications established in an application. As the saying
goes, no job is finished until the paperwork is done!
