SELECTION AND EVALUATION OF PACKAGING DATA

To this point retrospective validation has been discussed in the context of dosage
form manufacture. Some of the same concepts may be applied to validating a
packaging operation. Consider the following. Packaging lines are typically controlled
by making spot observations to confirm machinery performance and component usage. The frequency of the inspections and the number of samples
examined during each cycle are normally defined in a written procedure. Furthermore,
the results of each monitor are generally documented in an inspection
report, which becomes part of the packaging record for that lot of product. Also
available from the packaging record is the number of units produced, thus the
information needed to allow inferences about the reliability of a particular operation
is readily accessible.
If we can show that over an extended period of time an operation had a
certain reliability, it is not unreasonable to expect the same level of performance
for the future as long as the equipment is reasonably maintained. Conversely,
any conclusion reached by such a study would be invalidated by substantial
change to the equipment or its method of operation.
How many packaging runs must be examined to draw a sound conclusion
about the reliability of the operation? Unfortunately, no one answer is appropriate
for every situation, but there are some rules that will aid the decision
process. The sample size should be large enough to capture all variables normally
experienced; for instance, routine machine problems, shift and personnel
changes, component vendor differences, and seasonal conditions. Furthermore,
the sample must be of sufficient size to provide a high degree of confidence in
the conclusion. Ten thousand observations made over 6 to 12 months of continuous
production generally satisfy these requirements. For high-speed, multipleshift
operations the 10,000-observation figure is likely to be reached well before
sufficient time has elapsed to include all avenues of variability. In these cases,
time rather than units produced should be the first consideration.
To validate an aspect of the packaging operation retrospectively the following
information must be tabulated:
1. The total number of observations made for the quality attribute under
review
2. The total number of nonconformances detected by the inspection process