Anda di halaman 1dari 1

Peer-Reviewed: Medical Devices

3. Estimate the confidence level and evaluate the sample


plan. In most cases, if the p-value (the probability to
accept null hypothesis, i.e. reject the process) is less
then 5% at 95% confidence level, then the sampling
plan is adequate
4. Correct the sample plan and repeat the testing if
needed.
The following are considerations that need to be taken
into account:
Evaluate the measuring system. Is your testing method
and equipment adequate for the purpose of testing?
Sampling procedure. Different lots, different raw
materials lot, different equipment, different locations?
Random sampling. Random testing may be something along the lines of testing the 5th sample in
every 9th row.
Check the distribution. Most statistical test methods
are valid for the normal distribution only. Perform
the normality test, if applicable.
The following spreadsheet analysis tools can be used:
Tables (14, 15), Microsoft Excel, Minitab, SAS.
Frank provides another practical consideration for
sample size determination:
The sample size for each test should be established in
the test plan, and should be based on the criticality of the
process.
Sample size is normally based on the level of confidence
desiredtoensureacertainportionofthepopulationiswithin
the sample range. This may be calculated by the Wilks
equation:
npn-1 (n-1)pn = 1 - alpha
where:
n = sample size
p=proportionofpopulationcontainedwithinthesample
range
alpha = confidence level
Typically, a sample size of 30 is adequate for most testing,
which covers 90% of the population with 80% confidence.
Occasionally, a sample size as low as 3 may be used in
development activities, where 50% of the population is
covered with 50% confidence (3).
To provide a practical recommendation for the sample
size calculation, definitions for the main terms need to
be provided:
Null Hypothesisthe probability that the statistical
result is false. For example, when a process is changed,
we need to prove that the changes dont impact process
outputs. In case of a non-parametric output, where the
output is a variable, the appropriate tool to provide
90 Journal of Validation Technology [SPRING 2012]

statistical evidence that the process outputs have not


changed is the analysis of means. In this case, two
sample sets need to be collectedoutputs from the
original process and outputs from the modified process. The null hypothesis is that there is no difference
in means between the two sample sets.
Statistical Powerthe probability to reject the null
hypothesis when the null hypothesis is true. The more
samples that are collected, the greater the value of the
power. Usually, the power depends on the criticality
of the process, and the value can be 0.8 or 0.9.
Confidence levelthe level of confidence in the statistical results. The confidence level usually is selected
at 0.95 or 0.99 for critical parameters.
If the standard deviation of the process is not know, then
an assumption of its value can be made based on the results
of process monitoring and development activities. After
process data are collected, this assumption can be tested.
Another assumption that has to be made is the acceptable difference in means that indicates that the process
has not changed. That assumption can be based on the
specification limits, or accuracy of the test method. Usually,
5-10% of the specification can be considered as a reasonable
difference. Suppose in our case the allowable difference is
two standard deviations.
Figure 4 was calculated with Minitab, version 15. Alpha
is selected 0.05, which means that the confidence level is
0.95. The result shows that the sample size of 7 can be selected for the planned study. Also, the sample size depends on
power and allowable difference in means.

CONCLUSION
Process validation is one of the key components of a quality
system. The main constituents of the process are environment, buildings, equipment, personnel, and documentation. The existing guidances and standards cover different
aspects of process validation, which requires one to take
into consideration multiple sources for establishing a specific process validation methodology for each company.
FDA, as a government organization, strives to maintain
the public safety by providing guidances and regulations to
follow in order to produce safe products in qualified facilities. Individual companies establish validation processes to
conform to FDAs guidelines. Public complaints and FDA
inspections of manufacturing and research facilities allow the
government to control quality operations and development
activities in the pharmaceutical industry. Failure to follow
FDA guidances and regulations may lead to observations
and possible consent decrees, which are aimed at outlining corrective action procedures for companies to return
to a compliant good manufacturing practice environment.
ivthome.com

Anda mungkin juga menyukai