Back to the Top
All: I am reviewing calculations to assess incurred sample re-analysis, there are two popular
calculations and I would appreciate your opinions on those two as well as another.
Given: The actual value of the sample is unknown and ISR at this time only require a second
measurement of the sample in a separate plate or run.
The first is: (obs1-obs2)/((obs1+obs2)/2) the numerator is justified
but the denominator seems to put weight on the idea that the actual mean is between the two
numbers.
The second is ((obs2-obs1)/obs1)*100. this approach also puts weight on the initial value.
A third is of course standard deviation normalized to the mean of the two measures or %CV
SD (OBS1+OBS2).
Aside from any political stance, which would be most appropriate? I am leaning to the %CV (or SD
normalized to the mean). But I would appreciate your responses.
Thanks
Ed
Back to the Top
Ed-
Regarding ISR calculation, the one I have seen most often at both Pharma companies and CROs is the
first, with the difference in the 2 values divided by the mean of the two. Since there is no
inherent reason from a purely reproducibility standpoint (i.e., not a stability or degradation
issue), then there is no reason to indicate that the first value has more validity than the second.
At both Roche and Elan in the past, and at numerous CROs we used a similar calculation as part of
various decision trees in similar situations, such as using a second obtained value to confirm the
first as the first reportable value.
Since there are often statistical objections to calculating standard deviations from only 2 values,
the last would be a least-preferred choice.
-Tom
Thomas L. Tarnowski, Ph.D.
Bioanalytical Consultant
ttarnowski1.aaa.aol.com
Back to the Top
You may find that in many regulated Bioanalytical labs, sadly the science of what would be the
correct calculation must come second to what the regulator tells you to do. The FDA just published
a draft of the Bioanalytical Validation guideline that mandates your first calculation, hence that's
what we have to use.
Paul.
--
Paul Hurley
Back to the Top
Thanks Tom: But the SD and CV from two values is a cornerstone in run acceptance where the values
of two wells are used to generate a mean. Why would it be more appropriate to use it in sample
acceptance ( 2 wells, CV<=20%) and less appropriate to use it to compare two values obtained after a
latency of time (2 measures (CV<=20%)?
Back to the Top
Hi Ed,
your first formula (difference divided by mean) is stated in the
"Crystal City 3.5" Workshop Report (Fast et al., The AAPS J 11(2):
238-41, 2009). The same formula is referred in EMA's guideline on
bioanalytical method validation (2011) and FDA's recent draft guidance
(2013).
Helmut
--
Ing. Helmut Schuetz
BEBAC - Consultancy Services for
Bioequivalence and Bioavailability Studies
Neubaugasse 36/11
1070 Vienna, Austria
web http://bebac.at/
forum http://forum.bebac.at/
Back to the Top
Well they are looking for comments and input!
|
Want to post a follow-up message on this topic?
If this link does not work with your browser send a follow-up message to PharmPK@lists.ucdenver.edu with "Sample re-analysis calculation" as the subject |
Copyright 1995-2014 David W. A. Bourne (david@boomer.org)