Back to the Top
Target 10.0
Bias
S1 7.75 -22.5
S2 7.92 -20.8
S3 9.98 -0.2
X 8.55
SD 1.24
CV 14.5
BIAS -14.5
I would appreciate input using the data above. The target is 10.0 ng/mL, CV and Bias at |20.0|. An
interpretation of the EMEA guidance, and several Sponsors insistence on nearly the same, would
indicate that the experiment- say a time point stability measure- would pass. I have a different
opinion and would suggest that it failed since two of the three measures did not meet acceptance.
I am interested in the opinion and justifications offered for that opinion of others.
Edward F. O'Connor
efoconnor.aaa.gmail.com
www.aegisbioconsult.com
Back to the Top
Dear Ed,
Regarding a situation where a majority (2 of 3) individual results in a replicate evaluation failed,
but the mean %bias and %CV passed, this is almost identical to a real situation I faced a few years
ago myself as the Sponsor when a CRO presented similar data (n = 6, instead of n = 3) in which a
majority of individual results failed but the means and %CVs were just inside acceptability.
The CRO maintained that everything was fine, because it met their SOP acceptance, which was based
only on means and %CV, and Guidance did not specifically address it. We reluctantly accepted it,
but in future work (which we terminated soon thereafter with this CRO) agreed by contract that a
minimum number of individual values must pass in such types of analyses that involve replicates.
Consider a similar hypothetical example:
Value Individual Bias
12.2 22%
10.0 0%
7.80 -22%
Mean: 10.0
SD 1.97
%CV 19.7%
%
Technically, the test based on the %bias of the mean and the %CV would pass the target of |20.0|,
but clearly something is not under the control expected for bioanalytical assays in general. For
such an example, whether the results are due to an instrumental fluke, analyst error, or an
inherently variable method, I some sort of investigation and follow up testing would be warranted.
Thomas L. Tarnowski, Ph.D.
Director, Bioanalytical Chemistry
Gilead Sciences, Inc.
ttarnowski1.-at-.aol.com
Back to the Top
Hi Tom: This was a real situation as well, the data has been altered somewhat but the Sponsor
insisted in this case before I could at least suggest using the number of samples passing rather
than the mean. Unfortunately the guidances are not specific or worse and actually encourage this
way of assessing performance. That lack of clarity was behind my asking at the 2013 crystal city
what or which mean to use!
|
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 "Means and means" as the subject |
Copyright 1995-2014 David W. A. Bourne (david@boomer.org)