Back to the Top
The following message was posted to: PharmPK
To anyone who can provide help:
I would like to know if there are recommendations for how to handle
assay values from pharmacokinetic samples that have been re-assayed? We
have a guideline on how to decide which samples should be re-assayed, but
no guide to what to do with the numbers once we get them back. Some
possibilities are: 1) if "significantly" different from the original then
take the re-assay value for PK analysis; 2) if not "significantly"
different from the original, then take the original value: 3) if
somewhere in-between then take the average. As you can see I am unclear
as to how to decide what constitutes significance here. Of course if
more than two re-assays have been done on a given sample then the
decision is more complicated and may involve taking the median value
rather than the mean or looking for clustering in the re-assays.
--
Dan Combs
Clinical and Experimental Pharmacology and Pharmacokinetics
Genentech Inc.
voice: (650) 225-5847
fax: (650) 225-6452
email: combs.dan.-at-.gene.com
Back to the Top
The following message was posted to: PharmPK
IF the re-assay value is significantly different from the initial
assayed value you could not select either. You woulld need a third
analysis and pray that it agrees with one of the earlier ones. If it
does not, you have three different values and a real problem. If there
are reasons that provide justification for rejecting the value, from the
PK or form the analysis, it might be best to do so rather than
re-analyze.
Back to the Top
[Two replies - db]
From: Norma Rohde
Date: Mon, 22 Oct 2001 07:45:05 +1000
To: david.-a-.boomer.org
Subject: RE: PharmPK What to do with re-assay results
The following message was posted to: PharmPK
Daniel,
I would suggest looking at the FDA guidance on dealing with Out of
Specification results - in your case it's more out of trend than
'specification' presumably. From your method validation you should know
what variance you expect. If the numbers are within that range then there
is no down-side to averaging. If not, then there needs to be a reason to
eliminate an assay result - lab error, sample handling(?), equipment fault,
other(?) and that should be noted. The question needs to be asked why you
were re-assaying the sample - it's generally easier to do the checking
before resorting to reassay.
Norma Rohde
AMRAD Operations Pty. Ltd.
---
From: Russell Reeve
Date: Mon, 22 Oct 2001 10:23:16 -0400
To: david.-at-.boomer.org
Subject: RE: PharmPK What to do with re-assay results
Dan,
Depends on the reason for the re-assay. If re-assaying because your SOP
demands that x% randomly picked samples are re-assayed, the best answer is
the average, assuming both assays passed QC checks. If there was an issue
with that sample (possibly bad pipetting, QC sample out of range) and if the
difference is large, use the second. If the difference is not large, use the
average. Using only the first makes sense only if the second assay is
suspect. The reason: Let X1 be the result from the first assay, X2 be the
result from the second, and assume the assay process was the same for the
two assays. Using the first only yields Var{X1} = v; using average yields
Var{(X1+X2)/2}=v/2, while E{X1} = E{(X1+X2)/2}. Using average doesn't bias
you, but gives a more precise estimate of analyte "amount".
How to tell if significantly different? If |(X1-X2)/sqrt(2*v)| > 1.96, then
they're significantly different at the 5% level.
Russell Reeve, PhD
Scientific Software Strategist
email: rreeve.at.pharsight.com
phone: 919-852-4615
Back to the Top
The following message was posted to: PharmPK
Hi
One more comment. In addition to Russell's comment below to use the
average of the analytes (presuming of course that you have ruled out
some obvious lab/phlebotomist error etc).
> Depends on the reason for the re-assay. If re-assaying
> because your SOP demands that x% randomly picked samples are
> re-assayed, the best answer is the average, assuming both
> assays passed QC checks.
The alternative is to use both assay results. If you model the data
then the residual variance model can be structured so that the
information from both samples are used. This has advantages over using
the average as the model is not restricted to weighting both samples
equally (using the average automatically implies that each sample
provides equal information). I hope this is of interest.
Regards
Steve
Stephen Duffull
School of Pharmacy
University of Queensland
Brisbane, QLD 4072
Australia
Ph +61 7 3365 8808
Fax +61 7 3365 1688
http://www.uq.edu.au/pharmacy/duffull.htm
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (david@boomer.org)