Back to the Top
I have come across a data which fails to meet the US-FDA criteria for
Cmax (80.00-125%), when one of the subject (with a very high T/R
ratio)is included in the PK and Statistical Analysis. When this
particular subject is excluded from the PK and Stats analysis then
the CI is within the acceptance criteria.
Will anybody let me:
What statistical tests can I use to determine whether that particular
subject in question is an outlier or the T/R ratio is an abberant value
Can I go for a repeat study with the same subject including few
control subjects from the same study. what should be the statement
(or acceptance criteria) in the protocol to prove that the subject in
question has a abberant T/R ratio.
Regards:
Shaikh Feroz Ibrahim
Back to the Top
The following message was posted to: PharmPK
Hi Shaikh,
You asked: What statistical tests can I use to determine whether that
particular
subject in question is an outlier or the T/R ratio is an aberrant value
There are many tests to identify outliers. I like the Grubb's test
because
it tends to identify outliers that look suspect and it only
identifies one
outlier. This prevents misuse of outlier identification. The FDA doesn't
recommend any test in particular, as long as the test is identified
in your
SOPs and/or protocol. A description of the Grubb's test can be found
here:
http://www.graphpad.com/quickcalcs/GrubbsHowTo.cfm
You asked: Can I go for a repeat study with the same subject
including few
control subjects from the same study. what should be the statement
(or acceptance criteria) in the protocol to prove that the subject in
question has a aberrant T/R ratio.
The FDA does not provide pass or fail criteria for the mini study you
are
planning to run. Since you have the freedom to choose, you can use your
judgement so long as you set concrete limits. Either way, come up with
limits beforehand so that upon redosing the subject in question you can
provide an "answer" once you receive their PK data.
One possible scenario is this (and this is just a suggestion):
(1) Whatever statistical test you use to identify the anomalous
subject in
question, run it again on the original data set with the anomalous
subject's mini-study data replacing their original study data. If
they are
no longer identified as an outlier, then they are "normal" looking (i.e.
they belong to the study population) and their data may be removed
from the
original study.
(2) The control subjects are there to confirm the validity of the mini
study. In order for the mini-study to be unbiased they should be
selected
randomly. I would advise though (and here is where I go back on myself)
against selecting a control subject on the very edge of the dataset
without
the anomalous subject. If their redosed data do not fall within the
limits
of the original study dataset without the anomalous subject (say,
plus or
minus an arbitrary safety buffer if you wish), then they cannot
confirm the
validity of the mini-study and further investigations would be required.
Hope this helps,
Dave Dubins
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@boomer.org with "Repeat BE Study" as the subject | Support PharmPK by using the |
Copyright 1995-2011 David W. A. Bourne (david@boomer.org)