Back to the Top
The following message was posted to: PharmPK
Dear PharmPK user,
The number of patients involved in our phase III trial is huge. The
proposed sampling scheme is sparse. however, it is impossible to
collect samples from all patients ($$?). We want to randomly select a
subpopulation to collect PK samples. I was wondering what is the
"typical" number of PK samples and subjects reported in phase III
PopPK study?
We have already developed the Pop PK/PD model based on the Phase I
and II study. According to FDA popPK guidance, the sample scheme can
be multiple trough points or optimal sample points. I was wondering
which one is better.
Your suggestion will be highly appreciated!
Regards,
Mei
Back to the Top
The following message was posted to: PharmPK
:Mei,
If the population of patients cannot be separated by a particular
covariate
into subgroups, an informative sparse sampling scheme can be applied
to a
sub-sample of the population to obtain PK samples that would enable the
characterization of the pop. PK of the drug. All study sites might
not have
the ability to do PK sampling. Thus, PK sampling could be done mostly in
those sites that can offer such a service. You would want to be
certain that
the sites where samples are collected are not different in some
significant
way from sites where samples are collected. Thus, if you are
collecting data
at multiple international sites, the various regions of the world
should be
represented.
If for instance the study population is 400, 200 subjects could be
selected
for the estimation of the pop. PK of the drug. Sample sizes from 100 and
above have been shown to yield efficient estimates of pop. PK
parameters if
the inter-subject variability is moderate (approx. 45%CV). It would be
advisable to use 200 subjects for your PK if the total sample size is
400
and 150 if the sample size is 300. If the sample size is 200 I'd
suggest you
sample all subjects.
A simple and pragmatic approach that can be used to develop an
informative
sampling scheme for efficient pop. PK parameter estimation is the
informative profile (block) randomized (IBR) design developed by Ette
et al.
in 1994. The approach combines information theory (D-optimality) and
randomization enabling the characterization of the population profile
and
protects against model misspecification. This approach was compared
with the
population Fisher information approach in a recent publication and
found to
perform similarly with the latter if not better. It is easier to
implement.
It involves obtaining informative times with average population
parameters
using the SAMPLE routine in the ADAPT software and creating sampling
windows
around the informative times. Subjects are then assigned to sample times
such a population PK profiles is created when the samples are
obtained. Here
is a list of publications that will be helpful in implementing this:
.. Roy A, Ette EI A Pragmatic Approach to the Design of Population
Pharmacokinetic Studies. AAPS J 2005; 7 (2): E408 - E420.
.. Fadiran EO, Jones CD, Ette EI. Designing population pharmacokinetic
studies: performance of mixed designs. Eur J Drug Metab Pharmacokinet
2000;
25: 231 - 239.
.. Ette EI, Sun H, Ludden TM. Balanced designs in longitudinal
population pharmacokinetic studies. J Clin Pharmacol 1998; 38: 417 -
423.
.. Jones CD, Sun H, Ette EI. Designing cross-sectional pharmacokinetic
studies: implications for pediatric and animal studies. Clin Res Regul
Affairs 1996; 13 (3&4): 133-165.
Finally, the advantage to obtaining a sample from all subjects is
that if
the trial has an unexpected outcome the concentration-time profile
(exposure) can be reconstructed for each patient. Thus if the unexpected
outcome is a function of exposure it may be detected or elucidated.
Cheers!
Paul J. Williams, PharmD, MS, FCCP, FCP
Anoixis Corporation
1918 Verdi Ct
Stockton, CA 95207-5318
Back to the Top
The following message was posted to: PharmPK
Paul
I was interested in your discussion of popPK sampling.
> A simple and pragmatic approach that can be used to develop
> an informative sampling scheme for efficient pop. PK
> parameter estimation is the informative profile (block)
> randomized (IBR) design developed by Ette et al.
> in 1994. The approach combines information theory
> (D-optimality) and randomization enabling the
> characterization of the population profile and protects
> against model misspecification. This approach was compared
> with the population Fisher information approach in a recent
> publication and found to perform similarly with the latter if
> not better. It is easier to implement.
I was hoping to elucidate your meaning on the above points. (I have not
[unfortunately] as yet read the publication that you quoted.).
Nevertheless... It seems you might have used Adapt to compute the
Fisher
information matrix? I understand that Adapt is not capable of
handling a
full population information matrix - can you confirm this? Have you
compared the results obtained in your IBR versus those obtained with the
full popn information matrix (say using WinPOPT, PFIM_OPT, POPED or some
other software)? You suggest that IBR was better - can you share a few
thoughts on how it was better?
As a note - determination of sample size for a popPK, popPKPD or popKPD
study can be evaluated easily using programs that compute the population
Fisher information matrix (PFIM).
Cheers
Steve
--
Professor Stephen Duffull
Chair of Clinical Pharmacy
School of Pharmacy
University of Otago
PO Box 913 Dunedin
New Zealand
E: stephen.duffull.-a-.otago.ac.nz
P: +64 3 479 5044
F: +64 3 479 7034
Design software: www.winpopt.com
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (david@boomer.org)