Back to the Top
Dear members,
Googling on bioequivalence, I found the following link:
http://findarticles.com/p/articles/mi_qa3833/is_199907/ai_n8856521/pg_4
[Direct link to the full article
http://www.ajpe.org/legacy/pdfs/aj630215.pdf
- db]
In this article, the author states that the FDA is considering a draft
guidance to address the concern about confidence intervals for
bioequivalence testing of highly variable drugs that have been defined
as those demonstrating intrapatient variabilities greater than 30
percent in metrics of the extent and rate of absorption (e.g., AUC and
Cmax).
On the FDA site however, I can not find any info about this subject.
Does anyone know more about this?
Kind regards Bert
Back to the Top
The following message was posted to: PharmPK
Hi Bert,
The FDA has been talking about HVDs for some time. Here is a
(somewhat) recent key article:
Sam H. Haidar, Barbara Davit, Mei-Ling Chen, Dale Conner, LaiMing Lee,
Qian H. Li, Robert Lionberger, Fairouz Makhlouf, Devvrat Patel, Donald
J. Schuirmann, and Lawrence X. Yu. "Bioequivalence Approaches for
Highly Variable Drugs and Drug Products". Pharmaceutical Research,
Vol. 25, No. 1, January 2008.
The article comes with the proviso that it doesn't necessarily reflect
the views or policies of the FDA, but the author list reads like an
FDA who's-who. Various approaches are discussed. If my interpretation/
description of them is off, I apologize in advance, please see the
article itself for a first-hand description.
A) Direct expansion of BE limits (like the EMEA does, e.g. 75-133% for
the 90%CI of Cmax)
B) Expansion of BE limits based on fixed sample size:
A BE study is run comparing the reference product with the reference
product, with a given sample size. The resulting 90%CI would act as
the goalposts for a BE study with the test product vs. the reference,
using the identical sample size. This would require 2 BE studies to be
run, and in addition if you had more drop-outs than the original
study, this could jeopardize the use of the 90%CI (recall that the
90%CI usually shrink as n increases). My hunch is that industry
wouldn't be nuts about this approach, but on the other hand, it makes
wonderfully inherent sense since the reference product should by
definition be bioequivalent to itself. Question: how do you define the
sample size? If I were a generic company, I would say the smaller the
better, because that would give me wider goalposts.
C) The BE limits are widened based on the variability of the reference
drug. There are three methods proposed:
Method 1: Using the within-subject standard deviation and a scaling
factor:
[Lower, Upper] = exp(+/-k*sigma_wr), where k is the p-th percentile of
the standard normal distribution, and sigma_wr is the within-subject
standard deviation.
Method 2: Scaled BE approach using the within-subject variance
(uT - uR)^2 / sig_wr^2 < theta
Method 3:
[(uT - uR)^2 + (sig_wt^2 - sigwr^2) + sigD^2] / sig_wr^2 < theta
It's a test of the difference, where sigD^2 is the subject-by-
formulation interaction, and theta is the BE limit.
D) Expansion of BE limits based on sample size and scaling:
[Lower, Upper] = exp((+/- (talpha,2n-2 + t(beta/2),
2n-2)*(n^-0.5)*sig_wr)
Somewhat more complicated, this method relies on the power of the
study and the Type I error, the sample size, and the within-subject
standard deviation of the reference. I'm not really used to beta
coming into BE limits considering it is usually only used to project
sample sizes for the next study. Let's calculate what the BE limits
might look like for a drug with 80% power to determine BE with a ratio
between 95-105%, an alpha of 5%, a sample size of 66(=2n) subjects in
a crossover study, and a reference intraCV of 40% (note, 66 subjects
should provide 80% power with regular BE limits 80-125%):
sig_wr = 0.385 (=SQRT(LN((IntraCV)^2+1))
t0.05,64 = 1.669013
t0.20,64 = 1.29492
BE limits: [81.984-121.975%]
(I hope I'm doing the math right, t-values and me don't like eachother
much).
Given the exact same drug, run with 50 subjects instead of 66:
sig_wr = 0.385 (=SQRT(LN((IntraCV)^2+1))
t0.05,64 = 1.701131
t0.20,64 = 1.312527
BE limits: [74.099-134.955%]
I might be off in left field here, it looks like the less subjects you
have in your study, the better (however, the 90%CI would widen as well
as n is reduced). The paper goes on to define a 3-way replicate as the
design of choice for highly variable drugs (IntraCV > 30%), using C
(Method 2) as the preferred approach based on simulations.
B, C, and D have the common feature that in order to scale the BE
limits by using the within-subject variance, a good estimate of
variance is required, and it is obtained from the study itself. On one
hand, it makes total sense. If a drug is really variable, then why
can't more "slack" be given to the generic product? On the other hand,
the "little regulator" in me doesn't like it. If I really want to pass
my drug and the reference is highly variable, every mistake that
happens in the clinic works in my favour; i.e. anything that happens
that increases the variability of the reference drug in the trial will
help the test product pass. Devil's advocate, every mistake in the
clinic should have a paper trail and end up as a deviation. Many of
these deviations end up being reported as "no clinical relevance" or
"not expected to impact the results of the study." However, with BE
limits scaled using the variability of the reference product, there
seems to be a
special reward for Subject 9 not completing their entire meal in
Period 2, or Subject 35 taking a concomitant medication which is not
expected to have a pharmacokinetic interaction.
There's another can of worms here: BE limits will be SOP dependent,
because of the way subjects are included/excluded in the PK & stats
analysis. So the same data set analyzed at one company could end up
with different bioequivalence limits when analyzed at another company,
due to differences in data handling, ANOVA modeling, etc. I wonder
what the impact of an anomalous subject would be with the scaled
methods? It just might be the opposite effect to regular studies.
Could it be that if a formulation failure were to occur in the
reference product, the bioequivalence limits would widen to let
anything through? What if you have two studies and the reference
product looks more variable in your second study? They would have
different BE limits. The within-subject variance spit out by the ANOVA
is an estimate of the true variance based on who walks in the door. It
also seems to me that the FDA would have to keep a sharp eye on the
high variability studies to
make sure the BE limits were calculated properly.
I feel that I'm being too critical, I haven't had a chance to run any
simulations myself or think the whole issue through. Plus, I'm still
tripping through the math. My bias would be Method A, because it's
simple and it seems to work well for the EMEA. I don't envy the FDA in
selecting a method, it's no easy task. I just hope that when they come
out with the draft guidance, they include an example analysis run
through it from plasma data to final conclusion.
Hope this helps,
-Dave
--
David Dubins, Ph.D., B.A.Sc.
Global Bioequivalence Consulting
Assistant Professor, Leslie Dan Faculty of Pharmacy
University of Toronto
Back to the Top
The following message was posted to: PharmPK
Thank you Dave for your comprehensive response! A lot of stuff to think
about. I'm working for a veterinary generic company, dealing with farm
animals and EMEA guidelinesJ. Your right, smaller sample size in BE
studies reduces costs and labor. For drugs, not known to us we perform
pilot studies in different target animal species to evaluate
variability, estimate sample size and design an optimal sampling
schedule. In our studies, the number of drop-outs is very limited (no
concomitant medication, no smokers, no alcoholics......)
Also thanks for the article, I will start to read.
Regards Bert
Back to the Top
The following message was posted to: PharmPK
Dear Bert,
see this page and references within:
http://bebac.at/news/2006-10-06.htm
Best regards,
Helmut
Ing. Helmut Schuetz
BEBAC - Consultancy Services for
Bioequivalence and Bioavailability Studies
Neubaugasse 36/11
1070 Vienna, Austria
e-mail helmut.schuetz.at.bebac.at
web http://bebac.at/
contact http://bebac.at/Contact.htm
forum http://forum.bebac.at
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 "CI and BE - FDA guidance" as the subject | Support PharmPK by using the |
Copyright 1995-2011 David W. A. Bourne (david@boomer.org)