Back to the Top
I will appreciate it if anyone could elaborate on the simulation needed
for the sample size, or point to a good reference. I have been reading
the CDER 1999 August guidance on BE study, and sort of got the feeling
that the average approach now has to be replaced by population/individual
approach. Is it correct? Thanks in advance.
Jun
Back to the Top
[A few replies - db]
Date: Wed, 26 Apr 2000 00:38:22 -0400
From: Laszlo Endrenyi
X-Accept-Language: en
To: PharmPK.-at-.boomer.org
CC: Multiple recipients of PharmPK - Sent by
Subject: Re: PharmPK Sample size estimate for Population/Individual approach
This is only a draft guidance and is marked as "not for implementation". FDA
is currently reviewing the comments that it has received. So, evaluation of
average bioequivalence is still the basis for submissions. This may well
change in the future if and when the guidance is implemented.
Laszlo Endrenyi
University of Toronto
---
Date: Wed, 26 Apr 2000 01:44:21 -0700 (MST)
X-Sender: ml11439.aaa.pop.goodnet.com
To: PharmPK.at.boomer.org
From: ml11439.at.goodnet.com (Michael J. Leibold)
Subject: Re: PharmPK Sample size estimate for Population/Individual approach
Jun,
The concept of statistical power is one way statistical computation
deals with the sample size. The probability of making a type I error
(i.e. detecting a significant treatment effect when there is none) is
considered less than .05 when alpha= .05. The probability of making a
type II error (not detecting a significant treatment effect when one
does exists) is beta. The chance of detecting a significant treatment
effect when one exists is equal to (1-beta) which referred to as
statistical power. (2)
Statistical power for a given statistical test can be estimated
from: 1)the standard deviation of the measured variable (treatment effect)
in the population of interest, 2) the size of the change in the measured
variable (treatment effect), 3) the level of significance chosen (alpha=.05),
4) the nature of the statistical test, and 5)the sample size. This power
analysis can be accomplished with the above information and a power chart.
The same power chart can also be used to estimate the necessary sample size
to obtain an acceptable level of statistical power, which should be in the
area of .80-.90. (1) Some texts also provide the formulae useful for
calculating statistical power and sample size.(2)
One standard example is detecting the effects of diuretics in a
population. If the standard deviation in urine ouput in the population
of interest is 200ml/day, and the expected change in urine output is
200ml/day with diuretic treatment, then the ratio phi= delta/sigma =1.0
is used with the power chart appropriate for the statistical test to
determine the statistical power.(1)
Mike Leibold, PharmD, RPh
ML11439.aaa.goodnet.com
References
1) Glantz, S.A., Primer of Biostatistics 4th ed., New York, McGraw-Hill
1997: 151-183
2) Freidman, L.M., Furberg, C.D., DeMets, D.L., Sample Size in:
Fundamentals of Clinical Trials, Littleton, PSG :Publishing Company
1985:83-107
---
X-Lotus-FromDomain: QUINTILES
From: peter.bonate.-at-.quintiles.com
To: PharmPK.at.boomer.org
Date: Wed, 26 Apr 2000 07:39:31 -0500
Subject: Re: PharmPK Sample size estimate for Population/Individual
approach
Jun,
The proposed guidance on BE has not been finalized. Word on the street is that
there was so much criticism to the original version that the FDA is completely
overhauling the guidance. To gain a good understanding of this approach see
Chen and Liu's second edition of their book.
Pete Bonate
Quintiles
Back to the Top
Thanks for your help. I do have the Chow and Liu's book pulished around
either 1991 or 1992, although I did not find much elaboration on the topic
of individual or population approach. I have also noticed Mr. Endreyni's
long list of publication in the draft guidance reference section. I will
definitely try to net the articles. I hope I did not misuse this PK
oriented forum by asking too much statistical question, please protest or
recommend a Pharmaceutical/clinical trial forum/listserv, I will
appreciate it in either case.
Now suppose estimating sample size for a BE study is of interest, as
specified in the 1992, a two one sided ttest procedure will be used to
test the ratio of averages, now do people even try to log transform
available AUC or Cmax data from previous studies before they work out the
inter and intra subject coefficient? I have seen example of doing the
sample size estimate on the original scale, but analyze AUC and Cmax on
the log scale, is it what people are doing? BE statistical analysis is
new to me, my major is Biometry and statistics, please bear with my
ignorance. Thanks a lot.
Jun
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (david@boomer.org)