Dear colleagues,Back to the Top
I would appreciate your advice on these BE issues - what is the best
option to consider if the confidence interval for Cmax and AUC fall slightly
above the upper limit of the 80-125% e.g. 130% ? How about if it falls
slightly below the lower limit e.g 78% ? For our bioequivalence (BE)
studies, we have standardized our sample size - 18 normal volunteers are being
used in all our BE studies. By the way, we normally use the randomized 2-way
cross-over single dose-design for most, if not all our BE studies.
Thank you.
Joy
Joy,Back to the Top
My view is that BE is not like handgrenades. Close does not cut it.
If you are not entirely within 80 to 125% then you have failed.
Probably not what you wanted to hear.
Pete bonate
Peter L. Bonate, PhD, FCP
Director, Pharmacokinetics
ILEX Oncology
4545 Horizon Hill Blvd
San Antonio, TX 78229
phone: 210-949-8662
fax: 210-949-8219
email: pbonate.-at-.ilexonc.com
If the BE study is being performed for US regulatory purposes, any CIBack to the Top
limit outside the bioequivalence interval indicates failure to
demonstrate equivalence. Assuming the appropriate statistical analyses
have been performed to establish the confidence intervals in the
existing data, there isn't much one can do with the existing data. If
the test to reference mean ratios are near one, the formulations may in
fact be equivalent, but the study was not powered to demonstrate this.
If the test formulation is believed to be equivalent to the reference
one could run an additional study with more subjects in order to power
the study appropriately for the variability of the specific compound.
Standardizing BE studies to 18 subjects in a two-way crossover design
does not acknowledge the range of intrasubject variability which exists
across compounds and the need to increase sample size to appropriately
power studies for products with greater variability.
Joel
Joel S. Owen, Ph.D.
Director PK/PD
Cognigen Corporation
395 S. Youngs Road
Buffalo, NY 14221
(v) (716) 633-3463 ext. 247
(f) (716) 633-7404
(e) joel.owen.-a-.cognigencorp.com
http://www.cognigencorp.com/
Dear Joy,Back to the Top
probably you just standardized the wrong thing.
The sample size in BE studies depends on following variables:
1) acceptance range for any given parameter (AUC, Cmax,...)
2) intrasubject variability
3) expected deviation of test from reference
4) alpha-risk
5) beta-risk (the chance to fail to show BE if it really exists)
where
1) generally 80% - 125%, but may be set in some jurisdictions narrower
or wider based on clinical grounds (safety, efficacy)
2) is obtained from literature or a reasonably sized pilot study
3) generally is set to \0xB15% (or an estimate from a pilot study; in
Canada \0xB10%!)
4) fixed to 5% (for a patient = 10% for the population)
5) power = 1 - beta; generally set to 80% or 90%
If your confidence interval is *outside* the a-priori set acceptance
range, but the point estimate is *within*, your study is underpowered,
i.e., you have inclosed fewer subjects than need to show equivalence.
Therefore it's a bad idea to *standardize the sample size* for any
arbitrary number of e.g., 18 subjects.
Best regards,
Helmut
Helmut Sch\0xB8tz
BABECon Biokinet GmbH / Dept Biostatistics
Neubaugasse 36/11 Nattergasse 4
A-1070 Vienna/Austria A-1170 Vienna/Austria
tel/fax +43 1 9713935 tel +43 1 4856969 62
no cell phone ;-) fax +43 1 4856969 90
http://www.goldmark.org/netrants/no-word/attach.html
Dear Joy,Back to the Top
I am probably wrong, but...
By saying "...we have standardized our sample size - 18 normal
volunteers are being used in all our BE studies...",
do you mean you are conducting all your BE studies with 18 subjects,
whatever the tested drug ?
If it is well the case, you are very likely to get unconclusive results
(point estimate within the BE range but limits of the 90% CI outside
this range) from a lot of your BE studies.
For example, considering a quite common situation with an intra-subject
CV for a drug equal to 25% and with a true T/R ratio for the parameter
of interest actually equal to 0.97, the power to demonstrate the BE
using a 0.8-1.25 interval is only 64%, meaning that 36% (i.e. more than
one third!) of the studies performed with 18 subjects will not be able
to conclude to the bioequivalence.
On another side, if you include 32 subjects, the chance to demonstrate
the 0.8-1.25 bioequivalence rises up to 90%.
Normally, the sample size should be calculated a priori, based on the
intra-subject variability observed in previous studies or found in the
litterature, on the expected T/R ratio and on the target power
(generally >80%).
This sample size is then to be justified in the protocol.
This is clearly stated in both US (Guidance for Industry - Statistical
Approaches to Establish Bioequivalence: \0xDF V.B and Appendix V) and EU
(Note for Guidance on the Investigation of Bioavailability and
Bioequivalence: \0xDF 3.1 Design) guidelines.
For sample size calculations, there are some references in the
litterature (e.g. Diletti et al. 2001, Int J Clin Pharmacol Ther
Toxicol, 29:1-8) or you can also consider the use of dedicated
softwares (e.g. nQuery Advisor).
Hope this helps,
Fabrice
Fabrice Nollevaux, M.Sc.
Senior Biostatistician
SGS Biopharma - Wavre - Belgium
www.sgsbiopharma.com
HiBack to the Top
I just want to know that is it necessary to have 90% confidence
intervals for Tmax for an immediate release dosage formulation to
establish bioequivalence. The guidelines says 90% confidence intervals
of the geometric mean ratio of three PK parameters i.e AUC(0-t) ,
AUC(0-infinity) and Cmax between the test and the reference fall
within 80-125%.
Kindly suggest
Regards
Yagna
Research officer
Glenmark Pharmaceuticals Ltd.,
Nashik.
Maharastra
INDIA
Back to the Top
Dear Yagna
Unlike CI on logtransformed data of Cmax, AUCt & AUCinf, statistical
test for Tmax to be done on nontransformed data by wilcoxin ranken mean
test, whose explanation is as follows:
Many biological data correspond more closely to a log normal
distribution. AUC and Cmax tend to be skewed and their variances
increase with the means.
Log transformation makes the variances independent of the mean and the
frequency distribution is made more symmetrical. The third measure of
bioavailability, Tmax, poses a somewhat different problem.
What makes Tmax different is the discreteness of its measurement, as
well as the fact that it is measured with an error that will depend on
the study's sampling times. While Hauck and Anderson contended that
none of the statistical procedures available for use in the analysis of
bioequivalence studies appear to be appropriate for Tmax, Westlake has
suggested that a C.I. on the difference of the Tmax for standard and
test formulations following an ANOVA on the untransformed data seems
appropriate.
However, the regulatory agencies recommend that differences in Tmax be
evaluated by non-parametric tests (Wilcoxon signed rank test, Wilcoxon
rank sum test) on the untransformed values (additive model).
Nonparametric methods for estimating a population median include the
Sign test and the Wilcoxan Signed-ranks test.
Thanking you and Regards
Vishwottam K N
Senior Scientist - Biopharmaceutical Research
SUVEN LIFE SCIENCES Ltd
Serene Chambers, Road # 7,
Banjara Hills, Hyderabad 500 034
Phone : 9140 2355 6039, 3092 1169
Fax : 9140 2354 1152
Back to the Top
Yagan,
Depending the FDA may request CI on Early Exposue (see below from
Guidance
on BA/BE)
For orally administered immediate-release drug products, BE can
generally be
demonstrated by measurements of peak and total exposure. An early
exposure
measure may be informative on the basis of appropriate clinical
efficacy/safety trials and/or pharmacokinetic/pharmacodynamic studies
that
call for better control of drug absorption into the systemic circulation
(e.g., to ensure rapid onset of an analgesic effect or to avoid an
excessive
hypotensive action of an antihypertensive). In this setting, the
guidance
recommends use of partial AUC as an early exposure measure. We recommend
that the partial area be truncated at the population median of Tmax
values
for the reference formulation. We also recommend that at least two
quantifiable samples be collected before the expected peak time to allow
adequate estimation of the partial area.
Art Straughn
Department of Pharmaceutical Sciences
College of Pharmacy
University of Tennessee
HiBack to the Top
Please interpret the following
"Test and reference product are equivalent with respect to the extent of
absorption but not with respect to the rate of absorption".
This was the case with an immediate release dosage form where 90%
confidence
intervals of AUC(0-t) , AUC(0-infinity) and Cmax between the test and
the
reference were well within 80-125%.But Tmax slightly differed, in
connection
with the same I would like to ask what is the criteria to establish Tmax
(for an immediate release dosage formulation) to be bioequivalent.
Warm Regards
Yagna
Research Officer
Glenmark Pharmaceuticals Ltd.,
Nashik,
Maharastra,
INDIA.
Back to the Top
"For T max - no criteria exist," but it should be comparable to the
reference product that means T max of Test product should not be
differed
much from Reference.
You can refer the USFDA guideline " Food effect Bioavailability and Feb
Bioequivalence studies - Page no 8" for further clarification.
Thanxs
KUM Prasad
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (david@boomer.org)