Back to the Top
I would like to know your feedback
about the explanations to the ANOVA
output of bioequivalence analysis.
For a bioequivalence study using two-period,
two-sequence crossover design, if one of
the effects shows significant difference
after the ANOVA analysis, my thought is:
Sequence: two groups of subjects were
not from a population;
Period: carry-over effect;
Formulation: two formulations were not
Your comments on the explanations
are much aprreciated.
Back to the Top
>The FDA has a GUIDANCE paper "STATISTICAL PROCEDURES FOR BIOEQUIVALENCE
STUDIES USING A STANDARD TWO-TREATMENT CROSSOVER DESIGN Was printed in
July 1992. Mostly based on work by Schuirmann DJ. I have also run into the
problem of non equivalence by ANOVA but practically there was no difference.
Essentially you reverse the null hypothesis from there is no difference to
the difference lies within an a priori determined range which is acceptable
and the alternate is the difference is greater than this range, and if not
proven different, then accepted as equivalent. The FDA generally agrees
that the range should be between 80 & 125% for the ratio of the products.
This is a gross generalization and the paper was developed specifically for
solid oral dosage form. However, in my experience the principles outlined
have been adopted/adapted by all groups.
The guidance paper has the SAS model outlined to accomplish the analysis.
The parameters of interest need to be selected for analysis with some
rational scientific thought.
Back to the Top
For those wanting more information:
Chow S and Liu J. Recent Statistical Developments in Bioequivalence
Trials -- A Review of the FDA Guidance. Drug Information Journal, Vol. 28,
pp 851-864, 1994
This refers to the original document which I have scanned and am sending as
an attached file (Needs a little more cleaning up but the section on SAS
procedure seems intact)
[db - not attached]
The underlying thought is that, traditionally the researcher sought to
compensate for bias by trying to prove the opposite of what their belief was
and if not successful, then the researcher accepted the alternate. In
bioequivalence testing, the researcher has a bias in favor of equivalence
and this is not considered in standard null hypothesis approaches. As I
have stated previously, the null hypothesis is "reversed" by trying to prove
there is a difference, that is non-equivalence, and if not successful, then
the alternate of equivalence is accepted. This somewhat negates the
researchers bias but more practically, reduces the 'n' needed for power.
If you are talking to the FDA & they refer to 90% confidence intervals, this
is where they are coming from even if it doesn't pertain to solid oral
dosage forms. Note also: WinNonlin's new PRO package has this methodology
built in (I don't have any relationship with WinNonlin & prefer to do my own
laundry if possible -- but the windows 95 environment with cut&paste makes
me love some of the features of the program -- convince wins in the end).
Thought provoking question of the day:
Is the 95% confidence interval more or less restrictive than the 90% CI ?
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (firstname.lastname@example.org)