Back to the Top
Dear All,
We have two dissolution profiles and want to check their similarity.
We can't use similarity factor technique as %RSD at initial time
points is > 15.
According to FDA a Multivariate approach is to be used. In that we
have to calculate MSD (Multivariate statistical Distance).
Now My Question is Can we use Hotelling T sq. test to check similarity
of two disso. Profiles when Data available is Conc. at 15, 3 0, 60,
90, 120minutes for all units, Mean conc. and %RSD at all time points?
If yes, Please suggest any software for doing this.
Looking for quick reply.
Thanks
Ravi
Back to the Top
The following message was posted to: PharmPK
Dear Ravi,
There are few assumptions related to Hotelling's T sq technique, such
as Multivariate Normality, Independence and homogenity of variance.
Probably the assumption of independence can be a concern as the data
generated in some order of time may be correlated and this may violate
the assumption of indepence. Hotelling's T-sq is generally not robust
to this violation.
Instead of Hotelling's t-sq may be Mahalabonis Distance can be a
better choice.
Regards,
Vikesh S
Back to the Top
The following message was posted to: PharmPK
Dear Vikesh S,
Thanks for reply.
Can U Suggest any software which can be used for calculating Mahalabonis
Distance and data inputs required for calculation?
Thanks
Ravi
Back to the Top
Hi Ravi,
I agree with Vikesh but I have a major concern which is:
As per your data set and objective its better to go for Hotelling T
square only. But for that you need to check for Independence and
normality of data. I think it wont take much time and won't require
much expertise (Can use SAS) and if it follows both the conditions you
can easily proceed. Also there is not possibilty of dependency in case
of data set you are having. The reason that I am not suggesting you
for Mahaloibis distance that Its quite complicated and the kind of
data set you are havine will not suffice the objective. Actually your
data set is kind of two sample data set and in this case it is
preferable to go for Hottteling t square only.
Best regards,
Nand Kishore Rawat
Email : nand.rawat.aaa.novartis.com
Back to the Top
Dear Ravi,
You might want to visit the www site:
http://www.uef.sav.sk/advanced.htm
where You can read our paper No.16, describing a new method for
testing similarity of dissolutions and comparing the new method with
a method based on the similarity factor f2.
Using the new method, you can test similarity of both: amounts of drug
dissolved, rates of drug dissolution (as functions of time).
A version of the software used in the given study is downloadable from
the same site.
With best regards,
Maria Durisova DSc (Math/Phys)
www.uef.sav.sk/durisova.htm
Back to the Top
The following message was posted to: PharmPK
Hi Ravi,
There is a statistic known as the Kolmogorov-Smirnnov statistic which
measures the distance between two curves. I am probably spelling the
two names incorrectly and describing what it does incorrectly and/or
incompletely. It is usually used when comparing two groups. Assume two
groups with 10 values each. Curves for each group are prepared with
proportion of values from 0.0 to 1.0 as the y-axis and the observed
values as the x-axis. The K-S statistic is applied to these curves to
find how far apart they are.
I do not know if the K-S statistic can be considered the Multivariate
Statistical Distance that FDA requires.
Regards,
Stan Alekman
[See
http://en.wikipedia.org/wiki/Kolmogorov-Smirnov_test
- db]
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 "Similarity of Dissolution Profiles" as the subject | Support PharmPK by using the |
Copyright 1995-2011 David W. A. Bourne (david@boomer.org)