Back to the Top
Hi, I am a graduate student in Pharmaceutics. I am studying a course
in Non-parametric statistical methods. I am curious to know if
anybody can suggest me some examples in Pharmacokinetics where
non-parametric statistics has been applied. Please also suggest any
possible examples in the PK area where these methods may be needed.
Regards to all.
Rahul Pradhan
Back to the Top
The following message was posted to: PharmPK
Dear Rahul,
Nonparametric statistics are employed in pkpd related areas reasonably
frequently but not as frequently as parametric statistics. I have seen
their use in:
1. Statistical testing to compare two distributions.
Example: Kolmogorov-Smirnov test
2. Population PKPD analyses.
Example: http://lapk3.hsc.usc.edu/lapk/
3. Frequency calculations for complicated scenarios, where the probability
is determined by shear computational force, especially when the underlying
statistical distribution is under question.
Example: Re-randomization/permutation tests, nonparametric bootstrap, etc.
Regards,
Joga Gobburu,
Pharmacometrics,
CDER, FDA
Back to the Top
[Two replies - db]
From: BEDDING_ALUN.aaa.LILLY.COM
Date: Mon, 15 Oct 2001 08:38:15 +0100
To: david.at.boomer.org
Subject: Re: PharmPK
Dear Rahul,
The most obvious use of non-parametric methods is in the analysis of
the time to maximum concentration (tmax), which is dependent on the
sampling interval and therefore rarely fufils the assumptions for a
parametric assessment.
A useful reference for the non-parametric assessment in a
bioequivalence study is:
Hauschke D, Steinijans VW, and Diletti E (1990) A distribution free
procedure for the statistical analysis of bioequivalence
studiesInternational Journal of Clinical Pharmacology, Therapy and
Toxicology, 28, 72-8
Kind regards,
Alun Bedding
Senior Statistician, Clinical Pharamacology Regulatory and Scientific Expert
Eli Lilly
---
From: "Federico Lerner, MD"
Date: Mon, 15 Oct 2001 09:55:05 -0300
To: david.at.boomer.org
Subject: RE: Applications of Non-parametric methods
The following message was posted to: PharmPK
Dear Rahul,
Non-parametrics statistics are appliable for the analyses of non-nermal
metabolites Pk profile. This is applicalbe in cases of different
metabilizers phenotype, fluoxetine as example. This is because at least two
different subpopulations are present.
Regards
Federico Lerner, MD
Back to the Top
The following message was posted to: PharmPK
Dear Rahul,
Nonparametrics are mandatory for measurements originating from discrete
distributions (e.g., tmax). This is also mentioned in the current EU-guidance
(July 2001) for the investigation of bioavailability and bioequivalence, which
can be obtained from EMEA:
http://www.emea.eu.int/pdfs/human/ewp/140198en.pdf
If you plan and evaluate a confirmatory study solely on nonparametrics (yes,
AUC, cmax,...), everything is fine too (our sponsors got approvals with about
300 BA/BE-studies).
Some more references:
Steinijans, V.W. and D. Hauschke;
Update on the statistical analysis of bioequivalence studies
Int. J. Clin. Pharm. Ther. Toxicol. 28, 105-110 (1990)
Vuorinen, J. and J. Turunen;
A Simple Three-Step Procedure for Parametric and Nonparametric Assessment of
Bioequivalence
Drug Information Journal 31/1, 167-180 (1997)
Nonparametrical methods may also be useful in all areas, where a
formal prove of
distributional assumptions may either not be possible, or lack of statistical
power (e.g., if sample sizes are below appr. 30). The assymptotic efficacy of
many nonparametric methods is 3/pi=95% if the underlying distribution
is normal.
If distributions are 'heavy tailed' (which is actually very often the case in
the 'real world'), nonparametric methods will perform superior to their
parametric analogues.
Yet another reference:
Abebe, A., Crimin, K. and J.K. McKean;
Rank-Based Procedures for Linear Models: Applications to Pharmaceutical Science
Data
Drug Information Journal 35/3, 947-971 (2001)
They have also run a very nice web-site, were you can input your own data...
http://www.stat.wmich.edu/slab/RGLM/
Best regards,
Helmut Schuetz
Head biometrics
Biokinet GmbH
Nattergasse 4
A-1170 Vienna/Austria
tel +43(0)1 4856969-77
fax +43(0)1 4856970-90
email helmut.schuetz.aaa.chello.at
Back to the Top
Dear Rahul:
Yes, nonparametric (NP) methods are very useful in population
PK/PD modeling, as they are not constrained by any assumptions of
normality or lognormality in the parameter distributions. Because of
this, they can detect unsuspected subpopulations such as fast or slow
metabolizers. The methods are also consistent - that is, the more you
sample from the population, the more the results approach the true
results.
NP models are also best suited for acting on the basis of the
population information and data, using the new method of "multiple
model" dosage design, which evaluates the predicted precision with
which any dosage regimen with achieve a desired target goal, and
specifically designs the regimen which achieves the goal with maximal
precision. This is especially useful then the parameter distributions
are not Gaussian, but have genetic polymorphism, as so many do.
More info is on our web site www.lapk.org, under teaching
topics, and under old workshop sand presentations. You can download a
lot of stuff there.
Very best regards,
Roger Jelliffe
Roger W. Jelliffe, M.D. Professor of Medicine, USC
USC Laboratory of Applied Pharmacokinetics
2250 Alcazar St, Los Angeles CA 90033, USA
Phone (323)442-1300, fax (323)442-1302, email= jelliffe.-a-.hsc.usc.edu
Our web site= http://www.usc.edu/hsc/lab_apk
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (david@boomer.org)