Back to the Top
Dear All,
I would like to know if anyone has generated pk data using different software? How different is the
data ? Can anyone share this information ?
Regards
Manisha Banavalikar
GM Clinical Research
Anazeal Analytical & Research Pvt. Ltd
Navi Mumbai
[I remember when Winnonlin was being developed it was compared with other programs including Boomer
;-) Peter Bonate has more recently done these studies. It seems that most give much the same result
if the weighting schemes (and models) are equivalent. Some software presents more or less details
and have different ways of presenting the 'answers' - db]
Back to the Top
Dear Manisha,
David is correct. ;-)
An example comparing different packages in PopPK was presented at the
PAGE meeting (Pamplona, Spain 2005):
http://www.page-meeting.org/page/page2005/PAGE2005O08.pdf
Sometimes surprises are possible. One software package failed to give
correct results in bioequivalence (both crossover and parallel designs):
http://dx.doi.org/10.1208/s12248-014-9661-0
http://dx.doi.org/10.1208/s12248-014-9704-6
Helmut
Back to the Top
Dear Manisha,
What do you mean by generating PK data??
WinNonlin, Kinetica etc can be used for that purpose, since the algorithm is same behind these
softwares the data is also pretty much closer
Regards
Harish
Back to the Top
Hi Manisha,
This is something I've been looking into recently as I've been
developing the PKNCA R package. The comparisons I've found suggest that
they are very similar, and discussions with regulators has indicated the
same opinion. A couple of the comparisons I'm aware of are:
Kinetica-WinNonlin [1]:
https://ir.library.oregonstate.edu/xmlui/bitstream/handle/1957/39168/KarnprachaChanida2013.pdf?
sequence=1
SAAM II-WinNonlin (from 1998): http://dx.doi.org/10.1021/js9603562
If anyone knows of more references for comparisons of the software or
publicly available descriptions of the specifics of the algorithms used,
I'd be very interested so that I can incorporate them into PKNCA.
[1] A real surprise here is that they don't get the exact same value for
Cmax (see page 35 of the linked .pdf file).
Thanks,
Bill
Back to the Top
Dear Harish,
I wanted a comparative table of same data analyzed on WinNonlin, Kinetica SAS and if the study is
borderline passing, will the software make difference?
Regards
Manisha
--
Dear Helmut,
As you have rightly said that there are differences between software, I think it would be very
crucial for studies who are passing with borderline values.
Regards
Manisha
Back to the Top
Hi Manisha,
I've been on holiday a few days but I think Helmut and Bill already covered this well. There
was a project by the AGAH group in Germany to compare WinNonlin and Kinetica back in, I think, 2005;
lead authored by Gudrun Würthwein:
"Comparison of pharmacokinetic software: WinNonlin 4.1 (Pharsight) vs Kinetica 4.2 (Innaphase) for
PK, PD and PK/PD modeling"
However I can't find a live copy on the web anymore and it is of course rather out dated now.
Having used both software and now working for Certara I believe the most important thing for you to
check, especially with NCA, is that any assumptions the software make in the case of missing data
and/or extrapolation tally with what your own SOP or analysis plan describe. There may well be
options you want to select that are different to the defaults to address certain requirements or
problems of your compound's data.
If you are trying to compare modelling approaches then it becomes very important that you are
comparing like with like e.g. using the same weighting scheme/error model and even then I would not
be surprised to see small differences. Maybe you need to be more specific in your question e.g.
what analysis techniques you want to use (NCA and/or modelling? Popualtion versus individual) and
what concerns you have?
Best regards,
Simon DAVIS
Senior Scientific Consultant
Back to the Top
I have a question along this line. In veterinary medicine population PK studies are very rare.
Small individual animal studies generating standard (nonpopulation) parameters are the norm. A lot
of software offer both standard and popPK options, but some do not. Monolix for example seems aimed
specifically at popPK. If you use a popPK program such as this w/o declaring covariates are the
results going to be similar to those produced by nonPopPK software, or should I be avoiding such
use?
Thanks,
Cory
--
Cory Langston, DVM, PhD
Diplomate, American College of Veterinary Clinical Pharmacology
College of Vet Med
240 Wise Center Drive
Miss. State, MS 39762
langston.-at-.cvm.msstate.edu
Back to the Top
Manisha,
I do not have additional sources for comparison of different PK analysis software. However, my
company (Validated Cloud Applications) has developed a cloud-based platform for executing R scripts
and one of the R script-based solutions that we have developed (and are still developing) is a
non-compartmental PK analysis tool that is called NCA-Plus. Our current implementation uses a
publicly available R package and wraps it with some data management before calling NCA and we have
already added some TLF output following the White Paper published by PhUSE. The R package that we
are current using creates pretty much the same parameters as WinNonlin. We have not done a complete
comparison with WinNonlin output (or output of other tools).
If you or anybody else is interested to learn more, just contact me.
Thanks
Peter (pschaefer.at.vca-plus.com)
Back to the Top
Simulations Plus will be releasing PKPlus 1.0 very shortly.
This is a next-generation NCA and compartmental analysis program designed
from the ground up for regulatory submissions, based on the original
PKPlus Module that has been in GastroPlus for about 16 years.
The new standalone PKPlus offers audit trail, validation, and a modern
user interface that enables you to complete most analyses without ever
touching the keyboard - using just point-and-click operations. Because it
is based on a relational database, the program automatically creates
recognizable Groups from your raw data file (e.g., Sex, Fed, Fasted, Dose,
Ethnicity, Study, Arm, Series, etc.), and you can create new Groups by
combining Filters (e.g., Female and Fed and 20 mg; Age<20 and BMI>=30 and
Oral and 100 mg, etc.) to provide rapid Exploratory Data Analysis. All
common options for AUC calculations are included, following the PhUSE
document. BLQs and Exclusions are easily set with an optional comment for
each. Terminal Slope Points can be Automatic (with user override via a
graphical interface), Fixed number of points, or none (i.e., no AUCextr).
Compartmental analysis include 1-, 2-, and 3-compartment (in any
combination) model fitting with an optional lag time, and with all popular
weighting schemes. In addition, compartmental model can be run with your
input set of parameters (no fitting).
Report generation (Tables, Figures, and Lists (TFLs)) via both Automatic
and Custom reports with available output as Microsoft Word(r) .docx files
as well as Microsoft Excel(r) .xlsx files is included.
The positive feedback we've received from our Beta testers has been very
encouraging. This program will be available in August. More information is
available here:
http://www.simulations-plus.com/Products.aspx?PKPlus&grpID=3&cID=16&pID=33
On a personal note and hoping to start a discussion: Reviewing some of the
references posted in this thread, I am always amused at seeing PK outputs
with as many as 5-8 significant digits, from data where even the 3rd might
be questionable. Does it make sense to see something like 10.946372 when
the 5 in 10.95 is a stretch? Does anyone have a justification for using
more than 4 significant digits in PK analysis? Would 10.946 be more
"correct" than 10.95? Or even 10.9? Do we inform decisions any better with
all the extra digits or are we just making tables and lists more difficult
to read and interpret with all the busy values?
Walt Woltosz
Chairman and CEO
Simulations Plus, Inc. (NASDAQ: SLP)
42505 10th Street West
Lancaster, CA 93534
Back to the Top
There is also Pmetrics, a fully functional, all-in-one, completely free R package for non-parametric
and parametric pop PK modeling and simulation. There is a small but growing and dedicated user
community, along with periodic workshops.
Our Laboratory of Applied Pharmacokinetics and Bioinformatics at the Saban Research Institute of the
Children's Hospital Los Angeles and the University of Southern California has been developing
population modeling algorithms and approaches, particularly for Bayesian adaptive control of
individual patient therapeutics (through our companion BestDose software) for over 4 decades.
You can get Pmetrics or BestDose at www.lapk.org.
--
Michael Neely, MD, MSc, FCP
Director, Laboratory of Applied Pharmacokinetics and Bioinformatics
The Saban Research Institute,
Acting Chief, Division of Infectious Diseases,
Children's Hospital Los Angeles
Associate Professor and Clinical Scholar,
Department of Pediatrics, Keck School of Medicine,
University of Southern California
4650 Sunset. Blvd, #MS 51
Los Angeles, CA 90027
[A user, for PopPK. Don't forget Boomer ;-) www.boomer.org - db]
| PharmPK Discussion List Archive Index page |
Copyright 1995-2014 David W. A. Bourne (david@boomer.org)