Back to the Top
Hello everyone,
I have performed intraperitoneal (IP) pharmacokinetic study in mice.
The data is as follows:
Time (h) Conc_ (ug/L)
0.5 267333.33
0.75 90433.333
1 44866.667
2 5493.3333
8 163.66667
12 140
24 80.8
Using Phoenix Winnolin, I did NCA with Dosing Type : IV bolus ( assumed bioavailability
equal to one) with Linear Trapezoidal and Linear Interpolation and got AUCINF_OBS 758404.86.
I have also tried other calculation methods like Linear Trapezoidal Linear/Log
Interpolation, Linear up log down, but all the methods resulted in similar values. In
addition to this, I also performed WNL 5 Classic PK model no. 7 and obtained AUC
=293061.38 ( less than half of value obtained in NCA ).
Based on the above context my queries are:
1. Why the values obtained by both the methods (NCA and WNL 5 Classic model) are so
different?
2. Which one is the correct approach/ which values should I report: NCA or PK model?
3. Does the Beta HL of PK model represent the same thing as the HL_lamdaz of NCA?
Any comments will be greatly appreciated.
Thanks and regards,
Chetan
Back to the Top
Dear Chetan, remember NCA, ie. NON-compartmental is actually assuming a one-comp model
with an APPARENT elimination slope (LZ). Since you are oberving an exponential decay it
makesmore sense to use an AUC calcualtion method that estimates LOG trapezoids post dose.
Going on to your compariosn of AUC from NCA with a compartmental anaylsis, why do you
think a 2 compartment model with IV bolus is the best model for your IP admin, what does
the fit look like, what estimates did you get.
As an example I've included an image here that the rest of the group won't get but
hopefully helps illustrate my comment.
I've plotted the Observed data, the Lz slope (from NCA Summary Table) the Predicted data
from a quick fit I did with the two compartmental model you said you fitted. Note the fit
of the blue line is bad, if your data is looking like this the n you may need to re-assess
your model. I would suggest you consider posting your project to the Pharsight user
forums; www.pharsight.com/extranet so we can see better what you're trying to compare.
In short thought AUC is a Secondary parameter from compartmental modelling and it's
important to understand that you may not be comparing like with like.
For your other questions,
2. Which one is the correct approach/ which values should I report: NCA or PK model?
What does your protocol state? What is the aim of your analysis ?
3. Does the Beta HL of PK model represent the same thing as the HL_lamdaz of NCA?
In a 2 compartment model that fits the data appropriately, yes.
Best regards,
Simon.
Senior Scientific Consultant
Back to the Top
Chetan:
I took a look at your data in an Excel spreadsheet. In answer to your
questions:
1. Why are the values obtained by both the methods (NCA and WNL 5 Classic
model) so different?
NCA uses a trap method (with ln transformed data to figure out the Lamda z)
to calculate AUC. The PK model uses nonlinear regression of a multi
exponential with integration. These are fundamentally very different
approaches. You have an extreme decay profile, where the first 30 minutes
contribute to about half of your AUC. How the C0 is calculated is critical.
Making no assumptions and equating C0 to Cfirst (a rectangle for
calculations), I get an AUClast of 240000. A regression analysis of the last
three points yields a Lamda z of 0.04451, so the tail adds only 1800 to the
AUC. So, check the values for C0 used.
2. Which one is the correct approach/ which values should I report: NCA or
PK model?
All things are relative. What method are you using for the rest of your
data? You need to be consistent. Simplest is best. From the Excel analysis,
I would say an AUC of 240000 and a Lamda z of 0.0445 (a 16 hr halflife) are
your answers.
3. Does the Beta HL of PK model represent the same thing as the HL_lamdaz of
NCA?
No, for the reasons I mention above.
If you have any questions, feel free to contact me. The Excel spreadsheet I
used is attached. [Not included - db]
Chris
Christopher J. Kemper, Ph.D.
Pharma Navigators, LLC
DMPK/Bioanalytical Consulting
Alliance Management/Business Analysis
Back to the Top
Hi chetan,
As I know the PK model is also performs the same NCA with log-linear trapozidal
method. If you received the same value in all other calculation using phoneiox
than it has to match with the winNonlin 5 too. Do check the number of point used
for the calculation of AUCs in both the cases. may give some sort of idea to
you.
Regards,
venkatesan
[Sorry if these replies are out of order. I managed to download them to different computers - db]
Back to the Top
Chetan,
Your observation of the absence of a significant difference between AUC values
calculated by the classical trapezoid method and other often assumed more
accurate methods, is not uncommon. One publication statistically accounted for
the similarity in results as due to the inherent variability in serial sampling
of biological samples - making the point that in this case, refined (esoteric)
approaches are not necessarily superior to classical (simple) approaches. Thus,
go with what you calculated!
Regards,
David
David S. Farrier, Ph.D.
www.SummitPK.com
Summit Research Services & PK Software
Montrose, CO 81401
[Chetan, if you calculate the AUC with a 0,0 point because it is IP you will get a different AUC than if you model the data with the wrong model, i.e. IV, which might include an area from a calculated Cp(0) (not equal to zero) to the first data point. - db]
Back to the Top
Chetan, I think the clue is in the AUC_%Back_Ext_obs at 85.821299 %
NCA just back extrapolates the first two points and is giving a much higher
estimate of Czero
2336158
versus the Fitted value which is less than a third of that;
724848
So that's the main source of your extra AUC; as indicated by the first shaded
area in the image below;
http://www.pharmpk.com/images/sd_25Jul12/image002.jpg
Remember it's a log scale so that's a large area.
http://www.pharmpk.com/images/sd_25Jul12/image003.jpg
Simon.
[The image from the previous message
http://www.pharmpk.com/images/sd_25Jul12/image001.png
- db]
Back to the Top
The following message was posted to: PharmPK
To all:
Simon's arguments sound good to me. The fundamental problem is that there is
no data before 30 minutes, which is THE most critical part of the curve as
far as AUC calculations. Not having all the needed data points is very
common in non-clinical in vivo studies. The lesson here is that using
extrapolated and/or model fitted data can be misleading.
Chris
Christopher J. Kemper, Ph.D.
Pharma Navigators, LLC
DMPK/Bioanalytical Consulting
Alliance Management/Business Analysis
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 "Difference between NCA and PK model for IP study in mice" as the subject | Support PharmPK by using the |
Copyright 1995-2011 David W. A. Bourne (david@boomer.org)