Back to the Top
Dear All,
I have noticed a decreased central compartment volume of distribution
with weight for the following study.
Study Data: IV infusion 0.5 hr (Meropenem), in obese patients. Drug is
extremely water soluble.
Analyzed the data using WINONLIN compartmental modeling (CM).
Statistical information (AIC, CV%) supported 1CM if the patient weight
(WT) is less than 150 Kg and 2CM if WT is more than 150 Kg.
WT<150 Vol of distribution increased with increase in weight correlation (R sq=0.91).
WT>150 Central compartment volume of distribution decreased with
increase in WT (R sq=0.94). AIC values decreased when 2CM is applied to these patients.
What this means physiologically for a drug which is extremely water soluble.
Thank you very much in advance.
--
Regards,
Shankar Lanke
Back to the Top
The following message was posted to: PharmPK
Shankar:
Statistics are just one method of checking structural model fit. Graphical
analysis is just as important. What do the residual plots look like? Is there a
pattern in the 1CM residuals that indicates poor fit for individuals > 150 kg?
What does the histogram of he predicted parameters especially the Volumes (all
subjects combined) look like, do they indicate 2 populations? Probability plots
are also useful to determine the presence of sub populations. As for the central
compartment getting smaller, that is typical when fitting a 2CM as the data will
keep the overall volume (Vss) similar to the 1CM.
As for a physiologic reason I can only surmise that the extra fat creates a
longer distributional component that only shows up once you get past 150 kg. Do
you have any females in your study?
Is the weight based difference really a gender difference? As has been discussed
in this forum before, weight can be a strange covariate, especially when dealing
with the seriously obese.
William R. Wolowich, Pharm.D., R.Ph.
Assistant Professor
College of Pharmacy
Nova Southeastern University
wwolowic.at.nova.edu
Back to the Top
Dear Dr. Wolowich,
Thank you very much for your reponse.
I checked the residual plots and other plots for any possible outliers.
I plotted all plots similar to any pop PK analysis. Since then number of
subjects in this study is small I have used WINONLIN instead of NONMEM.
The fits for subjects >150 Kg in 1CM is poor and supported by all
possible plots (WRES, RES, etc.,) hence I opted for 2CM for subjects
>150 Kg.
There is a clearly difference between sub populations in Vd.
All the subjects in this study are females. The subject's weight range
is in between 110-200 Kg.
Thank you once again.
Regards,
Shankar Lanke Ph.D.
Back to the Top
The following message was posted to: PharmPK
Dear Shankar,
A couple additional points to bring up - Do you have rich or sparse
sampling scheme? What method are you using to fit the data?
The one thing I question, is while adipose tissue would seem an obvious
culprit, a highly hydrophilic drug would not have extensive distribution
into adipose tissue, so major changes in fat amounts wouldn't
necessarily point to more distribution.
One thing you could easily explore is to stratify your patients in
overlapping 'bins' to identify if there is truly an abrupt change in the
distribution .-a-. 150 or if it is just more pronounced past that point.
For example, fit patients in bins of 100-130, 115-145, 130-160,.... and
see how your model fits are between groups.
Another issue could be with a posteriori identifiability of the
different phases, especially if you only have sparse sampling .at. earlier
time points per individual.
Without more information about patient distribution, sampling scheme,
etc much is just conjecture.
Best of luck!
Devin Pastoor
Clinical Research Specialist
University of Maryland Center for Translational Medicine
Back to the Top
Dear Devin,
I have a rich data (6/7 DV) for each patient. I used in built WinNonlin
compartmental models to fit the data.
Drugs half-life is less than 1 hr, I have 6 sample points in first 2.5 hrs and
one in between two to three hours.
I will bin the data to check for any possible model improvements.
Thank you very much.
Back to the Top
That sounds like it shouldn't be an identifiability issue in that case - I would
also mimic Dr. Wolowich's suggestions and take a look at the histograms of the
various potential covariates as well as evaluate their impacts on Cl and Vd.
Have you tried using weight as a covariate and seeing how the model diagnostics
look? I would try a weight covariate normalized to the median value of your
population to start. If there is a truly abrupt change around 150 kg you could
theoretically try a categorical covariate for patients above/below 150 kg to
describe the given data. The 'problem' is I do not know how valid this model
would be for extrapolation if that is your end goal.
I would also suggest continuing to explore potential physiological differences
between the 'sub' populations you seem to have identified - race, age, etc that
could potentially have an impact on their physiology (though these are tenuous
at best).
Devin Pastoor
Clinical Research Specialist
University of Maryland Center for Translational Medicine
Back to the Top
The following message was posted to: PharmPK
Dear Shankar,
You wrote:
> Since the number of subjects in this study is small I have used WINONLIN
> instead of NONMEM.
I don't understand the rationale behind this choice. I would suggest to use
a population analysis for any number of subjects (even for two subjects!),
an to use NONMEM rather than WinNonlin for population analysis. Do you have
a particular reason to prefer WinNonlin here?
best regards,
Hans Proost
Johannes H. Proost
Dept. of Pharmacokinetics, Toxicology and Targeting
University Centre for Pharmacy
Antonius Deusinglaan 1
9713 AV Groningen, The Netherlands
|
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@lists.ucdenver.edu with "Decreased central compartment volume of distribution with weight" as the subject |
Copyright 1995-2014 David W. A. Bourne (david@boomer.org)