From: peter.bonate@quintiles.com
Subject: One vs two compartments
Date: Fri, 27 Jul 2001 08:05:50 -0500
Dear All,
I have a question for the group that isn't specifically NONMEM related.
Suppose you have dense data and have sampled out far enough that you can be
confident that a 2-compartmental oral model is a good fit to the data. Now
suppose you do a different experiment on the same drug and this time you
didn't sample out far enough to get information on the second compartment -
the data looks like a 1-compartment. Does anyone know of any papers or
references that show what the bias in parameter estimates of the
1-compartment model will be compared to the values under the 2-compartment
model?
Thanks.
Pete Bonate
One vs two compartments
From: "Sam Liao" <sliao@pharmaxresearch.com>
Subject: Re: One vs two compartments
Date: Fri, 27 Jul 2001 10:34:29 -0400
Pete:
The bias obviously will depend on the contribution of the second phase to
the total AUC. I did a simulation of the same case a year ago. In this
simulation, frequent PK samples were collected only during the absorption
phase and alpha phase (over 24 hr). Non-compartmental method (NCA) was used
to calculate clearance (CL) and volume of distribution (V). The same data
were also analyzed by NONMEM using a two-compartment model and fixed the V3
and Q to the mean parameter estimates obtained from historical data. The
results (see figure1 attached) showed that CL was significantly
over-estimated by NCA method, but not by NONMEM.
Best regards,
Sam Liao, Ph.D.
PharMax Research
270 Kerry Lane,
Blue Bell, PA 19422
phone: 215-6541151
efax: 1-720-2946783
From: "HUTMACHER, MATTHEW [Non-Pharmacia/1825]"
<matthew.hutmacher@pharmacia.com>
Subject: RE: One vs two compartments
Date: Fri, 27 Jul 2001 14:36:58 -0500
Pete,
"Design evaluation for a population pharmacokinetic study using clinical
trial simulations: a case study" by Kowalski, K and Hutmacher, M in Stats in
Medicine, Volume 20, 15 January 2001, page 75-92, may help. This paper
deals with the design, power, and bias of using a one-compartment
approximation to the 2 compartment model.
Hope that helps.
Matt