Back to the Top
It has been a while since I used WinNonlin to do compartmental
analysis of data, but here is an issue I am running in to. I have data
from a clinical study where subjects were given three doses of the
investigational drug (single ascending dose). I have intensive data
from 24 subjects (8 per cohort at three dose levels). I am looking at
doing a cohort-by-cohort compartmental analysis. A preliminary non-
compartmental analysis has provided me with estimates of clearance and
initial parameters which can be used for compartmental analysis. Given
that weighting of concentrations can have a significant impact on the
goodness of fit, what would the expert modelers recommend I do? I
would really hate to do a subject-by-subject analysis because it may
be very inefficient. Any thoughts?
Back to the Top
Dear Martin,
Please confirm that the drug is following which order of Kinetics.
From the Initial non compartmental analysis what you think about the
best fit model (one compartment. two compartment ....) this will
depends on your data only. You can also try AIC to find best fit model.
With Regards,
Dr. Tushar Nahata
Back to the Top
The following message was posted to: PharmPK
Dear Martin,
Yes applying different weighting schemes to the concentrations could
influence the goodness of fit and more importantly the estimates.
May be you can run the selected model with different weighting schemes
depending on the observed variance in your data and look at the
residual graphs in addition to AIC values.
If the residuals don't appear to be randomly scattered or have a
positive/negative runs, this would suggest incorrectness in the model
or weighing scheme. Even the observed y vs the weighted predicted Y
should provide some information (it should be at 45 degree to the base
line).
Thanks
Ravi
[Ignore AIC if changing weights - db]
Back to the Top
The following message was posted to: PharmPK
Dear All:
About weighting schemes. It is not just about adjusting weights to
get a good fit. It is about being realistic when weights are assigned.
For
example, one does well to weight each assay measurement of serum
concentrations by the reciprocal of the variance with which the
measurement
is made. Knowing this error pattern, based in reality, one can then
estimate
the remaining environmental noise with either an additive (probably
best) or
a multiplicative model. In weighting data, be very careful not to mess
with
Mother Nature. Do it right, don't be shifty with weighting schemes. You
might look at:
Jelliffe RW, Schumitzky A, Van Guilder M, Liu M, Hu L, Maire P, Gomis P,
Barbaut X, and Tahani B: Individualizing Drug Dosage Regimens: Roles of
Population Pharmacokinetic and Dynamic Models, Bayesian Fitting, and
Adaptive Control. Therapeutic Drug Monitoring, 15: 380-393, 1993.
Very best regards,
Roger Jelliffe
|
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 "Compartmental Modeling Using WinNonlin" as the subject | Support PharmPK by using the |
Copyright 1995-2011 David W. A. Bourne (david@boomer.org)