NONMEM Users Network Archive

Hosted by Cognigen

RE: Your suggestions/thoughts needed on allometric base or final model

From: Stephen Duffull <stephen.duffull>
Date: Tue, 15 Jul 2008 08:09:52 +1200

Hi

Perhaps it is worth noting at this point that (as we know) not all
covariates are equal. Nick indicated that some covariates are fundamental
and hence should not be discarded or even tested for. Perhaps a slightly
more generic framework is to consider that covariates naturally align to a
hierarchical framework, for instance I think it is indisputable in this
community that dose and time are at the most important covariates and we
wouldn't want to construct a PK model without them. Other covariates, such
as extracorporeal elimination (when relevant) are likely to be of a higher
order of importance than covariates based on phenotype, then biologically
meaningful covariates such as renal function, allometric weight, lean body
weight are more important than empirical covariates such as age... And so
forth.

So, when adding covariates into the model the hierarchy in which they enter
should, under a bioligical framework, not necessarily be up for statistical
testing. Since all of our models are nonlinear and therefore most linear
operations such as forward addition/back elimination are not consistent then
we really have no choice but to construct our hierarchy and consider tests
for covariates from the same hierarchical level (but perhaps not between
them). So under this construct you wouldn't consider swapping dose for
weight if it turned out that weight was statistically better than dose and
you also wouldn't consider swapping weight for age if age was statistically
better ...

At what stage you consider a covariate to be "fundamental" meaning it should
never be excluded will depend on the circumstance.

Steve
--
Professor Stephen Duffull
Chair of Clinical Pharmacy
School of Pharmacy
University of Otago
PO Box 913 Dunedin
New Zealand
E: stephen.duffull
P: +64 3 479 5044
F: +64 3 479 7034

Design software: www.winpopt.com
  
Received on Mon Jul 14 2008 - 16:09:52 EDT

The NONMEM Users Network is maintained by ICON plc. Requests to subscribe to the network should be sent to: nmusers-request@iconplc.com.

Once subscribed, you may contribute to the discussion by emailing: nmusers@globomaxnm.com.