Paul,
Whatever you do, I would make sure that these extra compartments have no
effect on the central compartment. I would first model the plasma
concentrations. If you see the difference, you can add categorical
covariate ("pregnant") to better describe your subset of interest. Then
fix the model parameters and use individual predictions of plasma
concentrations as a driving force for the amniotic fluid concentrations
(like a two-stage PK-PD model). After you build these separate models
(plasma, and then the rest) you can try to re-fit them jointly. If you
have just one observation per embryon, it is unlikely that you can
distinguish between intra and inter subject variability, so I would put
everything to noise (sigma).
Leonid
--------------------------------------
Leonid Gibiansky, Ph.D.
President, QuantPharm LLC
web: www.quantpharm.com
e-mail: LGibiansky at quantpharm.com
tel: (301) 767 5566
Paul Hutson wrote:
Has anyone suggestions on how best to code for data that presents
maternal plasma and multiple amniotic fluid concentrations? In addition
to multiple other data points of plasma from various doses and routes
for this tox study, I have simultaneous concentration data from 3
mouse dams (plasma) and samples from the amniotic fluid of three of each
of their embryos.
Modeling each embryo as another compartment, especially with only the
one concentration, is cumbersome and obviously doesn't converge.
Should the different embryonic AF concentrations be modeled as
'repeated' samples at this same time point of a specific (AF)
compartment, and the differences between them modeling as
interoccassion or as an intrasubject variability (ERR)? The latter
approach is the one I have taken, but I seek other suggestions as well.
Thank you in advance!
--
Paul R. Hutson, Pharm.D.
Associate Professor
UW School of Pharmacy
777 Highland Avenue
Madison WI 53705-2222
Tel 608.263.2496
Fax 608.265.5421
Pager 608.265.7000, p7856
