NONMEM Users Network Archive

Hosted by Cognigen

Re: estimating Ka from dataset combining rich sample study and sparse sampling study

From: Ethan Wu <ethan.wu75>
Date: Thu, 18 Jun 2009 07:39:03 -0700 (PDT)

Dear all,    thanks to Mats suggestion, using full cov matrix =
did assign Eta more reasonably to Ka, with not very precised estimates =
  another suggestion is that, there may be some underline differe=
nce in the structure model between sparse MD data, and rich SD -- by visual=
 inspection of sparse MD data, I really can't think it can support more com=
plext model itself with 2-3 datapoints per subject in each study perio=
d (total of 2), with the sampling the same across subjects.   =
another suggestion is that to model study individually,  bot=
h Ka and IIV on Ka estimated from sparse sample study were m=
uch larger.    so I think I will further pursue exploration of=
 IOV, as Jakob pointed out. ___________________________=
_____ From: James G Wright <james
<stephen.duffull
 "Ribbing, Jakob" <Jakob.Ribbing
m>; Jurgen Bulitta <jbulitta
Cc: Roger Jelliffe <jelliffe
Sent: Thursday, June 18, 2009 6:56:57 AM Subject: RE: [NMusers] estimatin=
g Ka from dataset combining rich sample study and sparse sampling study =

re is some problem with the model (most likely a lack of exchangeability).=
  I think the ideas suggested are all good, but the first thing I woul=
d try is to separate residual noise for the two studies with an indicator v=
ariable.  It is likely that study procedures, precision of recorded sa=
mpling times etc. vary between the two studiess.    Best reg=
ards, James   James G WrightPhD Scientist Wright Dose Ltd Te=
l: 44 (0) 772 5636914   -----Original Message----- From: owner-n=
musers
ephen Duffull Sent: 18 June 2009 07:04 To: Mats Karlsson; 'Ribbing, Jak=
ob'; 'Ethan Wu'; 'Jurgen Bulitta'; nmusers
iffe'; 'Neely, Michael' Subject: RE: [NMusers] estimating Ka from dataset=
 combining rich sample study and sparse sampling study   Dear Etha=
n   I concur with Mats’s comments below.   =

ment cannot result in less precise parameter estimates under the assumption=
 that the individuals from the two data sets are exchangeable.  Under =
this assumption therefore the Sparse data should merely add information to =
the Rich data.  That the Sparse data is affecting the parameter estima=
tes from the Rich data suggests that the two data sets are not exchangeable=
 (different centre, different assay, different covariates ...).   =
Another possible way to investigate the differences between the two data se=
ts would be to analyse them sequentially, perhaps with consideration for us=
ing the analysis from the Rich data as an informative prior for the analysi=
s of the Sparse data and see where this leads you.   Kind regards=

macy School of Pharmacy University of Otago PO Box 913 Dunedin New =
Zealand E: stephen.duffull
 7034   Design software: www.winpopt.com     =
      From:owner-nmusers
owner-nmusers
18 June 2009 9:17 a.m. To: 'Ribbing, Jakob'; 'Ethan Wu'; 'Jurgen Bulitta'=
; nmusers
: RE: [NMusers] estimating Ka from dataset combining rich sample study and =
sparse sampling study   Dear Ethan,   Variances estimated=
 to be zero may result from fixing off-diagonal variances to zero (i.e. not=
 using BLOCKs in IIV). Here, however, it may be that there are systematic d=
ifferences between the sparse and the rich data experiments. Maybe fasting/=
fed status or something else is different. If the fit to the rich data is m=
arkedly worse when including the rich data, at least one parameter is diffe=
rent between the two situations. I would explore what parameter(s) that wou=
ld be. In addition to Jakob’s suggestions below, the two data sets =
together may indicate a more complex structural model that a single profile=
 indicated. Maybe you need to go to a two-compartment for example.  =

Pharmacometrics Dept of Pharmaceutical Biosciences Uppsala University=

1 4003   From:owner-nmusers
lobomaxnm.com] On Behalf Of Ribbing, Jakob Sent: Wednesday, June 17, 2009=
 10:43 PM To: Ethan Wu; Jurgen Bulitta; nmusers
r Jelliffe; Neely, Michael Subject: RE: [NMusers] estimating Ka from data=
set combining rich sample study and sparse sampling study   Hi Eth=
an,   If OMEGA(?) for KA is drastically reduced when including the=
 sparse data, then something is wrong with your model and in this case it i=
s not the estimation method or assumption on distribution of individual par=
ameter). Eta-shrinkage would not drastically reduce the estimate of OMEGA, =
since this estimate is driven by the subjects/studies which contain informa=
tion on the parameter.   If the sparse data is multiple dosing it =
may be that KA is variable between occasions, rather than between subjects =
(assuming the sparse data contain some information on KA). Or if the sparse=
 data is from a less well-controlled study or a different population, it ma=
y be that increased IIV in other parts of the model (e.g. OMEGA on V) is ma=
king IIV in KA appear low for the rich study, when fitting the two studies =
together. If you get the covariate model in place this problem will be solv=
ed. For the simple model you have it should be quick to start out assuming =
that most parameters (THETAs and OMEGAs) are different between the two stud=
ies and then reduce down to a model which is stable and parsimonious. Obvio=
usly, if you eventually can explain the differences using more mechanistic =
covariates than study number that is of more use.   Cheers =
Jakob
Received on Thu Jun 18 2009 - 10:39:03 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.