From: PING CHEN chen.1016@osu.edu Subject: [NMusers] Help on rounding error ( error=134) ! Date: Sun, 12 Nov 2006 17:24:52 -0500 Hi, I was trying to do the popPK for a unstable drug which will be converted into its metabolite quickly. We only measure the parent compound concentration which was far below the dose. For example, 120 umol was given as i.v infusion for 1 h but the plasma concentration of the parent compound is only around 0.02-1.5 uM. I just tried the simple ADVAN3 TRANS4 as follows, however, the result turns out to be a very large V2 and very small CL and keeping giving the rounding error ( error=134)! Does anyone can help me figure it out? $PK V1=THETA(1)*EXP(ETA(1)) V2=THETA(2)*EXP(ETA(2)) CL=THETA(3)*EXP(ETA(3)) Q=THETA(4)*EXP(ETA(4)) K=CL/V1 K12=Q/V1 K21=Q/V2 S1=V1 $ERROR IPRED=F W=F IRES=DV-IPRED IWRES=IRES/W Y= IPRED+W*EPS(1) $OMEGA 0.25 $OMEGA 0.5 $OMEGA 0.5 $OMEGA 0.5 $SIGMA 0.25 Thanks a lot. Ping _______________________________________________________ From: Nick Holford n.holford@auckland.ac.nz Subject: Re: [NMusers] Help on rounding error ( error=134) ! Date: Mon, 13 Nov 2006 11:50:39 +1300 Ping, Did you look at the model predictions? If the predictions look Ok then you can ignore the rounding error message. A visual predictive check would be a good way to decide if your model predictions are OK. You say CL is small and V2 is large. On what basis of prior knowledge do you judge them to be small and large? Unless you have some good reason e.g. CL much bigger than cardiac output then it isn't usually helpful to judge the numerical values to be small or large. Nick -- Nick Holford, Dept Pharmacology & Clinical Pharmacology University of Auckland, 85 Park Rd, Private Bag 92019, Auckland, New Zealand email:n.holford@auckland.ac.nz tel:+64(9)373-7599x86730 fax:373-7556 http://www.health.auckland.ac.nz/pharmacology/staff/nholford/ _______________________________________________________ From: "Toufigh Gordi" tgordi@Depomedinc.com Subject: Re: [NMusers] Help on rounding error ( error=134) ! Date: Sun, 12 Nov 2006 20:19:14 -0800 Hi Ping, I am guessing that you have looked at some graphs of your data and can clearly see a bi-phasic decline, or have some prior knowledge about the drug, and that's why you apply a 2-comp model. I would suggest that early on in the model-building, you start with one ETA (my suggestion would be CL or VC). If you find a model that does a relatively good job, you can start adding ETAs to see whether they improve the fit. I don't knwo about the NONMEM codes for various errors have have obviously not seen your data but your problem might be due to over-parametrization. Toufigh _______________________________________________________