From: "Gordi, Toufigh" Toufigh.Gordi@cvt.com
Subject: [NMusers] Indirect response model in NONMEM 
Date: Wed, August 25, 2004 5:04 pm

Dear all,

I am trying to fit an indirect response model (decrease in KIN) to some data. I have
fairly good data, including pre-dose observations. There were three different
dosings: oral, iv bolus and iv bolus+iv infusion. The PK part is fixed based on
individual PK estimates. This is a sample of the model code:

$SUBROUTINE ADVAN9 TRANS1 TOL=5
$MODEL
NCOMP=5
COMP = (GUT)
COMP = (CENT)
COMP = (PRIPH)
COMP = (PER2)
COMP = (EFF)

$PK
        CL=CL1
        ...
        IEFF=THETA(1)
        IC50=THETA(2)*EXP(ETA(1))
        KOUT=THETA(3)
        KIN=THETA(4)

        S2=VC
        D2=DUR
        F1=BAV
$DES
        ...
        CONC=A(2)/VC
        HIT=IEFF*CONC/(IC50+CONC)
        ...
        DADT(5)=KIN*(1-HIT)-KOUT*A(5)

Could anybody give me some guidance on how the data file should be constructed? And
how the data file and the control stream should be modified in case I want to
estimate a baseline value? I have found some previous comments on "initializing" the
effect compartment but have not been able to make it work.

Thank you!     

Toufigh Gordi
_______________________________________________________

From:"Luann Phillips" luann.phillips@cognigencorp.com
Subject: RE: [NMusers] Indirect response model in NONMEM
Date: Wed, August 25, 2004 6:10 pm 

Toufigh,

Question:  Can you actually measure cmt=5?

If yes:
Add a record to your dataset at time=0 (very first record) with cmt=5,
amt=first concentration value from cmt=5, and dv=missing. In the control
stream add F5=1*exp(eta(n)) to allow for estimation of measurement error
in the first concentration of cmt=5. You can keep the first
concentration from cmt=5 as an observation record in addition to using
it as a dose record but I usually delete it.

If no:
Add a record to your dataset at time=0 (very first record) with cmt=5,
set amt=1.
All later predictions of A(5) will be relative concentrations (ie.,
fractional changes)
If you can not measure A(5), you will never be able to estimate a
baseline value for the compartment.  You assume a beginning
concentration (amt=1) and then your model estimates the relative impact
that the drug has on CMT=5.


Note: In both cases, you should treat it as a bolus dose.  If whatever
A(5) represents is at steady-state prior to introducing the drug, then
also use SS=1. (Review the bug about ADVAN9 and the SS6 routine).

Also, make sure that IEFF is restricted between 0 and 1 to prevent your
input rate for cmt=5 from becoming negative.

Regards,

Luann
_______________________________________________________

From:"mathangi" mathangi@msn.com
Subject: RE: [NMusers] Indirect response model in NONMEM
Date: Wed, August 25, 2004 6:58 pm 

Hello Toufigh,
In order to initialize the response compartment to the  baseline value, you
can use the following addition in the code:

$PK
        CL=CL1
        ...
        IEFF=THETA(1)
        IC50=THETA(2)*EXP(ETA(1))
        KOUT=THETA(3)
        KIN=THETA(4)

        S2=VC
        D2=DUR
        F1=BAV

        F5=KIN/KOUT                            ;Baseline value

And accordingly, the data file is arranged by including AMT=1 for CMT=5 at
TIME=0.

      ID TIME AMT EFF CMT
      1 0 100 0 1
      1 0 1 0 5
      1 0 0 14.739 5
      1 0.25 0 12.633 5
      1 0.5 0 10.043 5
      1 0.75 0 9.5252 5
      1 1 0 8.9563 5
      1 1.5 0 6.5243 5
      1 2 0 5.2844 5
      1 2.5 0 4.6631 5
      1 3 0 4.2629 5
      1 4 0 3.5792 5
      1 6 0 3.1123 5
      1 8 0 3.275 5

Hope this helps

Mathangi Gopalakrishnan
_______________________________________________________