Re: [NMusers] Indirect response model

Dear Nishit,

If I understand correctly, you are using concentration (COP) of
your drug as a time dependent covariate which is then used as
forcing function for your PD model.

As concentrations change over time, you probably need the
CALLFL = 0 option ($PK CALLFL=0) to read in the concentration
at every new time. I would write out COP in $TABLE in order to
check, if COP changes over time as it should.

This should work much better, but it will still give you a piecewise
constant concentration profile. This may cause numerical problems.
Instead, I would include the differential equations for your PK
model. This should give you better numerical stability and more
correct concentration predictions. You could start with reading in
the individual PK parameters (IPP approach, see reference below)
and then go to more complex PKPD analyses.

You might try the MATRIX=S statement in $COV, if you like to get
the covariance step to work.

Hope some of this works.

Best regards
Juergen

Reference:
Zhang, L., S. L. Beal, and L. B. Sheiner. 2003. Simultaneous vs.
sequential analysis for population PK/PD data I: best-case performance.
J Pharmacokinet Pharmacodyn 30:387-404.


-----------------------------------------------
Juergen Bulitta, PhD, Post-doctoral Fellow
Pharmacometrics, University at Buffalo, NY, USA
Phone: +1 716 645 2855 ext. 281, j_at_bulitta.com
-----------------------------------------------



-----Ursprüngliche Nachricht-----
Von: "Modi, Nishit [ALZUS]" <NMODI_at_ALZUS.JNJ.COM>
Gesendet: 15.05.07 18:31:57
An: nonmem_at_optonline.net
CC: nmusers_at_globomaxnm.com
Betreff: [NMusers] Indirect response model



I am conducting a sequential pharmacokinetic-pharmacodynamic model. The p=
harmacokinetic fits look good and I was using an indirect response model. =
 The PD model is that the drug inhibits clearance of the analyte (PD respo=
nse), thus one expects that the response increases with increasing drug (M=
odel II). There is a baseline measured (=Kfor/Kcl) and a dummy dose=1 uni=
t given. It seems despite trying various permutations of the model, eta1 s=
eems to be very small and no covariance step is conducted. The model and =
data for the first 3 subjects are reproducted below. Any assistance would=
 be appreciated. Note that since conc (COP) are read in, the model only r=
equires a single differential equation. Any insight would be appreciated.=



Nishit


$PROBLEM PD - ADVAN6

$DATA C:\PDDATA.CSV

$INPUT ID TIME DV AMT=DOSE COP MDV

; data are subject ID, Time, DV=PD response, Amt (dummy dose of 1 inserted=
), COP=plasma conc which drive PD model, MDV

$SUBROUTINES ADVAN6 TOL=6

$MODEL

   COMP=(EFFECT, DEFDOSE, DEFOBS)

 

$PK

   KFOR = THETA(1)

   KCL = THETA(2)*EXP(ETA(1))

   IC50 = THETA(3)

   IMAX = THETA(4)

   F1 = KFOR/KCL

   COEF = IMAX*COP/(IC50+COP)


$DES

   DADT(1) = KFOR-KCL*(1-COEF)*A(1)

$ERROR

     W = F

     Y = F*EXP(ERR(1))

     IPRED = F

     IRES = DV-IPRED

     IF (W.LE.0.) W=1

     IWRES = IRES/W


$THETA (0,0.3)

$THETA (0, 0.003)

$THETA (0,10)

$THETA (0, 0.3, 1)


$OMEGA 0.01

$SIGMA 0.5


$ESTIMATION METHOD=1 MAXEVAL=5000 PRINT=20

$COVR

$TABLE ID TIME PRED IPRED IRES KFOR KCL IC50 IMAX

    NOPRINT ONEHEADER

    FILE=C:\PD.TAB


1001 0 . 1 0 1

1001 0 98.3 . 0 0

1001 168 90.6 . 122.44 0

1001 840 92.8 . 183.69 0

1002 0 . 1 0 1

1002 0 105.1 . 0 0

1002 840 88.5 . 61.253 0

1002 842 106.7 . 106.8 0

1002 844 122.1 . 116.4 0

1002 1848 129.1 . 121.46 0

1002 1850 160.4 . 212.63 0

1002 1852 157.1 . 231.89 0

1101 0 . 1 0 1

1101 0 68.1 . 0 0

1101 840 88.1 . 0.13884 0

1101 842 105.5 . 0.12987 0

1101 844 108.8 . 0.12147 0

1101 1848 113.3 . 227.79 0

1101 1850 62.6 . 379.54 0

1101 1852 138.7 . 412.18 0



Received on Tue May 15 2007 - 13:04:42 EDT