Re: PK-PD, using 2 step approach
From VPIOTROV@janbelc1.ssw.jnj.com Mon Jul 29 03:47:08 1996
Subject: Re: PK-PD, using 2 step approach
Hello Mark,
The best way to get reliable estimates of EC50 and other PD parameters is to fit both PK and PD data simultaneously using the method developed by L. Sheiner and S. Beal and presented at the last NONMEM workshop in Uppsala. Your data file should contain both Cp and effect observations and should be organized in the following way: when DV is a CP observation, CMT = 2; when DV is an effect observation, CMT= 4. The following initial (without covariates) control stream could be proposed:
$PROBLEM PK-PD data: effect compartment model
$INPUT ID TIME DV DOSE=AMT CMT
$DATA yourdata
$SUBROUTINES ADVAN5 TRANS1
$MODEL
COMP=(DEPOT) ;1
COMP=(CENTRAL DEFOBS NOOFF) ;2
COMP=(PERIPH) ;3
COMP=(EFFECT) ;4
$PK
K12= THETA(1)*EXP(ETA(1)) ;Absorption
K20= THETA(2)*EXP(ETA(2)) ;Elimination
K23= THETA(3)*EXP(ETA(3)) ;To peripheral
K32= THETA(4)*EXP(ETA(4)) ;From peripheral
K24= .001*K20 ;To effect
K40= THETA(5)*EXP(ETA(5)) ;keo
S2= THETA(6)*EXP(ETA(6)) ;Central V
S4= S2*K24/K40 ;Preserves Cess = Cpss
EMAX=THETA(7)*EXP(ETA(7)) ;Emax
EC50=THETA(8)*EXP(ETA(8)) ;EC50
$ERROR
Y1=F*EXP(ERR(1))
Y2=EMAX*F/(EC50+F)*EXP(ERR(2))
Q=1 ;Cp
IF (CMT.EQ.4) Q=0 ;Effect
Y=Q*Y1+(1-Q)*Y2
$THETA
...
$OMEGA
...
$SIGMA
...
etc.
You can re-parameterize the PK model in terms of CL, Q and V1, V2. You can also add fixed effects of covariates and reduce the number of ETAs. Try and let me know if it works.
Yours,
Vladimir
----------------------------------------------------------------------
Vladimir Piotrovskij, Ph.D.
Janssen Research Foundation
Clinical Pharmacokinetics
B-2340 Beerse
Belgium
Fax: +32-14-603768
Email: vpiotrov@janbe.jnj.com