From: musor000@optonline.net
Subject: [NMusers] associatin/dissociation model
Date: Wed, 05 Oct 2005 22:16:54 -0400
Hello Team,
I try to fit an association/dissociation model. It does not converge. I tried
different models of errors and different starting values. Here are questions:
1. Is it better to use moles or grams? I use grams. Obviously, model will
look slighly different if we use moles.
2. If there is huge difference between Kon and Koff, can it cause problems?
Thanks!
Pavel
$PROBLEM Population SC dosing.
$INPUT ID TIME DOSE=AMT CP=DV MDV EVID CMT
;CMT: 1-INJ SITE, 2-T, 3-V, 4-TV
;EVID: 0-OBSERVATION, 1-DOSE, 4-DOSE & RESET.
$DATA PKCADI01.DA IGNORE #
$SUBROUTINES ADVAN8 TOL=8
;TOL - THE NUMBER OF ACCURATE DIGITS
$MODEL COMP=(INJ INITIALOFF)
COMP=(T)
COMP=(V)
COMP=(TV)
COMP=(AUC)
$EST MAXEVAL=9999 SIG=3 NOABORT PRINT=30
METHOD=1 POSTHOC INTERACTION ;REPEAT
$PK
VD =THETA(1)*DEXP(ETA(1))
KA =THETA(2)
KON =THETA(3)
KOFF =THETA(4)
RIN =THETA(5)
CLV =THETA(6)*DEXP(ETA(2))
CLT=CLV
;CLT =THETA(7)*DEXP(ETA(3))
CLTV=CLT
KD=KOFF/KON
KET =CLT /VD
KETV =CLTV/VD
KEV =CLV /VD
;HL =0.693147/KE
QT=0.912
QV=1-QT
F3=RIN/KEV
S4=VD
S2=VD ;The amount A in the observation compartment
;at the time of observation, divided by the
;value of a parameter S, is used as the prediction.
$DES
DADT(1) = - KA*A(1) ; Inj Site
DADT(2) = - KET *A(2) - QT*KON*A(2)*A(3) + QT*KOFF*A(4) + KA*A(1) ; T
DADT(3) = RIN - KEV *A(3) - QV*KON*A(2)*A(3) + QV*KOFF*A(4) ; V
DADT(4) = - KETV*A(4) + KON*A(2)*A(3) - KOFF*A(4) ; TV
DADT(5) = A(2) ; AUC
AUC=A(5)
IF (EVID.EQ.4) AUC=0
A1=A(1)
A2=A(2)
A3=A(3)
A4=A(4)
$THETA
(1,80,300) ;VD - VOLUME OF DISTRIBUTION
(0.1,3,30) ;KA - ABSORBTION COEFFICIENT
(0.1,10,1000) ;KON - ASSOCIATION RATE CONSTANT
(0.1,1,1000) ;KOFF - DISSOCIATION RATE CONSTANT
(0.002,0.04,0.1) ;RIN - PRODUCTION RATE OF V
(0.05,5,30) ;CLV - CLEARANCE
;(0.05,0.3,5) ;CLT - CLEARANCE
;(0.05,0.3,5) ;CLTV - CLEARANCE
$ERROR
CONC=F
Y=CONC+ERR(1)
;Y=CONC*EXP(ERR(1)) + ERR(2) ;CONCENTRATION ERROR
;IF (CMT.EQ.2) Y=CONC*EXP(ERR(1)) + ERR(2) ;CONCENTRATION ERROR
;IF (CMT.EQ.4) Y=CONC*EXP(ERR(1)) + ERR(3) ;CONCENTRATION ERROR
IPRE=CONC ;individual-specific prediction
IRES=DV-IPRE ;individual-specific residual
IWRE=IRES/CONC ;individual-specific weighted residual
$OMEGA DIAGONAL(3) 50 0.6 0.3 ;VARIANCE OF THETA
$SIGMA 50 ;0.5 ;VARIANCE OF EPS
associatin/dissociation model
From: "Perez Ruixo, Juan Jose [PRDBE]" JPEREZRU@PRDBE.jnj.com
Subject: RE: [NMusers] associatin/dissociation model
Date: Thu, 6 Oct 2005 14:09:32 +0200
Dear,
The model you described has been also called "target mediated drug
disposition model". From the information you provide, it's very difficult to
provide you with a clear answer. In fact, I've more questions than
clarifications.....:-). Hope you don't mind.
In my experience, it's very difficult to have enough information in the data
to estimate independently Kon, Koff, KETV and the turnover parameters of the
target (or receptor). Did you perform a sensitivity analysis to evaluate
what are the parameters of your model that can be identified given the
information you have?
I would suggest you to include IV data (if available). Just SC data might
not be good enough to fully characterize this complex model.
Also, I assume the bioassays used allow you to quantify the free drug. Do
you have observations from the A(4)? If that's the case, it should be
incorporated in the database and fit together with the free drug
concentrations. This information is critical to estimate the receptor
turnover.
Do you have prior information about the target (receptor) binding and
elimination of the drug receptor complex? If you have it, you could try to
fix Koff and/or KETV to the values obtain from previous (in vitro)
experiments. However, I'm not a big fan of that approach.
I personally prefer the "quasi-equilibrium" approach that has been recently
published [Mager DE, Krzyzanski W. Quasi-equilibrium pharmacokinetic model
for drugs exhibiting target-mediated drug disposition. Pharm Res. 2005
Oct;22(10):1589-96]. With that approach, NONMEM runs are faster and the
difference in parameter estimates and model predictions between the full
model and the quasi-equilibrium model are acceptable in most of the cases.
Hope it helps.
Juan Jose Perez Ruixo, PhD.
Principal Scientist. Advanced PK/PD Modelling & Simulation,
Global Clinical Pharmacokinetic and Clinical Pharmacology,
Johnson & Johnson Pharmaceutical Research & Development,
a Division of Janssen Pharmaceutica, NV.
* Turnhoutseweg 30, B-2340 Beerse, Belgium.
* jperezru@prdbe.jnj.com
From: musor000@optonline.net
Subject: RE: [NMusers] associatin/dissociation model
Date: Thu, 06 Oct 2005 16:58:24 -0400
Hello Perez,
Your advise is really valuable. I ordered the article.
We may have reasonable estimates of Kd, but estimates of Koff and Kon look unreliable. I
am trying to incorporate Kd and calculate Koff=Kd*Kon. My concerne is that Kd is really small.
Can NONMEM handle values such as 10e-10? It cannot read constants if they are too small...
KETV is almost the same as KET.
I did not do any sensetivity any analysis because I did not design the study.
We measure both T (A2) and TV (A4). We do not have V.
The model is nonlinear and it is not clear how bioavailability can affect the parameters. Hopefully
VD will "absorb" bioavailability as it happens in linear models, i.e. I will estimate VD/Bioavailability.
IV data can be available from another study. I will try to add IV data.
One thing I just noticed is that I use amounts, but some terms in the model should be concentrations!!!
Thank you!
Pavel
From: "Perez Ruixo, Juan Jose [PRDBE]" JPEREZRU@PRDBE.jnj.com
Subject: RE: [NMusers] associatin/dissociation model
Date: Thu, 6 Oct 2005 23:14:55 +0200
Hi Pavel,
I'm not sure what you mean with the model is non-linear. Do you mean the
disposition or the bioavailability? The first is included in the model, but
the second is not.
If you have IV and SC data you could do an exploratory deconvolution
analysis and explore if the dose normalized cumulative input change with the
dose. If that's the case probably the absolute bioavailability would
contribute to the nonlinearity as it happens for some biologicals.
Regards,
Juanjo.
_______________________________________________________