From: "Abouel Leil, Tarek M.   Ph.D." AbouelLeil.Tarek@mayo.edu
Subject: [NMusers] NONMEM Modeling of Plasma + Urine Data
Date: Tue, 21 Nov 2006 09:37:00 -0600

Hello,

I am trying to simultaneously model the PK of a drug and its metabolite using NONMEM. 
I have plasma observations for the parent compound (17-AAG) and the metabolite (17-AG). 
I also have 24hr urine collection data for both compounds.  When I tried to model this
using ADVAN7 in NONMEM, it stops almost immediately with the following error:

MONITORING OF SEARCH:

0PRED EXIT CODE = 1                                                                                                                 

0INDIVIDUAL NO.    1   ID=0.10100000E+03   (WITHIN-INDIVIDUAL) DATA REC NO.   4                                                     

 THETA=                                                                                                                             

  1.43E+04   2.42E+04   8.44E+04   8.96E+04   1.24E+04   2.13E+04   5.00E+04                                                        

 NUMERICAL DIFFICULTIES OBTAINING THE SOLUTION.                                                                                     

 THE COEFFICIENT MATRIX IS ALGORITHMICALLY SINGULAR.                                                                                

0PROGRAM TERMINATED BY OBJ

 MESSAGE ISSUED FROM ESTIMATION STEP

 AT INITIAL OBJ. FUNCTION EVALUATION

1THERE ARE ERROR MESSAGES IN FILE PRDERR                                                                 

2006.11.20.1820.57

Here is the NONMEM control stream that I am using:

$INPUT ID AMT EVID RATE TIME DV UVOL CMT BW BSA SEX AGE CRCL BR AP AST DOSE

$DATA ..\data.dta IGNORE=#

$SUBROUTINE ADVAN7

$MODEL

   COMP=(CENTRAL, DEFDOSE, DEFOBS)

   COMP=PERIPH

   COMP=METAB

   COMP=("URINE", NODOSE, INITIALOFF)

   COMP=("METABUR", NODOSE, INITIALOFF)

$PK

   TVCL = THETA(1)

   CL = TVCL*EXP(ETA(1))

   TVV = THETA(2)

   V = TVV * EXP(ETA(2))

   K14 = CL/V

   S1 = V

   TVQ = THETA(3)

   Q = TVQ * EXP(ETA(3))

   TVV2 = THETA(4)

   V2 = TVV2 * EXP(ETA(4))

   K12 = Q/V

   K21 = Q/V2

   TVCLM = THETA(5)

   CLM = TVCLM * EXP(ETA(5))

   K13 = CLM/V

   TVCLMO = THETA(6)

   CLMO = TVCLMO * EXP(ETA(6))

   TVV3 = THETA(7)

   V3 = TVV3 * EXP(ETA(7))

   K35 = CLMO/V3

   S3 = V3

   S0=UVOL

$ERROR

   ASY1=0

   IF (CMT.EQ.1) ASY1=1

   ASY2=0

   IF (CMT.EQ.3) ASY2=1

   ASY3=0

   IF (CMT.EQ.4) ASY3=1

   IPRED = F

   W = 1

   IF (F.EQ.0) W=1

   IRES = DV-IPRED

   IWRES = IRES/W

   Y=IPRED+ASY1*ERR(1)+ASY2*ERR(2)+ASY3*ERR(3)+(1-ASY1)*(1-ASY2)*(1-ASY3)*ERR(4)

$THETA (0,14300.,) ; TVCL

$THETA (0,24200.,) ; TVV

$THETA (0,84400.,) ; TVQ

$THETA (0,89600.,) ; TVV2

$THETA (0,12400.,) ; TVCLM

$THETA (0,21300.,) ; TVCLMO

$THETA (0,50000.,) ; TVV3

$OMEGA 0.25  ; ETATVCL

$OMEGA 0.25  ; ETATVV

$OMEGA 0.25  ; ETATVQ

$OMEGA 0.25  ; ETATVV2

$OMEGA 0.25  ; ETATVCLM

$OMEGA 0.25  ; ETATVCLMO

$OMEGA 0.25  ; ETATVV3

$SIGMA 100  ; ERRSD17AAG

$SIGMA 100  ; ERRSD17AG

$SIGMA 100  ; ERRSD17AAG URINE

$SIGMA 100  ; ERRSD17AG URINE

$EST METH=0 MAXEVAL=9999 NOABORT POSTHOC PRINT=5 SIGDIGITS=5

$COVARIANCE

The data file looks like this:

#ID     AMT     EVID    RATE    TIME    DV      UVOL    CMT    
101     31500000        1       31500000        0       0       0       1      
101     0       2       0       0       0       0       4      
101     0       2       0       0       0       0       5      
101     0       0       0       0.5     20      0       3      
101     0       0       0       0.5     231     0       1      
101     0       0       0       0.92    51      0       3      
101     0       0       0       0.92    300     0       1      
101     0       0       0       1.08    47      0       3      
101     0       0       0       1.08    146     0       1      
101     0       0       0       1.25    40      0       3      
101     0       0       0       1.25    109     0       1      
101     0       0       0       1.5     55      0       3      
101     0       0       0       1.5     121     0       1      
101     0       0       0       2       42      0       3      
101     0       0       0       2       119     0       1      
101     0       0       0       3       25      0       3      
101     0       0       0       3       75      0       1      
101     0       0       0       5       17      0       3      
101     0       0       0       5       25      0       1      
101     0       0       0       9       0       0       1      
101     0       0       0       9       0       0       3      
101     0       0       0       17      0       0       1      
101     0       0       0       17      0       0       3      
101     0       0       0       24      1993.956        904     -4     
101     0       0       0       24      609.6032        904     -5     

I was able to model the above data without urine observations, so I feel there must be
something wrong with the way I am coding the urine compartments.  I would really
appreciate any advice.

Thanks,

Tarek A. Leil, Ph.D.
Department of Oncology, Guggenheim 1311
Mayo Clinic
200 First Street SW
Rochester, MN 55905
ph: 507-538-4227
fax: 507-266-5146
pgr: 08189

_______________________________________________________

From: "Nandy, Partha [PRDUS]" PNandy@prdus.jnj.com
Subject: RE: [NMusers] NONMEM Modeling of Plasma + Urine Data
Date: Wed, 22 Nov 2006 19:23:38 -0500

Hi Tarek,
 
I am assuming that the drug is given as an infusion from the "RATE" column in your data file.
You mentioned that you ran the model with just the parent & metabolite data in plasma.  Did you
use ADVAN7 for that as well? The formation of metabolite is not well defined..Is it a saturable
process?  If so, Km and Vmax should be implemented.
 
Also, TVCL is total clearance as shown below:
TVCL = THETA(1)

CL = TVCL*EXP(ETA(1)) So, it is not clear how the following is implemented..
TVCLM = THETA(5)

CLM = TVCLM * EXP(ETA(5))

Usually Total Clearance is modeled as CL=CLRenal + CLMetab
Just quickly browsing through the data I do not see rate being declared in your control file
as R1=RATE.  Curious about how the RATE was specified in the control file when the parent & metab
concentration in plasma was modeled. I also see that V1 and V3 was scaled but not V2.
 
Your error structure is quite complex.  It is not clear to me what the purpose of ERR(4) is. 
Also your initial conditions for SIGMA seems to be very high! 
 
Typically the ERROR could be modeled as (There are many ways to do this..).
 
I1=0
I2=0
I3=0
I4=0
 
IF (CMT.EQ.1) I1=1
IF (CMT.EQ.3) I2=1
IF (CMT.EQ.4) I3=1
IF (CMT.EQ.5) I4=1
 
Y=IPRED + I1*ERR(1) + I2*ERR(2) + I3*ERR(3) + I4*ERR(4)
 
Instead of just additive model, did you try other ERROR models for e.g. ADD+PROP etc?
 
Hope this helps.
 
Partha
_______________________________________________________