Population pharmacokinetics model analysis
Objective and introduction
Objectives
- To provide practical experience of performing population pharmacokinetic analysis.
- To learn how PREDPP PK models are specified using NM-TRAN.
Introduction
A data set is provided which was simulated using NONMEM. It consists of sparse observations (4 per subject) in 100 subjects after a single oral dose. The subject weight, age and sex are recorded.
NONMEM
Note: All files should be loaded from and saved to your Pharmacometrics Data\Population PK folder for this assignment.
- Use Windows Explorer to explore the My Pharmacometrics\Pharmacometrics Data\Population PK folder.
- Open the ka1.ctl file and look at the NM-TRAN control stream code.
- Start a NONMEM run using the ka1.ctl NM-TRAN control stream with:
nmgo ka1
- When the run is finished the results will be shown in a table format extracted from the NONMEM output listing.
- If necessary copy the vpc folder from M:\Apps\My Pharmacometrics\Pharmacometrics Data\Population PK to P:\My Pharmacometrics\Pharmacometrics Data\Population PK
- Open the ka1.ctl file in the vpc folder on your P: drive and change the parameter estimates to those you obtained from using nmgo.
- Save the ka1.ctl file.
- Change directory to the vpc folder from the DOS prompt.
- Create a visual predictive check by running the nmvpc command.
-
nmvpc
- Look at the VPCs in the vpc\CP.pdf folder. Do you think you have got a good fit?
- Make a copy of ka1.ctl file so that you can test a lag time model.
copy ka1.ctl ka1L.ctl
- Edit ka1L.ctl using a text editor (e.g. EditPlus) or notepad so that it looks like Figure 1..
start notepad ka1L.ctl$PROBLEM First order absorption without lag (WFN extended format)Figure 1. Code for ka1L.ctl.
$INPUT ID TIME AMT WT AGE SEX DV
$DATA tgsparse.dat IGNORE=#
$ESTIM METHOD=ZERO POSTHOC NOABORT
$COV
$SUB ADVAN2 TRANS2
$THETA
(0,1) ; POP_CL L/h
(0,10) ; POP_V L
(0,0.5) ; POP_KA 1/h
(0,0.5) ; POP_TLAG h
$OMEGA
0.09 ; PPV_CL
0.09 ; PPV_V
0.09 ; PPV_KA
0.09 ; PPV_TLAG
$SIGMA
0.01 ; RUV_CV
0.25 ; RUV_SD mg/L
$PK
CL = POP_CL*EXP(PPV_CL)
V = POP_V*EXP(PPV_V)
KA = POP_KA*EXP(PPV_KA)
TLAG = POP_TLAG*EXP(PPV_TLAG)
ALAG1 = TLAG
S2 = V
$ERROR
;CP=F
CP=A(2)/V
Y=CP*EXP(RUV_CV) + RUV_SD
$TABLE ID TIME WT AGE SEX CL V KA TLAG Y
ONEHEADER NOPRINT FILE=ka1L.fit
- Run ka1L.ctl and examine the results.
nmgo ka1L
- Make a copy of the better
model control file so that you can see if sex affects clearance:
copy ka1L.ctl ka1L_sexcl.ctl
- Edit ka1L_sexcl.ctl.
Add a new parameter, FFEMCL, to make CL in females a fraction
of that in males (Figure 2).
$PROBLEM First order absorption without lag (WFN extended format)Figure 2. Code for ka1L_sexcl.ctl.
$INPUT ID TIME AMT WT AGE SEX DV
$DATA tgsparse.dat IGNORE=#
$ESTIM METHOD=ZERO POSTHOC NOABORT
$COV
$SUB ADVAN2 TRANS2
$THETA
(0,1) ; POP_CL L/h
(0,10) ; POP_V L
(0,0.5) ; POP_KA 1/h
(0,0.5) ; POP_TLAG h
(0,1,) ; FFEMCL
$OMEGA
0.09 ; PPV_CL
0.09 ; PPV_V
0.09 ; PPV_KA
0.09 ; PPV_TLAG
$SIGMA
0.01 ; RUV_CV
0.25 ; RUV_SD mg/L
$PK
IF (SEX.EQ.0) THEN
FSEXCL=FFEMCL
ELSE
FSEXCL=1 ; male or missing
ENDIF
CL = FSEXCL*POP_CL*EXP(PPV_CL)
V = POP_V*EXP(PPV_V)
KA = POP_KA*EXP(PPV_KA)
TLAG = POP_TLAG*EXP(PPV_TLAG)
ALAG1 = TLAG
S2 = V
$ERROR
;CP=F
CP=A(2)/V
Y=CP*EXP(RUV_CV) + RUV_SD
$TABLE ID TIME WT AGE SEX CL V KA TLAG Y
ONEHEADER NOPRINT FILE=ka1L_sexcl.fit
- Run ka1L_sexcl.ctl and examine the results.
- You will notice that the objective function is worse than the Ka1L model. This means NONMEM is in a local minimum.
- Copy the final parameter estimates back to the ka1L.ctl file with the nmctl command:nmctl ka1L
- Open ka1L.ctl and copy the final parameter estimates to ka1L_sexcl.ctl then save ka1L_sexcl..ctl as ka1Lsexcl_b.ctl
- Run ka1L_sexcl_b.ctl and examine the resutlts.
- You can compare the results of all your runs so far with the commands:
nmobj Summary of obj values.
nmmbt Summary of one or more runs. Open the nmmbt_*.txt file with Excel.
- Experiment with changing the model and explore the influence of weight, age, and sex on the model parameters.
Learning
- Describe the models you used to describe the fixed effects of covariates.
- Justify in biological terms the influence of each covariate you found to be important.
- List the objective function values for each model with a brief description of each model.
- Describe how you decide on the best model.