Phenobarbital_PML

Phenobarbital PML example

This example illustrates simple PK modeling for a one-compartment model with Cl (clearance) and V (volume) parameterization, and multiple doses per subject.

Data

The subject data are contained in an ASCII file (*.dat). The headers and the first subject data read as follows:

        pheno.dat 
          ## xid time dose  wt  apgr yobs
              1   0.0 25.0  1.4  7     .
              1   2.0   .   1.4  7   17.3
              1  12.5  3.5  1.4  7     .
              1  24.5  3.5  1.4  7     .
              1  37.0  3.5  1.4  7     .
              1  48.0  3.5  1.4  7     .
              1  60.5  3.5  1.4  7     .
              1  72.5  3.5  1.4  7     .
              1  85.3  3.5  1.4  7     .
              1  96.5  3.5  1.4  7     .
              1 108.5  3.5  1.4  7     .
              1 112.5   .   1.4  7   31.0

Column mappings

The ASCII file containing column mappings reads as follows.

          colspheno.txt 
          id(xid)
          time(time)
          dose(a <- dose)
          covr(wt <- wt)
          obs(cObs <- yobs)

The model

The PML model file reads as follows.

        phenophxexam.mdl 
          # One compartment model with IV bolus dosing
          # Cl, V parameterization with differential equation 
          # formulation
          Pheno(){
             dosepoint(a)
             covariate(wt)  #wt is a covariate for Cl and V
             deriv(a=-a*Cl/V)
             c=a/V
             fixef(
                tvCl=c(0, 0.0001,)
                wtCl=c(0, 0.005,)
                tvV=c(0, 0.1,)
                wtV=c(0, 1,)
                )
             ranef(
                diag(nCl, nV)=c(0.1, 0.1,)
                )
             stparm(
                Cl=(tvCl+wtCl*wt)*exp(nCl)
                V=(tvV+wtV*wt)*exp(nV)
                )
          #NONMEM initial estimate for sigma variance is 10
          #Phoenix initial estimate represents a standard deviation
          #rather than a variance, so the equivalent initial 
          #estimate is
          #eps1=sqrt(10)=3.16
          #  error(eps1=3.16)
          #  observe(cObs=c+eps1) 
          }

NONMEM control file

The equivalent model, written as a NONMEM control file (*.ctl), would read as follows.

        $PROBLEM PHENOBARB SIMPLE MODEL
        $INPUT ID TIME AMT WGT APGR DV
        $DATA pheno.dat
        $SUBR ADVAN6 
        $MODEL COMP=(CENTRAL,DEFOBS,NOOFF)
        $PK
           TVCL=THETA(1)+WGT*THETA(2)
           TVV=THETA(3)+WGT*THETA(4)
           CL=TVCL*EXP(ETA(1))
           V=TVV*EXP(ETA(2))
           S1=V
        $THETA(0, 0.0001,)(0, 0.005,)(0, .1,)(0, 1,)
        $OMEGA .1 .1
        $DES 
           DADT(1)=-CL/V*A(1)
        $ERROR
           Y=F+EPS(1)
        $SIGMA 10
        $ESTIMATION METHOD=1 PRINT=5 POSTHOC