Consider the following extension of the population model from introduction. Each of 59 neonates was given some loading dose of the drug, had a plasma drug level measured about 2 hours later, was given maintenance doses about every 12 hours thereafter, had possibly one trough level measured during maintenance dosing, and certainly one level measured some time after the last maintenance dose. Due to the long half-life of the drug, for the purposes of analysis all doses can be regarded as intravenous bolus doses. The weight and APGAR score of each individual are available as concomitant information. The APGAR score (1-10) is a measure of a neonate’s health at birth.

The one compartment linear model is implemented using ADVAN1. TRANS2, thus Cl and Vd are used for PK parameters and are described using an exponential model, with typical values proportional to weight. For Vd, the proportionality constant takes two possibly values depending on the APGAR score with threshold APGAR=2.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21

  $PROB PHENOBARB POPULATION DATA
  $DATA DATA1 (6F7.0)
  $INPUT ID TIME AMT WT APGR CP=DV
  $SUBROUTINES ADVAN1 TRANS2
  $PK
  ;# CLEARANCE AND VOLUME PROPORTIONAL TO WEIGHT
  ;# PROPORTIONALITY CONSTANT FOR VOLUME DEPENDS ON APGAR
  CALLFL=1
  CL=THETA(1)*WT*EXP(ETA(1))
  TVVD=THETA(2)*WT
  IF (APGR.LE.2) TVVD=THETA(3)*TVVD
  V=TVVD*EXP(ETA(2))
  S1=V
  $ERROR
  Y=F*EXP(EPS(1))
  $THETA (0,.0047) (0,.99) (0,1.0)
  $OMEGA .05 .03
  $SIGMA .02
  $ESTIM MAXEVAL=500 PRINT=5
  $COVAR
  $TABLE ID TIME AMT WT APGR

Note that the reserved variable assignment CALLFL=1 denotes that the PK routine is called only once for each individual record, at the first event record of the individual. This is sufficient as the weight is constant for each individual, and more efficient than the default setting that evaluates $PK for every data record. This is an example of calling protocol. A more verbose alternative is to specify

1
2

 $PK (ONCE PER INDIVIDUAL RECORD)
;#...

See mixture model for a modeling approach when assuming the weight is not measured.