**From gonv14@udcf.gla.ac.uk Thu Jul 6 08:01:55 1995
**

Dear NONMEM users,

Is it possible to instruct NONMEM/PREDPP to work from a concentration event at time zero, rather than a dose event? This would be useful in cases where prior dosage history is unknown.

Also, I was interested to note question on choosing between FO and FOCE method. If no obvious improvement in fit can be seen from plots, how else would you choose.

Thanks,

Nikki J.Bruce

Dept. Medicine and Therapeutics

University of Glasgow,

Scotland

E-mail: gonv14@udcf.gla.ac.uk

******
**

**From alison Thu Jul 6 13:19:22 1995
**

Nicola Bruce asks:

> Is it possible to instruct NONMEM/PREDPP to work from a concentration

> event at time zero, rather than a dose event? This would be useful in

> cases where prior dosage history is unknown.

PREDPP can indeed be given a concentration at time zero, when the prior dosage history is unknown. Here is an example that involves prior doses, as well as endogenous drug.

------------------------------------------------------------------------

CONTEXT: EXAMPLE

DISCUSSION:

In this example, an oral "drug" is an endogenous substance, and there is an unknown dosing history prior to the observation period (i.e., prior to time zero). This example illustrates how three sources of drug can be modelled: pre-existing endogenous drug, pre-existing drug from an unknown prior dosing history, and drug from known doses. Any combination of the three could be modelled without the others.

Endogenous drug production is assumed to be constant; that is, no feedback control of production is assumed. Thus endogenous drug is at steady-state, and, with linear kinetics, its effect is simply to add a constant increment to drug in the sampled compartment (the increment is modeled as theta(7)).

For the unknown dosing history, it is assumed that the subject is at steady state for this existing drug. This part of total drug is modelled by a Steady state infusion dose into the oral compartment, ending at time 0, and having an unknown rate. The result of the SS dose is to introduce drug into all compartments of the system, not just the central compartment, as it, unlike endogenous production, will be subject to subsequent decay. The unknown rate is modelled as theta(5). NONMEM will adjust theta(5) to best fit not only the "base- line" observation at time 0, but also the later observations.

Note that if samples are not taken sufficiently long after the time of the last dose ( > 4 half-lives), then theta(7) and theta(5) will not be separately identifiable. That is, the value of theta(7) is fixed by the residual concentration after all exogenous drug has disappeared.

The combined additive and ccv error model is used. Theta(8) estimates the ratio of the C.V. of the proportional component to the standard deviation of the additive component.

Any ADVAN/TRANS combination could be used. Population data could also be modelled in this manner.

$PROBLEM Example of pre-existing drug.

$INPUT ID TIME DV AMT SS II RATE

$DATA DATA1

$SUBROUTINES ADVAN4 TRANS5

$PK

AOB=THETA(1)

ALPHA=THETA(2)

BETA=THETA(3)

KA=THETA(4)

R1=THETA(5)

S2=THETA(6)

$ERROR

FP=THETA(7)+F ; adds endogenous component

W=(1+THETA(8)*THETA(8)*FP*FP)**.5

Y=F+W*ERR(1)

Note that, if there are other doses into compartment 1 with modelled rates, it is necessary to assign a value to R1 conditionally. E.g.,

IF (TIME.EQ.0) THEN

R1=THETA(5) ; rate for SS infusion record at time 0

ELSE

R1=....; rate of other kind of dose

ENDIF

Note also that the combined additive and ccv error model can also be modelled using two random variables:

Y = F*(1+ERR(1)) + ERR(2)

A fragement of the data follows. Record 1 is the SS infusion for the pre-existing drug, which ends at time 0. Record 2 is the baseline observation. Record 3 is an oral bolus dose. Record 4 is an observation.

1 0 . 0 1 0 -1

1 0 62.2 . . . .

1 0.01 . 95 . . .

1 0.50 235.93 . . . .

REFERENCES: Guide V, section 8 (p. 80)

REFERENCES: Guide VI, section III.F.5 (p. 26)