**From: KOLLIAG@bms.com
**

Dear NONMEM users,

I would like to use NONMEM to do some population modeling and I have a couple of questions, since this is a problem that does not seem to be typical of the examples in the books that I have available.

First, is the data item specified correctly for the following multiple dose steady state, infusion type data? Second what kind of structural model would be appropriate for these data, and is it reasonable to assume that the ADVAN1-2 /Trans1-2 could give an adequate model for such data? The final objective is to do population modeling, after combining 5 similar data sets, to estimate clearance and volume of distribution, and investigate the possible effects of covariates (age, gender, concomitant medication).

For about 15 subjects, we have transdermal patches applied daily for six weeks (one patch per subject per day). The patch is removed at the end of the day and a new patch is applied. At the end of each week, when steady state has been attained, the last patch of that week is removed and PK measurements are obtained. Then a set of new patches (possibly different size) are applied for another week, and pk measurements are obtained at the end of the second week, and so on until the last week. So, the size of the patches (and doses to be delivered) remain the same within each week, but can differ in different weeks.

I treat these data as a constant rate infusion problem with multiple doses, and I run it as steady state data I used the folowing data item for one subject: DV is concentration in ng/mL, and time is in Days.

```
-------------------------------------------------------------
```

Subject Dosemg Days DV AMT RAteperday ADDL II EVID
-------------------------------------------------------------
1 1.25 1 . 1.25 1.25 6 1 1
1 1.25 8 0.46 . . . . 0
1 1.25 8 . 1.25 1.25 6 1 1
1 1.25 15 0.48 . . . . 0
1 1.25 15 . 1.25 1.25 6 1 1
1 1.25 22 0.17 . . . . 0
1 2.50 22 . 2.50 2.50 6 1 1
1 2.50 29 0.90 . . . . 0
1 3.75 29 . 3.75 3.75 6 1 1
1 3.75 36 0.68 . . . . 0
-------------------------------------------------------------

and tried some of the basic models but I am not getting good fits, so I question whether the way the data step was specified was correct.

If someone has any advice, comments, on the above specification, or suggestions for how to model such data, I would really appreciate it.

Thanks in advance.

Georgia Kollia, Ph.D.

Biostatistics and Data Management

Bristol-Myers Squibb,

Princeton, NJ

******
**

**From: alison@c255.ucsf.EDU (ABoeckmann)
**

RE: mail from Georgia Kollias:

The doses in Georgia's data file are multiple transient infusions, not steady state doses (as PREDPP understands such doses). If she believes that they lead to steady-state, then she can describe them as SS doses this way:

```
------------------------------------
````ID `

DOSE TIME DV RATE EVID SS
------------------------------------
1 1.25 8 . 1.25 1 1
1 1.25 8 0.46 . . 0
1 1.25 15 . 1.25 1 1
1 1.25 15 0.48 . . 0
1 1.25 22 . 1.25 1 1
1 1.25 22 0.17 . . 0
1 2.50 29 . 2.50 1 1
1 2.50 29 0.90 . . 0
1 3.75 36 . 3.75 1 1
1 3.75 36 0.68 . . 0
------------------------------------

Note that a type 1 ("reset") SS record is present at the same time as each observation, implying that dosing with this rate has been going on"forever", and therefor the calculations are be done using steady state formulas.

I tried 2 little control streams, one for her original data, and one for the above, and got virtually the same predictions.

I don't say that ADVAN1 is the right model for a transdermal patch - I am only looking at the dosing here.

With Georgia's original data file:

$PROB not

$INPUT ID DOSE TIME DV AMT RATE ADDL II EVID

$DATA datanotss

$SUBR ADVAN1

$PK

K=1

$ERROR

Y=F+ETA(1)

$TABLE TIME

With the above data file:

$PROB not

$INPUT ID DOSE TIME DV RATE EVID SS

$DATA datassx

$SUBR ADVAN1

$PK

K=1

$ERROR

Y=F+ETA(1)

$TABLE TIME