From: "Miguel Alexandre Soares" <mbraga@ff.ul.pt>
Subject: Dose event problem
Date: Mon, 25 Oct 1999 11:32:26 +0100

Dear Collegues
I am trying to use the ADDL and II data item yet, I am not sure if I am using them correctly. Could you comment the lines 4 to 6 where I specify 3 observations but I don't define any of the 14 doses between them? And about lines 7 and 8 where I try to describe a regimen with to different dosages (175mg in the morning and 150mg at night), are they correct?
line ID Date Time Dose cp ADDL II alt

1 17 7-8-96 10:00 200 7 12

2 17 7-12-96 9:30 198.5

3 17 7-12-96 10:00 175 13 12

4 17 7-19-96 9:30 202.4 14

5 17 7-26-96 9:30 233.6 14

6 17 8-2-96 9:30 211.4

7 17 8-2-96 10:00 175 10 24

8 17 8-2-96 22:00 150 10 24

9 17 8-13-96 9:30 203.1

10 17 8-13-96 10:00 150 39 12

 

 

*****

 

 

Date: Mon, 25 Oct 1999 12:01:32 -0700 (PDT)
From: ABoeckmann <alison@c255.ucsf.edu>
Subject: Re: Dose event problem

It is a little difficult to see what is being done because the email was a bit garbled. I've tried to clean it up and omit the irrelvant "alt" data item. Everything looks ok. Line 3 specifies a total of 14 doses, 12 hours apart, starting at 10:00 on 7-12. I think the last dose will occur at 22:00 on 7-18, which is actually before the observations on lines 4-6. It would have been ok to specify more doses on line 3. The implied dosing is permitted to overlap the observations. As for lines 7 and 8, this appears correct.

When checking a complicated dosing scheme, I suggest that you (and all NONMEM users) perform a run with *NO* $ESTIMATION record. Look at a table of PRED vs. TIME with the initial estimates of theta. Make sure the predictions have the pattern of peaks and troughs and exponential decay that you expect to see. If necessary, insert some extra data records at various time points to see if the predictions follow the pattern that you expect. Such records must have EVID=2 ("other type") if you leave them in the data set during estimation, but if they are only there for a run with no $ESTIMATION, then they can look like observation records and have any value of DV, because the DV values are irrelevant when there is no estimation.

Alison Boeckmann