Pop PK Modeling: Etas for patients w/o Conc but with Dosing

Phoenix Certara Forums Modelling and Simulation
1

Dear,

I am doing a POP PK model (QRPEM) with a number of pts. Some pts have NO PK conc measured yet, but they do have dosing records in the dosing input sheet.

I have random effects on some PK parameters. No I realize looking at the eta table that NLME calculates for all of these pts the same random effects. As a consequence, some eta distributions get totally non-normal, see attached picture. All the nks on the left are from pts with no PK data but with dosing.

I have the following Qs

  • Why and how does NLME at all calculate etas for these pts, since they have no observed PK conc?

  • Will that not screw up my model and make it unsuitable for simulation as the variance of parameters gets inflated?

  • I cannot really delete these pts without PKobs, since in a 2nd step I want to go for a PK/PD model and thus I need these pts (they have PD data!)

Help would be greatly appreciated.

Thanks

Frank

Capture.PNG
Capture.PNG455×384 4.75 KB

2

Those Patients with zero concentration contributing to the PK model i.e. doses only and all PK records missing, non existent, or excluded by MDV =1 flag should have etas zeroed.
Can you share the project?

3

Hi smouksassi1,

thanks for the reply. I cannot share the project.

I checked over and over again, there is no PK record for these 16 pts. They do not appear in the PK conc input sheet.

However they do have dosing records and all of these 16 pts get some strange etas assigned.

Any idea what could be the problem?

4

How about emailing your project to support ?

5

I would first need to ckeck internally, the data is sensitive. I was hoping to get it solved w/o emailing/sharing the projcet.

6

how about submitting a project with obfuscated / invented data that reproduce the issue ? otherwise double check if you have some zero concentration ( zero and missing are not the same thing) if you want zero in your pk obs sheet use mdv =1 otherwise keep it empty as pml knows how to handle empty values.