Jump to content


Extravascular Model - Bioavailability Clarification and PML Code

extravascular bioavailabilty

  • Please log in to reply
4 replies to this topic

#1 Charvi



  • Members
  • PipPip
  • 19 posts

Posted 29 August 2018 - 06:39 PM

Dear All,


I know there have been several discussions around extravascular modeling and bioavaliability. However, I would like to clarify and confirm my understanding on how Phoenix handles bioavailability parameter:


- While using the in-built model for extravascular dosing - say for oral or SC absorption, unless the user specifically edits graphically or textually and inserts the bioavailability (F) parameter, Phoenix assumes it to be 1 or 100%?


- If edited graphically, i.e. say checking the box for bioavailability in the absorption compartment the parameter has to be specified in the parameter block in the code by editing it textually as well.


- And lastly, instead of checking the bioavailability box for the absorption model, can we code it in the equations directly? Can you share the example equations as to how will that look say for 1 CM model with oral absorption? For example one could multiply the input/dose with F; however with Phoenix since we generally use a dosing sheet, how would this work?



Thank in advance for your response and patience in addressing these most likely repetitive questions.






#2 Simon Davis

Simon Davis

    Advanced Member

  • Administrators
  • 1,160 posts

Posted 30 August 2018 - 08:18 AM

Charvi - you are correct that when using the in-built models for extravascular dosing bioavailability is essentially ignored unless you define an (F) parameter.


- If edited graphically, i.e. say checking the box for bioavailability in the absorption compartment the parameter has to be specified in the parameter block in the code by editing it textually as well. > YES.


When in both built in and graphical mode you can see the underlying code by looking at the model text tab - this is sa very useful learning optiona s you can see how various options are coded as you toggle them on and off.   You should also work through the NLME examples (even if you don't use/have an NLME license 95% = of the code/exercise is the same in individual modelling using PML in the the Phoenix model object.). See Help> Documents > NLME user guide.


For more information I would look at the PML and NONMEM to NLME forum that give some more detailed examples of code.


#3 smouksassi1


    Advanced Member

  • Members
  • PipPipPip
  • 179 posts
  • LocationMontreal

Posted 30 August 2018 - 02:05 PM

I found it more intuitive to code the model in PML like the following
fcovariate(route) # route 1 = oral route 0 = iv
dosepoint(Aa, bioavail = (frel))
stparm(frel = route== 1 ?  tvfrel : 1 )
the above code is saying use a column named route from my data as a covariate named route ( mapping will occur automatically if the name in your data is the same as the name you define here)
I have a dose into a compartment named Aa and it has a bioavailibility parameter that I want to name frel
stparm is saying:  if route equals 1 (oral) bioavailibility = tvfrel if not take it as 1 (reference IV by definition =1 )
( of course depending on your situation you might want to make sure tvfrel stay between 0-1 using an ilogit or other transformation.
if your reference is not IV then FREL can be higher than 1 if you add random effect on frel you need to constrain the whole thing.

#4 Charvi



  • Members
  • PipPip
  • 19 posts

Posted 30 August 2018 - 05:24 PM

Thank you Simon and Samer for your responses. They both were quite helpful.


I had gone through the NLME examples (I have and use the NLME licence)  , and hence wanted to confirm some of the observations that are specific to running models in Phoenix. I went through some of them in more detail and I found the Midazolam model example to be the most useful in terms of where and how to define F when writing a custom code etc. I just wanted to share, in case other people in the forum have similar questions. Samer, thank you for sharing the example as well, it makes things more clear - and I agree, I find it more intuitive to code in general and this provides another example.


Many Thanks,

#5 Gilles TUFFAL



  • Members
  • PipPip
  • 14 posts

Posted 17 May 2021 - 09:39 AM

Dee Samer, dear Simon,

I come back to forum discussion of way of integrating bioavailbility in a PML code as Theta (see Extravascular Model - Bioavailability Clarification and PML Code _ forum discussion Aug 2018 commented by you too)

 I had the opportunity to model SC/IV data intending to determine F through modeling and the model involved Tlag for SC. The job was done in NONMEM and you’ll find the rep file + the final table of parameters as joined doc. Of note it is a closed form model with Advan 4  since such model can be built without requiring to differential equation accepting ADVANs. Let’s consider it as reference model

I replicated this into Phoenix with the same inputs either as data or as initial value for Theta. I have some difficulties replicating NONMEM output. Basically, I have no covariance step despite in the base PML run , the theta tv were not that much far NONMEM ones. When I tried changing modeling setting (Advanced,  things went even worse or I didn’t find the right parameter to change.

Maybe can someone have a look at this. Is it a matter of code (I feel no not absolutely sure) or modeling settings. Which of them may influence covariance step success such as TOL, R/S matrix used in NONMEM




Attached Files

0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users