34 replies to this topic

[amh79@georgetown.edu]

Hello all,

 

Relatively new user....

 

I have ETAs from a previous analysis and I want to simulated different dose regimens in this population. 

 

I froze all of the fixed effects from the previous analysis. Structural model was selected etc.

 

I have loaded my ETA worksheet into the random effects tab. I am sort of stuck from there. I followed the help guide for the PKPD and got only so far. 

 

What do I need in my main/dosing sections? 

 

How do I use ADDL in the main data sheet to get to 2-3 weeks of dosing?

 

When I go to predictive run tab; do I just need to simulate C and Cobs (what I am looking for here)?

 

 

Thanks,

-A

[smouksassi1]

It is unclrea what you are trying to do and there is several questions in your post:

If you want to reuse Etas ( or individual patient paramters) this means you are asking the question:
what would have happened in these specific patient if I change the dose.

 

etc.

if you use the simulation capabilties with monte carlo etas this is answering the question about the distribution of the outcome overall not in these specific patients.

 

 

Sharing a project would help also mentioning which version are you using 

 

bests,

Samer


 

 

[amh79@georgetown.edu]

Let me rephrase my question - 

 

I have a structural model, thetas, and etas etc. 

 

CAN I simulate concentration data from this with varying dosing regimens?

[Simon Davis]

Short answer is yes, can you post your project and clarify your version as Samer asked please?

[amh79@georgetown.edu]

I am using V7.0 .

 

I don't have a program...yet...I deleted most of my attempts since they didnt really generate any results.

 

I have tables of eta, theta, sigma etc. 

 

Is there a resource somewhere that outlines this better than the help manual?

[amh79@georgetown.edu]

Here is a model that will run, but wont generate anything close to reasonable results (all C <<<<<<<< 1).

 

 

Attached Files

•  Dose Sim.phxproj   357.27KB   390 downloads

[smouksassi1]

Dear Amh,

I opened the project and it seems the problem is that you have ADDL with empty Aa and mdv 1 you need one line like this no mdf phoenix know about this automatically. You can keep Cobs empty as a placehold no need to put zero
ID Time        Aa  ADDL

11006 0 85 14

11006 168 85

 

When I do this I get below

 

note that this is telling the dosing algorithm that you 

 

had a dose of 85 at time zero and then you had 14 other doses of 85 every II column or time interval constant you specify assuming you want BID This will result in doses at the folliowng times:


12 24 36 48 60 72 84 96 108 120 132 144 156 168

 
now you add one more doses at 168 so you will have double dose here if not intended then make ADDL 13 or remove this line.

if you want just to simulate at SS you can also do that using the SS flag.

 

and AUCss does not need all this it is just Dose/CL and it is not defined anywhere in your code.




 

 

Samer 

Attached Thumbnails

• 
• 

[amh79@georgetown.edu]

When I use my full data set of ETAs, it doesnt change the estimated dosing for each individual subject. Am I not mapping the random effects correctly? Do I need to specify the Omega matrix?

[smouksassi1]

You data input as it is, apply the same dosing to all subjects if you want several dosing regimens then each id should have a different dose/ADDL etc.

 

it is not clear what you are trying to do and what you are expecting. Omega will vary the randomly generated eta in simulation mode but it will not change your dosing scheme.

 

 

 

Samer

[amh79@georgetown.edu]

I have a model that has been validated, random effects and fixed effects determined. 

 

I want to look at what a different dosing regimen than what was used for the model - based on the model parameters - would look like in terms  of population exposures.

[smouksassi1]

ok so your question is you want to see how different dosing regimen behave in the same subjects:
Use whatever etas / individual pk parameters but changing the dosing scheme.

 

for this you can use individual pk parameters as a covariates and then change the dose to what you need

 

 

you can also test what would the response be using a different dosing regimen in a random sample of the same population ( simulate using new random effects using  omega matrix)

 

for this use a simulation/vpc run in addition to changing the dosing scheme so you generate new etas.

[amh79@georgetown.edu]

Is there an example of this somewhere?

[smouksassi1]

can you share your project with dummy data we will walk you through it

[amh79@georgetown.edu]

See attached.

Attached Files

•  QD.phxproj   3.46MB   393 downloads

Edited by amh79@georgetown.edu, 01 February 2018 - 02:02 PM.

[smouksassi1]

I have added a model that use the eta's as covaraite so you end up using each individual parameters to conduct simple simulation for the 24 subjects ( no random simulation of etas just using the etas you wanted).

 

You will get 24 different pk profile given the dosing regimen you have specified.

 

 

Attached Files

•  QD.phxproj   4.4MB   469 downloads

[Frank Gibbons]

This was an incredibly useful example. Thanks to OP ("A") and Samer for working it out.

 

I work in (preclinical) discovery, and think this is actually an *extremely* common situation. Perhaps the majority of Phoenix users may be trying simply to fit their clinical data, understand the effects of covariates, etc. But in discovery, we're still trying to work out how it all works: running short experiments, building up a picture over time, testing hypotheses. You might call it "counter-factual simulation" or "what if we had done this instead/next time?". Too bad it takes what seems like software contortions to get PHX to do it. Would be great to see this kind of thing covered in the manual, either as shown here, or as described by Samer, with resampling from omega.

 

Thanks both!

[smouksassi1]

you are most welcome. to be able to help and improve we need data related stories like these so we can come up with a scenario that walk you through a useful and satisfactory solution. I made note that we need some reference / teaching materials that showcase this workflow.

[Phennecat]

Hi,

 

I have a very similar question to this.  I have estimates from previous modelling. However, when I try to run the simulation with these estimates, I only yield data values of "-1" for all simulated concetrations at all time points.

 

See attached. The settings are trying to simulate one infusion dosing of 160 mg through 672 hours, but I will be using other dose levels and other dosing frequencies, once the "-1" issue is fixed.

 

Thanks,

Jerry

 

Edit: I figured it out. This seems to be a WNL naming issue.  Under "Run Options," the Y variable has to be the input name for C in the desired compartment.  Additionally, the input name cannot have an underscore (_), or else the program will truncate the name and read it incorrectly.  In the attached, I fixed this by renaming the central compartments from "C_par" and "Central_M1" to "Cpar" and "CentralM1"; then, I adjusted the Run Option Y variable to "Cpar, CentralM1." 

 

Edit2: removed the attached WNL file.

Edited by Phennecat, 24 October 2018 - 06:34 PM.

[sundar]

I have added a model that use the eta's as covaraite so you end up using each individual parameters to conduct simple simulation for the 24 subjects ( no random simulation of etas just using the etas you wanted).

 

You will get 24 different pk profile given the dosing regimen you have specified.

Hi, 

 

Apologies for reopening an old thread, could you please re-share the attachment -   QD.phxproj. 

 

Since i am getting error message while try to open. 

[smouksassi1]

what version of phoenix are you using I could open this project on my end.

the pml of the model is below.

 

 

 

test(){

cfMicro(A1, Cl / V, Cl2 / V, Cl2 / V2, first = (Aa = Ka))

dosepoint(Aa, tlag = Tlag)

C = A1 / V

error(CEps(freeze) = 1.47)

observe(CObs = C + CEps)

fcovariate(nCl)

fcovariate(nV1)

fcovariate(nCl2)

fcovariate(nV2)

fcovariate(nKa)

fcovariate(nTlag)

stparm(Ka = tvKa * exp(nKa))

stparm(V = tvV * exp(nV1))

stparm(V2 = tvV2 * exp(nV2))

stparm(Cl = tvCl * exp(nCl))

stparm(Cl2 = tvCl2 * exp(nCl2))

stparm(Tlag = tvTlag * exp(nTlag))

 

fixef(tvKa = c(11, 13.1, 15.1))

fixef(tvV = c(47.8, 139.4, 713.5))

fixef(tvV2 = c(180.7, 30786, 878120))

fixef(tvCl = c(330, 954.2, 4015))

fixef(tvCl2 = c(102.9, 4659, 15160))

fixef(tvTlag = c(0.031, 0.034, 0.041))

#ranef(diag(nCl, nV, nCl2, nV2, nKa, nTlag) = c(0.329, 0.35, 2.45, 4.87, 0.036, 0.01))

ranef(diag(ndumm) =c(0.0001))

}