15 replies to this topic

[LLLi]

Hi,

 

I am trying to use the graphical model of NLME to recreate PK5 application in Dan Weiner's book but got nothing. I can understand the ASCII code in the book but the graphical model of NLME is confusing for me. Can someone help me? Thanks!

 

The raw data is attached. function 1 means plasma and function 2 means urine. There is only 1 subject.

 

Attached Files

•  plasma and urine.xlsx   8.23KB   370 downloads

Edited by LLLi, 18 August 2016 - 05:42 PM.

[serge guzy]

Dear LLi

I fit your model to the data you sent using Phoenix graphics.

The project is attached.

What you need to remember is that the urine compartment is in amount and therefore we need to define an expression for the urine concentration. I assume the volume added at each time is the same. Volume of urine is assumed to be a structural parameter. Without that I do not know how you can handle the problem. I am used to get concentration in urine after washout but here it cumulates.

Let me know if it looks good to you.

best Regards

Serge

 

PS: You did not give me the dose. I assume a dose of 1  PK_plasma_cum_urine_data.phxproj   470.7KB   410 downloads

 

 

[serge guzy]

Dear LLLi

I have a quesiotn about the problem you asked an asnwer for.

The volume for urine should not be a volume fo dsitirbution but you should get at each collection time the informaution about the volume collected.

What duid you get here?

At each time I see the urine concentration increasing which is usaully what I would see if we would measure the cumulative amount. Then I am confused as I do not know exactkyt what you measured.

Whatever it is, what I showed you in the project is how you can mnage this kind of problem.

 

Serge

[serge guzy]

Sorry for the typo errors. I am after 35 hours flight.

Serge

[LLLi]

Dear Serge,

 

Thank you for your prompt reply after such a long flight.

 

Actually the urine data is amount in urine. Sorry for the confusion. So I revised your model (delete S1; use amount instead of concentration;use the actual dose; add Fe) and it worked perfectly (please see the attachment). 

 

But I still have some questions:

1) In your model, the green box CObsurine is an continuous block and the Curine white box is an expression box for input? For each obs block added, there must be an input expression for it? I do not see an expression block for CObs. 

 

2) A silly question. I can insert an observation box and an expression box but can not connect them. How to connect them just like you did in your model?

 

Thank you very much!

 

LLLi

Attached Files

•  PK_plasma_cum_urine_data LLLi.phxproj   1.06MB   407 downloads

[serge guzy]

Dear Serge,

 

Thank you for your prompt reply after such a long flight.

 

Actually the urine data is amount in urine. Sorry for the confusion. So I revised your model (delete S1; use amount instead of concentration;use the actual dose; add Fe) and it worked perfectly (please see the attachment). 

 

But I still have some questions:

1) In your model, the green box CObsurine is an continuous block and the Curine white box is an expression box for input? For each obs block added, there must be an input expression for it? I do not see an expression block for CObs. 

 

2) A silly question. I can insert an observation box and an expression box but can not connect them. How to connect them just like you did in your model?

 

Thank you very much!

 

LLLi

 

Dear LLLI

Originally I had to define an expression to get the urine concentration because you gave me cumulative concentrations in rine which of course do not make any sense. The urine compartment is in amount and the observation was in concentration. Therefore you cannot connect amount with concentration.

Therefore an expression was needed to define the response in the same units as the observation.

 

Then you could connect the predicted response with he observed response which is what is needed.

 

Now that the response is in amount, you can link directly the A0 compt with the observation which is what I did in the attached project.

I get the same answer and all is good.

 

You can connect an expression to an observation. This is what I did.

I do not understand. Pu the cursor on the expression until you see Aurine , left click , keep the mouse pushed and drag to the observation box where you see written Curine and release the mouse.

Best

Serge

Attached Files

•  PK_plasma_cum_urine_data.phxproj   1.37MB   421 downloads

[LLLi]

Dear LLLI

Originally I had to define an expression to get the urine concentration because you gave me cumulative concentrations in rine which of course do not make any sense. The urine compartment is in amount and the observation was in concentration. Therefore you cannot connect amount with concentration.

Therefore an expression was needed to define the response in the same units as the observation.

 

Then you could connect the predicted response with he observed response which is what is needed.

 

Now that the response is in amount, you can link directly the A0 compt with the observation which is what I did in the attached project.

I get the same answer and all is good.

 

You can connect an expression to an observation. This is what I did.

I do not understand. Pu the cursor on the expression until you see Aurine , left click , keep the mouse pushed and drag to the observation box where you see written Curine and release the mouse.

Best

Serge

Dear Serge,

 

Hope you had a good rest during the weekend : )

 

Thank you for your reply. Your new model is simple but works perfect.

 

Here I attached a project which is from Dan's book (PK10) since it is also about simultaneously fitting data  graphically (iv plus po data). PO is a 2 compartment model with Tag and Bioavailability (F1). I can run the model but the results are confusing. The F1 is more than 1 and the some results are also different from those in the book. Could you please take a look at my project?

 

Thank you very much!

 

LLLi

Attached Files

•  PK10 sim po iv.phxproj   523.06KB   385 downloads

[serge guzy]

Dear Serge,

 

Hope you had a good rest during the weekend : )

 

Thank you for your reply. Your new model is simple but works perfect.

 

Here I attached a project which is from Dan's book (PK10) since it is also about simultaneously fitting data  graphically (iv plus po data). PO is a 2 compartment model with Tag and Bioavailability (F1). I can run the model but the results are confusing. The F1 is more than 1 and the some results are also different from those in the book. Could you please take a look at my project?

 

Thank you very much!

 

LLLi

Dear LLLi

I fixed your model.

You had 1 error

1: The Aa dose must be at t=0

2: I add a concentration box for C2 as I fit also using clearance parametrization. It is OK for micro constants not to use the concentration box as the mass balance does not include V2.

best Regards

Serge

Attached Files

•  PK10 sim po iv.phxproj   2.05MB   398 downloads

[serge guzy]

Last model in the project I attached just now is the same model you wanted to use (micro constants and F without transformation).

best

Serge

[LLLi]

Dear Serge,

Thank you for your reply.

I added bio=ilogit expression into the last model as you did in another model and ran the new model. The bio expression had no effect on the results. So I have a question. What is the purpose of the bio expression?

Thank you!

LLLi

[serge guzy]

Dear Serge,

Thank you for your reply.

I added bio=ilogit expression into the last model as you did in another model and ran the new model. The bio expression had no effect on the results. So I have a question. What is the purpose of the bio expression?

Thank you!

LLLi

Dear LLLi

If you have a population model with variability in bioavailability, the ilogit transformation as F=ilogit(flofit+nflogit) which transform a normal distributed parameters into a 0-1 bounded one (F) will insure that each individual will have its F between 0 and 1.

If you sue the individual mode, you have no random effects and often there is no need to use the ilogit function because you estimate only the fixed effect which will be between 0 and 1 if your model enable estimating F.

Best

Serge

[LLLi]

Dear Serge,

 

Thank you for your reply about the Bio expression issue.

 

Another question. Why we need insert a parameter (F1) for bioavailability but do not need for Fe in NLME?

 

Thank you!

LLLi

Edited by LLLi, 29 August 2016 - 09:31 PM.

[serge guzy]

Dear LLI

Fe is also constrained between 0 and 1 and therefore you could use the same ilogit expression for Fe.

It is useful if Fe is a population parameter as you cannot control the variability to result in a 0 to 1 number while you can control the fixed effect (mean here) by putting boundaries in the fixed effect tab (except QRPEM that does not allow boundaries)

Best

serge

[LLLi]

Hi serge,

 

I have some questions about the Fe estimate (again ). My model results show Fe is larger than 1. I tried to use Bio expression as you mentioned below. But it is till larger than 1.

 

Background: compound X, 500 mg, data (the plasma concentration and amount of parent drug in urine, and amount of metabolite in urine), MM metabolism.

 

I tried PHX NLME and got the parameters (Clr, V, Vmax, Km). Please see the 1st model in the project attached.

Then in the second model, I added Fe for the parent drug but the Fe is larger than 1. I don't know what is the problem.

 

Would you please take a look at my project?

 

Thank you!

LLLi

Attached Files

•  MM kinetics Drug&Metabolite in urine.phxproj   692.56KB   344 downloads

Edited by LLLi, 02 December 2016 - 08:44 PM.

[serge guzy]

Dear LLLi

I made some changes in your model to make sure Fe will be bounded between 0 and 1.

Look at the last model.

I created an expression I called Fe Fe=ilogit(Felogit)

Felogit is the parameter you estimate and Fe is calculated.

It comes out that Fe=1 as Felogit is estimated to be about 18 which means Fe=1.

If this is not the right value, then that is another problem.

 

Best Regards

Serge

Attached Files

•  MM kinetics Drug&Metabolite in urine.phxproj   1MB   377 downloads

[LLLi]

Dear serge,

 

Thank you for your reply.

 

Why did you connect Fe expression with Pu elimination compartment? Because there are two elimination compartments (Mu and Pu)? I tried to elimination the connection and rerun the model, the tvFelogit is about 21 and tvFe is 1.

 

As my understanding, we need to check Fe for Pu elimination compartment, then add a Felogit as a parameter (not Fe), and next add an ilogit expression. Please correct me if I am wrong.

 

Thank you!

LLLi

Edited by LLLi, 05 December 2016 - 03:49 PM.