Jump to content


Photo

Population PK Post Hoc Analysis

population pharmacokinetics

  • Please log in to reply
4 replies to this topic

#1 csheme

csheme

    Advanced Member

  • Members
  • PipPipPip
  • 59 posts

Posted 01 April 2018 - 04:18 PM

Hello,

 

I have a 2-cmpt pop PK model, and have generated the post-hoc output table by individual subject, which includes each of the thetas by ID. I also do have a corresponding eta table by ID.

 

My question is how do I setup my model to simulate exposures for each of the subjects in the study based on the individual post-hoc estimates for each subject. I would like to setup a fixed dose (126 mg given q3w for 6 cycles with PK simulations every 2 hours) for each of the subjects based on their (actual theta and eta) post-hoc estimates.

 

Previously I believe I was doing clinical trial simulations by accepting all fixed and random effects with my model, and adding a simulation table to include 50 random individuals, I added the dosing sheet, and put in seq (0,2260,2) with CObs as the output, I would then do descriptive stats on CObs and plot the median and also the shaded area band from the 5th to 95th percentiles (90% confidence). Feel free to see the image below.

 

zyemc5.png

 

This time around I am interested in the post-hoc analysis, but am not familiar with the approach for this. I would like to see the entire profiles, and generate PK parameters such as AUCtau for each actual subject ID based on the actual post-hoc theta and eta estimates (and not a random population of 50 theoretical subjects with the fixed population PK estimates as I had done before in the image above). You can see I also added the actual observations into the plot.

 

I imagine I will get a table (DV vs CObs) for each of the actual subject IDs, which I could then do a simple NCA with descriptive stats based on the post-hoc results.

 

Thanks a lot for your input, please see my model below. If you could please describe the process and also help me set this up in my model I would be much obliged.

 

https://www.dropbox....is.phxproj?dl=0


Edited by csheme, 01 April 2018 - 04:40 PM.


#2 soujanya234

soujanya234

    Newbie

  • Members
  • Pip
  • 4 posts

Posted 02 April 2018 - 03:44 PM

Hello,

 

I have a 2-cmpt pop PK model, and have generated the post-hoc output table by individual subject, which includes each of the thetas by ID. I also do have a corresponding eta table by ID.

 

My question is how do I setup my model to simulate exposures for each of the subjects in the study based on the individual post-hoc estimates for each subject. I would like to setup a fixed dose (126 mg given q3w for 6 cycles with PK simulations every 2 hours) for each of the subjects based on their (actual theta and eta) post-hoc estimates.

 

Previously I believe I was doing clinical trial simulations by accepting all fixed and random effects with my model, and adding a simulation table to include 50 random individuals, I added the dosing sheet, and put in seq (0,2260,2) with CObs as the output, I would then do descriptive stats on CObs and plot the median and also the shaded area band from the 5th to 95th percentiles (90% confidence). Feel free to see the image below.

 

zyemc5.png

 

This time around I am interested in the post-hoc analysis, but am not familiar with the approach for this. I would like to see the entire profiles, and generate PK parameters such as AUCtau for each actual subject ID based on the actual post-hoc theta and eta estimates (and not a random population of 50 theoretical subjects with the fixed population PK estimates as I had done before in the image above). You can see I also added the actual observations into the plot.

 

I imagine I will get a table (DV vs CObs) for each of the actual subject IDs, which I could then do a simple NCA with descriptive stats based on the post-hoc results.

 

Thanks a lot for your input, please see my model below. If you could please describe the process and also help me set this up in my model I would be much obliged.

 

https://www.dropbox....is.phxproj?dl=0

Hi,

Try using individual eta estimates as the covariates instead of population eta, that way you would get Cobs adjusted to their eta for each individual subject- see the sample code:

test(){
 cfMicro(A1, Cl / V, Cl2 / V, Cl2 / V2, first = (Aa = Ka))
 dosepoint(Aa)
 C = A1 / V
 error(CEps = 0.268573)
 observe(CObs = C * (1 + CEps))
 stparm(Ka = tvKa * exp(nKa))
 stparm(V = tvV * exp(nV))
 stparm(V2 = tvV2 * exp(nV2))
 stparm(Cl = tvCl * exp(nCl))
 stparm(Cl2 = tvCl2 * exp(nCl2))
 fcovariate(nV)
 fcovariate(nCl)
 fcovariate(nKa)
 fcovariate(nV2)
 fcovariate(nCl2)
 fixef(tvKa = c(, 0.00871573, ))
 fixef(tvV = c(, 6.42957, ))
 fixef(tvV2 = c(, 900.051, ))
 fixef(tvCl = c(, 0.0973144, ))
 fixef(tvCl2 = c(, 0.0582446, ))
 #ranef(diag(nV, nCl, nKa, nV2, nCl2) = c(0.46607611, 0.4057042, 0.044551053, 0.016906062, 0.46273008))
 ranef(diag(ndumm) =c(0.0001))
 
}

 

hope this helps!

Thanks,

Soujanya



#3 csheme

csheme

    Advanced Member

  • Members
  • PipPipPip
  • 59 posts

Posted 02 April 2018 - 04:51 PM

Thanks Soujanya for the quick reply. 

 

What about the theta for each individual? The way the model is written above is using the fixed population estimates.

 

I would like to simulate based off the post-hoc estimates for each individual, theta + eta respectively (i.e. volumes and clearances for each subject are unique). 

 

Thanks greatly.



#4 soujanya234

soujanya234

    Newbie

  • Members
  • Pip
  • 4 posts

Posted 02 April 2018 - 07:06 PM

Thanks Soujanya for the quick reply. 

 

What about the theta for each individual? The way the model is written above is using the fixed population estimates.

 

I would like to simulate based off the post-hoc estimates for each individual, theta + eta respectively (i.e. volumes and clearances for each subject are unique). 

 

Thanks greatly.

Ok.. I tried following code as test case and it worked- 

I have used the structural parameters from the post hoc estimates table from your model and removed the pop. estimates.

 not sure if it is accurate or not though.

test(){
cfMicro(A1, Cl / V, Cl2 / V, Cl2 / V2, first = (Aa = Ka))
dosepoint(Aa)
C = A1 / V
error(CEps = 0.268573)
observe(CObs = C * (1 + CEps))
stparm(Ka = Ka)
stparm(V = V)
stparm(V2 = V2)
stparm(Cl = Cl)
stparm(Cl2 = Cl2)
#fixef(tvKa = c(, 0.00871573, ))
#fixef(tvV = c(, 6.42957, ))
#fixef(tvV2 = c(, 900.051, ))
#fixef(tvCl = c(, 0.0973144, ))
#fixef(tvCl2 = c(, 0.0582446, ))
#ranef(diag(nV, nCl, nKa, nV2, nCl2) = c(0.46607611, 0.4057042, 0.044551053, 0.016906062, 0.46273008))
ranef(diag(ndumm) =c(0.0001))
}


#5 csheme

csheme

    Advanced Member

  • Members
  • PipPipPip
  • 59 posts

Posted 06 April 2018 - 12:37 AM

Thanks Soujanya for the the quick follow-up.

 

Could anyone confirm if this is the correct way to do post-hoc analysis of population pharmacokinetic data using NLME?

 

I am looking to simulate distinct individuals with the population PK model based on the individual post hoc PK parameters.

 

Thanks all, much obliged.







Also tagged with one or more of these keywords: population pharmacokinetics

0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users