6 replies to this topic

[Graham Blakey]

I am undertaking individual compartmental modelling of plasma concentration data using the phoenix model object. I would like to adjust the resultant parameter estimates for CL & V for body weight, additionally I would like to investigate the output in terms of sex.  All these data are included in the source concentration datafile, I am though struggling to see how I can get these data to carry over into the model output files.  Is it possible, if so how is it achieved?

[bwendt@certara.com]

Hi Graham,

what you describe is essentially the general scheme of population modeling: you want to take a population estimate for each parameter and adjust for covariates like body weight. This is not possible with individual estimates since they have been fitted using single profiles where any adjustment of the final estimates with regards to e.g. body weight cannot be made or don't make sense. If you want to adjust you need to have a reference. There is no reference with individual fits.

Not sure if this answers your question, let me know if there are more.

 

Bernd 

[serge guzy]

Dear colleague

You have an option when clicking on "run option" which is "add table" that enables you to ask as one option "when covariate is set", there write all the covariate you want and fill all the other options that include observed data, for variable, the predicted level etc..

If you are trouble, let me know and I will prepare for you a project that shows you how to do it.

best

Serge

[serge guzy]

Are you talking about population fit or individual fit? The best if population fit is as a the next step to account for covariate relationship during the population stage. All this can be done with the interface but only in the population mode.

Let us know what you are trying to do.

Serge

[Graham Blakey]

Hi Serge

 

I am not doing population analysis though I agree that this is the way to explore covariate relationships in a more formal manner. The work I am doing involves fitting individual subject profiles. In a very much 'look & see' way I wish to summarise the estimated parameters by sex and other demographic characteristics.  This subject information is in the source datafile. What would be beneficial is these data being carried along with the subject through the analysis so that it is outputted as a column in the PK parameter (theta) output file.  It is possible to merge these demographic characteristics with the output file after the analysis though I was hoping there was a more elegant solution whereby it could be 'pre-programmed' prior to the analysis.

 

Kind regards

Graham

[Simon Davis]

Graham, a simple 'carry' for Phoenix model, particularly when running in individual mode is noted as an enhancement; QC5868.

 

   Simon.

[mittyright]

Hi Serge

 

I am not doing population analysis though I agree that this is the way to explore covariate relationships in a more formal manner. The work I am doing involves fitting individual subject profiles. In a very much 'look & see' way I wish to summarise the estimated parameters by sex and other demographic characteristics.  This subject information is in the source datafile. What would be beneficial is these data being carried along with the subject through the analysis so that it is outputted as a column in the PK parameter (theta) output file.  It is possible to merge these demographic characteristics with the output file after the analysis though I was hoping there was a more elegant solution whereby it could be 'pre-programmed' prior to the analysis.

 

Kind regards

Graham

Dear Graham, Dear All,

 

Please note that  even NCA carryover is not all-mighty, in case of time-dependent columns they will not carry at all

By the way I think we have very flexible workaround to do this in PHX with secondary parameters option

 

Please take a look at my example

test(){

  deriv(Aa = - Ka * Aa)

  deriv(A1 = Ka * Aa - Ke * A1)

  dosepoint(Aa)

  C = A1 / V

  error(CEps = 1)

  observe(CObs = C + CEps)

  stparm(Ka = tvKa)

  stparm(V = tvV)

  stparm(Ke = tvKe)

  fcovariate(wt)

  fixef(tvKa = c(, 2, ))

  fixef(tvV = c(, 1, ))

  fixef(tvKe = c(, 1, ))

 

  double(wt_sequence)

  sequence{

    wt_sequence = wt

    while(wt>0){

      wt_sequence = wt

      sleep(1)

    }

  }

  secondary(wt_secondary=wt_sequence)

  # secondary(wt_secondary=wt)

  secondary(tvKa_secondary=tvKa)

  secondary(tvV_secondary=tvV)

  secondary(tvKe_secondary=tvKe)

}

 

So I just copied all typical values to the secondary parameters. The things are going to be interesting with time-dependent covariates. Using sequence statement we can do amazing things, assigning to the secondary parameter the covariate with requested conditions. The code is presented here to use the last observation. By default the first observation will be used (uncomment the line 'secondary(wt_secondary=wt)' and comment the sequence statement to see this; you need a data set with with wt changing in time)

I was using Theo Model from the Examples folder with edited wt column.

 

All your theta values could be in secondary sheet with desired demographic data

Edited by mittyright, 28 October 2016 - 09:13 AM.