7 replies to this topic

[joybaker]

Hi,

 

I have a question related with the Enterohepatic Recycling (PML School Lesson 7), I have a model with OCA and its two metabolites which both have EHC. I used normal model (nonEHC) and EHC in different format (simply EHC and EHC with bile/gut compartments). However, the fitting always gives negative values though in the structure they all suppose to be positive (exp sturcture). I tried to give 0 limit but other parameter will turn to negative instead.

 

I have attached the file  Test drug_test sample.phxproj   3.84MB   224 downloads. Apologies it may look over parameterized. The model is also referenced from the FDA review model and I have scale down the dataset with only five typical subjects so it won't take a long time to finish the fitting but the problems stay in a similar way. I would like to figure out why the parameter is negative and how can I improve it. The FDA example  OCA POPPK model.pdf   588.45KB   363 downloads is also attached.

 

Thank you

Jo

 

[smouksassi1]

link to the full fda review doc:

https://www.accessda...0ClinPharmR.pdf

I was part of the team who did the OCA consulting project but this was not in the public domain.

We can help in specific software related questions

[joybaker]

Thank you for the information. The data and file are not from the this FDA  review though. And the questions are connected with the software about how I can set the parameters so that it won't give negative estimation for convergence. I hope this clarifies:)

link to the full fda review doc:

https://www.accessda...0ClinPharmR.pdf

I was part of the team who did the OCA consulting project but this was not in the public domain.

We can help in specific software related questions

[kniefort]

Thank you for the information. The data and file are not from the this FDA  review though. And the questions are connected with the software about how I can set the parameters so that it won't give negative estimation for convergence. I hope this clarifies:)

 

Hi Joy, with respect to the negative parameters, I believe the issue is caused by the way volumes for compartments A2 and A3 are specified:

 

V2_gly = V_OCA * KOG / KGO

V3_tau = V_OCA * KTO / KOT

 

Using the estimates from the paper,  the product of V_OCA multiplied by the ratio of rate constants gives V2 and V3 estimates of ~2200 and ~35 respectively, nowhere close to the reported values which are closer to V_OCA.  

 

I recoded your model to estimate V2_gly and V3_tau and it ran with no negative parameter values.  I did not go any further, but agree with you that your model is probably overparameterized.  Perhaps you can fix V2 and V3 and/or reparameterize from microconstants to clearances, or maybe others have ideas for how to code this.  

 

Hope this helps!

 

Keith

Attached Files

•  Test drug.phxproj   3.92MB   219 downloads

Edited by kniefort, 13 January 2022 - 06:47 PM.

[joybaker]

Thank you Keith, I think the major problem is over parameterization.

 

Another question to ask, if I have two switch point say have means at 4 and 12 h after dosing, is the following setting right?

 

 

double(Switch)

            Ri=4

            Ri1=12

            K_BT = Switch/TAU

            K_BG = Switch/TAU 

                       

                        sequence{

                        Switch=0;

                        sleep(Ri);

                        Switch=1;

                        sleep(TAU);

                        Switch=0;

                          sleep(Ri1);

                         Switch=1;

                         sleep(TAU);

                         Switch=0;

                        }

                   

     TAU=1.5  #TAU as 90 min

 

 

 

Jo

Hi Joy, with respect to the negative parameters, I believe the issue is caused by the way volumes for compartments A2 and A3 are specified:

 

V2_gly = V_OCA * KOG / KGO

V3_tau = V_OCA * KTO / KOT

 

Using the estimates from the paper,  the product of V_OCA multiplied by the ratio of rate constants gives V2 and V3 estimates of ~2200 and ~35 respectively, nowhere close to the reported values which are closer to V_OCA.  

 

I recoded your model to estimate V2_gly and V3_tau and it ran with no negative parameter values.  I did not go any further, but agree with you that your model is probably overparameterized.  Perhaps you can fix V2 and V3 and/or reparameterize from microconstants to clearances, or maybe others have ideas for how to code this.  

 

Hope this helps!

 

Keith

 

[bwendt@certara.com]

Jo,

 

it is not clear what you want to achieve with the switch construction, perhaps you can explain in more detail. Right now it is switching on both rate constants from 4-5,5 and 17,5-19 hours. I have no idea if that would be of any use.

 

Bernd

[joybaker]

The switch setting is based on the  PML school setting  (Lesson 7 ) but with a small twist that it removes the gut chamber.

 

Originally the meal was taken at 4 h after dosing, so the code is like this :

 

double(Switch)

 Ri=4

  K_BT = Switch/TAU

             

                           

                            sequence{

                            Switch=0;

                            sleep(Ri);

                            Switch=1;

                            sleep(TAU);

                            Switch=0;

                            }

                       

         TAU=1.5  #TAU as 90 min

 

The "switch" was only on at 4 h for the duration of TAU.

So what if the meal was taken at 4 h and 12 h after dosing, and the EHC is initiated at both 4 h and 12 h for the duration of 90 min?

 

Thank you,

Jo

 

Jo,

 

it is not clear what you want to achieve with the switch construction, perhaps you can explain in more detail. Right now it is switching on both rate constants from 4-5,5 and 17,5-19 hours. I have no idea if that would be of any use.

 

Bernd

Edited by joybaker, 22 January 2022 - 04:17 AM.

[bwendt@certara.com]

Jo,

 

you need to factor in the Tau period when calculating the intervals of meals.

Here is a code that I would use for that:

 

test(){
 double(Switch)
 deriv(A1 = - (Cl * C)- (Cld * (C - Ct))- (A1 * K1g) + (Agut * Ka))
 urinecpt(A0 = (Cl * C))
 deriv(A2 = (Cld * (C - Ct)))
 deriv(Abile = (A1 * K1g)- (Abile * Rate))
 deriv(Agut = (Abile * Rate)- (Agut * Ka))
 C = A1 / V
 double(dosetime)
 dosepoint(A1, dobefore={dosetime=t},idosevar = A1Dose, infdosevar = A1InfDose, infratevar = A1InfRate)
 tad=t-dosetime
 Ct = A2 / Vt
 error(CEps = 0.2)
 observe(CObs = C * (1 + CEps))
 Ri1=4
 Ri2=12
 rest=period-Ri1-Ri2-Tau
 Rate = Switch / Tau
 sequence{
  while(1)
  {
  Switch=0
  sleep(Ri1)
  Switch=1
  sleep(Tau)
  Switch=0
  sleep(Ri2-Tau)
  Switch=1
  sleep(Tau)
  Switch=0
  sleep(rest)
  }
 }
 stparm(V = tvV*exp(nV))
 stparm(Cl = tvCl*exp(nCl))
 stparm(Vt = tvVt)
 stparm(Cld = tvCld)
 stparm(K1g = tvK1g)
 #stparm(GBr = tvGBr)
 stparm(Ka = tvKa)
 fixef(tvV = c(, 9, ))
 fixef(tvCl = c(, 3.5, ))
 fixef(tvVt = c(, 21, ))
 fixef(tvCld = c(, 15, ))
 fixef(tvK1g = c(, 1, ))
 #fixef(tvGBr = c(, 1, ))
 fixef(tvKa = c(, 1, ))
 fixef(Tau = c(, 1.5 ,))
 fixef(period = c(,24,))
 ranef(diag(nV,nCl) = c(0.3,0.1))
}

 

Here is the outcome: