1 reply to this topic

[Teodora Dumitrescu]

Hello,

 

I am trying to construct a poppk model using a two compartment model with catenary chain absorption (n=2) and first order elimination. I want to set the rate constants (Ka1 and Ka2) in the catenary chain equal and add eta on one of the Kas. First I simply wrote into the code: Ka=Ka1:

test(){

    deriv(A = - (Cl * C)- (Cl2 * (C - C2)) + (Atr * Ka1))

    urinecpt(A0 = (Cl * C))

    deriv(Aa = - (Aa * Ka))

    deriv(A2 = (Cl2 * (C - C2)))

    deriv(Atr = (Aa * Ka)- (Atr * Ka1))

    Ka=Ka1

    C = A / V

    dosepoint(Aa)

    C2 = A2 / V2

    error(CEps = 1)

    observe(CObs = C * (1 + CEps))

    stparm(V = (tvV))

    stparm(Cl = (tvCl) *exp(nCl))

    stparm(V2 = (tvV2))

    stparm(Cl2 = (tvCl2))

    stparm(Ka = (tvKa) * exp(nKa))

    stparm(Ka1 = (tvKa1))

    fixef(tvV = c(, 360, ))

    fixef(tvCl = c(, 58, ))

    fixef(tvV2 = c(, 8537, ))

    fixef(tvCl2 = c(, 298, ))

    fixef(tvKa = c(, 0.64, ))

    fixef(tvKa1 = c(, 0.64, ))

    ranef(diag(nCl, nKa))

}

But when I run it, Ka and Ka1 come out with different values.

 

Second step was to just use a single Ka throughout the differential equations:

 

test(){

    deriv(A = - (Cl * C)- (Cl2 * (C - C2)) + (Atr * Ka))

    urinecpt(A0 = (Cl * C))

    deriv(Aa = - (Aa * Ka))

    deriv(A2 = (Cl2 * (C - C2)))

    deriv(Atr = (Aa * Ka)- (Atr * Ka))

    C = A / V

    dosepoint(Aa)

    C2 = A2 / V2

    error(CEps = 1)

    observe(CObs = C * (1 + CEps))

    stparm(V = (tvV))

    stparm(Cl = (tvCl) *exp(nCl))

    stparm(V2 = (tvV2))

    stparm(Cl2 = (tvCl2))

    stparm(Ka = (tvKa) * exp(nKa))

    fixef(tvV = c(, 360, ))

    fixef(tvCl = c(, 58, ))

    fixef(tvV2 = c(, 8537, ))

    fixef(tvCl2 = c(, 298, ))

    fixef(tvKa = c(, 0.64, ))

    ranef(diag(nCl, nKa))

}

In this case, the value of nKa is pretty decent (~0.16)

However, from what I understand, In NONMEM, if you give the two absorption constants the same name and add eta on it, the eta exponentially increases.

 

I also tried the following model:

 

test(){

    deriv(A = - (Cl * C)- (Cl2 * (C - C2)) + (Atr * Ka1))

    urinecpt(A0 = (Cl * C))

    deriv(Aa = - (Aa * Ka))

    deriv(A2 = (Cl2 * (C - C2)))

    deriv(Atr = (Aa * Ka)- (Atr * Ka1))

    C = A / V

    dosepoint(Aa)

    C2 = A2 / V2

    error(CEps = 1)

    observe(CObs = C * (1 + CEps))

    stparm(V = (tvV))

    stparm(Cl = (tvCl) *exp(nCl))

    stparm(V2 = (tvV2))

    stparm(Cl2 = (tvCl2))

    stparm(Ka = (tvKa)*exp(nKa))

    stparm(Ka1 = (tvKa))

    fixef(tvV = c(0, 100, 3600))

    fixef(tvCl = c(0, 58, 580))

    fixef(tvV2 = c(0, 8537, 85370))

    fixef(tvCl2 = c(0, 298, 2980))

    fixef(tvKa = c(0, 0.64, 6.4))

    ranef(diag(nCl, nKa))

}

This way,keeping the names of the absorption constants different, but setting them equal by using stparm(Ka1 = (tvKa)), allows me to add eta on only one of the two absorption transit constants. However, the nKa increases to 0.56.

 

Since I don't have a very good understanding of how Phoenix works, could you tell me which one of these models is correct?

 

Thanks!!

[Emily Colby]

Dear Dora,

 

I suggest building the model graphically to make sure it has the structure you're looking for (e.g., two absorption compartments), and switching to text for further refinement if necessary. It is hard to say which one of your models is "correct" based on the information given. Your first model does not appear to be doing what you want it to - it has two different fixed effect Ka's even though you set them equal, but only one random effect like you wanted. The second model appears to have only one Ka, which is a function of one fixed and one random effect. The third model appears to match the description of what you're looking for, as it has two Ka's described by one common fixed effect, with a random effect on one of the Ka's.

 

Best regards,

Emily