lag time in model

[filippo-ciceri]

Hi,

I am just wondering how to incorporate lagtime into the model.
The lag, alag commands seem to work only when referenced to the central compartment (lag(central) <- lagtime), instead the depot one (as shown here).

Also, which are the unit of measure of the estimated parameter? Hours?

Code example here below:

two.cmt.allo.lag <- function() {
  ini({

    tk01 <- log(1)
    tcl <- log(500)
    tv <- log(5e+004)
    tk12 <- log(1)
    tk21 <- log(1)
    
    lagtime <- log(50)
    
    eta.cl ~ 0.5
    eta.v ~ 0.5
    prop_err <- c(0,0.5)
    
  })
  model({
    
    D <- Dose
    s <- Study
    
    cl <- exp(tcl + eta.cl) * ((WT/70) ** 0.75)
    v <- exp(tv + eta.v) * (WT/70)
    
    k01 <- exp(tk01) 
    k10 <- cl/v
    k12 <- exp(tk12)
    k21 <- exp(tk21) 
    
    d/dt(cent) = k10 * depot - (cl/v + k12) * cent + k21 * per
    d/dt(depot) = -k01 * depot
    d/dt(per) = k12 * cent - k21 * per
    
    lag(cent) = exp(lagtime)
    
    cp = cent/v
    cp ~ prop(prop_err)
    
  })
}

PKmodel_12 <- nlmixr(two.cmt.allo.lag, data.modelling, est="focei")

Thank you very much for the help!!
nlmixr has been so far invaluable for my work!

Filippo

[mattfidler]

Hi @filippo-ciceri lag(cent) or alag(cent) should be correct. The time units of the lag time would be the same time units as the input time.

[filippo-ciceri]

Just to be 100% clear: I am talking about lag time for absorption of the compound. For this reason I am a bit confused that lag(central is correct instead of lag(depot).

[mattfidler]

I thought you were talking about syntax. Most often oral absorption has a lag. So a lag time on depot would make more sense.

[filippo-ciceri]

Thank you. I was a bit confused because the model reaches a good fit with lag(cent), however I was expecting lag(depot) to be more appropriate.

[mattfidler]

Well, lag() will only affect doses in whatever compartment you use. The saem algorithm may allow this to be optimized though it doesn't make much sense in the model itself.

[filippo-ciceri]

Yes, I forgot to mention that I was modeling an oral dosing. This is why I thought lag(depot) was the better option.
The second part of your answer is a bit more obscure. What exactly can be optimized using the saem algorithm?