issue with transit compart model

[lirui0321]

Hi Matt,

If transit compartment is model with your approach #1, then dosing into another compartment will not work. However, if I use build-in transit function, I can dose into a different compartment. Here is what I mean.

This does not work

mod <- function() {
  model({
    ## Table 3 from Savic 2007
    cl = 17.2 # (L/hr)
    vc = 45.1 # L
    ka = 0.38 # 1/hr
    mtt = 0.37 # hr
    bio=1
    n = 20.1
    k = cl/vc
    ktr = (n+1)/mtt
    ## note that lgammafn is the same as lgamma in R.
    d/dt(depot) = exp(log(bio*podo(depot))+log(ktr)+n*log(ktr*tad(depot))-
                        ktr*tad(depot)-lgammafn(n+1))-ka*depot
    d/dt(cen) = ka*depot-k*cen
  })
}

et <- et(0, 7, length.out=200) %>%
  et(amt=20, time=0, cmt=2, evid=1)

transit <- rxSolve(mod, et)

This works

mod <- function() {
  ini({
    ## Table 3 from Savic 2007
    cl  <- 17.2 # (L/hr)
    vc  <- 45.1 # L
    ka  <- 0.38 # 1/hr
    mtt <- 0.37 # hr
    bio <- 1
    n   <- 20.1
  })
  model({
    k           <- cl/vc
    ktr         <- (n+1)/mtt
    d/dt(depot) <- transit(n,mtt,bio)-ka*depot
    d/dt(cen)   <- ka*depot-k*cen
  })
}

et <- et(0, 7, length.out=200) %>%
  et(amt=20, time=0, cmt=2, evid=1)

transit <- rxSolve(mod, et)

I am assuming these two models are the same. Why one allows me to dose into central compartment, but another does not?

BTW, does the input n for function transit(n,mtt,bio) need to be an integer, or is any positive number fine?

Thanks!

Rui

[lirui0321]

I add an if statement, now it works. I guess when podo(depot) is zero, it causes some problems.

mod <- function() {
  model({
    ## Table 3 from Savic 2007
    cl = 17.2 # (L/hr)
    vc = 45.1 # L
    ka = 0.38 # 1/hr
    mtt = 0.37 # hr
    bio=1
    n = 20.1
    k = cl/vc
    ktr = (n+1)/mtt
    ## note that lgammafn is the same as lgamma in R.
    trans = 0
    if(podo(depot) > 0) trans = exp(log(bio*podo(depot))+log(ktr)+n*log(ktr*tad(depot))-
                         ktr*tad(depot)-lgammafn(n+1))
    d/dt(depot) = trans -ka*depot
    d/dt(cen) = ka*depot-k*cen
  })
}

et <- et(0, 7, length.out=200) %>%
  et(amt=20, time=0, evid=1, cmt=2)

transit <- rxSolve(mod, et)

[lirui0321]

One more question, with evid = 7, can I do a repeated dose like the following?

et2 <- et(0, 72, length.out=2000) %>% et(amt=20, evid=7, addl=2, ii=24)

[mattfidler]

Hi Rui,

BTW, does the input n for function transit(n,mtt,bio) need to be an integer, or is any positive number fine?

Any number should be fine.

One more question, with evid = 7, can I do a repeated dose like the following?

Yes, though I believe there is an underlying assumption that the absorpton through the transit compartment is (mostly) complete before the next dose starts.

[mattfidler]

I am assuming these two models are the same. Why one allows me to dose into central compartment, but another does not?

The reason is you use tad() which gives NA before a dose, if you use tad0() which gives 0 before a dose this gives the correct model:

library(rxode2)
#> rxode2 5.0.1 using 11 threads (see ?getRxThreads)
#>   no cache: create with `rxCreateCache()`

mod <- function() {
  model({
    ## Table 3 from Savic 2007
    cl = 17.2 # (L/hr)
    vc = 45.1 # L
    ka = 0.38 # 1/hr
    mtt = 0.37 # hr
    bio=1
    n = 20.1
    k = cl/vc
    ktr = (n+1)/mtt
    ## note that lgammafn is the same as lgamma in R.
    d/dt(depot) = exp(log(bio*podo0(depot))+log(ktr)+n*log(ktr*tad0(depot))-
                        ktr*tad0(depot)-lgammafn(n+1))-ka*depot
    d/dt(cen) = ka*depot-k*cen
  })
}

et <- et(0, 7, length.out=200) %>%
  et(amt=20, time=0, cmt=1, evid=7)

transit <- rxSolve(mod, et)
#> ℹ parameter labels from comments are typically ignored in non-interactive mode
#> ℹ Need to run with the source intact to parse comments

plot(transit, depot, cen)

^{Created on 2025-12-10 with reprex v2.1.1}

As a note, the dosing compartment should be 1 still with a "phantom" evid.

[mattfidler]

You can still dose into other compartments, though; there is no reason why you couldn't have a mix of iv and oral with transit compartment:

library(rxode2)
#> rxode2 5.0.1 using 11 threads (see ?getRxThreads)
#>   no cache: create with `rxCreateCache()`

mod <- function() {
  model({
    ## Table 3 from Savic 2007
    cl = 17.2 # (L/hr)
    vc = 45.1 # L
    ka = 0.38 # 1/hr
    mtt = 0.37 # hr
    bio=1
    n = 20.1
    k = cl/vc
    ktr = (n+1)/mtt
    ## note that lgammafn is the same as lgamma in R.
    d/dt(depot) = exp(log(bio*podo0(depot))+log(ktr)+n*log(ktr*tad0(depot))-
                        ktr*tad0(depot)-lgammafn(n+1))-ka*depot
    d/dt(cen) = ka*depot-k*cen
  })
}

et <- et(0, 7, length.out=200) %>%
  et(amt=20, time=0, cmt=1, evid=7) %>%
  et(amt=20, time=2, cmt=2)

transit <- rxSolve(mod, et)
#> ℹ parameter labels from comments are typically ignored in non-interactive mode
#> ℹ Need to run with the source intact to parse comments

plot(transit, depot, cen)

^{Created on 2025-12-10 with reprex v2.1.1}

[mattfidler]

This is also the reason why the first model didn't work at all because NA breaks the ODE

[lirui0321]

Thank you for your help, Matt!

If I'd like to reset and dose with a transit compartment, since there is not an option like EVID of 4 for the typical dosing, I guess I will need to do something like this. Do I get this right?

Rui

et <- et(0, 7, length.out=200) %>%
 et(time = 0, evid=3) %>%
  et(amt=20, time=0, cmt=1, evid=7) 

[mattfidler]

The transit model really just uses tad0() so if you use reset and dose with a dose of 0, it should reset the time after dose as well as the system.

I don't know offhand if evid=3 by itself resets the tad0() function.

[lirui0321]

Thanks, Matt!

I am trying to model a few individuals in a crossover study, so will need to reset the model before the 2nd dose. You mentioned reset and dose, do you mean using amt = 0, evid = 4?

[mattfidler]

amt=0 and evid=4 should work followed a little later (like 1e-6 or something) by a dose.

[lirui0321]

amt=0 and evid=4 should work followed a little later (like 1e-6 or something) by a dose.

Got you. Thanks, Matt!