Combining two NONMEM models with the same PK using nonmem2rx

[harroj02]

I'm combining two different models from nonmem outputs. Each of them is a PK/PD model. Model 1 tracks PD1, and model 2 track PD2. Both are using post hoc estimates for the PK portion and are the same. I've used nonmem2rx to read in and valiate both independently. What is the best way to combine these models to simulate based on post hoc estimates?

What I tried:

• Read in PK/PD1 using nonmem2rx() to create mod1
• Create a table of posthoc estimates of PK/PD1 (ds1) based on data read in from mod1
• Manually create the portion of PKPD2 including only the PD2 portion (mod2) assuming that PK will come from mod1
• I renamed parameters, iiv terms, etc in mod2 to avoid collisions.
• Manually read in the post hoc estimates associated with PD2 (ds2) and rename them to be consistent with mod2
• Append mod1 and mod2
• Combine ds1 and ds2

I think this should work but I'm getting the following error:

Error in `rxSolveSEXP()`:
! when specifying 'thetaMat' the parameters cannot be a 'data.frame'/'matrix'.
Hide Traceback
    ▆
 1. ├─rxode2::rxSolve(...)
 2. └─rxode2:::rxSolve.rxUi(...)
 3.   ├─base::do.call("rxSolve.default", .lst)
 4.   └─rxode2:::rxSolve.default(...)
 5.     └─base::tryCatch(...)
 6.       └─base (local) tryCatchList(expr, classes, parentenv, handlers)
 7.         └─base (local) tryCatchOne(expr, names, parentenv, handlers[[1L]])
 8.           └─value[[3L]](cond)
 9. 

Since I'm using post hoc estimates, I don't think I need the thetaMat. Can I delete it? How do I do that?

Alternative, is there a better way to merge PKPD2?

[mattfidler]

You are right, since nonmem2rx also imports the covariance matrix (thetaMat) it interferes with the simulations that you are performing.

You can delete anything inside the model by:

1. Making sure that the model is a decompressed UI object
2. Removing it from the function
3. (Optionally) re-compressing the UI object.

In the case of nlmixr2/rxode2 5.0 this is simply changing the object from a list to a environment, but in earlier versions this was changing binary to environment.

A version compatible method to do this is below:

library(nlmixr2)
#> ── Attaching packages ───────────────────────────────────────── nlmixr2 5.0.0 ──
#> ✔ lotri         1.0.2     ✔ monolix2rx    0.0.6
#> ✔ nlmixr2data   2.0.9     ✔ nlmixr2lib    0.3.1
#> ✔ nlmixr2est    5.0.0     ✔ nlmixr2rpt    0.2.2
#> ✔ nlmixr2extra  5.0.0     ✔ nonmem2rx     0.1.8
#> ✔ nlmixr2plot   5.0.0     ✔ posologyr     1.2.8
#> ✔ rxode2        5.0.0     ✔ shinyMixR     0.5.2
#> ✔ babelmixr2    0.1.9     ✔ xpose.nlmixr2 0.4.1
#> ✔ ggPMX         1.3.2
#> ── Optional Packages Loaded/Ignored ─────────────────────────── nlmixr2 5.0.0 ──
#> ✔ babelmixr2        ✔ nonmem2rx    
#> ✔ ggPMX             ✔ posologyr    
#> ✔ monolix2rx        ✔ shinyMixR    
#> ✔ nlmixr2lib        ✔ xpose.nlmixr2
#> ✔ nlmixr2rpt
#> ── Conflicts ───────────────────────────────────────────── nlmixr2conflicts() ──
#> ✖ rxode2::boxCox()     masks nlmixr2est::boxCox()
#> ✖ rxode2::yeoJohnson() masks nlmixr2est::yeoJohnson()

one.cmt <- function() {
  thetaMat <- lotri(tka +tcl ~ c(1,
                                 0.05, 1))
  ini({
    ## You may label each parameter with a comment
    tka <- 0.45 # Log Ka
    tcl <- log(c(0, 2.7, 100)) # Log Cl
    ## This works with interactive models
    ## You may also label the preceding line with label("label text")
    tv <- 3.45; label("log V")
    ## the label("Label name") works with all models
    eta.ka ~ 0.6
    eta.cl ~ 0.3
    eta.v ~ 0.1
    add.sd <- 0.7
  })
  model({
    ka <- exp(tka + eta.ka)
    cl <- exp(tcl + eta.cl)
    v <- exp(tv + eta.v)
    linCmt() ~ add(add.sd)
  })
}

f <- one.cmt()

f$thetaMat
#>      tka  tcl
#> tka 1.00 0.05
#> tcl 0.05 1.00

f <- rxUiDecompress(f) # don't know if this is needed for 5.0

rm("thetaMat", envir=f$meta)

f <- rxUiCompress(f)

f$thetaMat
#> NULL

f

^{Created on 2025-12-02 with reprex v2.1.1}

[harroj02]

I'll give that a shot. Is there a way to remove ODEs from a model using piping?

[mattfidler]

I believe there is

[billdenney]

Yes, you can remove ODEs, if you know what you want to remove, but your model may not work the way you expect afterward:

library(nlmixr2)
library(nlmixr2lib)

mod_onecmt <- readModelDb("PK_2cmt_des")
mod_onecmt |>
  model(-d/dt(peripheral1))
#> ℹ add covariate `peripheral1`

^{Created on 2025-12-03 with reprex v2.1.1}

[mattfidler]

I think the issue with removing ODEs is you may need to remove groups of ODEs which are currently handled line by line, but need to be handled group by group for it to be more useful.