Import estimates from multiple NONMEM runs

The $NMEXT and $NMXML blocks allow you to import estimates (THETA, SIGMA, and OMEGA) from a NONMEM run. So if you are running a population PK model in NONMEM and translating to mrgsolve, you should never need to copy the final estimate but rather read them in using these blocks.

If you are running sequential PK/PD or parent/metabolite models, you might need to pull estimates from multiple NONMEM runs. This is easy to do with multiple $NMEXT (or $NMXML) blocks.

1 Import the first (PK) model

First, grab the estimates from the PK model; let’s call it run 101. The $NMEXT block could look like this

$NMEXT
run = 101
project = "../model/nonmem"
root = "cppfile"

This will find the file 201.ext in the ../model directory which is taken to be relative to the location of the mrgsolve model file (using the root argument). If the paths are a little confusing, you can try installing the here package and using

$NMEXT
run = 101
project = here::here("model/nonmem")

This would require an Rstudio project file in the (project) root directory and then we would locate the NONMEM run relative to that project root (e.g. in <project-root>/model/nonmem).

Once the first model is imported, we now have access to THETA1, THETA2 etc from the PK model.

2 Import the second (PD) model

Now, we import estimates from the PD model run (let’s call it 201) using another call to $NMEXT

$NMEXT
run = 201
project = here::here("model/nonmem")
tname = "theta"

This is the second import in the same mrgsolve model file, so we change the handle for referencing the THETA estimates: now the PD model uses theta1, theta2 etc so that we can retain the THETA numbering for both the PK and PD models, remembering that the PD model uses theta1 (lower case) while the PK model uses THETA1 (upper case). Note that we have to change the handle (to something like theta) otherwise mrgsolve will throw an error because there would be two THETA1 in the parameter list.