How to implement B2 method in mrgsolve ?

[Ravua1992]

Hello,

B2 method is inclusion of the observed baseline response as a covariate acknowledging the residual variability> when I tired to implement this in mrgsolve I simulated negative values.
Are there any good examples how to do this ?

This is how I am doing it currently:

mrgsolve in $MAIN:
double RTOT1 = RTOT1_OBS * (1.0 + ETA(15) * CCVRESi) + ETA(16) * ADDRESi;

nonmem

RTOT1 = RTOT1_OBS*(1.0 + ETA(15)*CCVRES) + ETA(16)*ADDRES

Thanks,

[kylebaron]

Hi @Ravua1992 -

I'm assuming that, when ETAs are 0, RTOT1 comes out to be non-negative, right?

If that's the case, and simulating ETA results in (some) negative RTOT1, then you can resimulate ETA until RTOT1 is positive.

The code might look like:

double RTOT1 = RTOT1_OBS * (1.0 + ETA(15) * CCVRESi) + ETA(16) * ADDRESi;

while(RTOT1 <= 0) {
  simeta();
  RTOT1 = RTOT1_OBS * (1.0 + ETA(15) * CCVRESi) + ETA(16) * ADDRESi;
}

You will have to make sure your code is structured so that the most-recently simulated set of ETA are used for the rest of the ETAs.

Kyle

[curtisKJ]

Hey @Ravua1992

Just a note, I think you need to be careful to realize that NONMEM is allowing for negative predicted/simulated values. To then set up the mrgsolve model in a way that precludes negative values will induce bias at the near zero values and won't be consistent with what NONMEM would simulate/predict if you used that instead.