QRPEM covariates and BQL

Phoenix Certara Forums Modelling and Simulation
1

Good morning

I have some basic questions regarding QRPEM, please

I am trying to model a dataset that has parent drug and metabolites all in plasma

I have been trying to use QRPEM but I have couple of issues to confirm

  1. How would QRPEM handle BQL? would it be similar to Laplacian in terms of M3 method? or what concept would it rely on while handling BQL?

  2. Would it be possible to run step wise covariate search for categorical covariates? because for some reason, when I have categorical covariates, the engine assumes step 0 to have -2LL of zero and thus no covariates would be identified

Thanks a lot

2
  1. How would QRPEM handle BQL? …

M3 method is about handling BQL data during LL calculation for given parameters of some subject, So both Laplacian and QRPEM are applicable (the model is not Gaussian in such case)

  1. Would it be possible to run step wise covariate search for categorical covariates?

yes, it is possible. in built-in mode enable(N) statement is added automatically for each covariate fixed effect irrespective of the covariate type.

step 0 is no covariate scenario, in other steps the covariate effects are added sequentially

3

thanks Simon for your reply

If I may follow up on couple of points pertaining to the second question please

  • When I am trying to map the categorical covariates using QRPEM engine, If I added +1 for those, the model won’t run, while if I chose yes instead, it would run fine. Any reason why? or how shall I address such issue?

  • for the continuous covariates, for some reason, unlike Laplacian, QRPEM is not allowing me to have different values of the continuous covariates for each patient. So, for example, If I am trying to add creatinine as a covariate, and I had different measurements on different days, the model won’t run. It will only run if I had those values set to same number (like baseline creatinine and have this number copied over for all cells). How can I address such issue?

Appreciate your help and support

Best,

4

for qrpem the equation need to be easily separable into mu (mean) and variance

in the pmx jargon you might find something called mu modeling

qrpem would work with the interface with equations that are automatically mu modeled

you can type in different equations but then you will have to tell qrpem how to handle it or break your equation into different lines ( go into pml editing mode):

QRPEM can run with time varying covariate like this:

CLnocov =tvCLnocov*exp(nCL)

CL=CLnocov *(any equation with time varying or not, power hockey stick etc)