I have a question concerning the setup of a simulation after a successfull parameter estimation. Which “options” exactly do I have to pick in Phoenix 7.0 to start a pop simulation taking into account all parameters variabilities that were estimated before?
So what I did: i copied a pop model after successfull parameter estimation and pasted it to the workflow again.
In the pasted model I checked “Sim. / Pred. Check” in run options and entered the number replicates. Then I executed the model. Among other outputs I got a VPC - no warnings.
Here, I am wondering if, for the simulations, Phoenix takes into account the distribution of the estimated theta, the theta correlation/covariance, variance/covariance matrix, the Omega and so on??
If not, do I have to select some options?
In general, how could I check myself which sources of variability Phoenix uses in the pop simulation.
As you may have noticed, I am quite new to Phoenix and also POP PK modeling… my background is PBPK & QSP modeling using tools like PK-Sim/MoBi, Simcyp and Matlab.
Please note that the VPC is ignoring the theta distribution variance/covariance matrix it only uses the Omega and the epsilon estimates to do monte carlo simulation.
If you want to simulate with uncertainty i.e with the uncertainty (standard error on theta, omega, epslion) this is a long process you need to construct say a 1000 set of parameters form your theta var covar matrix and assing them to fake ID and do a simulation ( to start you can ignore the error on omega and epslion)
Trial Designer ( upcoming software) can simulate with uncertainty but I agree that the vpc should have this option included i.e. the user can ask whether to do a real posterior predictive check versus the dgenerativepredictivee check ignoring uncertainty. ( currently supported option)
There is a lot of ways to do it so no script that will run on all cases you will need to develop and tweak to fit your purpose.
Uncertainty is especially important for trial design so it won’t affect your submission per se most if not all VPC are done without uncertainty but when it comes the time to ask what if scenarios and to design new trials in new population ignoring uncertainty might be costly as your simulations might give you false impression that your trial will be highly successful.
