I am trying to build a PKPD model and I have few questions: At each time point 6-9 mice are sacrificed, PK and PD samples are collected for each mouse. So in the end there is no longitudinal data for each mouse subject, but there is corresponding PK and PD data from each mouse at every time point. There is high variability from mouse to mouse especially in PD data (CV% 30-80%) at each time point. There are data for 3 dose levels. I am using Monolix to model and the following are ways I think to I should handle the data, but it is not working.
A. I ignore individual mouse subject and rather map dose-level to ID. This seems to work okay, with decent individual fitting. However, individual parameters estimated for each dose level are a bit different and SD of random effects (omega) have high relative standard error (RSE)%. Additionally, population prediction does not fit the data at each dose level.
B. To be able to simultaneously fit the data and estimate the same parameters for both dose levels, someone suggested I remove random effects on parameters. This results in worse data fitting.
I have the following questions:
- Is there a way to handle this kind of data so that the inter-mouse variability is reflected, and estimate relevant parameters for 2 (or more) dose levels at the same time?
- Is there a way to match each mouse’s PK to its PD observation even when there is no longitudinal data for each mouse?