Dear NMusers,
I am developing a PK model for a drug. Previous studies showed that the PK of
this drug was bested described by a 3-compartment model. Due to clinical
limitation, the sample schedule in my study was not long enough to capture the
whole PK (Last blood samples were around 50-90 min while the elimination t1/2
from literatures is about 2-4 hours).
First I tried a 2-compartment model, the model is stable, but I found obvious
bias in CWRES vs Time plot. Trend line is higher than zero in 5-25 min, lower
than zero in 25-60 min, and then slightly higher at the end. I tried to add
covariates but this problem remained.
Then I tried a 3-compartment model, there was no obvious bias in all goodness
of fit plots. But the model became unstable (parameters change significantly
when adding ETAs or covariates), maybe a little over-parameterized (rounding
error, high shrinkage)? I can add one or two ETAs on peripheral compartment
parameters (Q3 or V2), but adding ETA for CL or V1 will result in significant
parameter change.
I am wondering if the three compartment model with only one/two ETAs on
peripheral compartment parameters (Q3 or V2) acceptable? If we want to explore
covariates for this model, especially on parameter V1 and CL, what should we do
if ETAs for CL/V1 were not included in the model?
Thanks,
Wen