Dear Nonmem User's
I am analyzing data from multiple studies (7 studies). The data is one
compartment model with first order absorption (Advan2 trans 2).
When modeling the data I added Study as a covariate on Vd
While reading the nonmem users archive I found the topic about having third
level of random effects -- the inter study variability (ISV).
http://www.cognigencorp.com/nonmem/nm/99may311999.html
after reading the paper by Silvy Laporte and Pascal Girard JPS 89, 2000.....
I used Study as Fixed effect (FSE) since I have 7 studies (<20)
However I could'nt quite follow the logic of adding study on IIV (or how the
assumptions differed from adding study as a covariate on parameters (Vd) or
on IIV.
I did code both ways (on Vd and on IIV) and found that the OBFV improved
better with ISV on IIV than adding study as a covariate on Vd.
Next thing is that I already have age as a covariate on CL.
The studies are in people with different age groups and there is high
correlation amongst the covariates AGE, CRCL and STUDY in my dataset.
However since the drop in OBFV is significant (p< 0.01) after adding STUDY
as covariate (with AGE altready in the model) either on Vd or IIV, I am not
sure whether I should add Study as covariate in my final model ??
The fits dont improve a lot after adding study as covariate.
But when I add Study on IIV, IIV splits and becomes smaller (33% from 41%),
but no change in IIV when Study is added on Vd.
To summarize, whats the difference between adding Study on Vd or on IIV on
Vd ??
And whether it is reasonable to just exclude Study as covariate since I
already have AGE in my model (AGE and STUDY highly correlated) ???
Any inputs are appreciated.
TIA
--
--Navin