Hi all, could you help me explain how to calculate the percentage of explained and unexplained variability in IIV that has been reported in many publications?
example see this paper *Evaluation of covariate effects on pharmacokinetics of monoclonal antibodies in oncology (nih.gov)
for example if final model Volumn is calculated in this way:
V = tvV*exp(nV) * (WT/medianWT)^dVdWT * (AGE/medianAGE)^dVdAGE
if tvV=2
exp(nV) = 0.7
(WT/medianWT)^dVdWT for a median WT is 0.2
(AGE/medianAGE)^dVdAGE for a median AGE is 0.1
The explained variability by percentage is (0.2+0.1)/(0.2+0.1+0.7) = 30%?
and unexplained by 70%
is that correct?
Then how to account for a categorical factor ie. gender?
what are your omega V at base model and after inlcusion of covaraites ?
e.g. Base model:
V = tvV*exp(nV)
Omega =0.15
CV % = sqrt(exp(omegav)-1) = 40.2 %
e.g. weight added
V = tvV*exp(nV) * (WT/medianWT)^dVdWT
Omega V =0.09
CV % = sqrt(exp(omegav)-1) = 30.7 %
so weight explained 31 % of the BSV 40/30 ~ 1.31
The percent of explained variability has nothing to do with the coefficient of your covariate and whether it is continous or categorical
you can do the computation after the additiona of each covariate or at the end to see how much total variability you have explained.
other related topics
https://www.page-meeting.org/pdf_assets/9033-poster.pdf
https://pubmed.ncbi.nlm.nih.gov/22767340/
https://link.springer.com/article/10.1007/s10928-012-9273-1