Dear All,

I have a question related to selection of covariates in ANCOVA analysis.
I understand that baseline covariates can be selected based on their possible impact on the primary endpoint. I have a situation where treatment can impact another non-baseline variable (which can change during treatment) during the study which can also impact the primary endpoint?
Can I use that non-baseline variable as a covariate?

For example, if my primary endpoint is a continuous variable 'x' which is actually AUC of a parameter through the study. I know that my treatment also changes another variable 'y' which also can be measured as an AUC through the study. I know that as y is indirectly correlated to x. can I use 'y' as a co-variate to show bioequivalence of 'x' using ANCOVA?

My statistician says we can not use a variable which is non-baseline and getting impacted by treatment as a co-variate. Is he right in saying that?