For the quantitative covariate, is it a between-subject variable (i.e., each subject has one value)? If so, you can take all the beta values from the TENT basis functions and treat them as the levels of a within-subject factor in the input for 3dMVM. Read the following paper (in press) in case you want to know more details about the approach:
[
afni.nimh.nih.gov]
Gang