direction of influence: is NIC → BMI
If this is the case, I recommend constructing two separate models: one for the effect of BMI and the interaction between NIC and BMI, and another for NIC only.
Model 1: Effect of BMI and NIC-BMI interaction
...
-bsVars 'BMI+NIC+inter'
-qVars 'BMI,NIC,inter'
-qVarCenters '0,0,0'
-mask BrainMask_MNI_222.nii
-num_glt 2
-gltLabel 1 'BMI_main' -gltCode 1 'BMI :'
-gltLabel 2 'BMI_NIC_interaction' -gltCode 2 'inter :'
...
Model 2: NIC only
...
-bsVars 'NIC'
-qVars 'NIC'
...
-num_glt 1
-gltLabel 1 'NIC_main' -gltCode 1 'NIC :'
...
The rationale for this modeling statregy is detailed further here.
I corrected but no result
When you say "no result," do you mean that 3dMVM
1) failed to execute, 2) produced no output, or 3) yielded unexpected results?
Gang Chen