I presume you have created the individual subject % change maps via 3dDeconvolve, etc.
There are two obvious approaches. One is to divide the subjects into some small set of groups (young, middle, old). Then do ANOVA on these groups, using 3dANOVA. The other is to do linear regression (say) on % changes in each voxel, using age as the covariate, using 3dRegAna.
For further details, either consult the relevant manuals for these programs, or post some more. However, this week is the AFNI bootcamp here at the NIH, so we are all involved in teaching AFNI to 40 victims. Replies will be sluggish for a while.
bob cox