Do you know what your dominant leakage mechanism is? Are you talking Vgs=0
or (in this analog section) subthreshold drain current? DIBL, which is worse at
extremely short channel but might respond to a moderate reduction in worst case
drain voltage? GIDL, which might only need you to not drive the gate so far
negative in an analog switch? Edge leakage which might want a channel-stop
or a closed geometry?
Well, to fix the "problem" (or natural characteristic) you have
to see or know what it responds to. Which might be predicted
if you knew the mechanism.