What proc you used?
You cluster for variables or observation?
And In the output there is some statistical index (such as CCC) will display the variance of this obs / variables accounts for whole data's variance.
Maybe you can use primary component analysis for the multi-variables data to see which variable is the most important.
BTW.Just as Patrick said why not post it in ' Statistical Procedure' forum,There is man named lvm will give you more help, He is an expert about these.
In my opinion.
There is ' proc varclus' ( the name I have forgetted) which can do cluster analysis for variables which has the same principle with proc cluster but its object analysised is variables not obs.
And you also can do primary component analysis for these variable to see which variables influent the whole data's variance.