I should use the estimate statements to estimate paired coefficients and judge them with t-value.

Here is an example with simulated data. group=1,2 indicates the first and second sample.

data t1;
   do i = 1 to 30;
       group=1;
           x=rannor(12345);
           y=1+1*x+rannor(12345);
           output;
       group=2;
           x=rannor(12345);
           y=1+2*x+2*rannor(12345);
           output;
        end;
run;

proc glm data=t1;
   class group;
   model y= group group* x/solution noint;
   estimate  'group1-group2' group 1 -1;
   estimate  'group1x-group2x' group*x 1 -1;
   run;
   quit;

thanks a lot