Using the include= option in the model statement will keep the first n variables listed.
With your code, for example, to keep the *kp_x1* and *kp_x2* variables, you would write:
proc logistic work.data;
model y (event ='1')= kp_x1 kp_x2 x3 x4 x5 x6 x7 / selection=score best=3 include=2;
run;