Are the observations correlated across the responses? ie, did you measure the 5 outcomes on the same experimental unit, with units divided into two group? A repeated measures model might work then. You can use PROC GLIMMIX, fitting a separate mean to each level of response and you could compare the means using an LSMEANS statement with SLICEDIFF= and the appropriate ADJUST=.
Something like
proc glimmix data=test;
class group resp;
model y=group*resp;
random int / subject=id;
lsmeans group*resp / slicediff=resp adjust=tukey;
run;
might work. MIXED does not have the ability to adjust across only the sliced tests.