You may want to look at the CLASSDATA=dataset option along with EXCLUSIVE. You use that data set to specify the combinations of class variables you want to appear in the data and EXCLUSIVE removes the rest.
An example:
data classdataexample;
sex= 'M';
do age = 13,14;
output;
end;
sex='F';
do age= 12, 13;
output;
end;
run;
proc tabulate data=sashelp.class
classdata=classdataexample exclusive;
class sex age;
var height;
table sex*height*(min mean max),
age
;
run;
The resulting table does show missing values of Height statistics for the age/sex combinations that result but don't match the restricted age/sex combinations for the class variables.
You'll want to start with small sets and few variables to build things as when you get lots of elements you may have a hard time predicting all the interactions.
But you may well need to use missing on the class somewhere because of your data contents.
You may have to provide a small example of your data with some of the missing values and what you expect the proc tabulate output to look like. Sometimes your data just won't support some results without modification.