Help with Defining the null hypotheses for Manova

Hi all,

I'm trying to test the combination of 4 different treatments therefore I have 16 different groups with different combinations of the treatment.
I have 4 DVs that represent a single construct.

I would like to proove (or statisticlally show):
1 - what is the best treatment
2 - that the combination of treatments is not linear - meaning that sometimes giving treatment A+B is worse that treatment A or Treatment B.

Since Manova null hypothesis is simply that the groups are identical, rejecting this hypothesis doesnt tell me which is best.
Can someone please help me define the hypothesis.

Many thanks,