Everything I have read about MOST, and all the presentations I have seen, treat the three phases as if they are all essential. Yet, I wonder whether it is always necessary to conduct an evaluation trial.
Recently, thinking has begun to change about this. Some people working in the optimization methodology area agree that there are likely to be times when an evaluation trial is not needed. The big and interesting question is how to go about deciding whether or not an evaluation trial is needed. This is an active research area.