The problem with MEND is that it requires a lot of effort and memorizing from the patients to follow the program. Many drop out for that reason. A simple pill is much more convenient.
In a post earlier this morning you said that the super responders (25%) were in a normal variation range implying that the high scores were immaterial relative to the low scores. All just normal variation within the group.
Now you say this: "So maybe they shouldn't be grouped together since it is quite different thing to have two points improvement than to have five point improvement."