The problem is how to measure improvement? I have personally seen cases where the person stopped treatment because of "no noticeable improvements", and it was disastrous. In certain conditions (bipolar, for example), just the fact that the patient is stable and somewhat functional is already a huge victory. Wouldn't that be an "improvement" over the alternative?