I sometimes wonder if the complexity of the human body doesn't stop us from seeing things that can have great positive effect on a set of people because it's counteracted by the effect on another set of people so the result in the whole is cancelled out. I now wonder if the statistic methods used in these studies take this into account.
All this to say that I approve of controlled self-experimentation, but you need to be very rigorous and brutally honest. Most people are not.
but it's tricky to figure out and i assume the consensus rules are good enough for most people