The trick in defining moderation is knowing what the extreme values are, which is why I find reports on optimal body mass indexes interesting.
We've done these experiments in animals and those experiments suggest an extreme starvation diet leads to the longest lifespan.
http://norvig.com/experiment-design.html
which mentions some of the trade-offs involved in different forms of study designs.
The history of the prescribed drug Rimonabant (Acomplia)
http://en.wikipedia.org/wiki/Rimonabant
illustrates some of the problems with using animal studies to establish safety of interventions for human beings. The animal studies have to be done, and they are generally very helpful, but some aspects of human behavior are sufficiently complicated and not analogous to animal behavior that what looks safe and effective for animals turns out not to be beneficial for human beings.