The idea is to see whether an already existing drug can reduce dependence on cannabis. Cannabis is effectively made up by two components: one (THC) is the substance that gets you high and stoned (and, incidentally, yes: it can lead to psychotic episodes); the other (CBD) is actually much less harmful (in fact, the main clinical hypothesis is that it's beneficial, at least for users with addiction problems).
Now: "industrial" cannabis is mass-produced, typically by criminals, using massive exposure to light. On the one hand, this increases production by many-fold; but on the other hand, too much light destroys all the "good" CBD. So, the reason why cannabis is addictive and potentially not-so-good for you is that most people buy the equivalent of mass-produced Tesco not-so-much beef burgers, while if they bought the organic stuff that Hugh Fearnley-Whittingstall grows in his garden (that's still a metaphor $-$ I don't mean to imply that HFW is a junkie!) they'd be all fine...
From the more technical point of view, I think the study it's interesting because we'll use a Bayesian adaptive design, which will allow us to look at the data (on 3 active dosages and placebo) at several interim points. Based on the predictive distribution that each is the most effective, we'll (hopefully!) get to the point where the others are dropped and the best one is tested against placebo. In addition to that, I managed to include an element of quality of life in the study, so that we could give potential indications on cost-effectiveness too. Should be fun!
Please comment on the article here: Gianluca Baio's blog