CannaBayes trial
Our paper on the CBD trial has just been published in The Lancet Psychiatry (and found its way through the media, for example here or here). Basically, the objective was to determine whether (a specific dose of) CBD was a safe treatment for cannabis addition (I’d already talked about this here and here).
This has been quite a long process — I think even before I got involved there has been lots of work put in. I came into this study in 2013(!), I think, as the funder of the research proposal had suggested a Bayesian adaptive design (so the study came my way — though I hadn’t done any Bayesian trial; but I was the Bayesian in the room…).
I’ve really enjoyed being part of the team (all super nice people) and also the whole process. In the study, we’ve started with three active doses and placebo, with a view of using posterior probabilities given interim data to modify the design (in the event, we dropped the lower dose because it showed a very low probability of success — that is being the most effective dose).
The modelling included two main outcomes — the number of days abstinent (self reported) and the levels of a given metabolite in urine (which can indicate actual consumption of cannabis). What I think was a nice feature of the modelling structure (and building on the stuff I do in health economics — you may think you’re looking at cost-effectiveness planes, here…), my main representation was in terms of joint posterior distributions for the outcomes. I think this is cool and actually very informative, because it tells the whole story in terms of how much better one dose can do in comparison to another — in this case there is also a dynamic element and the different shades of colouring indicate the weeks in the follow up.
I think the study will be taken forward perhaps — there is probably quite some potential for CBD as a drug treatment to prevent the worst complications of cannabis addiction. And we manage to get the journal on board with some of the technicalities (in terms of how the results have been presented, which is perhaps non-standard, from the usual perspective of the trials that they are used to reporting)…