Protecting the human rights of people who use psychedelics.

294 www.thelancet.com/psychiatry Vol 2 April 2015 reporting bias in treatment rankings. However, network meta-analysis does not make multiple comparisons that can be corrected in this way, and the study concerned does not in fact demonstrate any bias, but only shows that the posterior distribution of rankings is sensitive to the pattern of data–hardly a surprising result. In any case, conclusions on which treatment is best are never based on the posterior distribution of the ranks, which are notoriously unstable, but on the posterior expected eff ect size, which is what is reported in our study. Decisions based on imperfect evidence can always be wrong, but network meta-analysis is a method that helps identify the optimum decision in view of the evidence.