I commented on a network meta-analysis. A risk with these is that when all studies are put in a pot, also seriously flawed ones, such an analysis is flawed. In this case, and in Cipriani’s horrible network meta-analysis in Lancet, which I also describe, this is what happened. Studies with a drug-withdrawal comparison group (misleadingly called a placebo group, but they lead to abstinence depression, which is not a true depression but a drug harm) were included. Therefore, table 2 in this paper should be ignored (the one with a huge number of comparisons). The flawed studies contribute to overestimating the effect of pills and therefore also the effect of pills in the comparisons between pills and psychotherapy. What is indisputable is the authors’ conclusion that psychotherapy has an enduring effect, which pills do not have. This has also been shown in head-to-head meta-analyses where pills and psychotherapy have been compared directly (not in a network analysis). But the authors dared not conclude that we should not use pills.