The need to support clinical decision making and cost-effectiveness analyses in medicine, despite a dearth of head-to-head treatment comparisons, has encouraged the development of methods enabling indirect comparisons of treatment alternatives, including network meta-analysis (NMA). Valid application of NMA requires close similarity of compared trials, including their design, patient characteristics, and methods of diagnosis and symptomatic assessment. When biological or other objective measures of outcomes are not available, as is the case in psychiatric disorders, subtle differences in characteristics of trials or participants may lead to unrecognized incoherence within a network and thus to inconsistent results. By considering comparative-efficacy analyses of psychotropic drugs in major psychiatric disorders as working examples, we underscore the risks of violating the fundamental transitivity assumption in the context of NMA and suggest precautions for creating a coherent network. We conclude that with thoughtful and critical application, NMA can add useful information concerning the comparative benefits, risks, and costs of specific treatments in psychiatry.