In the absence of head-to-head randomized studies, often indirect treatment comparisons are performed for reimbursement submissions. Often it is mentioned that in order to obtain unbiased estimates based on indirect comparisons the distribution of characteristics of the patients included in the different trials needs to be similar, as well as the study design. By means of directed acyclic graphs (DAGs), which are often used in epidemiology for inferences, it is explained that indirect and mixed treatment comparisons are biased when differences in patient characteristics and trial design do act as an effect modifier of the treatment effect. Furthermore, the graphs can be used to differentiate between heterogeneity, selection, and confounding bias. DAGs for indirect comparisons of RCTs are compared with DAGs for head-to-head randomized designs and meta-analysis of RCTs.