The network meta-analysis (also called mixed comparisons of treatments) is a powerful statistical technique that combines different studies to realize the analysis of multiple treatments or to estimate an indirect effect in the absence of a direct comparison. These studies are carried out by the development of a network of analysis, allowing to calculate the relative effects of all the treatments or interventions included in the network simultaneously and using techniques that estimate the direct and indirect analysis of the evidence. Inductions are comparisons of different treatments using data from different studies and using a comparator in common, either because these studies are not available or are of poor quality or if one wishes to compare numerous alternatives. In the network meta-analysis the mixed treatment comparison is based on a closed- loop network, which provides much more information and is less biased than open loops. Currently in closed loops or cycles, several statistical methods are used for its analysis, but in each study a unique statistical approach is used. The most frequent to date are the Bayesian methods, therefore is more important the analysis of the process of research and network created that obtaining a single weighted final measure. The objective of this narrative review is to describe the fundamental concepts of the network meta-analysis, utility, methodological considerations, the fundamentals of the analysis, the conformation of the network and its main limitations.