Nodes of real complex networks are topologically organized in local clusters or modules. These groups are generally named communities and intuitively defined as sub-graphs with a density of internal connections larger than the one of external links. In spite of the intense research activity of this subject, a firm mathematical definition of community is still missing. In this work, we introduce a new measure aimed to quantify the statistical significance of communities in networks. Extreme and Order Statistics are used to predict the statistics associated with clusters in random graphs, including confidence intervals. This allows us to quantify the statistical significance of communities in networks, detecting ``artificial\'\' clusters arising as topological fluctuations in graphs without intrinsic structure. The method is successfully applied in the case of real-world networks for the evaluation of the significance of their communities.