There is great interest to study how co-expression gene networks change across tissues. However, the reproducibility assessment of these studies is challenged by a lack of fully confirmatory experiments from independent researchers. While an increment in the number of studies with expression data for several tissues is expected, statistical measures are still needed to assess the reproducibility between studies. We identified a gap in the statistical literature concerning the assessment of agreement between studies across numerous conditions. The gap precluded us to test, using standard statistics, the level of agreement between the GTEX (RNAseq) and BRAINEAC (microarray) studies to distinguish the structure of co-expression networks across four brain tissues. We propose a generalization of a classical measure of agreement, Cohen's kappa, derive its distributional characteristics and determine its reliability properties. In the gene expression studies, our generalization of kappa showed full agreement for genome-wide networks in BRAINEAC benchmarked against GTEX, and highest agreement for brain specific pathways. Our highly interpretable measure can contribute to anticipated efforts on reproducibility research.