With the aim of standardizing Ki-67 immunohistochemistry, we assessed interobserver and interlaboratory variability of the Ki-67 labeling index and Ki-67 score among eight general pathologists for 24 gastrointestinal stromal tumors (GISTs) and 12 leiomyosarcomas, which were predominantly of the gastrointestinal (GI) tract, mesentery and retroperitoneum, based on a review of a tissue microarrays subjected to immunohistochemistry with antibodies for Ki-67. For Ki-67 immunostaining of mesenchymal tumors of the GI tract, including GISTs, differences were seen in the scores given by regional hospitals. Conversely, for two categories of the Ki-67 labeling index, namely <10% and ≥10%, concordance of the Ki-67 score between microscopic observation and image analysis, and between the observers, was good, but it was not good for the other four categories of the index for <5%, 5–9%, 10–29%, and ≥30%. The concordance of the Ki-67 scores between the observers in two categories was higher using the Ki-67 pre-stained tissue microarrays (TMAs) within each participating institute than that using the Ki-67 stained TMAs between the participating institutes. The reproducibility of a 10% cut-off value for the Ki-67 labeling index to predict the prognosis of GISTs was relatively high, but there is an urgent need to standardize the staining technique.