S. Oberndorfer noted in 1907 that the multiple unusual tumors in the small bowel were distinct clinical entities and proposed the term “carcinoid” (“carcinoma-like”), emphasizing in particular their benign features [1, 2, 25]. However, malignancy of this tumor group has been confirmed based on metastatic or survival rates in long term studies [26–28]. Some studies have discussed the tumor origins and suggested that these neoplasms may arise from a type of endocrine cells, known as amine precursor uptake and decarboxylation (APUD) cells [2–4]. Although the APUD cells were initially suggested to be derived from neural crest cells, it is now generally recognized that gastroenteropancreatic APUD cells probably arise from the endoderm [2, 29]. Tumor definition was initially based on typical morphological characteristics, and subsequent studies demonstrated argyrophilia-staining properties of the tumors using silver impregnation techniques, such as the Grimelius method [2, 30–33]. Moreover, the neuroendocrine origin of the tumors was definitively identified and assessed in detail using stains for neuroendocrine differentiation markers, chromogranin A, and/or synaptophysin [10, 34]. After these investigations by prognostic or diagnostic procedures, the term "carcinoid" has been regarded as a misnomer . Since these tumors are considered as cancers of the neuroendocrine system, a more adequate term “neuroendocrine tumors” is now widely used. Based on these lines of evidence, the WHO 2000 classification provided a rational approach to the nomenclature and classification of NETs of the digestive system , which subdivided them into WDET, WDEC, and PDEC . NETs have historically been classified according to the foregut, midgut, or hindgut derivation, and they have been indicated the prognostic differences. Some studies have also demonstrated the prognostic potential of IHC and TNM classification [5, 6]. Since long-term follow up studies have indicated the malignancy of these tumors, in 2010, a revised version of the WHO classification appeared . The new classification defines the entire group of tumors as NENs, which have been confirmed to arise from the neuroendocrine cell system since they are shared with marker proteins of this system [9, 10]. NENs are further categorized into NECs and NETs. NECs are morphologically similar to small cell carcinoma and large cell carcinoma of the lung, while NETs encompass neoplasms that were previously termed “carcinoid” or “atypical carcinoid” [7, 9, 10]. The NETs are subdivided by their mitotic index or Ki67 index into either G1 or G2 NETs. This revised classification is a simple and useful grading system based on the proliferative activity. However, the assessment of tumors with Ki67 index that is greater than 2% and less than or equal to 3% is still unclear. Despite this, the inter-observer differences in mitotic counts are larger than that of Ki67 index, and moreover, it is difficult to scan routinely at least 50 high power fields (HPFs) (1 HPF = 2 mm2), that is required in the WHO 2010 classification for evaluation of the mitotic index. Thus, the validity and reproducibility of Ki67 index are superior to those of the mitotic index . Tumors with a Ki67 index of <2% are classified as G1 while those of 3—20% are classified as G2. Because one cutoff value is used to divide continuous values into two groups, we validated the Ki67 index criteria of gastrointestinal NETs of the WHO 2010 classification with the aim of clarifying the assessment of tumors with Ki67 index between 2—3%. We performed computer-assisted cytometrical analysis of Ki67 IHC, which was established in several of our past studies [12, 13], using the WinRooF image processing software. Although it was not considered generally, a concordance study using other softwares indicated that digital image analysis and manual count were highly concordant and were acceptable standards for Ki67 assessment . Furthermore, in our experience, the number of cells could be more accurately and objectively counted with some devices. In this study, several problems occurred which were similar to those in our previous studies. One of them was that the non-tumor cells, such as lymphocytes or histocytes, occasionally showed positivity for Ki67. To resolve this, the non-tumor cells were eliminated with a touch pen by introducing a liquid crystal touch panel as our previous reports. After that, we evaluated whether the grading system could accurately predict metastasis or recurrence. Our results confirmed that 2.8% (or approximately 3%) is the best Ki67 index cutoff value for predicting metastasis or recurrence. If the cases were classified based on 3%, the sensitivity was 28.6% and the specificity was 89.5%, which is meaningful for predicting metastasis or recurrence because of the high specificity. On the other hand, due to the low sensitivity, metastases may be observed not only in G2 but also in G1 regardless of biological malignancy. In recent studies, the location of the primary tumor, size, invasion depth, and multiplicity of tumors were associated with metastasis or recurrence [26, 35], and in our study, it was confirmed that tumor size was correlative. It was also reported that the presence of Circulating Tumor Cells (CTCs) with cellular expression of epithelial cell adhesion molecule (EpCAM) was more predictive of clinical outcomes than the WHO grading system . However, the methods to isolate CTCs for molecular characterization are still being developed, and there are few studies regarding CTCs in patients with NETs [36, 37]. Thus, at present, the WHO grading system may be the most useful classification to predict metastasis or recurrence. In summary, division of NETs into G1/G2 based on Ki67 index of 3% was appropriate to predict metastasis or recurrence (positive predictive value = 33.3%, negative predictive value = 87.2%). To the best of our knowledge, this is the first study to validate the numerical criteria of the WHO 2010 classification of NETs and to indicate the appropriateness of the Ki67 index. Further investigation is required in order to support this finding.