[Accurate grading of pancreatic neuroendocrine tumors with Ki-67 index in fine-needle aspiration specimens: a comparative cytologic and histologic study].

To study the cytomorphologic features and determine whether pancreatic neuroendocrine tumors (PanNET) sampled by fine-needle aspiration (FNA) can be accurately graded based on the Ki-67 index when compared to surgical samples. Corresponding intraoperative (19 cases) or endoscopic ultrasound-guided (3 cases) FNA cytology and surgical tissue specimens were obtained from 22 tumors, which were reviewed and stained for Ki-67 proliferation marker. The cytological samples included more than 200 tumor cells. Samples were graded by scoring the Ki-67 positive index in accordance with the 2010 WHO criteria. The grading scores assigned to the FNA cytology samples were compared with the scores assigned to the corresponding histological samples. Concordance was achieved by using 5% (instead of 2%) as a cut-off value for defining G2 tumors. One cytological sample included less than 500 tumor cells was excluded in the concordance calculation. The cytological smears consisted of uniform, monotonous and isolated cells, loose cellular aggregates and rosette-like formations. Some tumor cells clustered around segments of capillaries. The cells demonstrated distinct cytoplasmic and nuclear features. Mitoses and necrosis were rarely seen. When traditional 2% Ki-67 index cut-off value were used to classify G2 tumors, the majority (86.4%, =0.812, <0.01) of FNA cytology samples and corresponding surgical tissue specimens demonstrated concordance. When a 5% cut-off value was adopted, the concordance rate was 95.5% (21/22, =1.000, <0.01). Similar concordance rates between the cytological and histological grades were achieved with threshold value of cytological assessment material set at more than 500 or 200 cells. The cytological Ki-67 index in adequate material (>200 tumor cells) is useful in grading pancreatic neuroendocrine tumors, and a cut-off value of 5% showed better predictive value compared with that of 2%. Accurate grading of PanNET is critical for predicting tumor biology, patient prognosis, and making informed decisions regarding patient management and treatment.