To explore a simple and reliable non-invasive distinguishing system for the pre-operative evaluation of malignancy in pancreatic cystic neoplasm (PCN).This study first enrolled an observation cohort of 102 consecutive PCN patients. Demographic information, results of laboratory examinations, and computed tomography (CT) presentations were recorded and analyzed to achieve a distinguishing model/system for malignancy. A group of 21 patients was then included to validate the model/system prospectively.Based on the 11 malignancy-related features identified by univariate analysis, a distinguishing model for malignancy in PCN was established by multivariate analysis: PCN malignant score=2.967 × elevated fasting blood glucose (FBG) (≥6.16 mmol/L) ± 4.496 × asymmetrically thickened wall (or mural nodules ≥ 4 mm) ± 1.679 × septum thickening (≥2 mm)-5.134. With the optimal cut-off value selected as -2.8 in reference to the Youden index, the proposed system for malignant PCN was established: septum thickening (>2 mm), asymmetrically thickened wall (or mural nodules>4 mm), or elevated FBG (>6.16 mmol/L, accompanying commonly known malignant signs), the presence of at least one of these 3 features indicated malignancy in PCN. The accuracy, sensitivity and specificity of this system were 81.4

作者：Xiaoyong, Shen；Di, Lu；Xiao, Xu；Jianguo, Wang；Jian, Wu；Sheng, Yan；Shu-sen, Zheng

来源：European journal of radiology 2013 年 82卷 11期