ABSTRACT
Objective To investigate the predictors for the preoperative diagnosis of pancreatic mucinous cystic neoplasm with invasive carcinoma (MCN-IC).Methods Clinical data of 132 patients with pancreatic mucinous cystic neoplasm ( MCN ) who underwent surgery and were pathologically diagnosed in Shanghai Changhai Hospital and General Hospital of Xinjiang Military Region from August 2000 to December 2013, including gender, age, medical history, clinical presentations, laboratory examinations and imaging findings and etc , were retrospectively analyzed .All cases were classified into two groups:MCN with noinvasive carcinoma ( MCN-nIC, including MCN with low-or intermediate-grade dysplasia and MCN with high-grade dysplasia ) and MCN-IC.The univariate and multivariate logistic regression was used to analyze the differences on laboratory examinations and imaging findings and the like to identify the predictors for the preoperative diagnosis of MCN-IC.Receiver operator characteristic ( ROC) curve was used to evaluate fitting performance and Hosmer-Lemeshow test was performed to evaluate goodness of fit .Results Of the 132 patients, 115 (87.12％) were MCN-nIC, 17(12.88％) were MCN-IC.Univariate analysis identified old age(≥60 years), abdominal pain, anorexia, GLU elevated, CEA≥5 ng/ml, CA19-9≥37 U/ml, unclear border of tumor , thick wall (>2 mm) , presence of mural nodules and absence of the septa as independent predictors for MCN-IC. Multivariate analysis identified old age (≥60 years), abdominal pain, CA19-9≥37 U/mL, unclear border of tumor, presence of mural nodules and absence of the septa as the predictors for MCN -IC.The maximal area under ROC ( AUC) was 0.947, which indicated that the fitting performance of the model was satisfactory and the goodness of fit was better (P=0.056).Conclusions MCN-IC had a generally low prevalence .Old age (≥60 years), abdominal pain, CA19-9≥37 U/ml, unclear border of tumor, presence of mural nodules and absence of the septa may predict the diagnosis of MCN-IC.