BACKGROUND: Endoscopic ultrasound is increasingly used for evaluation of pancreatic cancer. The potential of sonographic morphology to differentiate histology type and biological behaviour of pancreatic lesions is doubtful. METHODS: We prospectively studied 115 patients with focal pancreatic lesions on endoscopic ultrasound. Morphology was assessed using Olympus UM3/20/200 echoendoscopes. Histologic confirmation of diagnosis was obtained in all patients. RESULTS: Endoscopic ultrasound correctly diagnosed 18/34 benign and 77/81 malignant lesions. Sensitivity, specificity, accuracy, PPV and NPV for diagnosing malignancy were 95%, 53%, 83%, 83% and 82%, respectively. Endosonographic diagnosis of the lesions (% correct) were: pancreatic cancer, 84 (63.3%); chronic pancreatitis, 14 (71.4%); ampullary cancer, 9 (77.8%); cystadenoma, 5 (80%); ampullary adenoma, 2 (50%); acute pancreatitis, 1 (0). In 13 patients of chronic pancreatitis, diagnosed as cancer, diagnosis was based on absence of sonographic features of chronic pancreatitis (7) or suspected involvement of adjacent structures (6). In 3 patients malignancy was missed owing to features of chronic pancreatitis. Non-suspected neuroendocrine tumours were misjudged in all 10 cases using morphologic criteria as pancreatic cancer (8), cystadenoma and chronic pancreatitis. Accuracy for prediction of metastatic lymph nodes and an advanced pancreatic cancer stage (TxN1 or T3Nx) was 61% and 75%, respectively. On retrospective analysis, a lesion >2 cm, vessel ingrowth, absence of cystic spaces and absence of diffuse pancreatitis were associated with pancreatic cancer. CONCLUSIONS: While overall sensitivity was high, specificity of endoscopic ultrasound for diagnosis of malignancy was low, especially in presence of chronic pancreatitis. In addition, endosonography had only a limited potential to predict the histological type of lesions.
BACKGROUND: Endoscopic ultrasound is increasingly used for evaluation of pancreatic cancer. The potential of sonographic morphology to differentiate histology type and biological behaviour of pancreatic lesions is doubtful. METHODS: We prospectively studied 115 patients with focal pancreatic lesions on endoscopic ultrasound. Morphology was assessed using Olympus UM3/20/200 echoendoscopes. Histologic confirmation of diagnosis was obtained in all patients. RESULTS: Endoscopic ultrasound correctly diagnosed 18/34 benign and 77/81 malignant lesions. Sensitivity, specificity, accuracy, PPV and NPV for diagnosing malignancy were 95%, 53%, 83%, 83% and 82%, respectively. Endosonographic diagnosis of the lesions (% correct) were: pancreatic cancer, 84 (63.3%); chronic pancreatitis, 14 (71.4%); ampullary cancer, 9 (77.8%); cystadenoma, 5 (80%); ampullary adenoma, 2 (50%); acute pancreatitis, 1 (0). In 13 patients of chronic pancreatitis, diagnosed as cancer, diagnosis was based on absence of sonographic features of chronic pancreatitis (7) or suspected involvement of adjacent structures (6). In 3 patientsmalignancy was missed owing to features of chronic pancreatitis. Non-suspected neuroendocrine tumours were misjudged in all 10 cases using morphologic criteria as pancreatic cancer (8), cystadenoma and chronic pancreatitis. Accuracy for prediction of metastatic lymph nodes and an advanced pancreatic cancer stage (TxN1 or T3Nx) was 61% and 75%, respectively. On retrospective analysis, a lesion >2 cm, vessel ingrowth, absence of cystic spaces and absence of diffuse pancreatitis were associated with pancreatic cancer. CONCLUSIONS: While overall sensitivity was high, specificity of endoscopic ultrasound for diagnosis of malignancy was low, especially in presence of chronic pancreatitis. In addition, endosonography had only a limited potential to predict the histological type of lesions.