Endosonographic diagnosis of solid pancreatic tumors: a retrospective analysis from a tertiary referral center

The sensitivity of endosonography for pancreatic tumors is good, but differentiation between malignant and benign lesions remains a challenge. We, therefore, analyzed the correlation between endosonographic findings and pancreatic carcinoma in a population with a high prevalence of chronic pancreatitis.

171 endosonographic examinations were retrospectively evaluated using 22 dichotomous criteria. Final diagnosis was defined by the results of pancreas resection or by clinical follow-up (median: 41 months). A sum score was established after multivariate analysis.

Final diagnosis was carcinoma, chronic pancreatitis, neuroendocrine tumor and normal pancreas in 67, 65, 9, and 38 subjects (prevalence 39 %, 38 %, 5 %, 22 % respectively) After multivariate analysis, carcinoma was significantly correlated with six endosonographic criteria and age, which resulted in the following score (yes = 1, no = 0): (dilated pancreatic duct) + (vascular infiltration) + (suspicious lymph nodes) + (tumor edge with pseudopodia-like extensions) + (age > 50 years) - (increased pancreas parenchyma echogenicity) - (tumor diameter < 10 mm) + 3. After receiver operating characteristic analysis, the area under the curve was 0.85. For score values of 5 (4) or more, sensitivity was 0.63 (0.93), specificity 0.91 (0.55), positive predictive value 0.82 (0.57), negative predictive value 0.79 (0.92), positive likelihood ratio 7.24 (2.05), and negative likelihood ratio 0.41 (0.14).

A simple and potentially useful score for the prediction of pancreatic cancer based on six endosonographic criteria and patient age was established. Distribution of final diagnoses in the patient population argues for the score's applicability in clinical practice of a tertiary referral center and warrants prospective validation.