Endoscopic ultrasound (EUS) guided cytology: correlation with histological findings

Abstract

Aim Endoscopic ultrasound guided cytology is a relatively new technique in the diagnosis of masses of the upper gastrointestinal tract and pancreas. We aim to correlate the accuracy of cytological diagnoses with histological findings and discuss reasons for non-correlation. Methods All cases of endoscopic ultrasound guided cytology received at our institution from August 2006 to September 2008 were retrospectively retrieved and compared to histology reports from biopsy or resection where available. Results Ninety-three specimens were received from 83 patients. Twenty-nine atients had histology available. The cytological diagnosis correlated exactly with the histological diagnosis in 15 cases. In a further six cases, the cytology diagnosis was sufficiently accurate (i.e., benign, atypical, malignant) that the correlation was considered true. The specimen was insufficient for diagnosis in four cases and inconclusive in one. In the remaining three cases, the cytological diagnosis was at variance with the histology. Conclusions The positive predictive value was 89% for all diagnoses (adenocarcinoma 71%, GIST 100%); sensitivity 74%; specificity 67%. The negative predictive value was 40%, reflecting the high number of false negatives but also selection bias. False negative results were usually due to sampling error. Contamination from the GI tract should be well recognised to avoid misinterpretation.