Tissue Microarray Immunohistochemical Analysis To Distinguish Contaminating Gastrointestinal Epithelium from Non-Malignant Branch Duct IPMN
MB Pitman, V Desphpande. Massachusetts General Hospital, Boston, MA
Background: Contaminating GI epithelium presents a diagnostic pitfall in the cytological diagnosis of pancreatic cysts in general, and branch duct (BD) IPMN in particular when sampled by EUS. Accurate diagnosis is vital for proper patient care, which is increasingly non-surgical. Morphology alone is often insufficient to distinguish GI contamination from non-malignant epithelium of these typically low grade mucinous cysts, and misinterpretation can lead to both false positive and negative diagnoses.
Design: Tissue microarrays were constructed using tissue cores from the stomach, duodenum and cyst lining of 15 BD IPMN (6 moderate dysplasia and 9 adenomas) analyzed in quadruplicate with immunohistochemical stains to B72.3, MUC1 core, MUC5ac, MUC2, MUC6, CA19-9, S100P, p16 and CEA. Each core of tissue was individually assessed to account for heterogeneity, and staining was assessed as either positive or negative. ROC curve analysis was performed to determine the optimal immunohistochemical panel to distinguish cyst lining from gastric and duodenal contamination. In ROC analysis perfect tests are associated with an area of 1.0 and completely random tests with an area of 0.5.
Results: The majority of IPMN cores stained positively for S100P (84%), CEA (86%), and p16 (78%), with the first two being excellent markers to distinguish duodenal epithelium from an IPMN. The specificity and sensitivity of S100P and CEA were 91%/ 86% and 91%/82%, respectively. In distinguishing gastric epithelium from IPMN, ROC analysis showed that all evaluated markers were rated as either fair or poor. Among these, CA19.9 stained 71% of IPMN cores, with a sensitivity and specificity of 72% and 78%, respectively. The table below lists the area under the curve for duodenum versus cyst in the first row and stomach versus cyst in the second row.
Area Under ROC Curve
Conclusions: Immunohistochemistry staining for S100P and CEA can accurately distinguish non-malignant cyst lining of BD IPMN from duodenal contaminating epithelium, but no marker tested can reliably distinguish gastric epithelium. Given that the majority of BD IPMN occur in the pancreatic head using a transduodenal approach, these data prove useful in the pre-operative differential diagnosis of non-malignant BD IPMN.
Wednesday, March 11, 2009 1:00 PM
Poster Session VI # 59, Wednesday Afternoon