Endoscopic ultrasound is less accurate in pancreatic cancer staging than previously thought: A blind analysis of videotapes

Background: Previous studies showing very high accuracy rates of endoscopic ultrasonography (EUS) in pancreatic cancer staging could not be confirmed in other trials. We therefore used completely blind videotape analysis of pancreatic head cancer (PHC) patients to reassess this issue. Patients and Methods: 67 patients (42 male, age 41-77 years) were included whose PHC was either resected (n=37) or operatively explored (n=9) or who had a positive angiography (n=21). Examinations had been done by 4 examiners and completely blind videotape analysis was performed by one of the authors. Results: 35 patients had portal venous infiltration (stage T3), 32 not (3 T1, 29 T2). T staging (all pts.) and N staging (37 resected tumors) was correct in 70% and 57% of cases, respectively. Using "lack of a vessel wall at the border to the PHC" as a parameter (additionally to "tumor in vascular lumen" and "complete obstruction"), the accuracy was only 70%, sensitivity and specificity being 74% and 75%, respectively. Omitting this parameter, sensitivity fell to 54%, but specifity rose to 88% and T staging accuracy to 78%. Of the 9 failures to detect venous invasion 2 were due to technical reasons (no instrument passage, tumor too large) and 3 due to lack of visualization of exclusive superior mesenteric vein involvement. Of the vessel negative cases, 75% had no direct tumor-vessel relationship, and 9% a lack of a vessel wall. This was also seen in 29% of cases with vascular infiltration, and 43% of these patients had direct endosonographic evidence of tumor in the vascular lumen. Conclusions: EUS is considerably less accurate in PHC staging than previously thought; the presently used criteria seem to be partially insufficient. Combination of radial and linear EUS instruments and/or use of the Doppler may be more helpful but have to be further evaluated.