4597 Interobserver variation in endosonographic cancer staging better results if endoscopy is included in the assessment.

Background and Aims : Whereas a multitude of literature exists on the accuracy of endoscopic ultrasonography (EUS) in cancer staging, data on reproducibility (interobserver variability, IOV) of this technique are sparse. We conducted a videotape evaluation of esophagogastric cancer cases who all had surgical confirmation. Methods : Well documented videotapes were selected of patients with esophageal and gastric cancer who underwent surgical resection between 1990-1996. After agreeing on EUS staging criteria by consensus on the basis of established and published criteria, 5 examiners independently looked at the full videotape sequences to diagnose the T and N category in 1997. Endoscopy videos were initially not made available. Thereafter, selected cases, where both endoscopy and EUS documentation were available, were again reviewed 1-2 years later (to avoid recognition of images) by 2 examiners (TR and HJD). Kappa values of >0.75 were regarded as excellent, 0.4-0.75 as fair/good and Results : Videotape sequences of 55 patients with esophageal and of 53 patients with gastric carcinoma were included in the IOV analysis. For T as well as for N staging, kappa values were poor in 50% of cases as outlined in the Table. In 62 selected cancer cases (esophageal cancer: n=30, gastric cancer: n=32), we compared the interobserver variability for the two main investigators only (HJD, TR): here, the kappa-value for T-staging of all categories T1-4 was 0.30 (poor), but increased to 0.48 (fair/good) when videotapes of endoscopic aspects were added in another analysis by HJD + TR. Conclusions : Reproducibility of EUS cancer staging results is rather poor, but can be significantly improved when endoscopy is added to the analysis. The factors influencing image analysis for scientific evaluation have to be studied further.