RESUMO
OBJECTIVES: Learning endosonography (EES) is known to be difficult, and the theory must be understood before performing EES routinely. The aim of the study was to evaluate the diagnostic indexes of EES in the staging of cardio-oesophageal cancer after a period of theoretical apprenticeship of EES and a learning period in a centre experienced in EES since 1989. METHODS: Five observers, having never used EES, followed to EES examinations for a 3-6 month period. They then reviewed the standardized records of 29 patients with cardio-oesophageal cancer. They had to evaluate the degree of tumour infiltration within oesophageal wall and the site of metastatic lymph nodes. Results were compared with the diagnosis of 5 experienced senior endoscopists. Interobserver agreement was estimated with kappa statistics and considered excellent for k > or = 0.75, good to moderate if 0.75 > k > or = 0.40, and poor if k < 0.40. RESULTS: Inter-observer agreement was poor for the topographic diagnosis of lymph nodes (kappa index from -0.09 to 0.33), lower to that of the 5 senior observers (0.33 to 0.77). It was satisfactory for degree of tumour infiltration (T1: k = 0.66; T2: k = 0.58; T3: k = 0.56; T4: k = 0.46). The individual sensitivities were weak according to lymph nodes site, but good for presence of lymph nodes (86 to 100%), with a specificity of 40 to 73%. CONCLUSIONS: After theoretical training of EES, agreement and diagnostic performances are good enough to diagnose pathological images (except for T4 tumors), and poor for localizing images in the mediastinum. This emphasizes the difficulties in learning echo-anatomy and gives useful guidelines for training programs in EES.