Interobserver variability in endoscopic ultrasonography: a prospective evaluation.

Abstract

BACKGROUND: Endoscopic ultrasonography (EUS) is an imaging modality that is now widely used to stage gastrointestinal malignancies. Few studies have addressed the issue of interobserver variability in the interpretation of EUS, particularly as it pertains to the staging of rectal carcinoma. METHODS: Twenty-six patients with a diagnosis of rectal carcinoma were evaluated prospectively by three endoscopists. One performed sigmoidoscopy, the second (primary endosonographer) performed an EUS staging examination with full knowledge of the patient history and sigmoidoscopic appearance of the lesion, and the third endoscopist (secondary endosonographer) performed EUS blinded to this information. The results of the respective examinations were then compared. RESULTS: When the EUS findings of the endosonographers were compared, T stage agreed in 88% of patients, with the following kappa coefficients: T1 (kappa = 0.00); T2 (kappa = -0.04); T3 (kappa = -0.05); T4 (kappa = 0.00). Interrator N stage agreed in 73% of patients (kappa = 0.42). CONCLUSION: Our study prospectively evaluated interobserver variation in staging rectal carcinoma by EUS. The protocol that was followed provides a useful template that eliminated potential observer bias. Fair agreement was demonstrated regarding lymph node assessment. Although the raters agreed in 88% of the patients, kappa statistic analysis did not reach significant agreement, due to this institution's preponderence of UT3 lesions. Thus, validation of our findings in a setting where a broader spectrum of disease is encountered is required.