Endoscopic ultrasound errors in esophageal cancer.

BACKGROUND Previous assessments of endoscopic ultrasound (EUS) classification of esophageal cancer are dominated by symptomatic patients with advanced stage disease. Fewer data exist on EUS errors in a cohort balanced between early and advanced disease. PURPOSE Assess EUS errors in classification of esophageal cancer in a more balanced cohort, and identify clinical and tumor characteristics associated with EUS errors. METHODS A total of 266 patients underwent EUS and esophagectomy without preoperative chemoradiotherapy. Pathologic classification of disease extent: 108 (41%) tumors were confined to the esophageal wall (pTis-pT2, pN0, pM0); 158 (59%) were advanced beyond (pT3-pT4, pN1, or pM1). Logistic regression analysis was performed to identify correlates of error in T classification and disease extent using 10 clinical and tumor characteristics (gender, age, dysphagia, weight loss, tumor length, location, traversability, morphology, histopathologic type, and histologic grade). RESULTS EUS erroneously predicted pathologic T (pT) in 119 patients (45%). When T classification was dichotomized into tumors whose depth of invasion was not beyond the muscularis propria (pTis-pT2) and those beyond (pT3-pT4), errors occurred in 42 patients (16%). EUS erroneously predicted N classification in 67 patients (25%), and was insensitive to the presence of distant metastases. EUS misclassified disease extent in 40 patients (15%). Logistic regression analysis indicated that weight loss and tumor length were the only clinical and tumor characteristics correlated with EUS errors; more weight loss was associated with decreased odds of misclassification, while the odds of misclassification were four to six times greater for intermediate length tumors than for shorter tumors. CONCLUSIONS EUS errors, particularly in predicting pT, are more frequent than previously reported. Weight loss and tumor length are the only clinical and tumor characteristics correlated with EUS errors.