Abstract
BACKGROUND: Over the last 15 years, large randomized controlled studies have validated the benefit of preoperative therapy for patients with resectable gastric cancer. Computed tomography (CT) and endoscopic ultrasonography (EUS) are commonly used to select patients for preoperative treatment, but studies of preoperative staging accuracy that focus on patient selection for preoperative therapy are rare; therefore, whether CT or EUS can reliably identify patients eligible for preoperative therapy is still unclear. Our purpose was to determine the accuracy of EUS and CT for preoperative staging of gastric cancer and to identify factors that may affect their usefulness in selecting patients for preoperative therapy. METHODS: We reviewed the medical records of 8,260 patients with gastric or gastroesophageal adenocarcinoma treated at our institution from 1995 to 2013, identifying those who underwent gastrectomy without preoperative treatment. We compared T stage and N status from preoperative EUS and CT reports with those drawn from surgical pathology reports. Clinicopathologic and demographic variables associated with incorrect preoperative staging were investigated using univariate and multivariate analyses. RESULTS: We identified 187 patients who underwent preoperative staging by EUS (n=145) and/or CT (n=134) before gastrectomy. The accuracy, sensitivity, and specificity of EUS in distinguishing stage T1 from more advanced tumors were 82%, 78%, and 85%, respectively. Variables associated with underestimation of EUS T stage were lymphovascular invasion [odds ratio (OR), 7.51; 95% confidence interval (CI), 1.91-29.50; P<0.01] and white race (OR, 3.75; 95% CI, 1.31-10.75; P=0.01). The accuracies, sensitivities, and specificities for determining N status were, respectively, 65%, 49%, and 79% with CT and 66%, 29%, and 95% with EUS. Lymphovascular invasion was associated with a false negative result (OR, 3.79; 95% CI, 1.34-10.70; P=0.01), and well- or moderately differentiated histology was associated with a false positive result for CT N status (OR, 7.14; 95% CI, 2.00-25.44; P<0.01). CONCLUSIONS: EUS is accurate in distinguishing T1 from T2-T4 lesions; both CT and EUS have low sensitivities and high specificities in determining N status. These accuracies and variables associated with inaccurate staging, including race, should be considered when selecting gastric cancer patients for preoperative therapy.