Background: Recent studies demonstrated the positive impact of neoadjuvant treatment in locally advanced gastric cancer. Objective: To assess the accuracy of endoscopic ultrasound (EUS) in the selection of patients with gastric adenocarcinoma for neoadjuvant therapy (T ! 2 and/or Nþ). Methods: Retrospective analysis of patients with an anatomopathological diagnosis of gastric adenocarcinoma between January 2011 and June 2017, who had EUS for staging and underwent surgery as a first therapeutic attempt. The concordance (k) and accuracy (area under the curve (AUC)) of EUS for T ! 2 and/or Nþ were assessed using the anatomopathological staging of the resected surgical specimen as the gold standard. Results: The final sample included 152 patients (66.4% male, 67.1 AE 12.2 years). The concordance, accuracy, sensitivity and specificity of the EUS for T ! 2 and/or Nþ were 0.72, 0.86 AE 0.03, 88.5% and 83.1%, respectively. The results were higher in proximal (k ¼ 0.93, AUC ¼ 0.96 AE 0.05, sensitivity (S) ¼ 99.0% and specificity (E) ¼ 90.9%) compared with distal lesions (k ¼ 0.67, AUC ¼ 0.84 AE 0.04, S ¼ 85.7% and E ¼ 81.5%), and in intestinal subtype (k ¼ 0.77, AUC ¼ 0.88 AE 0.04, S ¼ 92.6% and E ¼ 84.1%) compared with diffuse (k ¼ 0.58, AUC ¼ 0.79 AE 0.10, S ¼ 85.0% and E ¼ 72.7%) or mixed-subtype tumours (k ¼ 0.65, AUC ¼ 0.84 AE 0.10, S ¼ 76.9% and E ¼ 90.0%). Conclusion: In one of the largest series of patients, we showed that EUS has overall high agreement and accuracy in the selection of patients with gastric adenocarcinoma for neoadjuvant therapy, although the agreement and accuracy are greater for proximal and intestinal lesions.