From 1983 to 1989, 95 patients with carcinoma of the esophagogastric junction underwent resection. Overall hospital mortality rate was 6.2% (6/95). Actuarial survival analysis showed 5- and 10-year survivals of 33% and 31%, respectively. Five- and 10-year survivals of patients according to TNM stages were as follows: stage I (n = 13), 90% at both 5 and 10 years; stage II (n = 13), 70% at both intervals; stage III (n = 28), 28% at both intervals; and stage IV (n = 40), 11% and 8%, respectively. For patients with undiseased nodes (n = 26), 5- and 10-year survivals were 72% and 72%, compared with 18% and 16% for patients with diseased nodes (n = 68; p < 0.005). In patients who had involvement of both the abdominal and thoracic lymph nodes (n = 28), 5- and 10-year survivals were 13% and 13%, compared with 26% and 26% if metastases were confined to the abdomen (n = 37; p > 0.05). Grouping patients with diseased intrathoracic nodes together with patients with N2 abdominal nodes showed survivals of 14% at both 5 and 10 years. When tumors were staged as an esophageal carcinoma, classification of individual patients changed, as did the 5- and 10-year survivals. Five- and 10-year survivals were as follows: stage I (n = 8), 100% for both 5 and 10 years; stage II (n = 18), 68% for both 5 and 10 years; stage III (n = 27), 37% for both 5 and 10 years; and stage IV (n = 41), 10% for 5 years and 6% for 10 years. These data indicate that tumors of the esophagogastric junction tend to spread to both abdominal and thoracic nodes. However, reasonably good 5- and 10-year survivals can be obtained even in patients with nodal metastases in both areas. We suggest that N2 labeling be included for thoracic node metastases instead of the actual M+Ly label, because the N2 label better reflects the potential for curative surgery. Finally, staging tumors as gastric or esophageal carcinoma makes no significant difference in survival analysis, which raises the question whether these tumors behave more like esophageal carcinoma than gastric carcinoma.