Background: Nodal disease in esophageal and gastric cancer is associated with poor survival. Objectives: To determine the critical level of lymph node involvement where survival becomes significantly compromised. Methods: Survival analyses using multivariable Cox regression and receiver operator characteristics (ROC) were undertaken to determine what number of positive lymph nodes were most sensitive and specific in predicting survival. Results: A total of 317 patients underwent esophagectomy (n = 190, 59.9%) and gastrectomy (n = 127, 40.1%) for adenocarcinoma. At multivariable analyses, four nodes positivity (irrespective of T-category) was associated with nearly a fivefold increased risk of mortality when compared to node-negative patients (hazard ratio [HR], 4.9; interquartile range 2.0-11.5; P < .001). A positive ratio of up to 50.0% was not associated with worse survival than having four nodes positive (HR, 4.6; 95% confidence interval, 2.6-8.1; P < .001). ROC analysis demonstrated four lymph nodes positive to have a sensitivity of 80.5%, a specificity of 60.1%, and an accuracy of 77.8 (P < .001). Conclusion: The absolute number of nodes positive for cancer is more important than the proportion of positive nodes in predicting survival in esophageal/gastric cancer. Four positive lymph nodes are associated with a fivefold increase in mortality. Beyond this, increasing numbers of positive lymph nodes make no appreciable difference to survival.