Although current guidelines suggest that 16 or more lymph nodes (LNs) are required for the appropriate staging of gastric cancer, the effect that the minimum number of examined LNs (eLNs) in the different types of gastrectomy has on survival remains unclear. This study retrospectively analyzed 2662 patients who underwent curative gastrectomy with D2 lymphadenectomy for gastric cancer at Fujian Medical University Union Hospital from January 2000 to December 2010 and randomly divided them into development (70 %, n = 1863) and validation (30 %, n = 799) data sets. An additional external validation was performed using the data set (n = 285) collected during the same period from the Sun Yat-sen University Cancer Center in Guangzhou, China. A hypothetical tumor-node-metastasis (TNM) classification (hTNM) was proposed based on eLNs and survival. The mean numbers of nodes removed during radical distal and total gastrectomy were respectively 26 +/- 9.6 and 29 +/- 10.7 (p < 0.01). The optimal LN-count thresholds were determined to be 16 for patients who underwent curative distal gastrectomy and 21 for patients who underwent total gastrectomy. The hTNM staging system had higher linear trend and likelihood ratio chi (2) scores and lower Akaike information criterion (AIC) values than the seventh American Joint Committee on Cancer (AJCC) TNM classification. Thus, the hTNM staging system exhibited superior prognostic stratification. Similar results were found in the two validation data sets. A harvest of at least 21 LNs may represent a superior threshold for radical total gastrectomy (RTG) and could yield a better prognosis. For patients undergoing RTG, the hTNM staging system may predict survival more accurately and discriminatively. However, a validation from a Western institution is warranted.