Several publications have showed that the number of metastatic lymph node (LN) should be taken into consideration in nodal category of esophageal cancer, but seldom considered extent of involved regional LNs. The aim of this study is to evaluate the significance of the extent of regional LN metastasis on survival in patients with esophageal cancer. A total of 245 thoracic esophageal cancer patients underwent transthoracic esophagectomy with standard lymphadenectomy between January 2000 and December 2006 were included in the study. Data including demographic factors, pathologic findings, LN parameters and survival outcomes were collected. The survival experience was depicted using Kaplan-Meier method. A multivariate Cox proportional hazard model was used to screen the significant prognostic factors. The univariate analysis to further explore the significant prognostic factor was done by log-rank test. After a median follow-up of 53.2 months, the 5-year survival rate was 46.3% for the entire cohort. Cox model regression indicated that the LN status and perigastric nodal status, aside from residual tumor status, histological tumor type and depth of invasion, were the independent prognostic factors. Patients without LN metastasis had better 5-year survival than those with positive nodes (64.2% vs. 18.9%, X2=35.875, P<0.001). However, For those patients with nodal involvement, there was no difference in 5-year survival between patients with involved nodes<3 and >or=3 (27.8% vs. 0%, X2=0.925, P=0.336). When considering the location of LN metastasis, patients could be further stratified according to whether the perigastric nodes were involved or not (37.5% vs. 10.0%, X2=4.295, P=0.038). In conclusion, involved LN number had no prognostic implication in nodal involved patients based on our data. Whereas, perigastric nodal involvement should be used to refine the N category (N0, no nodal metastasis, N1, non-perigastric node metastasis, N2, perigastric node metastasis) for the future esophageal cancer staging criteria.