Related Experiment Video
Updated: May 22, 2026
Comparison of Predictive Performance of Three Lymph Node Staging Systems in Colorectal Signet Ring Cell Carcinoma Based on Machine Learning Model
Published on: April 18, 2025
Spatial nodal pattern refines disease-free survival stratification beyond pathological N stage in resected esophageal
Yongkang Bai1,2,3,4, Xiaorui Dong1,2,3, Yuxian Qian1,2,3
1Department of Thoracic Surgery, Nanjing Medical University Affiliated Cancer Hospital & Jiangsu Cancer Hospital & Jiangsu Institute of Cancer Research, Nanjing, P. R. China.
Background:
Lymph node metastasis is a major determinant of recurrence in esophageal squamous cell carcinoma (ESCC), but conventional pathological N staging is based mainly on positive-node counts and may not capture the anatomical heterogeneity of nodal spread. We developed and evaluated a station-level Spatial-N classification for postoperative 3-year disease-free survival (DFS) stratification in resected ESCC.
Methods:
We retrospectively analyzed 1,276 patients with pathologically confirmed ESCC who underwent upfront curative-intent esophagectomy between January 2019 and December 2024. Spatial-N pattern was assigned hierarchically using station-level pathological nodal mapping: Pattern 0, N0; Pattern 1, single-region metastasis; Pattern 2, recurrent laryngeal nerve (RLN)-chain high-risk station involvement, defined as 106recR and/or 106recL positivity without cross-region metastasis; and Pattern 3, cross-region metastasis. DFS was analyzed within a prespecified 36-month framework using Kaplan-Meier analysis and Cox regression. Predictive performance was compared among the 8th AJCC pathological N stage model, the Spatial-N model, and an integrated Spatial-N model incorporating Spatial-N pattern, pathological T stage, lymphovascular invasion, and adjuvant therapy.
Results:
Among 1,276 patients, 745 (58.4%), 250 (19.6%), 180 (14.1%), and 101 (7.9%) were classified as Patterns 0, 1, 2, and 3, respectively. Spatial-N pattern significantly stratified 3-year DFS in the overall cohort and within the pN1 subgroup. In multivariable analysis, Spatial-N pattern remained independently associated with DFS, with adjusted hazard ratios of 3.14, 7.15, and 10.25 for Patterns 1, 2, and 3 versus Pattern 0, respectively. Compared with 8th AJCC pN staging, the Spatial-N model improved discrimination (C-index, 0.731 vs 0.699; 3-year AUC, 0.775 vs 0.742). The integrated model achieved the highest discrimination (C-index, 0.747; 3-year AUC, 0.794). Adding positive lymph node count, lymph node ratio, or positive nodal station count did not materially improve performance.
Conclusions:
Spatial-N pattern provides anatomically organized prognostic information beyond count-based nodal staging and refines postoperative 3-year DFS stratification in resected ESCC. This station-level framework may complement 8th AJCC pathological N staging and support individualized postoperative risk assessment, pending external validation.
