Prognostic impact of examined lymph-node count for patients with esophageal cancer: development and validation prediction model

Shasha Yuan, Wei Chen, Mengyu Wang, Wenying Deng, Chi Zhang, Ning Li, Suxia LuoScientific 2023 · DOI

Abstract

Esophageal cancer (EC) is a malignant tumor with high mortality. This study aimed to identify the optimal examined lymph node (ELN) count threshold and develop a model to predict survival after radical esophagectomy. Analyzing a training cohort (n=734, Chinese registry) and an external testing cohort (n=3208, SEER registry), we found that a higher ELN count significantly correlates with improved survival. The optimal cut-off was determined to be 18 nodes. A Random Survival Forest (RSF) model was developed, demonstrating high predictive accuracy (AUC: 87.5 in training, 79.3 in testing) and low Brier Scores. These findings suggest that an ELN count exceeding 18 is a critical prognostic factor and that the RSF model provides a robust tool for individual risk assessment.

Introduction

Esophageal cancer (EC) remains one of the most aggressive malignancies globally, ranking sixth in cancer-related mortality. While radical esophagectomy is the cornerstone of treatment, predicting long-term survival is challenging. The traditional TNM (Tumor-Node-Metastasis) staging system often fails to account for the significant variability in outcomes among patients within the same stage.

A critical surgical consideration is the extent of lymphadenectomy. The number of examined lymph nodes (ELN) is vital for accurate staging; however, the "optimal" number remains a subject of debate. Insufficient dissection leads to "under-staging" (stage migration), while excessive dissection may increase surgical morbidity. This study sought to define a definitive ELN threshold and leverage machine learning to improve prognostic accuracy.

Methods

The study utilized two large-scale datasets: a training cohort of 734 patients from the Affiliated Cancer Hospital of Zhengzhou University and an external validation cohort of 3,208 patients from the U.S. SEER registry.

To move beyond the limitations of linear statistical models, researchers employed:

  1. Lasso-Cox Regression: Used for variable selection to identify clinical factors (e.g., age, sex, tumor site) with the most significant impact on survival.
  2. Random Survival Forest (RSF): An ensemble machine learning algorithm that builds multiple decision trees to capture complex, non-linear interactions between variables.
  3. Performance Metrics: Model discrimination was evaluated using the C-index and Area Under the Curve (AUC), while prediction error was assessed via the Brier Score.

Results

The Prognostic Significance of the 18-Node Threshold

Statistical analysis identified 18 as the optimal ELN count threshold. In both the training and testing cohorts, patients with more than 18 ELNs demonstrated significantly better overall survival (OS) and disease-free survival (DFS).

Survival curves stratified by ELN count (≤18 vs. >18) in training and testing cohorts.

The Hazard Ratio (HR) for ELN count was 0.98 (P < 0.01) in both cohorts, indicating that each additional node examined slightly reduces the risk of mortality. This suggests that thorough lymphadenectomy may offer a direct therapeutic benefit by removing occult micrometastases.

Machine Learning Model Performance

The RSF model consistently outperformed traditional TNM-based and Lasso-Cox models. In the training cohort, the RSF model achieved an AUC of 87.5, while in the external SEER testing cohort, it maintained a robust AUC of 79.3.

ROC curves comparing the RSF model (model 4) against other predictive models.

Furthermore, the RSF model exhibited the lowest Brier Scores (0.122 in training, 0.152 in testing), confirming its reliability and superior predictive power compared to conventional methods.

Discussion

The identification of 18 ELNs as a critical threshold provides a clear target for surgeons to ensure adequate staging and potentially improve patient outcomes. By reducing "stage migration"—where metastatic nodes are missed—this threshold ensures patients receive appropriate adjuvant therapies.

The success of the RSF model highlights the potential of machine learning in oncology. Unlike traditional formulas, RSF handles the "messy" nature of biological data more effectively. However, limitations remain, including the retrospective nature of the data and the "black box" aspect of machine learning, which can make clinical interpretation difficult. Future work should focus on integrating genomic and inflammatory biomarkers into these models to further refine personalized prognosis.

Conclusion

An ELN count of at least 18 is significantly associated with improved survival in esophageal cancer patients following radical esophagectomy. The developed RSF model offers a high-accuracy tool for individual prognosis, outperforming traditional staging systems and providing a foundation for more personalized clinical decision-making.

Summary updated

Abstract

Esophageal cancer (EC) is a malignant tumor with high mortality. We aimed to find the optimal examined lymph node (ELN) count threshold and develop a model to predict survival of patients after radical esophagectomy. Two cohorts were analyzed: the training cohort which included 734 EC patients from the Chinese registry and the external testing cohort which included 3208 EC patients from the Surveillance, Epidemiology, and End Results (SEER) registry. Cox proportional hazards regression analysis was used to determine the prognostic value of ELNs. The cut-off point of the ELNs count was determined using R-statistical software. The prediction model was developed using random survival forest (RSF) algorithm. Higher ELNs count was significantly associated with better survival in both cohorts (training cohort: HR = 0.98, CI = 0.97-0.99, P < 0.01; testing cohort: HR = 0.98, CI = 0.98-0.99, P < 0.01) and the cut-off point was 18 (training cohort: P < 0.01; testing cohort: P < 0.01). We developed the RSF model with high prediction accuracy (AUC: training cohort: 87.5; testing cohort: 79.3) and low Brier Score (training cohort: 0.122; testing cohort: 0.152). The ELNs count beyond 18 is associated with better overall survival. The RSF model has preferable clinical capability in terms of individual prognosis assessment in patients after radical esophagectomy.
BibTeX
@inproceedings{yuan2023prognostic,
  author = {Shasha Yuan and Wei Chen and Mengyu Wang and Wenying Deng and Chi Zhang and Ning Li and Suxia Luo},
  title = {{Prognostic impact of examined lymph-node count for patients with esophageal cancer: development and validation prediction model}},
  booktitle = {Scientific},
  year = {2023},
  doi = {10.1038/s41598-022-27150-6},
}
Figure