Scholarlink

Objective <p>Conventional TNM staging systems exhibit limited prognostic accuracy in gastric cancer patients following neoadjuvant therapy (NAT), creating an urgent need for novel risk-stratification tools. This study aimed to develop a machine learning-based nomogram integrating lymph node tumor regression grade (LN.TRG) and other clinicopathological features to improve recurrence prediction in this setting.</p> Methods <p>A cohort of 112 gastric cancer patients receiving NAT was retrospectively analyzed. Clinical and pathological features, including lymph node tumor regression grade (LN.TRG) were evaluated. Univariable analysis and Least Absolute Shrinkage and Selection Operator (LASSO) regression were employed to identify recurrence-free survival (RFS)-associated predictors. A predictive nomogram was developed using multivariable Cox regression and validated via ROC curves, decision curve analysis (DCA), and calibration curves. Additional analyses explored the prognostic impact of metastatic lymph node location, residual tumor ratio in lymph nodes, and lymph node dissection count.</p> Results <p>Six key predictors—including LN.TRG and lymph node-positive rate—were selected via Least Absolute Shrinkage and Selection Operator (LASSO) regression. The model exhibited robust discrimination in both training (AUC: 0.836–0.886) and validation (AUC: 0.907–0.957) sets. DCA and calibration curves confirmed clinical utility and stability. Risk stratification via ggrisk plots and Kaplan-Meier analysis revealed significant separation between high- and low-risk groups (<i>P</i> &lt; 0.001), outperforming traditional staging systems (ypStage/ycStage). A fitted curve demonstrated a negative correlation between recurrence risk and lymph node dissection count (<i>P</i> &lt; 0.05). Location-specific analysis identified higher residual tumor ratios and hazard ratios (HR) in station No. 5 and No. 6 lymph nodes.</p> Conclusion <p>The LN.TRG-integrated model enables precise risk stratification for post-NAT gastric cancer. Proactive dissection of lymph nodes, particularly station No. 5 and No. 6, may improve patient outcomes. This study offers insights to enhance survival and quality of life.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Integrating spatial lymph node patterns and multimodal clinicopathological features to predict post-neoadjuvant recurrence in gastric adenocarcinoma: a machine learning nomogram

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月7日

查看全文
  • Zhao Fazhi,
  • Hu Shangzhi,
  • Ding Zhi,
  • Liu Yang,
  • Xiao Shuomeng,
  • Zhou Yehan

摘要

Objective

Conventional TNM staging systems exhibit limited prognostic accuracy in gastric cancer patients following neoadjuvant therapy (NAT), creating an urgent need for novel risk-stratification tools. This study aimed to develop a machine learning-based nomogram integrating lymph node tumor regression grade (LN.TRG) and other clinicopathological features to improve recurrence prediction in this setting.

Methods

A cohort of 112 gastric cancer patients receiving NAT was retrospectively analyzed. Clinical and pathological features, including lymph node tumor regression grade (LN.TRG) were evaluated. Univariable analysis and Least Absolute Shrinkage and Selection Operator (LASSO) regression were employed to identify recurrence-free survival (RFS)-associated predictors. A predictive nomogram was developed using multivariable Cox regression and validated via ROC curves, decision curve analysis (DCA), and calibration curves. Additional analyses explored the prognostic impact of metastatic lymph node location, residual tumor ratio in lymph nodes, and lymph node dissection count.

Results

Six key predictors—including LN.TRG and lymph node-positive rate—were selected via Least Absolute Shrinkage and Selection Operator (LASSO) regression. The model exhibited robust discrimination in both training (AUC: 0.836–0.886) and validation (AUC: 0.907–0.957) sets. DCA and calibration curves confirmed clinical utility and stability. Risk stratification via ggrisk plots and Kaplan-Meier analysis revealed significant separation between high- and low-risk groups (P < 0.001), outperforming traditional staging systems (ypStage/ycStage). A fitted curve demonstrated a negative correlation between recurrence risk and lymph node dissection count (P < 0.05). Location-specific analysis identified higher residual tumor ratios and hazard ratios (HR) in station No. 5 and No. 6 lymph nodes.

Conclusion

The LN.TRG-integrated model enables precise risk stratification for post-NAT gastric cancer. Proactive dissection of lymph nodes, particularly station No. 5 and No. 6, may improve patient outcomes. This study offers insights to enhance survival and quality of life.