Prognostic value of LODDS in late-onset gastric adenocarcinoma: analysis of SEER, TCGA, and external multicenter cohorts.

INTRODUCTION

Accurate survival prediction is crucial for optimizing individualized treatment and follow-up in patients with late-onset gastric adenocarcinoma (LOGA). Traditional lymph node staging systems such as N-stage, positive lymph node (PLN), and lymph node ratio (LNR) have limitations in predictive accuracy, especially in cases with inadequate lymph node dissection. The log odds of positive lymph nodes (LODDS), a novel nodal staging metric that incorporates both positive and negative lymph nodes through a log-transformed ratio, has shown potential advantages by providing a more stable and refined assessment of nodal involvement.

MATERIALS AND

METHODS

This study included 10,361 LOGA patients from the SEER database, 135 from TCGA, and 252 from two medical centers. A novel prognostic model was constructed based on a training cohort from SEER and validated using internal (SEER testing set) and external (TCGA and hospital datasets) cohorts. The model incorporated age, gender, grade, size, chemotherapy, and LODDS. Four staging systems (TNM-stage, PLN-stage, LNR-stage, LODDS-stage) were compared using the Akaike Information Criterion (AIC), Concordance Index (C-index) and time-dependent Area Under the Curve (AUC). LODDS-stage model, the most effective model, was used to build nomograms for overall survival (OS) and cause-specific survival (CSS). Model performance was evaluated using calibration curves, Decision Curve Analysis (DCA), and Kaplan-Meier analysis.

RESULTS

Univariate and multivariate Cox regression identified age, gender, grade, tumor size, chemotherapy, and LODDS-stage as independent prognostic factors. Among the four models, the LODDS-based model showed the highest discrimination and best calibration for predicting OS and CSS at 1, 3, 5, and 10 years. Nomograms incorporating these variables exhibited excellent predictive accuracy in both internal and external cohorts. Survival risk classification based on model scores effectively stratified patients into high- and low-risk groups, with significantly different survival outcomes across all datasets (p < 0.05).

CONCLUSIONS

The LODDS-based prognostic model outperformed traditional nodal staging systems in survival prediction for LOGA patients. This model showed high accuracy and consistent performance across different datasets, indicating its potential to support personalized treatment and long-term follow-up strategies for elderly patients with gastric cancer.