Comparison of three lymph node classifications for survival prediction in distant metastatic gastric cancer.

BACKGROUND: The optimal lymph node (LN) classification system for prognostic assessment in distant metastatic gastric cancer (DMGC) patients who undergo LN dissection remains unclear. Therefore, we compared the prognostic performance of positive LN (PLN), LN ratio (LNR), and log odds of positive LNs (LODDS) in DMGC patients.

METHODS: A total of 1999 DMGC patients who underwent lymphadenectomy recorded in the Surveillance Epidemiology and End Results database from 2004 to 2012 were reviewed.

RESULTS: Univariate analyses showed that the PLN, LNR and LODDS systems were all significantly correlated with cancer-specific survival (CSS). However, only the LODDS classification remained an independent prognostic factor through the multivariate analysis. Furthermore, this classification could efficiently discriminate survival outcomes in patients within the same positive PLN category, as well as in patients with no positive node involvement. Both the LODDS and LNR classifications had better discriminatory ability, monotonicity, and homogeneity of prognostic stratification, as well as more accurate 1 or 2-year CSS prediction, than the PLN classification. The performances of the LNR and LODDS classifications were similar. Additionally, we found that inclusion of PORT carried a survival benefit across all LODDS intervals except the "LODDS ≤ -1.0" subgroup.

CONCLUSION: Our findings indicate that the LODDS classification is the most optimal system for prognostic assessment in DMGC patients. Incorporating LODDS into the staging system of DMGC patients will enable clinicians to more accurately predict prognosis and guide regional therapy regimen decisions in DMGC patients.