C-Reactive Protein-Based Composite Indices for Predicting Tumor Overgrowth Restenosis After Partially Covered Duodenal Stenting in Gastric Cancer: A Cohort Study.

Abstract

BACKGROUND/OBJECTIVES: Partially covered duodenal stents rapidly relieve malignant gastric outlet obstruction, but tumor overgrowth restenosis limits durability. We compared six preprocedural inflammatory, nutritional, and immune indices for predicting this outcome in patients with gastric cancer. METHODS: We retrospectively analyzed 68 consecutive patients from a prospectively maintained cohort. The primary endpoint was endoscopically or radiologically confirmed tumor overgrowth restenosis. Discrimination was assessed using receiver operating characteristic curves (pairwise DeLong tests with Bonferroni correction) and time-dependent areas under the curve (AUCs) accounting for death as a competing risk. Multivariable cause-specific Cox models were restricted to preprocedural covariates; post-stenting chemotherapy or radiotherapy was examined in time-dependent sensitivity analyses. RESULTS: Technical and clinical success rates were 100% and 94.1%. Seventeen patients (25.0%) developed restenosis at a median of 66 days. The C-reactive protein-albumin-lymphocyte (CALLY) index showed the highest AUC (0.859), followed by the C-reactive protein-to-albumin ratio (CAR; 0.822) and neutrophil-to-lymphocyte ratio (0.774); these three indices did not differ significantly. At an exploratory, internally derived cutoff of ≤0.110, CALLY had 76.5% sensitivity and 84.3% specificity and remained associated with restenosis after adjustment for stenosis site and stent length (adjusted hazard ratio, 12.30; 95% confidence interval, 3.89-38.83; C-index, 0.836), with consistent results in continuous, time-dependent, and tumor-covariate-adjusted analyses. The 180-day cumulative incidence was 52.4% with low CALLY versus 4.3% with high CALLY. CONCLUSIONS: C-reactive protein-based indices showed the highest numerical discrimination, although pairwise differences among CALLY, CAR, and NLR were not statistically significant. CALLY is a promising exploratory biomarker for restenosis risk stratification, but its cutoff should not guide clinical decisions until externally validated.