Endoscopic ultrasound vascular parameter-based stratification for precise treatment of esophagogastric varices: a study on endoscopic ultrasound-guided tissue glue obliteration.

OBJECTIVES

To establish a hierarchical model based on endoscopic ultrasound (EUS) vascular parameters for predicting outcomes following EUS‑guided tissue glue obliteration in patients with esophagogastric varices (EGV), and to develop a precise treatment strategy.

METHODS

Clinical and EUS data from 56 patients with F2-F3 EGV were retrospectively analyzed. Patients were classified into mild, moderate, and severe abnormality groups according to EUS‑derived vascular parameters. The primary outcome was rebleeding within 12 months. Multivariate logistic regression was used to identify independent risk factors. Internal validation was performed using bootstrap resampling and cross‑validation. The incremental predictive value of the stratification system was assessed using net reclassification improvement (NRI).

RESULTS

The overall rebleeding rate at 12 months was 25.00%, with a stepwise increase across the mild (5.56%), moderate (28.00%), and severe (46.15%) groups (P = 0.033). The variceal remission rate at 6 months showed a corresponding inverse gradient (83.33%, 64.00%, and 38.46%, respectively; P = 0.036). Portal vein flow velocity was identified as an independent protective factor against rebleeding (OR 0.60, 95% CI 0.39-0.84, P = 0.009). The stratification system demonstrated incremental predictive value beyond Child-Pugh, MELD, and Baveno criteria (NRI = 0.243, P < 0.001). Internal validation confirmed model stability without overfitting (optimism‑corrected AUC = 0.88).

CONCLUSIONS

The EUS‑based vascular parameter stratification system effectively predicts rebleeding risk and treatment outcomes in patients with EGV undergoing endoscopic therapy. Portal vein flow velocity is a key independent predictor. This stratification approach may serve as a practical tool to guide individualized treatment decisions, though external validation in larger cohorts is warranted.