Abstract
Background/Objectives: Gastrointestinal bleeding (GIB) in hemodialysis (HD) patients carries substantial mortality risk. The A4C and CHAMPS scores are novel risk stratification tools, while CAGIB was developed for cirrhosis-associated GIB. We compared the discriminative performance of these scores in HD patients with acute GIB, stratified by variceal and non-variceal etiology. Methods: We conducted a retrospective cohort study of 57 HD patients with acute GIB (January 2020-December 2024) following STROBE and TRIPOD guidelines. Patients were stratified as non-variceal (n = 42) or variceal (n = 15). The primary outcome was 30-day mortality; secondary outcomes included ICU admission, rebleeding, and transfusion requirements. A4C, CHAMPS, CAGIB, ABC, AIMS65, and Glasgow-Blatchford scores were compared using AUROC analysis. Results: Mean age was 45.8 ± 13.2 years. Non-variceal GIB (73.7%) was predominantly caused by angiodysplasia (28.6%) and peptic ulcer disease (23.8%); variceal GIB (26.3%) was mainly from esophageal varices (80.0%). Overall 30-day mortality was 17.5%, significantly higher in variceal (26.7%) versus non-variceal GIB (14.3%, p = 0.048). For non-variceal GIB, CHAMPS demonstrated excellent mortality discrimination (AUROC 0.91), significantly outperforming CAGIB (AUROC 0.68, p = 0.02). Conversely, for variceal GIB, CAGIB showed superior performance (AUROC 0.89) compared to CHAMPS (AUROC 0.72, p = 0.04). A4C performed consistently for transfusion prediction across both groups (AUROC 0.75-0.78). Conclusions: Optimal risk stratification in HD patients with GIB requires etiology-specific scoring: CHAMPS for non-variceal and CAGIB for variceal bleeding. This complementary performance reflects distinct pathophysiological mechanisms underlying mortality. Prospective validation in larger multicenter cohorts is warranted.