Document resource
Background The optimal timing of endoscopy in critically ill patients with upper gastrointestinal bleeding remains controversial. While early endoscopy is widely recommended, causal evidence for mortality benefit in intensive care unit populations is limited.Methods We conducted a retrospective cohort study using the Medical Information Mart for Intensive Care IV (MIMIC-IV) database. We included adult ICU patients (≥18 years) with upper gastrointestinal bleeding who underwent endoscopy within 72 hours of ICU admission. The primary exposure was early endoscopy (≤12 hours vs >12 hours). The primary outcome was in-hospital mortality; secondary outcomes included 30-day mortality, rebleeding, and blood transfusion requirements. We used Double Machine Learning with cross-fitting to estimate causal treatment effects while adjusting for 11 baseline confounders including age, vital signs, laboratory values, and comorbidities. We assessed treatment effect heterogeneity and conducted extensive sensitivity analyses across alternative time thresholds, outcomes, and subgroups.Results Among 584 ICU patients with upper gastrointestinal bleeding (mean age 64.4 years, 59.8% receiving early endoscopy), the overall mortality rate was 9.8%. Using Double Machine Learning, we found no significant effect of early endoscopy on in-hospital mortality (average treatment effect: +1.45 percentage points; 95% CI: -3.08 to +5.99; p=0.530). However, sensitivity analyses revealed concerning safety signals: early endoscopy was associated with significantly increased rebleeding (+6.18 percentage points; 95% CI: +1.75 to +10.61; p=0.006) and blood transfusion requirements (+9.77 percentage points; 95% CI: +3.78 to +15.76; p=0.001). Heterogeneous treatment effect analysis demonstrated significant age-dependent effects, with older patients (≥median age) experiencing higher mortality with early endoscopy (+7.77 percentage points; 95% CI: +0.44 to +15.10; p=0.038), while younger patients showed trends toward benefit ( IDDF2026-ABS-0241 Figure 1. Forest plot of treatment effects across all outcomes). Post-hoc power analysis revealed only 9% statistical power to detect the observed mortality effect, requiring approximately 29,000 patients for 80% power.Conclusions In this hypothesis-generating study, early endoscopy in critically ill patients with upper gastrointestinal bleeding showed no significant mortality benefit but was associated with increased rebleeding and transfusion requirements. Significant harm was observed in older patients, suggesting that routine early endoscopy may not be appropriate for all ICU patients. These findings challenge current practice paradigms and highlight the need for risk-stratified approaches and adequately powered prospective trials.Abstract IDDF2026-ABS-0241 Figure 1