Prevalence, etiology, and prognostic significance of upper gastrointestinal hemorrhage in diabetic ketoacidosis

Abstract

We reviewed the discharge records of all diabetic ketoacidosis hospitalizations over 30 months for the presence of clinically significant upper gastrointestinal hemorrhage. Of 284 hospitalizations in 193 patients, hemorrhage occurred in 26 hospitalizations (9%) in 25 patients (13%). None required invasive therapy to achieve hemostasis, and there were no bleeding recurrences and no deaths due to bleeding. Endoscopy in eight revealed esophagitis in all (five had erosions or ulcerations), one Mallory-Weiss tear, five with gastritis (mild in four), four with duodenitis (one erosive), one duodenal ulcer, and no gastric ulcers. Hemorrhage patients had a longer diabetes duration (14.85 vs 9.16 years, P < 0.02), and more nephropathy (40% vs 11%, P < 0.001), retinopathy (28% vs 12%, P < 0.03) and gastroparesis (36% vs 10%, P < 0.002) than those without hemorrhage. Ulcer medication (42% vs 23%, P < 0.03) or anticoagulant (12% vs 1%, P < 0.005) but not nonsteroidal antiinflammatory drug usage (12% vs 12%) was higher in the hemorrhage group. Admission glucose (P < 0.02), BUN (P < 0.04), and creatinine (P < 0.02) levels were higher in hemorrhage patients, but arterial pH, serum ketones, hemoglobin, platelet count, and coagulation values were not. Hemorrhage patients required more blood transfusions (27% vs 10%, P < 0.003) and intensive care unit admissions (69% vs 43%, P < 0.009). Total (15% vs 3%, P < 0.003) and intensive care unit mortality (22% vs 6%, P < 0.026) were higher in the hemorrhage group. We conclude that upper gastrointestinal hemorrhage complicates 9% of diabetic ketoacidosis hospitalizations. Blood transfusion may be required, but the bleeding is self-limited and not severe. The most common lesion is erosive esophagitis. Hemorrhage correlates with glucose level, admission to the intensive care unit, duration of diabetes, the presence of diabetic complications, and portends a high non-bleeding-related mortality.