PREVALENCE, ETIOLOGY, AND PROGNOSTIC-SIGNIFICANCE OF UPPER GASTROINTESTINAL HEMORRHAGE IN DIABETIC-KETOACIDOSIS

We reviewed the discharge records of all diabetic ketoacidosis hospita lizations over 30 months for the presence of clinically significant up per gastrointestinal hemorrhage. Of 284 hospitalizations in 193 patien ts, hemorrhage occurred in 26 hospitalizations (9%) in 25 patients (13 %). None required invasive therapy to achieve hemostasis, and there we re 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 year s, P < 0.02), and more nephropathy (40% vs 11%, P < 0.001), retinopath y (28% vs 12%, P < 0.03) and gastroparesis (36% vs 10%, P < 0.002) tha n those without hemorrhage. Ulcer medication (42% vs 23%, P < 0.03) or anticoagulant (12% vs 1%, P < 0.005) but not nonsteroidal antiinflamm atory drug usage (12% vs 12%) was higher in the hemorrhage group. Admi ssion glucose (P < 0.02), BUN (P < 0.04), and creatinine (P < 0.02) le vels were higher in hemorrhage patients, but arterial pH, serum ketone s, hemoglobin, platelet count, and coagulation values were not. Hemorr hage 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 gas trointestinal hemorrhage complicates 9% of diabetic ketoacidosis hospi talizations. Blood transfusion may be required, but the bleeding is se lf-limited and not severe. The most common lesion is erosive esophagit is. Hemorrhage correlates with glucose level, admission to the intensi ve care unit, duration of diabetes, the presence of diabetic complicat ions, and portends a high non-bleeding-related mortality.