Luo JC, Leu HB, Hou MC, et al. Nonpeptic ulcer, nonvariceal gastrointestinal bleeding in hemodialysis patients. Am J Med 2013;126:e25–32.
The prevalence of chronic kidney disease (CKD) and subsequent need for dialysis has been on the rise in the United States. The number of patients enrolled in the end-stage renal disease Medicare-funded program increased from approximately 10,000 beneficiaries in 1973 to 547,982 as of December 31, 2008 (Am J Kidney Dis 2002;39[2 Suppl 1]:S1–266). As a consequence, the number of patients requiring dialysis also escalated. In a survey using the United States Renal Data System database, the number of patients who initiated dialysis over the age of 80 increased from 7054 patients in 1996 to 13,577 in 2003 (Ann Intern Med 2007;146:177–183), corresponding to a 57% increase in rate of dialysis initiation after adjustment for population growth.
Prior studies have suggested that the presence of CKD is a risk factor for gastrointestinal (GI) hemorrhage, mainly secondary to GI angiodysplastic lesions (AVMs) in addition to erosive esophagitis (Ann Intern Med 1985;102:588–592). The presence of hemodialysis has been associated with GI hemorrhage, likely owing to intermittent usage of heparin for dialysis treatments and the presence of uremia-induced platelet dysfunction (Am J Med 1985;79:552–559).
In this population-based cohort study, investigators from Taiwan compared the occurrence of nonpeptic ulcer, nonvariceal GI hemorrhage in patients with CKD with or without requirement for dialysis with controls presenting with GI hemorrhage. To perform the study, the investigators analyzed the Longitudinal Health Insurance Database of the National Health Research Institute (NHI) in Taiwan containing a dataset of 1 million randomly sampled individuals who were alive in 2000, traced retrospectively to 1996, and followed to the end of 2006. The authors also used a database of patients on dialysis who applied for catastrophic illness certificates from the Bureau of NHI in Taiwan containing information on inpatient and outpatient claims. The data collected included enrollment files, claims data, catastrophic illness files, and drug prescriptions. Patients with primary hospitalization claims before and at study enrollment for GI malignancy, inflammatory bowel disease, and GI bleeding were excluded, as well as patients who received renal transplantation or peritoneal dialysis before or after enrollment. A control group was matched to the hemodialysis cohort for age, gender, and time of enrollment.
The study endpoint was the occurrence of administrative claims for nonpeptic ulcer, nonvariceal hemorrhage or lower GI hemorrhage as the main diagnosis during a hospitalization after January 1, 2000. Multiple regression analysis was conducted to identify potential risk factors for episodes of GI hemorrhage. The study included 20,830 cases of which 8210 were in the hemodialysis cohort, 4190 in the CKD group, and 8430 controls. Compared with controls, patients with CKD were more likely to have comorbid conditions including hypertension, diabetes, coronary artery disease, heart failure, cirrhosis, and prior history of uncomplicated peptic ulcer disease in addition to the usage of multiple medications including aspirin, nonsteroidal anti-inflammatory drugs (NSAIDs), steroids, clopidogrel, warfarin, and selective serotonin reuptake inhibitors (SSRIs).
During the follow-up period of 7 years, there were 934 cases of nonpeptic ulcer, nonvariceal hemorrhage in the cases (647 in the dialysis group and 189 in the chronic renal group) compared with 98 in the control cohort. Using Kaplan–Meier analysis, the authors demonstrated that the hemodialysis patients had a higher hazard of bleeding compared with both controls and patients with chronic renal disease; in addition, CKD patients had a greater risk of bleeding compared with the controls. On univariate analysis, risk factors for GI bleeding included older age, the presence of hypertension, diabetes, coronary artery disease, and other comorbid conditions; and usage of aspirin, NSAIDs, and SSRIs. After adjustment for the presence of comorbid conditions, age, and usage of medications, multivariate analysis concluded that independent risk factors for nonpeptic ulcer, nonvariceal hemorrhage included older age, diabetes, cirrhosis, chronic obstructive pulmonary disease, history of uncomplicated peptic ulcer disease, chronic renal disease, hemodialysis, and use of SSRIs. The hazard ratio was 9.4 for the hemodialysis group compared with 5.2 for the CKD cohort.
The definition of nonpeptic ulcer, nonvariceal GI bleeding in this study included bleeding from other major causes such as angiodysplasia, diverticular bleeding, Dieulafoy's lesions, erosive esophagitis, gastritis, and/or duodenitis, other causes of lower GI bleeding including anal and rectal sources, and other unspecified sites of hemorrhage including obscure GI bleeding, defined as hemorrhage with negative upper and lower endoscopy findings. Because the performance of video capsule endoscopy and/or enteroscopy is not covered by the National Health Service in Taiwan, most patients with obscure bleeding did not undergo these subsequent examinations after normal upper and/or lower endoscopic examinations; therefore, small bowel pathology was not identified.
Marcuard and Weinstock described a potential association with renal failure and GI angiodysplasia in 1988 (J Clin Gastroenterol 1988;10:482–484). The authors followed 47 patients with CKD on dialysis >5 years and found that 15 of the 47 (32%) experienced GI hemorrhage over the time period secondary to angioectasias found on upper or lower endoscopy. Active bleeding from the lesions was observed in 60% of the patients and 10 (67%) experienced >1 bleeding episode. Erosive gastritis and/or duodenitis was the second most common cause of bleeding.
Chalasani et al investigated the role of CKD in patients presenting with upper GI bleeding (UGIB) in 1996 (Am J Gastroenterol 1996;91:2329–2332). CKD was defined as the presence of a creatinine level >2 mg/dL ≥6 months before and after a UGIB episode. Of the 727 patients with UGIB over a period of 50 months, 60 (8%) had CKD with a mean creatinine level of 5.6 mg/dL. Peptic ulcer disease was the most common cause of bleeding (gastric ulcer in 37% and duodenal in 23% of the cohort) and angioectasias were present in 13%. AVMs were more likely to be present in patients with CRF compared with controls (13% vs 1%; P < .01). The presence of AVMs was correlated with duration of renal failure (P = .004) and need for dialysis (P < .0001).
In 1998, Chak et al (Gastrointest Endosc 1998;47:18–22) reported results from 129 patients who underwent push enteroscopy for obscure GI bleeding. The presence of small bowel angiodysplastic lesions was associated with age >65 years, chronic renal failure, and a history of prior angioectasias on upper or lower endoscopic examinations on univariate analysis. Renal failure was present in 7 of the 13 patients (54%) with small bowel angioectasias (compared with 27% without CKD; P < .05).
A 2003 study published from Iran assessed endoscopic findings in patients with CKD pre-renal transplantation who were receiving weekly dialysis. (Endoscopy 2003;35:502–505). Seventy-four percent of the cohort was asymptomatic; the major symptoms in the remaining 26% of the patients included nausea, vomiting, heartburn, and abdominal pain. The mean hemoglobin level was 9 g/dL and 94% of the cohort was anemic. Upper endoscopy was normal in 36%; the major findings in the remainder of the cohort included duodenal and gastric erosions and antral erythema with detection of Helicobacter pylori infection in 59% of the patients.
In 2005, Lepere et al from Belgium described findings on push enteroscopy in patients presenting with either melena or hematochezia who were considered to have obscure GI bleeding after conventional upper and lower endoscopic examinations (Gastrointest Endosc 2005;61:709–714). The overall diagnostic yield of push enteroscopy was 34% (62/182). Chronic renal failure, defined by a creatinine level >150 mmol/L for >6 months, was observed in 29 of 182 patients (16%), including 13 on hemodialysis. On univariate analysis, the presence of CKD in addition to the presence of melena, hemoglobin level <7 g/dL, requirement of >4 U of blood transfusion, use of anticoagulant therapy, and/or age >65 years were found to be significant predictors for the presence of small bowel lesions. On multivariate analysis, only the risk of CKD and melena were considered to be significant factors. When the authors considered the presence of jejunal lesions only, chronic renal failure was the only predictive factor associated with positive findings on push enteroscopy.
Karagiannis et al in Greece reported usage of wireless capsule endoscopy (WCE) in patients with obscure overt and/or obscure GI hemorrhage comparing a cohort with CKD with controls with normal renal function (World J Gastroenterol 2006; 12:5182–5185). The authors compared WCE findings for 17 patients with CKD compared with 51 patients with normal renal function. WCE identified abnormalities in the small bowel for 71% of patients with CKD compared with 41% of the cohort with normal renal function (P < .05) with angiodysplastic lesions found in 47% of the CKD cohort compared with 18% of controls. There were no differences between cohorts regarding patient age, presence of comorbid conditions, mean hemoglobin levels, or number of blood transfusions. Overall, the presence of CKD was a significant risk factor for small bowel angiodysplastic lesions (odds ratio, 4.1; 95% confidence interval, 1.3–14).
Based on this literature and the current study from Taiwan, the presence of CKD seems to be an independent risk factor for GI hemorrhage. Patients requiring dialysis for CKD seem to be at additional increased risk. The presence of peptic ulcerations is the most common etiology, and the risk of ulcer rebleeding is increased in this cohort (CMAJ 2011;183:E1345–E1351; Gut 2011;60:1038–1042). The current study and prior literature also demonstrates that GI angiodysplastic lesions seem to be the second most common cause of bleeding, and are found to be more prevalent in patients with CKD presenting with GI bleeding compared with control populations.
Reasons for the increase in GI hemorrhage may be related to several factors. First and most important, patients with CKD have underlying platelet dysfunction secondary to the uremic state (Am J Med Sci 1998;316:94–104; Semin Dial 2006;19:317–322). Abnormalities associated with uremia have included defects in platelet function, and platelet–vessel wall metabolism resulting in decreased von Willebrand factor activity. These patients also demonstrate abnormalities in blood coagulation, and a bleeding diathesis resulting in chronic anemia (Kidney Int 2003;64:1455–1461). In addition to physiologic factors associated with uremia, the usage of aspirin and NSAIDs (J Nephrol 2009;22:502–507) and administration of anticoagulant therapy (Am J Med 2013;126:e25–32) increase bleeding risks.
Given the increased prevalence of angiodysplastic lesions in this cohort, future research should focus on the optimal management of these lesions. Although endoscopic therapy is a reasonable treatment option, many of these patients are poor candidates for endoscopic therapy secondary to the presence of CKD and other comorbid conditions. In addition, the AVMs can be multiple and recur despite endoscopic therapy (Endoscopy 2011;43:759–765; J Dig Dis 2013;14:113–116) Medical options for patients with recurrent bleeding can include somatostatin analogs (Dig Dis Sci 2010;55:2129–2134) and/or angiogenic agents including thalidomide (Gastroenterology 2011;141:1629–1637). Further research investigating the role of renal transplantation in this population would be useful. Gastroenterologists should be aware of the increased risk of GI hemorrhage in the patient population with chronic renal disorders.
Published online: August 23, 2013
Philip S. Schoenfeld, Section Editor, John Y. Kao, Section Editor
© 2013 AGA Institute. Published by Elsevier Inc. All rights reserved.