Lower gastrointestinal bleeding (LGIB) is a common

CLINICAL GASTROENTEROLOGY AND HEPATOLOGY 2004;2:485 490 Early Predictors of Severe Lower Gastrointestinal Bleeding and Adverse Outcomes: A Prospective Study FERNANDO S. VELAYOS,* ANN WILLIAMSON, KAREN H. SOUSA, EDWARD LUNG,* ALAN BOSTROM, ELLEN J. WEBER, JAMES W. OSTROFF,* and JONATHAN P. TERDIMAN* Divisions of *Gastroenterology; Nursing; and Emergency Medicine; and Department of Epidemiology and Biostatistics, University of California San Francisco, San Francisco, California; and Division of Nursing, Arizona State University, Tempe, Arizona Background & Aims: Unlike in upper tract bleeding, prognostic factors for ongoing or recurrent bleeding from the lower gastrointestinal tract have not been welldefined. The aim of this study was to identify risk factors for severe lower gastrointestinal bleeding and for significant adverse outcomes. Methods: All patients seeking attention at a university emergency department for gastrointestinal bleeding were prospectively identified during a 3-year period. Ninety-four of 448 (21%) admitted patients had lower gastrointestinal bleeding. Clinical predictors available in the first hour of evaluation were recorded. The primary outcome,severe lower gastrointestinal bleeding,was defined as gross blood per rectum after leaving the emergency department associated with either abnormal vital signs (systolic blood pressure < 100 mm Hg or heart rate > 100/min) or more than a 2-unit blood transfusion during the hospitalization. Significant adverse outcomes,including death,were tabulated. Results: Thirty-seven patients (39%) had severe lower gastrointestinal bleeding. Independent risk factors for severe lower gastrointestinal bleeding were initial hematocrit <35% (odds ratio [OR], 6.3; 95% confidence interval [CI],2.2 16.7); presence of abnormal vital signs (systolic blood pressure < 100 mm Hg or heart rate > 100/min) 1 hour after initial medical evaluation (OR,4.3; 95% CI,1.4 12.5); and gross blood on initial rectal examination (OR,3.9; 95% CI,1.2 13.2). Nineteen patients (20%) experienced a significant adverse outcome,including 3 deaths. The main independent predictor of adverse outcomes was severe lower gastrointestinal bleeding (OR,5.3; 95% CI,1.7 16.5). Conclusions: Risk factors are available in the first hour of evaluation in the emergency department to identify patients at risk for severe lower gastrointestinal bleeding. Severe lower gastrointestinal bleeding is a significant risk factor for global adverse outcomes. Lower gastrointestinal bleeding (LGIB) is a common medical condition accounting for 250,000 500,000 admissions to acute care hospitals in the United States each year. 1 Unlike upper gastrointestinal bleeding, predictors of poor outcome are not well defined, 2 5 and the diagnostic and therapeutic approach is not well-standardized. 6 Distinguishing patients at high risk for severe LGIB and adverse outcomes can be difficult, 7 because although most episodes of bleeding will stop spontaneously, bleeding can be intermittent and severe. 8 Differentiating between high- and low-risk patients is essential if care is to be appropriate and cost-effective. Defining prognostic factors offers the potential to use objective criteria instead of the current subjective and individualized criteria in making major diagnostic and therapeutic decisions. 9 High-risk patients would preferentially receive directed therapy with the goal of altering the natural history of the bleeding and improving outcome. 6,10 The aims of this prospective study were to identify predictors for severe LGIB by using data available in the first hour of evaluation in the emergency department (ED) and to identify risk factors associated with adverse outcomes in LGIB patients. Materials and Methods Patient Population The research protocol was approved by the Institutional Review Board of the University of California, San Francisco. Because the source of bleeding is not always evident on initial presentation, 11 patients with any evidence of gastrointestinal bleeding were identified. Specifically patients aged 18 years and older presenting to the ED at the University of California, San Francisco between April 1996 and March 1999 with a chief complaint of bloody emesis or passage of blood per rectum within the previous 48 hours were prospectively iden- Abbreviations used in this paper: CI,confidence interval; DBP,diastolic blood pressure; ED,emergency department; HR,heart rate; LGIB,lower gastrointestinal bleeding; OR,odds ratio; SBP,systolic blood pressure. 2004 by the American Gastroenterological Association 1542-3565/04/$30.00 PII: 10.1053/S1542-3565(04)00167-3

486 VELAYOS ET AL. CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 2,No. 6 tified by the treating emergency physician or consulting gastroenterologist. Patients transferred from outside institutions or evaluated for stable anemia or trivial bleeding were excluded. Trivial bleeding was defined as a chief complaint of blood on the toilet paper or hematochezia with heme-positive brown stool on rectal examination in the absence of abnormal vital signs or abnormal hematocrit. A total of 550 patients were identified, 532 agreed to participate, and 448 were admitted and prospectively followed. Bleeding was categorized as either proximal (upper) or distal (lower) to the ligament of Treitz on the basis of clinical presentation and the results of diagnostic or therapeutic tests and procedures. Patients with hematochezia and hypotension underwent esophagogastroduodenoscopy to exclude an upper source of bleeding. A total of 354 patients were categorized as having upper gastrointestinal bleeding. The remaining 94 patients (21%) were categorized as having LGIB on the basis of a lower source identified on tagged red blood scan, angiography, colonoscopy, or surgery or by the exclusion of an upper source of bleeding on the basis of clinical presentation and the absence of signs or symptoms of the upper tract (melena, hematemesis) or the presence of a normal esophagogastroduodenoscopy. Predictor Variables and Definitions All data were recorded on a standardized instrument and prospectively collected through patient interview and medical chart review during the patient s hospitalization. Variables recorded included the following: age; sex; comorbid history (cardiac disease, chronic obstructive pulmonary disease, advanced liver disease, or renal failure); history of prior gastrointestinal bleeding; use during the previous 7 days of at least 2 doses of aspirin or a nonsteroidal antiinflammatory agent, 10 mg of prednisone, or any dose of warfarin; consumption of 3 alcoholic drinks per day; history of abdominal pain; initial heart rate (HR) and blood pressure; presence of abnormal vital signs (systolic blood pressure [SBP] 100 mm Hg or HR 100/min) 1 hour after initial medical evaluation; presence on initial medical evaluation of gross blood on rectal examination or presence of abdominal tenderness; initial hematocrit, platelet count, international normalized ratio, and creatinine. For this study, cardiac disease was defined as any history of myocardial infarction, angina, or congestive heart failure; chronic obstructive pulmonary disease as any regular use of bronchodilators, steroids, or prior hospitalizations for chronic obstructive pulmonary disease; advanced liver disease as any documented portal hypertension, encephalopathy, impaired synthetic function (albumin 3.3, or international normalized ratio 1.5); and renal failure as regular dialysis. Other data collected during the hospitalization included the number of packed red blood cells transfused and the results of all endoscopic, nuclear medicine, angiographic, and surgical procedures. Similar predictors were evaluated for adverse outcomes. Severe LGIB was also analyzed as a predictor of adverse outcomes. Outcome Variables and Definitions The outcome of interest for this study was severe LGIB defined as ongoing bleeding or rebleeding after leaving the ED. This was defined operationally as clinically evident gross rectal bleeding at any point after leaving the ED associated either with abnormal vital signs (SBP 100 mm Hg or HR 100/min) or with a transfusion of more than 2 units of packed red blood cells during the hospitalization. Those patients meeting the transfusion requirement after admission but without further evidence of bleeding were not categorized as severe LGIB. An in-hospital adverse outcome was defined as death, myocardial infarction, development or exacerbation of congestive heart failure, precipitation or worsening of dysrhythmia, stroke, onset of respiratory failure, development of an infection requiring intravenous antibiotics, onset of delirium/encephalopathy, or onset of any acute medical condition prolonging hospital stay after admission for gastrointestinal bleeding. Statistical Analysis For the assessment of the prognostic significance of individual factors, continuous variables were first categorized into dichotomous variables, and their predictive value in identifying severe LGIB or adverse outcome was evaluated by using Fisher exact test. Individual odds ratios (ORs) and their 95% confidence intervals (CIs) were computed for each variable. All univariate predictors with a P value 0.05 were evaluated by multiple stepwise logistic regression analysis to identify independent predictors of severe LGIB and adverse outcomes. Age and sex were incorporated in both final models as potential confounders. All reported P values are 2-sided. The SAS (Version 8.2; SAS Institute, Inc., Cary, NC) statistical software was used to perform statistical analysis. Results Patient Characteristics The 94 patients with LGIB ranged in age from 30 to 101 years of age (mean age, 69 years). A total of 50 patients (53%) were men. A total of 41 patients (44%) had at least 1 comorbid condition, and 33 of these patients (80%) were 60 years or older. Diagnostic Workup and Sources of Bleeding Tests performed included angiography in 13 patients (14%), radionuclide scan in 34 (36%), and esophagogastroduodenoscopy in 23 (24%). Lower endoscopy was performed on 79% of patients; 63 had a

June 2004 EARLY PREDICTORS IN LGIB 487 Table 1. Univariate Analysis Between Admission Predictors and Severe LGIB Parameter Severe LGIB (n 37) (%) No severe LGIB (n 57) (%) OR 95% CI P value Demographic Age 65 yr 29 (78) 33 (58) 1.9 0.8 4.8 ns Male gender 21 (57) 29 (51) 1.4 0.6 3.2 ns Medical history Cardiac history 17 (46) 17 (30) 1.8 0.8 4.2 ns Advanced liver disease 0 4 (7) ns COPD 2 (5) 8 (14) 0.3 0.1 1.6 ns Renal disease 0 5 (9) ns Gastrointestinal bleeding 17 (46) 22 (39) 1.1 0.5 2.6 ns Medication use ASA and/or NSAID 16 (43) 23 (40) 1.1 0.5 2.5 ns Prednisone 2 (5) 4 (7) 0.7 0.1 4.1 ns Warfarin 6 (16) 5 (9) 1.9 0.5 6.7 ns Alcohol use 1 (3) 3 (5) 0.5 0.1 4.7 ns Presenting symptoms and signs No abdominal pain 27 (73) 30 (53) 3.5 1.3 9.1 0.013 HR 100/min 14 (38) 14 (25) 1.7 0.7 4.1 ns SBP 100 mm Hg 7 (18) 5 (9) 2.2 0.7 7.7 ns DBP 60 mm Hg 15 (41) 10 (18) 2.9 1.1 7.1 0.03 Abnormal vital signs after 1 hr 17 (46) 7 (12) 5.9 2.1 16.6 0.0006 Gross blood on rectal examination 33 (89) 37 (65) 3.7 1.2 10.9 0.019 Nontender abdomen 30 (81) 35 (61) 2.0 0.8 5.0 ns Initial laboratory values Hematocrit 35% 28 (76) 25 (44) 4.3 1.7 11.1 0.0018 Platelets 150,000/ L 6 (16) 5 (9) 1.9 0.6 6.9 ns INR 1.5 7 (19) 6 (11) 2.0 0.6 6.6 ns Creatinine 1.5 mg/dl 8 (22) 7 (13) 1.9 0.6 5.8 ns ns, not significant; COPD, chronic obstructive pulmonary disease; ASA, aminosalicylic acid; INR, international normalized ratio. colonoscopy, and 11 had a sigmoidoscopy. Only 2 patients underwent urgent colonoscopy (less than 12 hours after admission) with a rapid purge. Therapeutic angiography was used in 3 cases of diverticular bleeding and 1 bleeding small bowel hemangioma. Therapeutic colonoscopy was used to treat 2 colonic ulcers, 1 vascular ectasia, 1 bleeding tumor, and 1 postpolypectomy bleed. Surgical intervention was required on 11 patients (12%) for bleeding from diverticulosis (n 5), small bowel lesions (n 5), and ulceration at an established surgical anastomosis in a patient with Crohn s disease. A total of 9 (9%) patients did not undergo any procedures. Sources of bleeding were determined after complete evaluation. The most common sources of bleeding were diverticulosis in 30 patients (32%) and colitis in 15 (15%). Colonic ulcers accounted for 4 cases (4%), as did bleeding from hemorrhoids, vascular ectasias, and inflammatory bowel disease. Colonic tumors accounted for 6 cases (6%), and postpolypectomy bleeding accounted for 2 cases (2%). No cause for bleeding was found in 10 cases (11%), and a confirmed small bowel source was seen in 6 patients (6%). Outcomes At least 1 unit of packed red blood cells was required by 60 patients (64%), whereas 43 patients (46%) required more than 2 units, 28 patients (30%) required more than 4 units, and 16 (17%) required more than 6 units. After 24 hours of hospitalization, 37 patients (39%) received at least 1 additional unit of packed cells. Severe LGIB occurred in 37 patients (39%). The most common sources of severe LGIB were diverticulosis in 17 patients (45%) and a small bowel source in 5 (13%). Severe bleeding from ischemic colitis, colonic ulcers, or cancer occurred in 3 patients each. One patient had severe bleeding from an arteriovenous malformation and another from Crohn s disease. No source for severe LGIB was found in 4 patients (11%). An adverse outcome occurred in 19 patients (20%), and 14 (74%) of these patients were in the severe LGIB group. Three patients (3%) died. Risk Factors for Severe LGIB Univariate analysis showed 5 predictors associated with severe LGIB (Table 1): lack of abdominal pain,

488 VELAYOS ET AL. CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 2,No. 6 Table 2. Multivariate Predictors for Severe LGIB Parameter OR 95% CI P value Initial hematocrit 35% 6.3 2.2 16.7 0.0007 Abnormal vital signs after 1 hr 4.3 1.4 12.5 0.01 Gross blood on initial rectal examination 3.9 1.2 13.2 0.03 initial diastolic blood pressure (DBP) 60 mm Hg, the presence of abnormal vital signs (SBP 100 mm Hg or HR 100/min) 1 hour after initial medical evaluation, gross blood on initial rectal examination, and initial hematocrit 35%. Multivariate logistic regression analysis identified 3 independent risk factors for severe LGIB: initial hematocrit 35%, presence of abnormal vital signs 1 hour after initial medical evaluation, and gross blood on initial rectal examination (Table 2). The incidence of severe LGIB increased with the number of independent risk factors. Severe LGIB occurred in 11 of 14 patients (79%) with 3 risk factors, 20 of 35 (57%) patients with 2 risk factors, 6 of 36 (17%) with 1 risk factor, and 0 of 9 patients with no risk factors. Risk Factors for Adverse Outcomes Among the risk factors obtainable on admission, initial hematocrit 35% was the only risk factor on univariate analysis associated with adverse outcomes (Table 3). Other factors associated with adverse outcomes, although not statistically significant, were age 65 years (P 0.07), initial DBP 60 mm Hg (P 0.09), and creatinine 1.5 mg/dl (P 0.05). Underlying comorbidity was not associated with adverse outcomes. Once admitted, having severe LGIB was significantly associated with the occurrence of an in-hospital adverse outcome (P 0.002). When analyzed along with admission predictors, severe LGIB was the only independent risk factor for this outcome (OR, 5.3; 95% CI, 1.7 16.5; P 0.004). Discussion Several studies have identified prognostic indicators of severity and mortality in upper gastrointestinal bleeding. 2 5 Similar research in LGIB has been less robust. We prospectively evaluated risk factors available within the first hour of evaluation in the ED to identify those patients at risk for severe LGIB. In addition, we analyzed risk factors responsible for in-hospital adverse outcomes. We found 3 variables that were independent predictors of severe LGIB on multivariate analysis: initial hematocrit 35%, presence of abnormal vital signs (SBP 100 mm Hg or HR 100/min) 1 hour after initial medical evaluation, and gross blood on initial rectal examination. The likelihood of severe bleeding increased with the number of risk factors. These factors are objective and can easily be obtained in the first hour of evaluation in the ED. Because morbidity and mortality are relatively uncommon in LGIB, 12 we used a combined end point for adverse outcomes that encompassed a diverse category of complications. Unexpectedly, we did not find an association between adverse outcomes and comorbidities, possibly in part because of the limited number of comorbid conditions analyzed. We did find Table 3. Univariate Analysis Between Predictor Variables and Adverse Outcomes Parameter Adverse outcome (n 19) (%) No adverse outcome (n 75) (%) OR 95% CI P value Demographic Age 65 yr 16 (84) 46 (61) 3.4 0.9 12.7 ns Male gender 10 (53) 40 (53) 0.9 0.3 2.5 ns Medical history Cardiac history 9 (47) 25 (34) 1.8 0.7 5.1 ns Advanced liver disease 0 4 (5) ns COPD 3 (16) 7 (9) 1.9 0.4 8.1 ns Renal disease 1 (5) 4 (5) 1.0 0.1 9.4 ns Presenting symptoms and signs HR 100/min 6 (33) 22 (30) 1.1 0.4 3.4 ns SBP 100 mm Hg 3 (16) 9 (12) 1.4 0.3 5.9 ns DBP 60 mm Hg 8 (42) 17 (23) 2.5 0.9 7.1 ns Initial laboratory values Hematocrit 35% 16 (84) 37 (49) 5.3 1.4 20.0 0.01 Creatinine 1.5 mg/dl 6 (32) 9 (12) 3.3 1.0 10.8 ns Severe LGIB 14 (74) 23 (31) 5.8 1.9 17.9 0.002 ns, not significant; COPD, chronic obstructive pulmonary disease.

June 2004 EARLY PREDICTORS IN LGIB 489 that severe LGIB was independently predictive of global adverse outcomes, analogous to the association seen between severe upper gastrointestinal bleeding and mortality. 2,5 Our results are comparable to a recent retrospective study by Strate et al. 13 They found ongoing bleeding, defined as bleeding per rectum in the first 4 hours of evaluation, to be one of several early predictors of severity. Other predictors such as hemodynamic instability (initial HR 100/min; SBP 115 mm Hg) and a nontender abdominal examination were also significant in their final model. In our study, analogous measures (DBP 60 mm Hg, and lack of abdominal pain) were significant on our univariate but not multivariate analysis. In contrast to Strate et al., 13 we did not find an association between aspirin use or comorbidity and bleeding severity, although we differed in the methods used to define these predictors. Prognostic factors for severity and mortality in upper gastrointestinal bleeding are used clinically to triage patients into inpatient versus outpatient care, regular floor versus monitored setting, and emergency versus elective endoscopy. 9,14,15 We identified 3 risk factors for severe LGIB after admission that might identify patients on admission who would benefit from more intensive medical care and more urgent endoscopic procedures, analogous to the management strategy used for upper gastrointestinal bleeding. We used urgent colonoscopy infrequently; therefore our ability to analyze its effect on adverse outcomes is limited. The infrequent use of urgent colonoscopy reflected a common management strategy during the time period of the study. Subsequently, interest in urgent colonoscopy has increased since Jensen et al. 10 demonstrated that its use decreased recurrent bleeding and need for surgery in patients with severe LGIB. A more recent randomized study of urgent colonoscopy compared to standard care in 100 consecutively admitted patients with LGIB showed urgent colonoscopy resulted in more definitively diagnosed sources of bleeding but no difference in mortality or rebleeding. 16 Interestingly, the standard care group required therapeutic intervention just as frequently in the form of angiography. We identified risk factors available in the ED associated with severe LGIB; determining the optimal management of LGIB patients and the role of urgent colonoscopy on the basis of risk stratification requires further study. The strengths of this study include the prospective design and the attempt to minimize the number of missed LGIB cases by identifying and following all patients with any evidence of gastrointestinal bleeding as they entered the ED. The main limitation of this study is the sample size. A larger sample size might have resulted in additional predictors achieving significance in our multivariate analysis of severe LGIB and adverse outcomes. Finally, the lack of a uniform definition of severe LGIB might affect the predictive value of these risk factors if a different definition is used. In summary, data are available in the first hour of evaluation in the ED that identify those patients at risk for severe LGIB. Severe LGIB is a significant risk factor for global adverse outcomes. The effect of management decisions such as more intensive medical care or early therapeutic intervention on decreasing adverse outcomes in this high-risk group has not been fully evaluated and merits additional investigation. References 1. Terdiman JP. Colonoscopic management of lower gastrointestinal hemorrhage. Curr Gastroenterol Rep 2001;3:425 432. 2. Clason AE, Macleod DA, Elton RA. Clinical factors in the prediction of further haemorrhage or mortality in acute upper gastrointestinal haemorrhage. Br J Surg 1986;73:985 987. 3. Provenzale D, Sandler RS, Wood DR, Levinson SL, Frakes JT, Sartor RB, Jackson AL, Kinard HB, Wagner EH, Powell DW. Development of a scoring system to predict mortality from upper gastrointestinal bleeding. Am J Med Sci 1987;294:26 32. 4. Rockall TA, Logan RF, Devlin HB, Northfield TC. Risk assessment after acute upper gastrointestinal haemorrhage. Gut 1996;38: 316 321. 5. Zimmerman J, Siguencia J, Tsvang E, Beeri R, Arnon R. Predictors of mortality in patients admitted to hospital for acute upper gastrointestinal hemorrhage. Scand J Gastroenterol 1995;30: 327 331. 6. Richter JM, Christensen MR, Kaplan LM, Nishioka NS. Effectiveness of current technology in the diagnosis and management of lower gastrointestinal hemorrhage. Gastrointest Endosc 1995; 41:93 98. 7. Vernava AM 3rd, Moore BA, Longo WE, Johnson FE. Lower gastrointestinal bleeding. Dis Colon Rectum 1997;40:846 858. 8. Zuckerman GR, Prakash C. Acute lower intestinal bleeding: part I: clinical presentation and diagnosis. Gastrointest Endosc 1998; 48:606 617. 9. Hussain H, Lapin S, Cappell MS. Clinical scoring systems for determining the prognosis of gastrointestinal bleeding. Gastroenterol Clin North Am 2000;29:445 464. 10. Jensen DM, Machicado GA, Jutabha R, Kovacs TO. Urgent colonoscopy for the diagnosis and treatment of severe diverticular hemorrhage. N Engl J Med 2000;342:78 82. 11. Zuccaro G Jr. Management of the adult patient with acute lower gastrointestinal bleeding: American College of Gastroenterology. Practice Parameters Committee. Am J Gastroenterol 1998;93: 1202 1208. 12. Longstreth GF. Epidemiology and outcome of patients hospitalized with acute lower gastrointestinal hemorrhage: a populationbased study. Am J Gastroenterol 1997;92:419 424.

490 VELAYOS ET AL. CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 2,No. 6 13. Strate LL, Orav EJ, Syngal S. Early predictors of severity in acute lower intestinal tract bleeding. Arch Intern Med 2003;163:838 843. 14. Longstreth GF, Feitelberg SP. Outpatient care of selected patients with acute non-variceal upper gastrointestinal haemorrhage. Lancet 1995;345:108 111. 15. Longstreth GF, Feitelberg SP. Successful outpatient management of acute upper gastrointestinal hemorrhage: use of practice guidelines in a large patient series. Gastrointest Endosc 1998; 47:219 222. 16. Green B, Rockey DC, Portwood G, Tamasky PR, Guarisco S, Branch MS, Leung J, Jowell P. Urgent colonoscopy for evaluation and management of acute lower gastrointestinal hemorrhage: a randomized controlled trial. Gastroenterology 2003;124:A17 A18. Address requests for reprints to: Jonathan P. Terdiman,M.D.,Box 1623,University of California,San Francisco,San Francisco,California 94143. e-mail: jterd@itsa.ucsf.edu; fax: (415) 502-2249. Poster presented at the American College of Gastroenterology,October 12 15,2003. The authors thank their research assistants,ilyn Ballesteros,Stacey Hopper,and Natalee Ernstrom,for their contributions to this study. Supported by the American Digestive Health Foundation Clinical Outcomes Research Award 1996 1997.