Critical Appraisal Of The Literature
Critical Appraisal Of The Literature
There is an abundance of literature regarding GI bleeding. The subset of this literature that is most relevant to emergency physicians involves the identification and assessment of risk factors for subsequent complications, including rebleeding, the need for surgery to control hemorrhage, the need for transfusion, and death. Risk stratification studies range from large, multicenter, prospective studies with thousands of patients to very small retrospective studies. There is no standardization of terms in the literature, and, at times, terms are used without specific definitions.
When this body of literature is reviewed, even given the inherent limitations and inconsistencies, some features of high-risk and low-risk patients are suggested. (SeeTable 1.)
In the past, GI bleeding had been an indication for admission to the hospital in all but the most trivial cases. Clinicians probably recognized that this approach was in all likelihood excessively risk averse, but given the mysterious nature of GI bleeding and medicolegal concerns, it was difficult to safely select patients for discharge from the ED. Over the past 20 years, a number of studies have been published addressing this problem. The typical approach has been to first identify risk factors either through a review of the available literature, a chart review, or retrospective study comparing risk factors with outcome. The specific outcomes of interest vary from paper to paper, but the emphasis on minimizing risk in the management of GI hemorrhage is a common theme.
A small number of studies looked at risk stratifying patients based solely on clinical features without including endoscopy findings. Bordley et al published one of the earliest such studies in 1985.1 The paper identified six "early predictors" defining the low-risk patient. The authors asserted that a patient was low-risk if he or she met the following criteria: 1) age less than 75 years; 2) no unstable comorbid illness; 3) no evidence of ascites on examination; 4) normal prothrombin time; 5) systolic blood pressure of 100 mmHg or greater; and 6) a nasogastric aspirate free of fresh blood. In the prospective arm of the study, two of the 52 patients (3.8%) defined as "low risk" had poor outcomes. One died and another required urgent surgery. The study was small (111 patients in the prospective arm), but it suggested that some patients with acute GI hemorrhage might indeed be managed selectively based on simple clinical features.
In 1991, Wrenn et al reported on a cohort of subjects who presented to the ED with hematemesis, were prospectively enrolled, had well-defined low-risk characteristics, and were discharged home from the ED without endoscopy.2 Although this study design should yield results directly relevant to practicing emergency physicians and no patients were found to have clinically significant complications, the study included only 33 patients, and two of them were lost to follow-up. Obviously, it is difficult to draw conclusions from such a small negative study.
In 1992, Harland et al reported on a retrospective study of subjects with hematemesis.3 They identified two groups and concluded that subjects who vomited only altered blood (coffee grounds) and who did not have melena, and patients who vomited only altered blood and had hemoglobin concentrations greater than 12 g/ dL, had a low risk of complications. Only 145 subjects (157 separate visits) were included in this study; 79 of them were in the low-risk groups.
In 1997, Kollef et al reported on a classification tool for risk stratifying patients hospitalized for acute upper and lower GI bleeding.4 Although this classification tool had a catchy name ("BLEED") and used simple clinical criteria, their logistic regression model may have been inappropriate due to variables that were not independent, and the test characteristics reported for their model were not very good (area under the receiver operating characteristic curve of only 0.72).
In 2000, Blatchford et al reported a prospective validation of a risk score based solely on clinical criteria before endoscopy.5 This study included 197 patients. Although reasonable test characteristics were presented for the calculated risk scores based on clinical criteria, the dispositions of the patients and the performance of endoscopy in these subjects were not adequately described.
The small sample sizes of these studies limit the general applicability of the results. Although it is appealing to potentially avoid the time, cost, and risk of endoscopy in some patients, the combination of endoscopy results and clinical features to risk stratify patients is much more common in the literature.
Prompt endoscopy is a key feature of most studies of patients with hematemesis, melena, or hematochezia. (See Table 2.) Several interesting risk stratification papers were published during the mid-to-late 1990s. The size of these studies varied. Large retrospective studies examined the risk factors for complications of acute upper GI hemorrhage. For example, Katschinski et al reported on 2217 patients,6 and Yavorski et al reported on 3294 patients.7 A different approach seemed to be taking form when Longstreth and Feitelberg reported on their experience using a practice guideline to select patients for prompt discharge from the hospital following endoscopy.8,9 The patient population in their study was drawn from a staff-model health maintenance organization where urgent endoscopy was readily available and patient tracking was excellent. In 1995 they reported on the development of their practice guideline and on 141 prospectively enrolled patients with acute nonvariceal upper GI bleeding.9 Of these patients, 34 were treated as outpatients, and one of these patients returned with rebleeding. Another study by Longstreth and Feitelberg reported on a consecutive series of 176 patients who met their practice guideline and were treated as outpatients after endoscopy.8 In this cohort, no patients died, but two were subsequently hospitalized and one had rebleeding. In the historical context of all patients with GI bleeding being hospitalized, the concept of treating some patients as outpatients was interesting. In 1999, however, Tham et al revealed some of the limitations to this approach for an urban teaching hospital.10 In their study population, only 18 of 145 patients presenting to the ED with acute upper GI hemorrhage met Longstreth and Feitelberg's criteria for outpatient care. Rockall et al developed a similar set of low-risk criteria and had much larger sample sizes.11,12 Unfortunately, a follow-up validation study failed to validate the Rockall risk scoring system.13 Hay et al produced similar results with a different practice guideline and published the results in 1996 and 1997.14,15 The focus of Hay et al's work was on decreasing the length of hospital stay and not on discharge from the ED following endoscopy.
Several large and interesting drug studies have been published. Derry and Loke presented a meta-analysis of 24 randomized, controlled trials and suggested that longterm aspirin use is associated with a significant increase in the incidence of GI hemorrhage.16 Ofman et al presented a meta-analysis of 16 placebo-controlled randomized clinical trials and about 30 other studies and reported that nonsteroidal anti-inflammatory drugs (NSAIDs) are associated with an increased risk of GI hemorrhage.17 Khuroo et al presented results from a double-blind, placebo controlled trial of omeprazole vs. placebo for bleeding peptic ulcers and demonstrated that omeprazole was better than nothing for preventing recurrent bleeding.18 Several well-designed randomized, controlled trials have demonstrated that octreotide is a useful adjunct to endoscopic sclerotherapy for preventing rebleeding from esophageal varices.19-22
John L. Westhoff; Kurtis R. Holt
March 1, 2004