Acute upper gastrointestinal (GI) hemorrhage is one of the commonest causes for hospitalization worldwide. Endoscopic therapy is effective in achieving primary hemostasis. The shift of management from the operating theater to the endoscopy suite has not changed the rate of mortality over the past 20 years. Several hypotheses are discussed that may account for the lack of improvement in the mortality resulting from bleeding peptic ulcer. One potential way to improve management is to identify those at risk for adverse outcomes, which may improve the initial triage, timing of primary endoscopic hemostasis, and postendoscopic management. Two adverse outcomes generally considered as significant for acute upper GI hemorrhage are rebleeding and mortality. Numerous clinical risk models have been developed to predict these outcomes, and this article reviews them.
Acute upper gastrointestinal (GI) hemorrhage is one of the commonest causes for hospitalization worldwide. In the United States, there are 250,000 to 300,000 hospital admissions and 15,000 to 30,000 deaths each year resulting from acute upper GI hemorrhage. In England it is a common medical emergency with an annual incidence of 100 per 100,000. Two meta-analyses conducted in the 1990s confirmed that endoscopic therapy is effective in achieving primary hemostasis. The shift of management from operation theater to endoscopy suite has not changed the rate of mortality over the past 20 years. The reported mortality for patients with bleeding peptic ulcer still amounts to 15%. Several hypotheses may account for the lack of improvement in the mortality resulting from bleeding peptic ulcer. Coinciding with the decline in the prevalence of Helicobacter -related peptic ulcer, the incidence of aspirin-related or other nonsteroidal antiinflammatory drug (NSAID)–related ulcer disease is on the rise. Patients taking aspirin or other NSAIDs are usually in the older age group and have a background of severe comorbidities such as ischemic heart disease. They are more vulnerable to a physiologic challenge from an acute bleeding episode. One of the potential directions toward improvement in the management of acute GI bleeding is to identify those at risk for adverse outcomes, which may improve the initial triage, timing of primary endoscopic hemostasis, and postendoscopic management. Two adverse outcomes generally considered as significant for acute upper GI hemorrhage are rebleeding and mortality, and numerous clinical risk models have been developed to predict these outcomes. The following section focuses on a review of the reported predictive models in the literature on mortality and rebleeding for acute upper GI hemorrhage.
Prediction of mortality for acute nonvariceal upper GI hemorrhage
Rockall and colleagues developed a well known prediction score for mortality from a prospective multicenter study of 4185 cases of acute upper GI hemorrhage. These included diagnosis of various etiologies, such as bleeding peptic ulcers, varices, esophagitis, Mallory-Weiss syndrome, and malignancy. This predictive scoring was subsequently accredited in another cohort of 1625 patients. The Rockall score consisted of age, shock, major comorbidities, the endoscopic diagnosis, and the presence of major stigmata of recent hemorrhage (SRH). All these factors were then combined and the significance of each factor was weighted according to their representation in the multivariate analysis ( Table 1 ). These showed that the risk for mortality increased when the score increased. The Rockall score was further validated in a Dutch group’s multicenter prospective cohort on prediction of both mortality and rebleeding in 951 patients. On analysis with the receiver-operator characteristics (ROC) curve for prediction of mortality, it showed an area-under-curve (AUC) of 0.73 in the Dutch sample and an AUC of 0.81 in the Rockall sample. This confirmed that Rockall score is satisfactory in predicting mortality among patients with acute upper GI hemorrhage. However, the prediction of rebleeding did not seem to fit in both groups of patients. As Rockall score was initially designed to evaluate the risk for mortality, it may not perform as well to predict rebleeding. Another study enrolled 211 patients who were prospectively collected in two randomized controlled trials, and they were retrospectively evaluated on the prediction of clinical outcomes using Rockall score. The authors found that those with a Rockall score of more than seven had a significantly higher risk for rebleeding and mortality. The study was limited by the sample size and retrospective nature. Zimmerman and colleagues analyzed the prognostic factors among 321 patients with acute upper GI hemorrhage from 1988 to 1991. Multivariate analysis showed that age older than 75 years, blood in stomach, high serum creatinine and aminotransferase, and persistent or recurrent bleeding were significant factors that predicted mortality. In this study, there was no validation cohort to reply and confirm the predictive power of these factors. Moreover, there was a high rate of 18% rebleeding after endoscopic therapy and adjuvant high-dose proton pump inhibitor (PPI) therapy was not used as postendoscopic treatment.
| Rockall et al (1995) | Zimmerman et al (1995) | Blatchford et al (1997) | Klebl et al (2005) | Imperiale et al (2007) | Marmo et al (2008) | Chiu et al (2008) | |
|---|---|---|---|---|---|---|---|
| Age | 2.8 (2.14–3.67) | 12.7 (1.9–84.4) | 4.5 (2.0–10) | 4.7 (2.3–7.0) | 1.47 (1.09–1.98) | ||
| ASA | 7.24 (4.0–10.4) | ||||||
| APACHE II | 3.11 (1.13–8.57) | ||||||
| No. of comorbidities | – | 1.70 (1.23–2.36) | |||||
| Type of comorbidities | – | 5.25 (2.34–11.79) | 4.12 (2.65–6.39) | ||||
| Cardiac failure | 7.73 (5.68–10.5) | 9.4 (3.2–28) | |||||
| Ischemic heart disease | 4.3 (3.17–5.81) | ||||||
| Liver cirrhosis/failure | 8.65 (6.01–12.5) | 43 (14–133) | 11.01 (7.5–14.5) | ||||
| Renal disease/failure | 10.3 (6.96–15.3) | 2.4 (1.3–4.2) | 5.29 (2.8–7.8) | ||||
| Diabetes mellitus | 0.5 (0.3–1.1) | ||||||
| Disseminated malignancy | 11.7 (8.28–16.6) | 3.8 (1.8–8.1) | 9.81 (7.3–12.3) | ||||
| Hematological malignancy | 6.23 (3.91–9.94) | ||||||
| Coagulopathy | 3.2 (1.8–5.7) | ||||||
| Systolic blood pressure | 9.8 (5.1–19) | 2.24 (1.65–3.05) | |||||
| Tachycardia | 2.17 (1.73–2.71) | 0.6 (0.43–0.91) | |||||
| Presentation of shock | 3.37 (2.58–4.39) | ||||||
| In-hospital bleeding | 4.8 (2.5–9.2) | 2.46 (1.78–3.39) | |||||
| Hematemesis | 2.0 (1.1–3.5) | 1.45 (1.02–2.05) | |||||
| Syncope | 3.0 (1.3–7.0) | ||||||
| Corticosteroid usage | 2.2 (1.0–4.3) | ||||||
| Hemoglobin | 1.94 (1.61–2.33) | 3.8 (2.4–6.2) | 4.7 (2.3–7.0) | ||||
| Urea | 18 (5.3–59) a | ||||||
| Creatinine level | 14.8 (2.6–83.5) | 5.8 (3.2–11) b | |||||
| Aminotransferase | 20.2 (2.9–140.4) | ||||||
| Diagnosis at endoscopy | 3.93 (2.8–5.5) c | 3.85 (1.31–11.3) | |||||
| SRH | 3.97 (3.08–5.1) | 5.28 (1.89–14.8) | |||||
| Blood in stomach | – | 18.9 (1.8–203.7) | |||||
| H pylori | 0.20 (0.13–0.33) | ||||||
| Rebleeding | 6.24 (5.15–7.56) | 57.3 (7.2–453.9) | 1.63 (1.09–2.41) | ||||
| Surgery | 4.60 (2.95–7.19) |
Blatchford and colleagues conducted a regional audit in the west of Scotland on the epidemiology and clinical outcomes of patients who were admitted to 19 regional hospitals with a complaint of acute upper GI hemorrhage. The overall incidence was 172 per 100,000 per annum, and it was 67% more than the peak incidence reported in the United Kingdom. In 6 months the mortality was 8.1% among 1882 patients in the study group. Multiple logistic regression showed that age, uremia, diastolic hypotension, cardiac and hepatic failure, pre-existing malignancy, major pre-existing diseases such as rheumatoid arthritis or inflammatory bowel disease, and presentation with hematemesis or syncopy were associated with increased risk of fatality. Klebl and colleagues reported a retrospective review of 454 patients with upper GI bleeding from 1992 to 1999. The main cause of upper GI hemorrhage in this group is still peptic ulcer, followed by varices and gastritis. Multivariate analysis showed that in-hospital bleeding, coagulopathy, renal disease, use of corticosteroid, cardiovascular disease, and diabetes mellitus were risk factors for mortality. The mortality rate in this group amounts to 26.5%, and 66.3% of these patients needed intensive care. The authors also analyzed the Rockall score for these patients and found that the mean Rockall score was 6.6 ± 1.3. This cohort most likely represented a selected group of patients who were at high risk and managed in a tertiary referral center. These risk factors identified were mostly related to significant comorbidities and background illnesses. Imperiale and colleagues derived and internally validated several clinical prediction rules for risk stratification of patients with acute upper GI bleeding in a prospective cohort involving Veterans Affairs Medical Centers in Durham, Indianapolis, and Seattle. The adverse outcomes were defined as: (1) GI hemorrhage–specific outcomes (rebleeding, need for surgery or hospital death) and (2) GI bleeding–specific outcomes plus new or worsening comorbidity. From a total of 391 patients enrolled, the rate of rebleeding was 4.6% and the overall mortality was 3.1%. From the derivation cohort, several factors were identified from multiple logistic regressions as predictors which included stigmata of recent hemorrhage, APACHE II score greater than 11, diagnosis of esophageal varices, and unstable comorbidity at hospital admission. These factors were then combined to classify the validation cohort into low-, intermediate-, and high-risk for adverse outcomes. The c-statistic for prediction of GI hemorrhage–specific outcomes was 0.83. This study defined outcomes as GI hemorrhage–specific outcomes, including rebleeding, need for surgery, or hospital death. Hence this model was not specific for predicting mortality. Moreover, the population from which this risk model was derived was US veterans, which led to a 99% preponderance of men. APACHE score had been studied by several other groups for risk assessment and prediction of death in acute upper GI hemorrhage. Wang and colleagues used APACHE II to differentiate patients at high risk with bleeding gastric ulcers. It was found that the mortality rate was 58% for those with an APACHE score of 15 or greater as compared with 5% for those with a score less than 15. The patients included in this study were those who had failed endoscopic treatment. Hence they represented a selected high-risk group of severe hemorrhage requiring surgical interventions, and their outcomes were expected to be poor. There may also be a logistic issue in clinical application of this risk model, as APACHE score will usually be available after the completion of the clinical management process. Marmo and colleagues reported a prospective multicenter study from Italy on patients with upper GI hemorrhage. Among a prospectively recruited cohort of 1020 patients, they showed an overall mortality of 4.5% and a rebleeding rate of 3.2%. Regression analysis showed that advanced age, presence of severe comorbidity, low hemoglobin levels at presentation, and worsening health status were independent factors that predicted 30-days mortality. The authors used a computer linked system for collection of data with independent data validation from a random set of samples for quality assurance. Ninety-seven percent of patients in this cohort received PPIs after endoscopic hemostasis. The use of H2 receptor antagonist was associated with a significantly increased 30-days mortality (15% versus 4%) while use of PPIs showed a protective effect with an absolute risk reduction of 11.5%. Leontiadis and colleagues reported a systematic review on the use of PPIs after endoscopic therapy for bleeding peptic ulcer and showed that PPIs significantly reduce rebleeding and need for surgery. The focus was on identification of risk factors for bleeding peptic ulcer–related mortality among 3220 patients as derivation cohort, and 634 patients as evaluation cohort. On multivariate analysis, age older than 70 years, presence of listed comorbidities, more than one comorbidity, hematemesis, initial systolic blood pressure <100 mmHg, and in-hospital bleeding were independent pre-endoscopic predictors, whereas rebleeding and need for surgery were postendoscopic predictors for in-hospital mortality. The resulting factors were grouped to a score and validated in a separate cohort of 634 prospective patients. The authors found that the risk for mortality increased when the composite score was ≧3. The model was limited by the fact that only people of Chinese ethnicity were included. Validation would be necessary if this predictive score is to apply to acute upper GI hemorrhage and other ethnicities.
Table 1 summarizes the reported predictive factors for mortality in patients presenting with acute upper GI bleeding. Each of the identified predictive factors was represented by the odds ratio and the 95% confidence interval. The most frequently reported factors included old age, presence of significant comorbidity including cardiac disease, liver and renal failure, disseminated malignancy, presentation with shock (in terms of low systolic blood pressure or tachycardia), in-hospital bleeding, low hemoglobin, diagnosis at endoscopy, presence of SRH, and development of rebleeding. These factors can be arbitrarily subclassified into pre-endoscopic and postendoscopic factors. Pre-endoscopic factors such as advanced age and significant comorbidities coupled with a severe episode of bleeding represented by significant hemodynamic changes or anemia upon presentation, lead to high mortality. A retrospective review from the authors’ institute from a cohort of 10,428 patients who presented with peptic ulcer bleeding showed that the majority of the mortality was because of non–bleeding-related death, including multiorgan failure, pulmonary conditions, or terminal malignancy. The clinical applications of these predictive factors were explored in another study which developed the Cedars-Sinai Medical Center predictive index based on an explicit review of the reported risk factors in the literature. The score includes clinical and endoscopic predictors including a predefined list of significant comorbidities, advanced age, endoscopic diagnosis, and evidence of active bleeding. Among a cohort of 500 patients, it was found that the application of this predictive score will allow a significantly shorter hospital stay with minimal complication.
Prediction of mortality for acute variceal hemorrhage
Most of the prediction factors for mortality among patients with nonvariceal upper GI hemorrhage were reviewed. Liver failure and cirrhosis were reported in several studies as important comorbidities to predict mortality. For patients with a background of liver cirrhosis presenting with acute variceal hemorrhage, the risk for mortality can be up to 50%. Lecleire and colleagues prospectively studied patients from four French geographic areas, with or without liver cirrhosis, who presented with acute upper GI hemorrhage. Variceal bleeding served as the most common cause (59.1%) for patients with cirrhosis, whereas bleeding peptic ulcer was the commonest diagnosis (41.8%) for patients without cirrhosis. Six independent predictors for mortality were identified for both groups, which included decreased prothrombin level, coexisting digestive carcinoma, use of corticosteroid, occurrence of upper GI hemorrhage in an inpatient, hematemesis, and age over 60 years ( Table 2 ). The prothrombin level directly correlated with the severity of liver cirrhosis. Le Moine derived predictive factors for early mortality in patients who had cirrhosis with bleeding varices among 121 patients. Three independent prognostic indicators were identified including encephalopathy, prothrombin time, and amount of blood transfused. The authors compared the predictive power of these factors to the Child-Pugh grading and found that the composite model achieved a better prediction. Two of the factors within the Child-Pugh grading, namely encephalopathy and prothrombin time, are important predictors for mortality. Sakaki and colleagues reviewed 98 patients who were on long-term endoscopic sclerotherapy and presented with acute variceal hemorrhage. It was found that the presence of hepatocellular carcinoma, bilirubin and albumin levels, and time to reach hospital were independent predictive factors for mortality. The conclusion coincided with the previous report that remnant hepatic function was an important prognostic indicator. Gatta and colleagues developed and validated a prognostic index on the basis of risk factors identified from 268 patients with cirrhosis. These factors included gender, presentation with hematemesis, serum creatinine, ascites, hepatocellular carcinoma, serum bilirubin, prothrombin index, and bleeding varices. This prognostic index was significantly better than Child-Pugh score and the Garden prognostic index in predicting mortality among patients with upper GI bleeding and cirrhosis. Afessa and Kubilis compared the Child-Pugh grading, Garden’ score, Gatta’s score, and APACHE II score in predicting the clinical outcomes for patients with cirrhosis and acute upper GI bleeding. Among 85 patients, the in-hospital mortality rate was 21% and the intensive care unit (ICU) admission rate was 71%. Child-Pugh grading showed an area-under-curve (AUC) of 0.76, which was similar to Garden’s (AUC of 0.70) and Gatta’s scores (AUC of 0.71). Although APACHE II achieved a similar AUC of 0.78, it tended to overestimate the risk for mortality. The new APACHE III system may improve the deficiency of APACHE II in predicting the outcome for patients with cirrhosis as its components include liver function. Among 101 patients with liver cirrhosis and esophageal variceal bleeding requiring ICU admission, total volume of ethanolamine injection during sclerotherapy, multiple blood transfusions, Glasgow Coma Scale (GCS), international normalized ratio (INR), and the presence of shock on admission were independent predictors for mortality. In this study, neither the APACHE II nor Child-Pugh grading was an independent predictor for mortality. The authors commented that both Child-Pugh and APACHE II were not predictors despite showing a difference in the outcomes by univariate analysis, GCS having a strong predictive value on multivariate analysis. GCS probably reflected the severity of hepatic encephalopathy as urea or creatinine levels showed the severity of hepatorenal syndrome.
| Garden et al (1985) | Le Moine et al (1992) | Lee et al (1992) | Gatta et al (1994) | Sakaki et al (1998) | Lecleire et al (2005) | |
|---|---|---|---|---|---|---|
| Age >60 | 1.83 (1.07–3.13) a | |||||
| Gender | 0.386 (0.043) | |||||
| Encephalopathy | <0.001 b | −2.748 (<0.01) | ||||
| GCS | 1.3 c | |||||
| Hepatocellular carcinoma | −0.842 (0.004) | 0.968 (0.0027) | 4.54 (2.07–9.95) a | |||
| Ascites | −0.306 (<0.001) | |||||
| In-hospital bleeding | 3.55 (1.92–6.58) a | |||||
| Hematemesis | 0.405 (0.037) | 2.66 (1.46–4.84) a | ||||
| Corticosteroid usage | 3.80 (1.32–10.9) a | |||||
| Time to hospital | 1.931 (0.0001) | |||||
| Shock | 3.7 c | |||||
| Blood transfusion (units) | −0.310 (<0.05) | 6.1 c | ||||
| Bilirubin | 0.005 b | −0.010 (0.001) | 1.463 (0.0001) | |||
| Prothrombin time/INR | 0.102 (<0.01) | 1.9 c | 0.036 (0.013) | 7.93 (4.58–13.7) a | ||
| Albumin | −1.068 (0.0033) | |||||
| Urea | 0.05 b | |||||
| Creatinine level | 0.001 b | −0.016 (<0.001) | ||||
| Varices as source of bleeding | −0.442 (0.020) | |||||
| Endoscopic sclerotherapy (volume of sclerosant injected) | 1.2 c |
Related posts:
The Role of Proton Pump Inhibitors in the Management of Upper Gastrointestinal Bleeding
Management of Massive Peptic Ulcer Bleeding
Stress-Induced Ulcer Bleed ing in Critically Ill Patients
Helicobacter pylori -Negative Nonsteroidal Anti-Inflammatory Drug-Negative Ulcer
Differences in Peptic Ulcer Between the East and the West
Balancing Risks and Benefits of Cyclooxygenase-2 Selective Nonsteroidal Anti-Inflammatory Drugs
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
