This prospective, single-center, observational study was conducted at a tertiary hospital in Daegu, Korea (Keimyung University Dongsan Hospital), and was approved by the institutional review board of the university (No. 11-294). Written informed consent was obtained from all of the subjects enrolled in the study. This study is registered in the World Health Organization International Clinical Trials Registry Platform (KCT0000514).

Adult patients (> 18 years), clinically diagnosed with gastrointestinal bleeding between April 2010 and November 2012, were enrolled. Patients were excluded if they did not complete at least 30 days of follow-up, if the bleeding source was varices or gastric cancer, if bleeding was associated with endoscopic procedures such as endoscopic mucosal resection, or if bleeding occurred in the lower gastrointestinal tract. The following information was documented prospectively: patient data (age, sex, date of admission, and endoscopy results); historical data (presenting symptoms, previous history of gastrointestinal bleeding, or peptic ulcer disease); social history such as current smoking status, alcohol consumption (heavy alcoholics were defined as women and men who consumed more than 40 and 60 g alcohol per day, respectively, for more than 5 years) [9]; physical findings (results of nasogastric tube aspiration, results of rectal examination, and initial hemodynamic status); initial laboratory data (hemoglobin and blood urea levels, platelet count, and nitrogen prothrombin time); and comorbidities including hypertension, diabetes mellitus (DM), ischemic heart disease, cerebrovascular disease, heart failure, liver cirrhosis (liver cirrhosis was defined according to clinical, laboratory, and radiologic data but not liver biopsy results because biopsy was not performed) [10], chronic kidney disease (defined as patients with an estimated glomerular filtration rate < 60 mL/min for at least 3 months calculated using the four-variable Modification of Diet in Renal Disease Study equation) [11], metastatic malignancy, and peripheral vascular disease. Hemodynamic instability was defined as tachycardia with a heart rate > 100 beats per minute and hypotension with a systolic pressure < 90 mmHg upon admission and during hospitalization. Medication history (antiplatelet agents, vitamin K antagonists, nonsteroidal anti-inflammatory drugs [NSAIDs], steroids, and proton pump inhibitors) was also recorded if patients had taken the aforementioned drugs 1 week before the first bleeding event. Patients were managed according to recent consensus recommendations, including endoscopic and pharmacologic management as well as transfusion [4]. All of the patients were intravenously administered proton pump inhibitors. In addition, each patient’s risk of bleeding was assessed by a Rockall score and a Blatchford score. The total Rockall score includes both endoscopic and nonendoscopic variables, whereas the Blatchford score is only based on nonendoscopic variables [6,7]. High-risk patients were defined as those with a total Rockall score ≥ 5, indicating an estimated mortality rate of 40% [6].

All of the endoscopic procedures for acute upper gastrointestinal bleeding (UGIB) were performed within 24 hours after arrival at the hospital. Procedures were performed by experienced endoscopists who had previously conducted > 1,000 endoscopies. Urgent endoscopy was defined as an endoscopy procedure conducted within 12 hours of admission [4]. High-risk patients underwent urgent endoscopy according to a strict hospital protocol. However, in patients who showed unfavorable clinical conditions for endoscopy or in those who arrived at night or during the weekend, the decision to perform this procedure was made by the individual endoscopist. Endoscopic variables included the cause of UGIB and location of the lesion. Lesions are either located in the body of the stomach, where the risk of bleeding is high due to numerous supplying vessels, or in other parts of the stomach [12]. Unidentified causes of bleeding were defined as the presence of blood in the stomach without any identifiable source of bleeding [3]. In patients with ulcers, bleeding activity at the ulcer base was classified according to the Forrest classification [13]. Forrest I, IIa, and IIb represent high-risk bleeding stigmata (HRBS) [3,6]. Endoscopic findings were reviewed and adjusted by two expert endoscopists (YJL and ESK) to improve interobserver agreement.

The primary outcome was mortality within 30 days of admission. Patients were advised to return for an examination at 4 and 12 weeks after discharge. For discharged or deceased patients, the relevant data from hospital records were extracted by investigators. Patients with hospital record data that did not satisfy the inclusion criteria were excluded. A 30-day rebleeding episode was defined as the onset of new hematemesis or hematochezia with hypovolemic shock, or a greater than 2 g/dL decrease in blood hemoglobin levels after a 24-hour period of stable vital signs within 30 days of admission [3,14]. All of the rebleeding episodes were confirmed with endoscopy. Bleeding-related 30-day mortality included death from irreversible hypovolemic shock during surgery for uncontrolled bleeding or endoscopy-related mortality occurring within 30 days of the index-bleeding episode. Nonbleeding-related 30-day mortality was defined as death from cardiac, pulmonary or cerebrovascular disease, liver or kidney failure, and malignant disease [15].

Statistical analyses were performed using SPSS version 18.0 (SPSS Inc., Chicago, IL, USA). The Student t test was used for comparison of continuous variables. Categorical variables were compared using Fisher exact test or a chi-square test. Independent risk factors for 30-day mortality were assessed by multivariate logistic regression analysis. An odds ratio (OR) and 95% confidence interval (CI) was calculated for each independent factor. A two-tailed p < 0.05 was considered statistically significant.

During the study period, 336 patients with gastrointestinal bleeding were admitted, and 184 (54.8%) met the inclusion criteria (Fig. 1). A total of 144 patients were excluded from the study because the source of bleeding was varices (82 patients), the lower gastrointestinal tract (59 patients), or gastric cancer (3 patients). Eight patients were lost to follow-up. Patient characteristics are shown in Table 1. The mean age was 59.81 years, and 79.3% of patients were male. The most frequently presenting symptom was hematemesis (50.0%). A total of 48 patients (26.1%) had a previous history of gastrointestinal bleeding, and 60 (32.6%) had a previous history of peptic ulcer disease. The most common comorbidity was hypertension (82 patients, 44.6%), followed by DM (44 patients, 23.9%). With regard to concomitant use of drugs that could have been related to bleeding, 53 patients (28.8%) used antiplatelet agents including aspirin, clopidogrel, or cilostazol, and 37 patients (20.1%) used NSAIDs. As outlined in Table 2, the mean serum level of hemoglobin upon admission was 8.97 g/dL. The percentage of patients with tachycardia (heart rate > 100 beats per minute) and hypotension (systolic pressure < 90 mmHg) during hospitalization was 33.7% and 22.8%, respectively. The percentage of patients with positive nasogastric tube aspiration and digital rectal examination was 68.9% (122/177) and 68.7% (123/179), respectively. Packed red blood cells were transfused in 111 patients (60.3%), and the mean number of units transfused was 2.36 ± 3.69. The mean total Rockall score was 4.62 ± 2.00, and 96 patients (52.2%) had a score ≥ 5, indicating a high risk of mortality. The mean Blatchford score was 11.49 ± 3.38.

An urgent endoscopy (< 12 hours after admission) was conducted in 121 patients (65.8%), whereas the remaining patients (63, 34.2%) underwent endoscopy 12 to 24 hours after admission. Peptic ulcer was the main cause of bleeding (78.3%); other causes of bleeding were Mallory-Weiss syndrome (12.5%), angiodysplasia (1.6%), and hemorrhagic gastritis (0.5%). The cause of bleeding in 13 patients (7.1%) could not be identified despite the presence of blood in the stomach. Of the 144 patients with peptic ulcer, 78 (54.2%) showed HRBS (Forrest I, 22.2%; IIa, 20.8%; and IIb, 11.1%) (Table 3).

The median follow-up period was 196 days (interquartile range, 77 to 404), and the overall number of deaths during this period was 38 (20.7%). The number of deaths within 30 days was 16 (8.7%) (Table 4). Fig. 2 shows the survival curve of patients with NVUGIB over a follow-up period of 30 days. Among patients who died within 30 days, half (8/16, 50.0%) died during the first 7 days; the 7-day survival probability was estimated to be 96.6%. The causes of 30-day mortality are presented in Table 5. Bleeding-related death within 30 days occurred in five patients (31.2%), and all of them died within the first 7 days. The deaths of the remaining 11 patients (68.8%) were associated with their comorbidities. Cardiovascular events including heart failure or acute myocardial infarction were the most frequent nonbleeding-related causes of 30-day mortality (seven patients, 43.8%), followed by liver failure (two patients, 12.5%), brain hemorrhage (one patient, 6.3%), and metastatic cancer progression (one patient, 6.3%). Overall, rebleeding occurred in 38 patients (20.7%) during the follow-up period, and rebleeding within 30 days of admission occurred in 27 patients (14.7%). The median length of hospitalization was 6 days.

According to univariate analysis, age ≥ 65 years (p = 0.009), DM (p = 0.004), chronic kidney disease (p = 0.004), metastatic malignancy (p = 0.023), heart failure (p = 0.016), Rockall score ≥ 5 (p = 0.003), tachycardia (heart rate > 100 beats per minute) during hospitalization (p < 0.001), hypotension (systolic pressure < 90 mmHg) during hospitalization (p < 0.001), and rebleeding within 30 days (p < 0.001) were significant risk factors for 30-day mortality. Multivariate logistic regression analysis identified the following variables as independent predictors of increased 30-day mortality in patients with NVUGIB: comorbidity of DM (OR, 12.67; 95% CI, 1.92 to 83.45; p = 0.008) or metastatic malignancy (OR, 29.24; 95% CI, 2.08 to 411.67; p = 0.012), age ≥ 65 years (OR, 5.06; 95% CI, 1.79 to 32.60; p = 0.048), and hypotension (systolic pressure < 90 mmHg) during hospitalization (OR, 16.63; 95% CI, 2.56 to 107.90; p = 0.003) (Table 6). Predictors for 30-day mortality were also analyzed after dividing patients into high- and low-risk bleeding stigmata groups. However, there were no independent risk factors for 30-day mortality according to risk stratification by Forrest classification (Supplementary Tables 1 and 2).