Evaluation of scoring systems without endoscopic findings for predicting outcomes in patients with upper gastrointestinal bleeding

Upper gastrointestinal bleeding (UGIB), a common medical emergency, is reported in the emergency room by 5% per year and accounts for 2~3% of hospitalization in developed countries [1]. Mortality for UGIB reportedly ranges from 2% to 15% and rebleeding is as high as 10–30% [2‐4]. Therefore, proper stratification for patients with UGIB can help to identify candidates for interventions including blood transfusion, endoscopic treatment, and radiologic or surgical intervention. Moreover, this process helps to decrease the 30-day mortality.

The 2010 International consensus guidelines recommended early risk stratification using proven factors for the management of patients with UGIB [5]. The recent American College of Gastroenterology practice guidelines for managing patients with ulcer bleeding also recommended that a risk assessment be conducted to stratify patients into high and low risk categories, which may assist in initial decisions such as timing of endoscopy, time of discharge, and level of care [6]. Early risk stratification in the emergency department facilitates proper treatment as well as rapid and accurate triage. This is important for timely administration of lifesaving treatments to patients and for reducing high medical care costs.

Several scoring systems have been developed to evaluate patients with UGIB [7, 8]. Decisions about urgent endoscopy are important in the diagnosis and treatment for patients with UGIB. The Rockall score (RS) is not suitable for the decision about urgent endoscopy since it requires endoscopic results. Whereas, the Pre-endoscopy Rockall score (Pre-E RS) and the Glasgow-Blatchford score (GBS) are suitable because they require only clinical and laboratory data. Therefore, they could be applied immediately without endoscopic findings. In addition, a modified GBS (mGBS), which eliminated subjective criteria from GBS, has been introduced [9].

In this study, we determined the immediate applicability without endoscopic findings of the GBS, mGBS, and Pre-E RS systems in estimating the need for interventions and predicting the 30-day mortality in patients with UGIB. Also, we investigated the potential cutoff scores to predict the need of interventions and 30-day mortality in applicable scoring systems.

In our study, the rate of subjects who received blood transfusion (76.1%) was higher compared with previously published studies [12, 13], because the rate of patients who showed hemoglobin under 8 g/dL was higher. On the other hand, 30-day mortality in our study was 4.2%, which was lower than that of previous report by Laursen et al. [12]. The reason of this difference was considered that the rate of UGIB caused by malignancy was lower in the present study. The other characteristics of patients such as the rate of patients who underwent endoscopic or radiologic interventions were consistent with the previous study [12].

Lin et al. [14] reported that cyclooxygenase-2 selective inhibitors (coxibs) significantly increased the incidence of UGIB, and various risk factors, including age, male gender, history of uncomplicated peptic ulcer disease, peptic ulcer bleeding, and H. pylori infection, contributed to the development of UGIB in coxibs users. The use of non-steroidal anti-inflammatory agents (NSAIDs) and anticoagulation agents associated with aging was considered to increase the development of UGIB by peptic ulcer. In our study, peptic ulcer disease including gastric ulcer and duodenal ulcer was the most common cause of UGIB, and 19.7% of patients took NSAIDs and anticoagulation agents. In our endoscopic findings, 2% of patients showing normal gastroscopic and colonoscopic findings were diagnosed with obscure gastrointestinal (GI) bleeding which is defined as persisting and/or recurrent GI bleeding of unidentified source after negative bidirectional endoscopic evaluation [15]. In case of obscure GI bleeding, we conducted capsule enteroscopy that is considered to be safe and effective for diagnosis of obscure GI bleeding [16].

Laine et al. [17] reported that patients who present with 6 to 12 h of the onset of UGIB symptoms show a significantly lower risk for transfusion owing to the higher hemoglobin level at presentation. On the other hand, patients delaying their presentation are more likely to have anemia, which increase transfusion requirement [17]. In the present study, we did not classify our study population according to time to presenation (rapid or delayed). Fortunately, Laine et al. [17] suggested that there is no apparent relationship between bleeding severity and time to presentation because there were no differences in mortality and requirement of hemostatic intervention with endoscopy, surgery or radiology among patients with rapid and delayed presentation.

When we planned this study, we considered a research involving only patients with non-variceal bleeding. However patients with symptoms and signs of UGIB who visited emergency department could not be distinguished clearly between variceal bleeding and non-variceal bleeding first. So, we analyzed patients with symptoms and signs of UGIB from either variceal or non-variceal source. However, when we analyzed data involving only patients with non-variceal bleeding, the result was similar with that of this study.

Several risk scoring systems have been produced to divide patients with UGIB into high- and low-risk categories. 2010 international consensus for UGIB recommended that the use of prognostic scales was needed for early stratification of patients with UGIB according to the degree of risk [5]. Early identification of patients who are likely to show the need of interventions or high mortality could improve efficiency of care. However, RS is not suitable for an early decision on need of urgent interventions in the management of patients with UGIB, since it requires endoscopic finding. For this reason, we compared 3 scoring systems (GBS, mGBS, and Pre-E RS) that could be applied immediately without endoscopic findings.

In this study, we used AUC to evaluate the 3 scoring systems for predicting the need of interventions and 30-day mortality in patients with UGIB. An arbitrary guideline proposed by Swets classifies scoring systems as non-informative (AUC = 0.5), less accurate (0.5 < AUC ≤ 0.7), moderately accurate (0.7 < AUC ≤ 0.9), highly accurate (0.9 < AUC < 1) and perfect tests (AUC = 1) [18]. AUC analyses in this study indicated that GBS and mGBS were moderately accurate (AUC = 0.727 and 0.733, respectively) and superior to Pre-E RS (AUC = 0.564) in predicting the need of interventions without endoscopic findings, consistent with prior studies. Cheng et al. [9] prospectively evaluated 199 patients with UGIB. They reported that mGBS (AUC = 0.85) performed as well as the GBS (AUC = 0.86, p = 0.81), and outperformed the Pre-E RS (AUC = 0.66, p < 0.0001) in predicting the need of interventions. Stanley et al. [19] reported that GBS was superior to Pre-E RS in predicting the need of interventions in a prospective study on 1555 subjects. On the other hand, the Pre-E RS is reported to be superior to GBS in predicting mortality of patients with UGIB [20]. Similarly, our study also showed that Pre-E RS outperformed GBS and mGBS in prediciting 30-day mortality. Many studies have reported that the RS was more closely associated with the probability of death than the chance of rebleeding [8, 21, 22]. Rockall et al. [23] reported that the RS was originally developed to predict the mortality rather than the prediction of rebleeding. The RS without endoscopy can identify the patients who are less likely to require intensive health care, and be used for endoscopic evaluation as outpatients, which improves the quality of patients’ care and allows substantial resource savings [24].

A recent retrospective study proposed that early gastroscopy should be performed for high-risk patients with GBS >12 to reduce mortality [25]. Our study indicated that the risk of significant bleeding to need interventions increased in patients with GBS > 9 or mGBS >9. Our results suggest that patients with GBS > 9 or mGBS >9 need hemostatic interventions and have to undergo early gastroscopy. Collectively, these findings advocate guidelines for early gastroscopy in high-risk patients with UGIB, which allow risk stratification and provision of appropriate endoscopic treatment [5]. Previous study showed that RS < 2 was indicative of low risk of mortality [10]. In our study, patients with Pre-RS > 4 showed high 30-day mortality, indicating that patients with Pre-RS > 4 should be managed with more immediate and intensive care to lower 30-day mortality. However, large prospective trials are needed for further evaluation of cutoff values.

This is the large-scale retrospective study to propose an appropriate scoring system without endoscopic findings for predicting the need of interventions and 30-day mortality in patients with UGIB. Our results may be useful to physicians in the emergency department for early decisions on interventions and prognosis. However, our study had some limitations. First, this was a single-center study from a teaching hospital, hence, our results cannot be generalized to all patients with the symptoms of UGIB. Second, there are generally some biases affecting the veracity of the retrospective study, such as selection bias, information bias and confounding, although all medical records used in this study were reviewed and entered into a database by one medical doctor for the quality of data. Finally, the rate of high-risk patients presenting with the symptoms of UGIB was high in this study. Therefore, further studies are required for accurate clinical applicability of results.

The GBS and mGBS are considered to be moderately accurate with good sensitivity and specificity in making an early decision about the need of interventions in patients with UGIB. Pre-E RS is considered to be highly accurate in early detection of patients at high risk for 30-day mortality without endoscopic findings. In addition, we suggested potential cutoff scores to predict the need of interventions in GBS and mGBS and 30-day mortality in Pre-E RS. These risk scoring systems predicting the need of intervention and/or 30-day mortality can contribute to saving lives and alleviating the severity of a patient’s condition. Especially, these scores may be useful in case of emergency where endoscopy are not available. In this sense, the verification of risk scoring systems through the substantial clinical data conducted in this study may be beneficial to both physicians and patients. However, the perfect score should be applicable in pre- and post-endoscopy. Further studies are needed to confirm the clinical applicability of results.