a b s t r a c t

Background: The clinical severities of Upper gastrointestinal bleeding (UGIB) are of a wide variety, ranging from insignificant bleeds to fatal outcomes. Several scoring systems have been designed to identify UGIB high- and low-risk patients. The aim of our study was to compare the Glasgow-Blatchford score (GBS) with the preendoscopic Rockall score (PRS) and the complete Rockall score (CRS) in their utilities in predicting clinical outcomes in patients with UGIB. Methods: We designed a prospective study to compare the performance of the GBS, PRS, and CRS in predicting primary and secondary outcomes in UGIB patients. The primary outcome included the need for blood transfusion, endoscopic therapy, or surgical intervention and was labeled as high risk. The secondary outcomes included rebleeding and 30-day mortality. The Area Under the Receiver Operating Characteristic Curve , sensitivity, specificity, and positive and negative predictive values for each system were analyzed. A total of 303 consecutive patients admitted with UGIB during a 1-year period were enrolled.

Results: For prediction of high-risk group, AUC was obtained for GBS (0.808), PRS (0.604), and CRS (0.767). For prediction of rebleeding, AUC was obtained for GBS (0.674), PRS (0.602), and CRS (0.621). For prediction of mortality, AUC was obtained for GBS (0.513), PRS (0.703), and CRS (0.620).

Conclusions: In detecting high-risk patients with acute UGIB, GBS may be a useful risk stratification tool. However, none of the 3 score systems has good performance in predicting rebleeding and 30-day mortality because of low AUCs.

(C) 2013

1. Introduction

Upper gastrointestinal bleeding (UGIB) remains the most common medical emergency managed by gastroenterologists, with an inci- dence of 50 to 172 per 100,000 of the population each year [1,2]. The

? Funding: This research was supported by Chang Gung Memorial Hospital, Chiayi, under contract no. CMRPG6A0171.

?? “Conflict of Interest” statement: There is no conflict of interest in this study.

? Author contributions: ICC, CHW, YWC, CJY, and YRY conceived the study and

designed the method. ICC, CHW, YWC, CJY, and YRY supervised the conduct of the data collection. CHW, YWC, and YRY undertook recruitment of participating centers and patients and managed the data, including quality control. ICC and CJY provided statistical advice on study design and analyzed the data; CHW chaired the data oversight committee. ICC, CHW, and YWC drafted the manuscript; and all authors contributed substantially to its revision. ICC takes responsibility for the paper as a whole.

?? Cheng-Hsien Wang and Yu-Wei Chen have the same contributions to the

manuscript and are co-first authors.

* Corresponding author. Department of Emergency Medicine, Chang Gung Memorial Hospital, Puzih City, Chiayi County 613, Taiwan.

E-mail address: [email protected] (I-C. Chen).

clinical severities of UGIB are of a wide variety, ranging from insignificant bleeds to fatal outcomes [3]. The indications for admission to the hospital are based on the need for blood transfusion, endoscopic intervention to control bleeding, or surgical intervention to control bleeding. However, most UGIB patients do not need blood transfusion or emergent endoscopic intervention [1]. About 23% of patients with UGIB need blood transfusion, and 14% require endoscopic or surgical intervention [4]. Patients at high risk of rebleeding or mortality are also of great concern because clinical treatment aims to prevent patients from dying and complications. Upper gastrointestinal bleeding causes about 2% to 15% mortality and 10% to 30% rebleeding [5-9]. Accurate identification of high-risk patients can help physicians decide about hospital admission or discharge, the level of assistance (early endoscopy or not), and the type of treatment (medical, endoscopic, or surgical intervention).

Several scoring systems have been designed to identify these high- and low-risk patients. The complete Rockall score (CRS) is one the most well-documented and validated system [8,10-15]. However, CRS is based on endoscopic findings; and it may not be so helpful in decision making when urgent endoscopy is not available. The

0735-6757/$ – see front matter (C) 2013 http://dx.doi.org/10.1016/j.ajem.2013.01.007

776 C-H. Wang et al. / American Journal of Emergency Medicine 31 (2013) 775–778

Table 1

Glasgow-Blatchford score

Admission risk marker Score component value Blood urea nitrogen (mg/dL)

>=18.2 b22.4 mg/dL 2

>=22.4 b28 mg/dL 3

>=28 b70 mg/dL 4

>=70 mg/dL 6

Hemoglobin for men (g/dL)

>=12 b13 g/dL 1

>=10 b12 g/dL 3

b10 g/dL 6

Hemoglobin for women (g/dL)

>=10 b12 g/dL 1

b10 g/dL 6

Systolic blood pressure

>=100 b109 mm Hg 1

>=90 b99 mm Hg 2

b90 mm Hg 3

Other markers

Pulse >=100/min 1

Presentation with melena 1

Presentation with syncope 2

Hepatic disease 2

Heart failure 2

*Range of scores = 0 to 23 (maximum score = 23); high risk N 0.

preendoscopic Rockall score (PRS) and the Glasgow-Blatchford score require only clinical and laboratory data, so they can be applied immediately without need for urgent endoscopy [16-20].

The aim of our study was to compare GBS with PRS and CRS in predicting clinical outcomes in patients with UGIB.

1. Methods
   1. Study design and population

This was a prospective cohort study. All nontrauma adult patients presenting UGIB admitted to the hospital via the emergency department (ED) from March 1, 2011, to March 1, 2012, were evaluated. Patients who refused endoscopic examination or without study informed consents were excluded.

This study was conducted at one university-affiliated teaching hospital that was located in the center of Taiwan with approximately 70,000 visits to the ED every year.

This study had been approved by our hospitals’ Institute Review- ing Board with serial number 99-2160B, and informed consents were obtained from all enrolled patients.

Survey content and administration

All nontrauma adult (ie, N 18 years old) patients with UGIB who visited our ED were evaluated. The diagnosis of UGIB was based on patients’ presentations, including coffee ground vomits, hematemesis, melena, and blood on nasogastric aspirate. All UGIB patients who received endoscopy were enrolled prospectively into our study after informed consents were obtained. Variables including age, sex, chief concerns, presenting illness vital signs, laboratory data, presence of comorbid medical conditions, findings of endoscopy, number of unit of blood transfusion, types of treatment of UGIB, and time from ED admission to endoscopy were recorded. Rebleeding was defined as any of the following: (1) repeated endoscopy within 3 days, (2) continuous blood transfusion for more than 3 days, and (3) surgical intervention to control bleeding within 3 days. The GBS, PRS, and CRS were calculated based on the criteria documented in the original articles [11,20] (Tables 1 and 2) and recorded for all enrolled patients. All clinical management decisions were left to the discretion of the main responsible attending physician. The standard management for all patients with nonvariceal UGIB in our ED was administration of an intravenous proton pump inhibitor before endoscopy. The choice of intermittent or continuous infusion was left to the discretion of the main responsible physicians. Intravenous terlipressin 2 mg bolus and 1 mg q6h was started on all patients with suspected variceal bleeding. Blood transfusion was indicated for UGIB patients with hemoglobin less than 10 g/dL or with signs of hemodynamic instability despite fluid resuscitation. The decision of transfusion was made by the main

responsible physicians.

The primary outcome was high-risk patient, which was defined as patients needing blood transfusion, therapeutic endoscopy to control bleeding, or surgical intervention to control bleeding. The secondary outcome was 30-day mortality and rebleeding.

Data analysis

Statistical analysis was carried out using the software SPSS 17.0 (SPSS Inc, Chicago, IL). Mann-Whitney U test, a Student t test, and ?² or the Fisher exact test were used to compare the statistically significant differences between high-risk and low-risk patients. The difference was considered significant if the P value was b.05. Sensitivity and specificity in primary and secondary outcomes were calculated for GBS, PRS, and CRS with confidence interval (CI). The cutoff values in detecting primary and secondary outcomes were GBS greater than 0, PRS greater than 0, and CRS greater than 2 [3,21]. Receiver operating characteristic curves were calculated to identify GBS, PRS, and CRS cutoff values for predicting primary and secondary outcomes.

Table 2

Rockall score

Variable	Score
	0	1	2	3
Age	b60 y	60-79	>=80 y
Shock		HR N 100	SBP b 100 mm Hg
Comorbidity			IHD, CHF, any	Renal failure, Liver failure,
			major comorbidity	metastatic malignancy
Diagnosis	Mallory-Weiss tear or	Peptic ulcer disease,	Malignancy of UGI tract
	no lesion observed	erosive esophagitis
Stigmata of recent	Clean-based ulcer,		Blood in UGI tract, clot,
hemorrhage	flat pigmented spot		visible vessel, bleeding

HR, heart rate; SBP, systolic blood pressure; IHD, ischemic heart disease; CHF, congestive heart failure.

*The PRS is calculated without endoscopic finding for each case based on points assigned for 3 clinical variables: patient age at presentation, shock status based on initial heart rate and systolic pressure, and presence of comorbid disease.

*The CRS is calculated for each case upon points assigned for each of the 3 aforementioned clinical variables plus 2 endoscopic variables: the endoscopic diagnosis and stigmata of recent hemorrhage based on the initial endoscopic examination.

*Patients with PRS greater than 0 and patients with CRS greater than 2 are considered to be at high risk for developing adverse outcomes (recurrent bleeding, death).

C-H. Wang et al. / American Journal of Emergency Medicine 31 (2013) 775–778 777

Table 3

Comparison of characteristics and outcomes in high-risk patients and non-high-risk patients

Table 5

Sensitivity and specificity of GBS greater than 0, PRS greater than 0, and CRS greater than 2 in detecting rebleeding

Non-high-risk patients	High-risk patients	P value
(n = 88) (%)	(n = 215) (%)			Sensitivity	Specificity	Positive	Negative

Cirrhosis	15 (17.0)	94 (43.7)	b.0001
UGIB history	53 (60.2)	108(50.2)	.13
Cancer	6 (6.8)	58 (27.0)	b.0001
Aspirin use	11 (12.5)	21 (9.8)	.54
clopidogrel use	4 (4.5)	2(0.9)	.06
NSAID use	5 (5.7)	19 (8.8)	.48
Anticoagulant use	10 (11.4)	13 (6.0)	.15

NSAID, nonsteroidal anti-inflammatory drug.

1. Results

During the study period, 324 patients presented with UGIB in the ED, 19 were managed as outpatient, and 305 were admitted to the hospital. Excluding 2 patients who refused endoscopy and informed consents, a total of 303 patients were enrolled in this study. All enrolled patients underwent endoscopy for UGIB. Of these patients, 22 (7.3%) died after a 30-day follow-up; and 40 (13.2%) rebled. The mean age of the enrolled patients was 64.3 +- 15.6 years. Melena (82.1%), Epigastric pain (46.2%), syncope (45.2%), hematemesis (29.7%), poor appetite (25.7%), blood on nasogastric aspirate (24.8%), and coffee ground vomitus (23.4%) were common symptoms. However, dyspnea (16.5%), Chest tightness (8.3%), and cold sweating (5.6%) were less common. esophageal varices or gastric varices were found in 58 (19.1%) patients. One hundred sixty-nine (55.8%) patients needed blood transfusion during ED or hospital stay. One hundred thirty-three (43.9%) patients needed endoscopic treatment to control bleeding, and 4 (1.3%) patients needed surgical treatment to control bleeding. Two hundred fifteen patients (71.0%) comprised the high- risk group (needed blood transfusion, endoscopic treatment, and surgical treatment). The mean time elapsed from ED triage to endoscopy was 12.5 +- 4.2 hours. All patients underwent endoscopy within 24 hours.

The demographic data are shown in Table 3. The patients were

categorized as either high risk or non-high risk. No specific differences were found in age and sex between the 2 groups. Hematemesis, syncope, and dyspnea were more common in the high-risk group, whereas epigastric pain was less common. There were significant differences in chronic hepatitis, cirrhosis, and cancer, whereas none was found in other underlying diseases, such as coronary artery disease, diabetes mellitus, chronic obstructive pulmonary disease, or previous UGIB history between the 2 groups.

The sensitivity and specificity of GBS greater than 0, PRS greater than 0, and CRS greater than 2 in detecting the high-risk group, rebleeding, and mortality were demonstrated in Tables 4, 5, and 6, respectively.

(AUC) was obtained for GBS (0.808; 95% CI, 0.756-0.86), PRS (0.604;

Age (mean +- SD) Male	64.6 +- 16.5 62 (70.5)	64.2 +- 15.2 152(70.7)	.84 1	(95% CI)	(95% CI)	predictive value (95% CI)	predictive value (95% CI)
Hematemesis	16 (18.2)	74 (34.4)	.005	GBS N 0 100 (91.1-100)	1.10 (0.25-3.30)	13.3 (9.71-17.7)	100 (30.5-100)
Syncope	31 (35.2)	106 (49.3)	.03	PRS N 0 100 (84.4-100)	3.04 (1.33-5.91)	13.6 (9.87-18.0)	100 (62.9-100)
Epigastric pain	50 (56.8)	90 (41.9)	.022	CRS N 2 100 (84.4-100)	6.46 (3.81-10.15)	14.0 (10.2-18.6)	100 (80.3-100)
Dyspnea	8 (9.1)	42 (19.5)	.027
Chronic	22 (25.0)	82 (38.1)	.033
hepatitis For prediction of the high-risk group, the area under the curve

95% CI, 0.536-0.673), and CRS (0.767; 95% CI, 0.713-0.821). The GBS

was superior to PRS (P b .00001) and similar to CRS (P = .353) in prediction of the high-risk group.

For prediction of rebleeding, the AUC was obtained for GBS (0.674; 95% CI, 0.591-0.757), PRS (0.602; 95% CI, 0.517-0.688), and CRS (0.621; 95% CI, 0.537-0.706). The GBS was similar to CRS (P = .45) and

PRS (P = .305) in prediction of rebleeding.

For prediction of mortality, the AUC was obtained for GBS (0.513; 95% CI, 0.388-0.618), PRS (0.703; 95% CI, 0.597-0.809), and CRS (0.620;

95% CI, 0.514-0.725). In contrast, the PRS was superior to GBS (P = .04) but similar to CRS (P = .366) in prediction of 30-day mortality.

1. Discussion
   1. GBS, PRS, and CRS in identification of the high-risk group of UGIB

In our study, the sensitivity of GBS, PRS, and CRS in detecting the high-risk group was 100%, 98.6%, and 96.8%, respectively. The AUC of GBS (0.808; 95% CI, 0.756-0.86) was greater than those of PRS (0.604; 95% CI, 0.536-0.673) and CRS (0.767; 95% CI, 0.713-0.821). These

results support previous findings from the United Kingdom [20-22], the Netherlands [23], and Denmark [24]. In a multi-institutional study, the GBS preponderated the PRS (AUC 0.944 vs 0.756, P b .00005) and CRS (AUC 0.935 vs 0.792, P b .00005) in predicting the need for transfusion [4]. In predicting endoscopic or surgical intervention, the GBS preponderated the PRS (AUC 0.858 vs 0.705, P b .00005) [4]. In one recent study, Cheng et al [25] confirmed that the GBS was more effective than the full RS and clinical RS in predicting the need for blood transfusions, as well as endoscopic or surgical interventions. As in the UK studies, we found that the GBS had an AUC higher than 0.8 in predicting the need for clinical intervention and a high sensitivity ranging from 99% to 100% [20-22].

GBS, PRS, and CRS in identification of 30-day mortality and rebleeding of UGIB

The results of our study did not find any difference between the performance of the GBS, PRS and CRS in prediction of rebleeding. The AUC of these 3 scoring systems was between 0.602 and 0.67, implying that the clinical application of these systems in predicting rebleeding was limited. As shown in Table 3, GBS, PRS, and CRS had a high sensitivity in predicting rebleeding; but because of the very low specificity of these systems, routine use in predicting rebleeding should be considered.

Table 4

Sensitivity and specificity of GBS greater than 0, PRS greater than 0, and CRS greater than 2 in detecting high-risk patients

Table 6

Sensitivity and specificity of GBS greater than 0, PRS greater than 0, and CRS greater than 2N0 in detecting 30-day mortality

	Sensitivity	Specificity	Positive	Negative			Sensitivity	Specificity	Positive predictive	Negative
	(95% CI)	(95% CI)	predictive value (95% CI)	predictive value (95% CI)			(95% CI)	(95% CI)	value (95% CI)	predictive value (95% CI)
GBS N 0	100 (98.3-100)	3.41 (0.75-9.650	71.67 (65.4-76.0)	100 (30.5-100)		GBS N 0	100 (84.4-100)	1.07 (0.23-3.09)	7.33 (4.65-10.9)	100 (30.5-100)
PRS N 0	98.6 (96.0-99.7)	5.68 (1.89-12.8)	72.0 (65.6-77.0)	62.5 (24.7-91.0)		PRS N 0	100 (84.4-100)	2.85 (1.24-5.53)	7.46 (4.73-11.1)	100 (62.9-100)
CRS N 2	96.7 (93.4-98.7)	11.4 (5.6-19.9)	72.7 (67.2-77.8)	58.8 (33.0-81.5)		CRS N 2	100 (84.4-100)	6.05 (3.57-9.51)	7.69 (4.88-11.4)	100 (80.3-100)

778 C-H. Wang et al. / American Journal of Emergency Medicine 31 (2013) 775–778

In agreement with 2 prospective studies in Italy and Denmark [12,24], our data revealed similar performance of the CRS to detect rebleeding. These authors found similarly low values of AUC. In a large validation study of the CRS from the Netherlands, low AUC (0.61) was also noticed [8].

In our study, all of the 3 scoring systems (GBS, PRS, and CRS) performed poorly in predicting 30-day mortality because of low AUCs, specificities, and positive predictive values. The application of GBS, PRS, and CRS in predicting 30-day mortality should be made with caution.

1. Limitation

There were several limitations of this study. First, our study included only a single institution. However, many critical cases of UGIB from neighboring cities and counties were sent to our ED, in which 24-hour emergent endoscopy and 24-hour surgeon were available. Moreover, most patients discharged from our institution would return to their gastroenterologists or our ED if the UGIB episode recurred; thus, more accurate records could be documented. Second, our study did not enroll UGIB patients managed as outpatient. During the study period, a total of 324 patients with UGIB presented in our ED; and only 19 patients were managed as outpatient. Another limitation was that we did not calculate the scores for separate subgroups of patients with different sources of bleeding [3,8,26] because of the relatively small number of enrolled patients. Our reports included all patients with UGIB and calculated all 3 scores only in patients having endoscopy confirming the UGIB diagnosis, similar to Camellini et al [12]. Thus, the CRS in our study may be more accurate because all of our enrolled patients received endoscopy.

1. Conclusion

In detecting high-risk patients (need for blood transfusion, endoscopic or surgical intervention) with acute UGIB, GBS may be a useful risk stratification tool. It has a higher sensitivity than PRS and CRS in identifying high-risk patients. However, none of the 3 scoring systems has good performance in predicting rebleeding and 30-day mortality because of low specificity and low positive predictive value.

References

1. Atkinson RJ, Hurlstone DP. Usefulness of prognostic indices in upper gastrointes- tinal bleeding. Best Pract Res Clin Gastroenterol 2008;22(2):233-42, http:

//dx.doi.org/10.1016/j.bpg.2007.11.004 [Epub 2008/03/19].

1. Longstreth GF. Epidemiology of hospitalization for acute Upper gastrointestinal hemorrhage: a population-based study. Am J Gastroenterol 1995;90(2):206-10 [Epub 1995/02/01].
2. Rockall TA, Logan RF, Devlin HB, Northfield TC. Selection of patients for early discharge or outpatient care after acute upper gastrointestinal haemorrhage. National Audit of Acute Upper Gastrointestinal Haemorrhage. Lancet 1996;347(9009):1138-40.
3. Stanley AJ, Dalton HR, Glasgow-Blatchford O, Ashley D, Mowat C, Cahill A, et al. Multicentre comparison of the Glasgow-Blatchford and Rockall Scores in the prediction of clinical end-points after upper gastrointestinal haemorrhage. Aliment Pharmacol Ther 2011;34(4):470-5, http://dx.doi.org/10.1111/j.1365- 2036.2011.04747.x [Epub 2011/06/29].
4. Laine L, Peterson WL. Bleeding peptic ulcer. N Engl J Med 1994;331(11):717-27, http://dx.doi.org/10.1056/NEJM199409153311107 [Epub 1994/09/15].
5. Katschinski B, Logan R, Davies J, Faulkner G, Pearson J, Langman M. Prognostic factors in upper gastrointestinal bleeding. Dig Dis Sci 1994;39(4):706-12 [Epub 1994/04/01].
6. Terdiman JP. Update on upper gastrointestinal bleeding. Basing Treatment decisions on patients’ risk level. Postgraduate medicine. 1998;103(6):43-7, 51-2, 8-9 passim. Epub 1998/06/20.
7. Vreeburg EM, Terwee CB, Snel P, Rauws EA, Bartelsman JF, Meulen JH, et al. Validation of the Rockall Risk scoring system in upper gastrointestinal bleeding. Gut 1999;44(3):331-5 [Epub 1999/02/23. PubMed PMID: 10026316; PubMed Central PMCID: PMC1727413].
8. van Leerdam ME, Vreeburg EM, Rauws EA, Geraedts AA, Tijssen JG, Reitsma JB, et al. Acute upper GI bleeding: did anything change? Time trend analysis of incidence and outcome of acute upper GI bleeding between 1993/1994 and 2000. Am J Gastroenterol 2003;98(7):1494-9, http://dx.doi.org/10.1111/j.1572-0241. 2003.07517.x [Epub 2003/07/23].
9. Rockall TA, Logan RF, Devlin HB, Northfield TC. Incidence of and mortality from acute upper gastrointestinal haemorrhage in the United Kingdom. Steering Committee and members of the National Audit of Acute Upper Gastrointestinal Haemorrhage. BMJ 1995;311(6999):222-6 [Epub 1995/07/22. PubMed PMID: 7627034; PubMed Central PMCID: PMC2550278].
10. Rockall TA, Logan RF, Devlin HB, Northfield TC. Risk assessment after acute upper gastrointestinal haemorrhage. Gut 1996;38(3):316-21 [Epub 1996/03/01. PubMed PMID: 8675081; PubMed Central PMCID: PMC1383057].
11. Camellini L, Merighi A, Pagnini C, Azzolini F, Guazzetti S, Scarcelli A, et al. Comparison of three different risk scoring systems in non-variceal upper gastrointestinal bleeding. Dig Liver Dis 2004;36(4):271-7, http://dx.doi.org/10. 1016/j.dld.2003.10.017 [Epub 2004/04/30].
12. Enns RA, Gagnon YM, Barkun AN, Armstrong D, Gregor JC, Fedorak RN. Validation of the Rockall scoring system for outcomes from non-variceal upper gastrointes- tinal bleeding in a Canadian setting. World J Gastroenterol 2006;12(48):7779-85 [Epub 2007/01/05].
13. Church NI, Dallal HJ, Masson J, Mowat NA, Johnston DA, Radin E, et al. Validity of the Rockall scoring system after endoscopic therapy for bleeding peptic ulcer: a prospective cohort study. Gastrointest Endosc 2006;63(4):606-12, http:

//dx.doi.org/10.1016/j.gie.2005.06.042 [Epub 2006/03/28].

1. Sanders DS, Carter MJ, Goodchap RJ, Cross SS, Gleeson DC, Lobo AJ. Prospective validation of the Rockall risk scoring system for upper GI hemorrhage in subgroups of patients with varices and peptic ulcers. Am J Gastroenterol 2002;97(3):630-5.
2. Gralnek IM, Dulai GS. Incremental value of Upper endoscopy for triage of patients with acute non-variceal upper-GI hemorrhage. Gastrointest Endosc 2004;60(1): 9-14 [Epub 2004/07/02].
3. Tham TC, James C, Kelly M. Predicting outcome of acute non-variceal upper gastrointestinal haemorrhage without endoscopy using the clinical Rockall score. Postgrad Med J 2006;82(973):757-9, http://dx.doi.org/10.1136/pmj.2006.048462 [Epub 2006/11/14. PubMed PMID: 17099097; PubMed Central PMCID: PMC2660506].
4. Chen IC, Hung MS, Chiu TF, Chen JC, Hsiao CT. Risk scoring systems to predict need for clinical intervention for patients with nonvariceal upper gastrointestinal tract bleeding. Am J Emerg Med 2007;25(7):774-9, http://dx.doi.org/10.1016/ j.ajem.2006.12.024 [Epub 2007/09/18].
5. Pang SH, Ching JY, Lau JY, Sung JJ, Graham DY, Chan FK. Comparing the Glasgow- Blatchford and pre-endoscopic Rockall score in predicting the need for endoscopic therapy in patients with upper GI hemorrhage. Gastrointest Endosc 2010;71(7):1134-40, http://dx.doi.org/10.1016/j.gie.2010.01.028 [Epub

2010/07/06].

1. Glasgow-Blatchford O, Murray WR, Glasgow-Blatchford M. A risk score to predict need for treatment for upper-gastrointestinal haemorrhage. Lancet 2000;356(9238):1318-21, http://dx.doi.org/10.1016/S0140-6736(00)02816-6 [Epub 2000/11/10].
2. Stanley AJ, Ashley D, Dalton HR, Mowat C, Gaya DR, Thompson E, et al. outpatient management of patients with low-risk upper-gastrointestinal haemorrhage: multicentre validation and prospective evaluation. Lancet 2009;373(9657):42-7, http://dx.doi.org/10.1016/S0140-6736(08)61769-9 [Epub 2008/12/19].
3. Srirajaskanthan R, Conn R, Bulwer C, Irving P. The Glasgow-Blatchford scoring system enables accurate risk stratification of patients with upper gastrointestinal haemorrhage. Int J Clin Pract 2010;64(7):868-74, http://dx.doi.org/10.1111/j. 1742-1241.2009.02267.x [Epub 2010/03/27].
4. Schiefer M, Aquarius M, Leffers P, Stassen P, van Deursen C, Oostenbrug L, et al. Predictive validity of the Glasgow-Blatchford Bleeding Score in an unselected emergency department population in continental Europe. Eur J Gastroenterol Hepatol 2012;24(4):382-7, http://dx.doi.org/10.1097/MEG.0b013e3283505965 [Epub 2012/01/10].
5. Laursen SB, Hansen JM, Schaffalitzky de Muckadell OB. The Glasgow-Blatchford score is the most accurate assessment of patients with upper gastrointestinal hemorrhage. Clin Gastroenterol Hepatol 2012;10(10):1130-1135e1, http:

//dx.doi.org/10.1016/j.cgh.2012.06.022 [Epub 2012/07/18].

1. Cheng DW, Lu YW, Teller T, Sekhon HK, Wu BU. A modified Glasgow-Blatchford Score improves risk stratification in upper gastrointestinal bleed: a prospective comparison of scoring systems. Aliment Pharmacol Ther 2012;36(8):782-9, http:

//dx.doi.org/10.1111/apt.12029 [Epub 2012/08/30].

1. 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(3):219-22.