Sei sulla pagina 1di 10

ORIGINAL ARTICLE: Clinical Endoscopy

Risk stratification in acute upper GI bleeding: comparison of the


AIMS65 score with the Glasgow-Blatchford and Rockall
scoring systems
Marcus Robertson, MBBS(Hons),1 Avik Majumdar, MBBS(Hons),1 Ray Boyapati, MBBS(Hons),1
William Chung, MBBS,1 Tom Worland, MBBS,1 Ryma Terbah, MBBS,1 James Wei, MBBS,1
Steve Lontos, MBBS, MD,1 Peter Angus, MBBS, MD,1,2 Rhys Vaughan, MBBS, PhD1,2
Heidelberg, Victoria, Australia

Background and Aims: The American College of Gastroenterology recommends early risk stratification in pa-
tients presenting with upper GI bleeding (UGIB). The AIMS65 score is a risk stratification score previously vali-
dated to predict inpatient mortality. The aim of this study was to validate the AIMS65 score as a predictor of
inpatient mortality in patients with acute UGIB and to compare it with established pre- and postendoscopy
risk scores.
Methods: ICD-10 (International Classification of Diseases, Tenth Revision) codes identified patients presenting
with UGIB requiring endoscopy. All patients were risk stratified by using the AIMS65, Glasgow-Blatchford score
(GBS), pre-endoscopy Rockall, and full Rockall scores. The primary outcome was inpatient mortality. Secondary
outcomes were a composite endpoint of inpatient mortality, rebleeding, and endoscopic, radiologic, or surgical
intervention; blood transfusion requirement; intensive care unit (ICU) admission; rebleeding; and hospital length
of stay. The area under the receiver-operating characteristic curve (AUROC) was calculated for each score.
Results: Of the 424 study patients, 18 (4.2%) died and 69 (16%) achieved the composite endpoint. The AIMS65
score was superior to both the GBS (AUROC, 0.80 vs 0.76, P < .027) and the pre-endoscopy Rockall score (0.74,
P Z .001) and equivalent to the full Rockall score (0.78, P Z .18) in predicting inpatient mortality. The AIMS65
score was superior to all other scores in predicting the need for ICU admission and length of hospital stay.
AIMS65, GBS, and full Rockall scores were equivalent (AUROCs, 0.63 vs 0.62 vs 0.63, respectively) and superior
to pre-endoscopy Rockall (AUROC, 0.55) in predicting the composite endpoint. GBS was superior to all other
scores for predicting blood transfusion.
Conclusion: The AIMS65 score is a simple risk stratification score for UGIB with accuracy superior to that of GBS
and pre-endoscopy Rockall scores in predicting in-hospital mortality and the need for ICU admission. (Gastroint-
est Endosc 2016;83:1151-60.)

Upper GI bleeding (UGIB) is the most common GI emer- exists.5,6 The American College of Gastroenterology and inter-
gency and is associated with substantial morbidity and mortal- national consensus guidelines recommend early risk stratifi-
ity. Traditionally quoted as between 3% and 10%,1,2 more cation in patients presenting with UGIB to facilitate accurate
recent studies have shown improved inpatient mortality rates triage and assist in decisions such as timing of endoscopy,
of 2% to 2.5%,3,4 although significant geographic variation discharge planning, and level of care.7-9

Abbreviations: AUROC, area under the receiver-operating characteristic Current affiliations: Department of Gastroenterology and Liver Transplant
curve; ED, emergency department; GBS, Glasgow-Blatchford score; ICU, Unit, Austin Hospital (1), Department of Medicine, The University of
intensive care unit; UGIB, upper GI bleeding. Melbourne, Austin Health (2), Heidelberg, Victoria, Australia.
DISCLOSURE: All authors disclosed no financial relationships relevant Reprint requests: Marcus Robertson, MBBS(Hons), Austin Hospital Liver
to this publication. Transplant Unit, 145 Studley Road, Heidelberg, Victoria, Australia 3084.
Copyright ª 2016 by the American Society for Gastrointestinal Endoscopy If you would like to chat with an author of this article, you may contact
0016-5107/$36.00 Dr Robertson at marcus.Robertson@austin.org.au.
http://dx.doi.org/10.1016/j.gie.2015.10.021
Received April 28, 2015. Accepted October 11, 2015.

www.giejournal.org Volume 83, No. 6 : 2016 GASTROINTESTINAL ENDOSCOPY 1151


Risk stratification in acute upper GI bleeding Robertson et al

Multiple algorithms have been developed to predict METHODS


outcomes in UGIB. Currently, the most widely used risk
stratification tools are the Glasgow-Blatchford score10 Patient selection and data collection
(GBS) and the Rockall11 scores (Table 1). The GBS ICD-10 (International Classification of Diseases, Tenth
(range 0-23) was developed in 2000 from a United Revision) diagnosis codes were used to identify patients
Kingdom patient cohort to predict the need for presenting with UGIB requiring endoscopy to the Austin
hospital-based intervention or death after UGIB (a com- Hospital, a tertiary referral center, over a 42-month period
posite of inpatient mortality, in-hospital rebleeding, sub- from January 2010 to June 2013. Patients were excluded if
stantial decrease in hemoglobin concentration after the data required for calculation of risk stratification scores
admission, endoscopic or surgical intervention, and blood were incomplete or if medical records revealed an alterna-
transfusion). However, the GBS has limitations: weighting tive diagnosis. Each medical record was independently
makes calculation difficult and some of the criteria from examined by 2 reviewers (W.C., R.T.) to confirm the pres-
the patient’s medical history lack clear definitions. ence of UGIB. Discrepancies between data collected by the
The Rockall score was developed in 1996 from a large 2 reviewers were referred to a third reviewer (M.R.). The
population-based study designed to identify factors that study was approved by the Human Research Ethics Com-
predicted mortality in patients presenting with UGIB. It mittee at Austin Health.
is calculated by using pre-endoscopic and endoscopic vari- For each patient, the following data were collected
ables12 (Table 1). The Rockall score is limited by weighting, through manual chart review: age, sex, length of hospital
which leads to complexity in calculation and contains stay, intensive care unit (ICU) admission, discharge desti-
subjective variables such as the severity of systemic nation, symptoms on admission (hematemesis, coffee
diseases. Additionally, it requires endoscopic data for ground vomiting, melena, syncope, lethargy, confusion),
completion, rendering it impossible to use at the time of medical comorbidities, medication use (including oral or
presentation. An abbreviated pre-endoscopy or admission intravenous proton pump inhibitor, histamine-2 receptor
Rockall score has also been developed; however, there antagonist, antiplatelet agents, anticoagulant therapy,
has been debate about its accuracy and clinical nonsteroidal anti-inflammatory agents), vital signs
applicability.13,14 including Glasgow Coma Scale score, laboratory results (al-
Despite recommendations to incorporate risk stratifica- bumin level, international normalized ratio, urea, creati-
tion scores in UGIB, neither the GBS nor the Rockall nine, hemoglobin), endoscopy findings, and the need for
scores have been routinely adopted into clinical practice.15 endoscopic intervention, blood transfusion, repeat endos-
One contributing factor is undoubtedly the cumbersome copy, embolization, or surgery. For patients with multiple
nature of score calculation. laboratory tests or vital signs collected in the emergency
The AIMS65 score is a new risk stratification score department, the most abnormal values were used.
derived and validated to predict in-hospital mortality in Per previous studies involving the AIMS65 risk stratifica-
patients presenting with UGIB.16 The AIMS65 score tion score, a patient was considered to have a change in
assigns 1 point for each of the following: albumin mental state if the Glasgow Coma Scale score on presenta-
level <30 g/L, international normalized ratio >1.5, altered tion to the emergency department (ED) was less than 14 or
mental status, systolic blood pressure <90 mm Hg, and a designation of disoriented, lethargy, stupor, or coma was
age older than 65 years (Table 1). Compared with documented in the chart by a treating physician.
existing scores, the AIMS65 score has the advantages of All patients were risk stratified by using the AIMS65,
being unweighted, simple to remember, and easy to GBS, Rockall and pre-endoscopy Rockall scores. The pri-
calculate with laboratory values routinely obtainable in mary outcome was in-hospital mortality. Secondary out-
the emergency department. It is also does not rely on comes were a composite endpoint of inpatient mortality,
poorly defined medical history criteria. In a recent study, in-hospital rebleeding; endoscopic, radiologic, or surgical
the AIMS65 score was shown to be superior to the GBS intervention; blood transfusion requirement; ICU admis-
in predicting in-hospital mortality.17 We present the first sion; rebleeding; and length of hospital stay.
comparison of the AIMS65 score with the GBS, the
pre-endoscopy Rockall score, and the full Rockall risk
stratification score. Statistical analysis
The study objectives were to validate the AIMS65 score Data analysis was primarily performed by using SPSS sta-
as a predictor of in-hospital mortality in patients presenting tistical software Version 22 (IBM, Armonk, NY). The area un-
with acute UGIB in an Australian patient cohort and to der the receiver-operating characteristic curve (AUROC)
compare the AIMS65 score with established pre- was calculated for each scoring system and binomial
endoscopy (GBS and pre-endoscopy Rockall score) and outcome. For each scoring system, a cutoff point was calcu-
post-endoscopy (full Rockall score) risk stratification lated that maximized the sum of sensitivity and specificity in
scores. predicting the primary and secondary endpoints. Patients

1152 GASTROINTESTINAL ENDOSCOPY Volume 83, No. 6 : 2016 www.giejournal.org


Robertson et al Risk stratification in acute upper GI bleeding

TABLE 1. Comparison of the AIMS65, Glasgow-Blatchford, and Rockall risk stratifications scores

Glasgow-Blatchford Risk Score


AIMS65 Score Risk factor Score

Albumin <3.0 mg/dL 1 BUN, mg/dL


INR >1.5 1 18.2 to <22.4 2
Altered mental status 1 22.4 to <28.0 3
Systolic BP <90 mm Hg 1 28.0 to <70.0 4
Age >65 y 1 70.0 6
Maximum score 5 Hemoglobin, men g/dL
12.0 to <13.0 1
10.0 to <12.0 3
<10.0 6
Hemoglobin, women g/dL
10.0 to <12.0 1
<10.0 6
Systolic BP, mm Hg
100–109 1
90–99 2
<90 3
Other markers
Pulse 100 (bpm) 1
Presentation with melena 1
Presentation with syncope 2
Hepatic disease 2
Cardiac failure 2
Maximum score 23

Rockall Score Preadmission Rockall Score

Age, y Age, y
<60 0 <60 0
60–79 1 60–79 1
>80 2 >80 2
Shock Shock
No shock 0 No shock 0
Pulse >100 bpm, systolic BP >100 mm Hg 1 Pulse >100 bpm, systolic BP >100 mg Hg 1
Systolic BP <100 mm Hg 2 Systolic BP <100 mm Hg 2
Comorbidity Comorbidity
No major 0 No major 0
CCF, IHD, or major comorbidity 2 CCF, IHD, or major comorbidity 2
Renal failure, liver failure, metastatic cancer 3 Renal failure, liver failure, metastatic cancer 3
Diagnosis Maximum score 7
Mallory-Weiss tear or no lesion and no stigmata 0
All other diagnoses 1
GI malignancy 2
Evidence of bleeding
No stigmata or dark spot on ulcer 0
Blood in upper GI tract, adherent clot, visible or spurting vessel 2
Maximum score 11
GBS, Glasgow-Blatchford score; BUN, blood urea nitrogen; INR, international normalized ratio; BP, blood pressure; bpm, beats per minute; CCF, congestive cardiac failure;
IHD, ischemic heart disease; y, years.

www.giejournal.org Volume 83, No. 6 : 2016 GASTROINTESTINAL ENDOSCOPY 1153


Risk stratification in acute upper GI bleeding Robertson et al

TABLE 2. Patient characteristics* RESULTS


Characteristic Patient characteristics
Overall 424 (100) ICD-10 codes identified 481 patients with a primary
Male 278 (65.6) diagnosis of acute UGIB, 424 of whom were included in
Comorbidities the study. Fifty-seven patients were excluded from the
study: review of the medical records of 46 patients showed
Ischemic heart disease 114 (26.9)
they did not experience UGIB and 9 patients were
Congestive cardiac failure 53 (12.5)
excluded due to incomplete available data to calculate
Chronic obstructive airways disease 51 (12.0) risk scores. The median age was 71 years (range 15-93
Cerebrovascular disease 56 (13.2) years), and 66% of the patients were male. A total of 293
Diabetes mellitus 115 (27.1) patients (69%) were using antiplatelet or anticoagulant
Malignancy 64 (15.1) medications on admission (154 [36%] aspirin, 48 [11%]
Chronic renal impairment 88 (20.8) clopidogrel, and 90 [21%] warfarin or clexane); 209 pa-
Liver disease 71 (16.7) tients (49%) were taking a proton pump inhibitor. Patient
characteristics are shown in Table 2.
Medications
Aspirin 146 (34.4)
Aspirin and dipyridamole 8 (1.9) Mortality
Clopidogrel 48 (11.3)
The overall in-hospital mortality rate was 4.2% (18 pa-
tients) with a median age of 71 years (interquartile range,
Warfarin 81 (19.1)
58-81 years). Variceal hemorrhage was the cause of death
Therapeutic low molecular weight heparin 9 (2.1)
in 3 of these patients. All patients were treated with intra-
Dabigatran 1 (0.2) venous proton pump inhibitor therapy, and all patients
Nonsteroidal anti-inflammatory drugs 34 (8.0) with variceal hemorrhage received vasoactive therapy
Proton pump inhibitor 209 (49.3) with either terlipressin or octreotide. There was no differ-
Histamine-2 receptor antagonist 5 (1.2) ence in antiplatelet or anticoagulant use between survivors
Score components and nonsurvivors.
Age at admission, y 71 (58–81)
Mortality increased with increasing AIMS65 score
(Fig. 1), and no deaths were recorded in patients with an
Albumin 32 (28–36)
AIMS65 score of 0. The AIMS65 score was superior to
Coffee ground vomiting 67 (15.8)
both the GBS and pre-endoscopy Rockall score in predict-
International normalized ratio 1.2 (1.1–1.6) ing in-hospital mortality, with an AUROC of 0.80 versus
Mental state change 40 (9.4) 0.76 (P Z .03) and 0.74 (P Z.001), respectively (Fig. 2).
Systolic blood pressure, mm Hg 107 (95–120) There was no significant difference between the AIMS65
Hematemesis 106 (25) and the full Rockall scores in predicting mortality
Hemoglobin 93 (76–118) (AUROC, 0.80 vs 0.78; P Z .18) (Fig. 2, Table 3).
The cutoff threshold that maximized sensitivity and
Heart rate, bpm 91 (80–105)
specificity and hence defined low- and high-risk groups
Melena 323 (76.2)
for mortality was 3 for the AIMS65 score (sensitivity,
Urea 10.8 (6.3–18.4) 0.72; specificity, 0.77; total, 1.49). The mortality rate in
Syncope 47 (11.1) the high-risk group was 12.1% compared with 1.6% in
Liver disease 71 (16.7) the low-risk group (P < .001) (Table 4).
Congestive cardiac failure 53 (12.5)
bpm, Beats per minute. Composite clinical endpoint
*Proportions are presented as number (percentage) and continuous variables are
presented as median (interquartile range). Sixty-nine patients (16.3%) achieved at least 1 compo-
nent of the composite clinical endpoint. Eighteen (4.2%)
patients died, 41 (9.7%) experienced rebleeding, 46
were stratified into low- or high-risk groups by using the cut- (10.8%) required repeat endoscopy, 8 (1.9%) required
off threshold. The Fisher exact test was used to compare radiologic embolization, and 4 (0.9%) required surgery.
low- and high-risk groups. All P values were 2 sided, with a The AIMS65, GBS, and full Rockall score could all predict
threshold of .05 denoting statistical significance. Somer’s the composite clinical endpoint, with AUROCs of 0.63,
D statistic was used to compare the predictive accuracies 0.62, and 0.69, respectively (Table 3). When comparing
of continuous variables such as length of hospital stay. the AUROCs of the AIMS65 score, GBS, and the full
AUROCs were compared with the DeLong method by using Rockall score for the composite clinical endpoint, no
XLSTAT software Version 2013.5 (Addinsoft, New York, NY). significant differences between the scores was noted

1154 GASTROINTESTINAL ENDOSCOPY Volume 83, No. 6 : 2016 www.giejournal.org


Robertson et al Risk stratification in acute upper GI bleeding

9 100.0%

8 90.0%

80.0%
7

70.0%
6
60.0%

Length of Stay (Days)


5

Mortality (%)
50.0%
4
40.0%
3
30.0%

2
20.0%

1 10.0%

0 0.0%
0 1 2 3 4 5
Median Length of Stay (Days) 3 4 5 7 8 *
Mortality (%) 0.0% 1.7% 2.2% 7.2% 26.1% 100.0%

AIMS65 score

Figure 1. Inpatient mortality rate and length of stay by AIMS65 score. Inpatient mortality and length of stay increased with increasing AIMS65 score.
*Only 1 patient presented with an AIMS65 score of 5, and death occurred within the first day.

(AIMS65 vs GBS, P Z .83; AIMS65 vs full Rockall, P Z .46). 0.66; P < .001). The cutoff threshold for high risk and low
The pre-endoscopy Rockall score did not achieve statistical risk for blood transfusion was 2 for the AIMS65 score, 10
significance in predicting the composite endpoint for the GBS, 4 for the pre-endoscopy Rockall score, and
(Table 3). 5 for the full Rockall score (Table 4).

ICU admission Rebleeding


The majority of patients were managed on a general The overall inpatient rebleeding rate was 9.7%
ward, with 56 (13.2%) requiring ICU admission. The (41 patients). The full Rockall (AUROC 0.64), GBS
AIMS65 score was superior to all other scores in predict- (AUROC 0.64) and AIMS65 (AUROC 0.61) scores were
ing the need for ICU admission: AUROC, 0.74 versus 0.70 predictors of rebleeding (Table 3). When the AUROCs
for GBS (P Z .005); 0.62 for the pre-endoscopy Rockall for these scores were compared, no significant
score (P < .001); and 0.71 for the Rockall score differences were noted (AIMS65 vs GBS, P Z .54;
(P < .001). The cutoff value for low- and high-risk groups AIMS65 vs full Rockall, P Z .35). The pre-endoscopy
was 2 for AIMS65 score, 9 for the GBS, 4 for the pre- Rockall score did not achieve statistical significance
endoscopy Rockall score, and 6 for the full Rockall score (AUROC, 0.58). For inpatient rebleeding, the cutoff point
(Table 4). that maximized the sum of the sensitivity and the speci-
ficity was 2 for the AIMS65 score, 11 for the GBS, and
Blood transfusion requirement 6 for the full Rockall score (Table 4).
A total of 264 patients (62.3%) required red cell transfu-
sion, with a median transfusion requirement of 2 units (in- Length of hospital stay
terquartile range, 0-4). The GBS was the best predictor The median length of hospital stay was 5 days (inter-
with an AUROC of 0.90 compared with the AIMS65 score quartile range, 3-10) and increased with increasing
(AUROC, 0.72; P < .001), the full Rockall score (AUROC, AIMS65 score (Fig. 1). The AIMS65 score was superior to
0.68; P < .001); and pre-endoscopy Rockall score (AUROC, the GBS, pre-endoscopy Rockall, and full Rockall scores

www.giejournal.org Volume 83, No. 6 : 2016 GASTROINTESTINAL ENDOSCOPY 1155


Risk stratification in acute upper GI bleeding Robertson et al

1.0 previous studies and in the largest cohort of patients stud-


AIMS65 ied since the AIMS65 score was developed. We found the
Pre-Rockall AIMS65 score to be superior to other pre-endoscopy risk
0.8 GBS
scores in predicting in-hospital mortality and, impressively,
as accurate as the best validated post-endoscopy risk score.
0.6
In addition, the AIMS65 score had superior predictive abil-
Sensitivity

ity in terms of length of hospital stay and need for ICU


admission and equivalent accuracy in predicting a compos-
0.4 ite clinical endpoint of in-hospital mortality, rebleeding,
and endoscopic, radiologic, or surgical intervention
(Table 3).
0.2 The findings of this study are in agreement with those
of a retrospective analysis by Hyett et al17 that found the
AIMS65 score to be superior to the GBS for predicting
0.0
0.0 0.2 0.4 0.6 0.8 1.0 in-hospital mortality and equivalent for predicting length
A 1 - Specificity of hospital stay, time to endoscopy, rebleeding, and ICU
admission. The AIMS65 score, but not the GBS, was also
1.0 found to be an independent predictor of overall survival
AIMS65 in a Japanese cohort of patients presenting with acute
Full Rockall GI bleeding. This study included patients with both lower
0.8
and upper GI bleeding and concluded that the AIMS65
score was the best risk stratification score in predicting
Sensitivity

0.6 the prognosis of patients with acute GI bleeding.18 A


further study by Yaka et al19 found the AIMS65 score to
be equivalent to the GBS in predicting in-hospital mortal-
0.4 ity; however, the GBS had superior sensitivity in identi-
fying patients who were not likely to require endoscopic
0.2 intervention. As some studies have previously shown
that the Rockall score is superior to the GBS in predicting
patient mortality,20-22 we elected to compare AIMS65
0.0 score with both the GBS and Rockall risk stratification
0.0 0.2 0.4 0.6 0.8 1.0
B 1 - Specificity
scores.

Figure 2. Receiver-operating characteristic curves (AUROCs) for risk


stratification scores as predictors of inpatient mortality. A, The AIMS65 Use of risk stratification scores
score compared with pre-endoscopy risk scores (Glasgow-Blatchford, Multiple guidelines and consensus statements recom-
pre-endoscopy Rockall). B, The AIMS65 score compared with the full mend the use of risk stratification tools to guide UGIB
Rockall score. AUROCs for the AIMS65 score was superior to other pre-
management. Despite these recommendations, use of
endoscopy risk scores and equivalent to the postendoscopy full Rockall
score. risk scores remains poor, and no single score has been
routinely adopted in the clinical setting, leading to varia-
tion in practice.15 Risk stratification scores must be
accurate but also easy to remember and calculate if there
in predicting length of hospital stay with Somer’s D statis-
is to be widespread adoption of them in clinical practice.
tics of 0.28 (P < .001), 0.21 (P < .001), 0.20 (P < .001), and
The ideal risk stratification score for UGIB should be
0.24 (P < .001), respectively.
simple and easily applied at the bedside, should make
use of clinical data that are routinely available at hospital
DISCUSSION presentation, and should offer independent prognostic
information.23,24 The best validated UGIB risk scores
The AIMS65 risk stratification score was derived and (GBS and Rockall) have limitations including complexity
validated as a predictor of in-hospital mortality by Saltzman in calculation, weighting, and the need for endoscopic
et al in 2011.16 Our study is the first to compare the data for completion, all of which contribute to poor physi-
AIMS65 score with both the most commonly used and cian uptake. Modified scores have been developed in an
best validated pre-endoscopy (GBS and pre-endoscopy attempt to address these issues. For example, a modified
Rockall score) and post-endoscopy (full Rockall score) GBS was developed by Romagnuolo et al,25 which aimed
risk stratification scores for UGIB. It also validates the to eliminate subjective elements of the full score.
AIMS65 score in a different population from that of However, this modified risk stratification tool is also

1156 GASTROINTESTINAL ENDOSCOPY Volume 83, No. 6 : 2016 www.giejournal.org


Robertson et al Risk stratification in acute upper GI bleeding

TABLE 3. Comparison of AIMS65, GBS, and Rockall scores with significant clinical endpoints

Outcomes AIMS65 GBS Pre-endoscopy Rockall score Full Rockall score

Mortality 0.80, P < .001 0.76, P < .001 0.74, P Z .001 0.78, P < .000
AUC (95% CI) (0.69–0.91) (0.64–0.89) (0.65-0.83) (0.69-0.88)
Composite clinical endpoint 0.62, P Z .002 0.62, P Z .01 0.55* 0.63, P Z .01
AUC (95% CI) (0.55–0.70) (0.55–0.69) (0.48–0.62) (0.56–0.71)
ICU admission 0.74, P < .001 0.70, P < .001 0.62, P Z .004 0.71, P < .001
AUC (95% CI) (0.68–0.80) (0.64–0.77) (0.55–0.69) (0.64–0.77)
Transfusion 0.72, P < .001 0.90, P < .001 0.66, P < .001 0.68, P < .001
AUC (95% CI) (0.67–0.77) (0.87–0.93) (0.61–0.72) (0.63–0.73)
Rebleeding 0.61, P Z .03 0.64, P Z .04 0.58* 0.64, P Z .003
AUC (95% CI) (0.51–0.70) (0.55–0.73) (0.50–0.67) (0.56–0.73)
Length of hospital stay, Somer’s D 0.28, P < .001 0.21, P < .001 0.20. P < .001 0.24, P < .001
GBS, Glasgow-Blatchford score; AUC, area under the curve; CI, confidence interval; ICU, intensive care unit.
*Did not reach statistical significance.

TABLE 4. Comparison of AIMS65, GBS, and Rockall scores with cutoff values, high-risk and low-risk groups

Total
Sensitivity Specificity (sensitivity D High Low
Score Cutoff value (95% CI) (95% CI) specificity) risk, % risk, % P value

AIMS65 score
Mortality 3 0.72 (0.49-0.88) 0.77 (0.73-0.81) 1.49 12.1 1.6 <.001
Composite 2 0.74 (0.62-0.83) 0.45 (0.40-0.50) 1.19 20.8 10.1 .003
ICU 2 0.88 (0.78-0.95) 0.47 (0.42-0.52) 1.36 20.4 3.4 <.001
Transfusion 2 0.71 (0.65-0.76) 0.63 (0.66-0.70) 1.34 75.9 43.6 <.001
Rebleeding 2 0.76 (0.60-0.86) 0.44 (0.39-0.49) 1.20 12.7 5.6 .019
GBS
Mortality 15 0.56 (0.34-0.75) 0.87 (0.83-0.90) 1.43 15.5 2.5 <.001
Composite 12 0.56 (0.44-0.67) 0.620 (0.56-0.66) 1.18 22.6 11.7 .003
ICU 9 0.88 (0.76-0.94) 0.44 (0.39-0.49) 1.32 27.6 10.9 .001
Transfusion 10 0.76 (0.71-0.81) 0.83 (0.77-0.88) 1.59 53.3 46.7 <.001
Rebleeding 11 0.63 (0.48-0.76) 0.61 (0.56-0.66) 1.24 15.3 5.7 .001
Pre-endoscopy Rockall score
Mortality 5 0.83 (0.60-0.95) 0.64 (0.59-0.69) 1.47 9.3 1.1 <.001
Composite 3 0.85 (0.75-0.92) 0.23 (0.19-0.28) 1.08 17.8 14.5 .150
ICU 4 0.77 (0.64-0.86) 0.42 (0.37-0.47) 1.19 16.7 7.8 .008
Transfusion 4 0.71 (0.65-0.76) 0.55 (0.47-0.62) 1.26 72.1 47 <.001
Rebleeding 3 0.73 (0.58-0.84) 0.41 (0.36-0.46) 1.14 11.6 6.6 .095
Full Rockall score
Mortality 7 0.67 (0.44-0.84) 0.80 (0.76-0.83) 1.47 8.5 1.5 <.001
Composite 7 0.41 (0.30-0.53) 0.82 (0.77-0.85) 1.23 30.9 12.1 .003
ICU 6 0.68 (0.55-0.79) 0.66 (0.61-0.79) 1.33 23 6.9 <.001
Transfusion 5 0.67 (0.61-0.73) 0.61 (0.54-0.69) 1.28 74.4 46.8 <.001
Rebleeding 6 0.44 (0.30-0.59) 0.80 (0.76-0.84) 1.24 19.1 7.0 .019
GBS, Glasgow-Blatchford score; CI, confidence interval; ICU, intensive care unit.

weighted, has not been widely adopted, and only 1 other scores, such as the CHADS2 score, which estimates the
study has examined its efficacy.26 risk of stroke in patients with nonrheumatic atrial fibrilla-
This lack of uptake in the use of UGIB risk scores con- tion. CHADS2 was acclaimed for its simplicity, practicality,
trasts with widespread acceptance of other clinical risk and accuracy, which led to its incorporation within several

www.giejournal.org Volume 83, No. 6 : 2016 GASTROINTESTINAL ENDOSCOPY 1157


Risk stratification in acute upper GI bleeding Robertson et al

widely distributed guidelines.27,28 It has become a highly those with a higher risk of death or adverse out-
successful and widely adopted risk score that is used comes.16,17,40,41 Our data support an AIMS65 value of 2
throughout the world.29 Subsequent attempts to improve as the threshold to define a patient as being at high risk
this risk stratification with a weighted, points-based algo- of adverse outcomes; this facilitates early triage to ensure
rithm by using the Framingham Heart Data30 was appropriate resource allocation and location of medical
accurate and uniquely provided a 5-year risk of stroke, care.
but was not widely adopted due to its complexity.
Similarly, CURB-65 is a risk stratification score validated
Generalizability of study findings
for predicting mortality in community-acquired pneu-
To be most useful in the clinical setting, a UGIB risk
monia.31 Like CHADS2, CURB-65 can be simply and easily
stratification score should be generalizable to any patient
calculated by using 5 routinely measured parameters. An
with UGIB regardless of etiology. For this reason, we chose
alternative risk stratification score for pneumonia, the
study inclusion criteria that were broad and simple. This
Pneumonia Severity Index, has been demonstrated to
study included all adults with confirmed UGIB presenting
have superior discriminatory power for short-term mortal-
to a metropolitan tertiary referral center in Australia with
ity.32 The Pneumonia Severity Index, however, is more
a catchment area of approximately 1 million people and
complicated requiring 20 variables for calculation,
thus represents real-world data. The only exclusion crite-
including arterial blood gas sampling. Multiple
rion was the absence of data to calculate risk scores, result-
international guidelines specifically note CURB65’s
ing in only 2.1% of patients being excluded from analysis.
simplicity and use of easily available clinical and
The majority of patients presenting to the ED with acute
biochemical data.33 The British Thoracic Society
UGIB are elderly, have comorbidities, and are taking mul-
guidelines for the assessment of severity in pneumonia
tiple medications. Our study population includes patients
state that in clinical practice, the major limitation of the
with a high comorbidity burden, the majority of whom
Pneumonia Severity Index with regard to its widespread
were taking an antiplatelet or anticoagulant agent on pre-
and routine adoption in primary care, ED, or medical
sentation. In addition, our cohort contained patients with
admission units is the complexity involved in the
combination of both variceal and nonvariceal bleeding.
calculation of the score.34
Use of a risk stratification score such as the AIMS65
score could overcome many of these issues and facilitate Study limitations
standardization in practice. AIMS65 is nonweighted and This study has certain limitations that warrant discus-
simple to remember and calculate and uses parameters sion. First, it is a retrospective, single-center study, and
routinely available in the ED. The AIMS65 score, like thus all risk score calculations, data collection, and
CHADS2 and CURB65, is also acronym based, making it outcome ascertainments were based on existing clinical re-
easy for clinicians to remember. Importantly, this study cords. Errors were minimized by using a small number of
supports previous evidence that the AIMS65 score is a data collectors who entered information into a standard-
highly accurate risk stratification score in patients present- ized database. Each medical record was independently re-
ing with UGIB. viewed by 2 researchers, and discrepancies were referred
to a third reviewer.
Early identification of low- and high-risk Second, only patients who underwent endoscopy were
patients by using risk stratification scores included in this study, which meant that patients who
It is recommended that all patients with UGIB, with the refused endoscopy or were discharged directly from the
exception of very low risk patients, are treated with hospi- ED were not captured. Third, the retrospective nature of
tal admission and endoscopy within 24 hours.9,35 There is this study resulted in multiple health care workers docu-
no clear evidence of benefit from earlier endoscopy,36-39 menting the data parameters used to calculate risk stratifica-
although this requires clinical judgment on a case-by-case tion scores in a nonstandardized manner. This could
basis. Risk scores in UGIB should ideally be able to differ- potentially lead to errors in risk score calculation, especially
entiate low-risk patients for safe early discharge from high- because the risk scores incorporate subjective parameters
risk patients who would benefit from aggressive medical that require some degree of interpretation. Four of the 5
therapy. components of the AIMS65 score involve objective data.
In our analysis, the AIMS65 score was the most consis- The only subjective component is mental status; however,
tent UGIB score in reproducing cutoff thresholds to differ- in the ED, the Glasgow Coma Scale score was recorded for
entiate high- and low-risk groups for each of the clinical most patients, and this was the primary method used to
endpoints (Table 4). An AIMS65 cutoff value of 2 was determine impaired mental status. The definitions of the
established to differentiate high- and low-risk groups for medical history elements that are incorporated into the
all clinical endpoints excluding mortality, for which a GBS and Rockall scores were more difficult to standardize.
threshold of 3 maximized sensitivity and specificity. Previ- To enable consistency and in accordance with previous
ous studies have used an AIMS65 cutoff score of 2 to define studies involving the AIMS65 score, we used the definition

1158 GASTROINTESTINAL ENDOSCOPY Volume 83, No. 6 : 2016 www.giejournal.org


Robertson et al Risk stratification in acute upper GI bleeding

that Blatchford et al10 used in their original work: the 9. ASGE Standards of Practice Committee; Hwang JH, Fisher DA, Ben-
presence of those elements in the medical chart. Menachem T, et al. The role of endoscopy in the management of acute
non-variceal upper GI bleeding. Gastrointest Endosc 2012;75:1132-8.
Finally, different risk stratification scores for UGIB were 10. Blatchford O, Murray WR, Blatchford M. A risk score to predict need for
developed by using different endpoints, and therefore treatment for upper-gastrointestinal haemorrhage. Lancet 2000;356:
direct comparison should be performed cautiously. The 1318-21.
AIMS65 and Rockall scores predict in-hospital mortality 11. Rockall TA, Logan RF, Devlin HB, et al. Risk assessment after acute up-
and 30-day mortality, respectively. In comparison, the per gastrointestinal haemorrhage. Gut 1996;38:316-21.
12. Rockall TA, Logan RF, Devlin HB, et al. Selection of patients for early
GBS predicts the need for hospital-based intervention or discharge or outpatient care after acute upper gastrointestinal hae-
death and includes the need for blood transfusion as part morrhage. National Audit of Acute Upper Gastrointestinal Haemor-
of the composite endpoint. The appropriateness of using rhage. Lancet 1996;347:1138-40.
transfusion requirements as an endpoint for UGIB is now 13. Pang SH, Ching JY, Lau JY, et al. Comparing the Blatchford and
controversial given its defined role in resuscitation rather pre-endoscopic Rockall score in predicting the need for endoscopic
therapy in patients with upper GI hemorrhage. Gastrointest Endosc
than intervention, as indicated by recommendations from 2010;71:1134-40.
the American College of Gastroenterology and interna- 14. Chen IC, Hung MS, Chiu TF, et al. Risk scoring systems to predict need
tional consensus guidelines.7,42 The paradigm shift in the for clinical intervention for patients with nonvariceal upper gastroin-
use of blood transfusion in UGIB toward a more conserva- testinal tract bleeding. Am J Emerg Med 2007;25:774-9.
tive approach also raises questions as to whether blood 15. Liang PS, Saltzman JR. A national survey on the initial management of
upper gastrointestinal bleeding. J Clin Gastroenterol 2014;48:e93-8.
transfusions should be included as an endpoint.43 16. Saltzman JR, Tabak YP, Hyett BH, et al. A simple risk score accurately
predicts in-hospital mortality, length of stay, and cost in acute upper
GI bleeding. Gastrointest Endosc 2011;74:1215-24.
CONCLUSION 17. Hyett BH, Abougergi MS, Charpentier JP, et al. The AIMS65 score
compared with the Glasgow-Blatchford score in predicting outcomes
in upper GI bleeding. Gastrointest Endosc 2013;77:551-7.
The AIMS65 score is a simple risk stratification score for
18. Nakamura S, Matsumoto T, Sugimori H, et al. Emergency endoscopy
UGIB demonstrating accuracy superior to that of the GBS for acute gastrointestinal bleeding: prognostic value of endoscopic he-
and pre-endoscopy Rockall scores for predicting in- mostasis and the AIMS65 score in Japanese patients. Dig Endosc
hospital mortality and the need for intensive care unit 2014;26:369-76.
admission. The AIMS65 score is easy to remember and sim- 19. Yaka E, Yilmaz S, Dogan NO, et al. Comparison of the Glasgow-Blatch-
ford and AIMS65 scoring systems for risk stratification in upper gastro-
ple to calculate by using parameters routinely available in
intestinal bleeding in the emergency department. Acad Emerg Med
the ED and importantly does not require endoscopic vari- 2015;22:22-30.
ables for calculation. If these results are confirmed in a pro- 20. Dicu D, Pop F, Ionescu D, et al. Comparison of risk scoring systems in
spective trial, the AIMS65 score should become the new predicting clinical outcome at upper gastrointestinal bleeding patients
standard of care for risk stratification of UGIB. in an emergency unit. Am J Emerg Med 2013;31:94-9.
21. Wang CH, Chen YW, Young YR, et al. A prospective comparison of 3
scoring systems in upper gastrointestinal bleeding. Am J Emerg Med
2013;31:775-8.
REFERENCES 22. Yang HM, Jeon SW, Jung JT, et al. Comparison of scoring systems for
nonvariceal upper gastrointestinal bleeding: A multicenter prospective
1. Button LA, Roberts SE, Evans PA, et al. Hospitalized incidence and case cohort study. J Gastroenterol Hepatol. Epub 2015 Jul 25.
fatality for upper gastrointestinal bleeding from 1999 to 2007: a record 23. Stanley AJ. Update on risk scoring systems for patients with upper
linkage study. Aliment Pharmacol Ther 2011;33:64-76. gastrointestinal haemorrhage. World J Gastroenterol 2012;18:2739-44.
2. Gralnek IM, Barkun AN, Bardou M. Management of acute bleeding 24. Timoteo AT, Papoila AL, Lopes JP, et al. Is it possible to simplify risk strat-
from a peptic ulcer. N Engl J Med 2008;359:928-37. ification scores for patients with ST-segment elevation myocardial infarc-
3. Abougergi MS, Travis AC, Saltzman JR. The in-hospital mortality rate tion undergoing primary angioplasty? Rev Port Cardiol 2013;32:967-73.
for upper GI hemorrhage has decreased over 2 decades in the 25. Romagnuolo J, Barkun AN, Enns R, et al. Simple clinical predictors may
United States: a nationwide analysis. Gastrointest Endosc 2015;81: obviate urgent endoscopy in selected patients with nonvariceal upper
882-8. gastrointestinal tract bleeding. Archives of internal medicine 2007;167:
4. Laine L, Yang H, Chang S-C, et al. Trends for incidence of hospitaliza- 265-70.
tion and death due to GI complications in the United States from 2001 26. Cheng DW, Lu YW, Teller T, et al. A modified Glasgow Blatchford Score
to 2009. Am J Gastroenterol 2012;107:1190-5. improves risk stratification in upper gastrointestinal bleed: a prospective
5. Hearnshaw SA, Logan RFA, Lowe D, et al. Acute upper gastrointestinal comparison of scoring systems. Aliment Pharmacol Ther 2012;36:782-9.
bleeding in the UK: patient characteristics, diagnoses and outcomes in 27. Fuster V, Ryden LE, Cannom DS, et al. ACC/AHA/ESC 2006 guidelines for
the 2007 UK audit. Gut 2011;60:1327-35. the management of patients with atrial fibrillation: a report of the Amer-
6. Williams JG, Roberts SE, Ali MF, et al. Gastroenterology services in the ican College of Cardiology/American Heart Association Task Force on
UK. The burden of disease, and the organisation and delivery of ser- Practice Guidelines and the European Society of Cardiology Committee
vices for gastrointestinal and liver disorders: a review of the evidence. for Practice Guidelines (Writing Committee to Revise the 2001 Guide-
Gut 2007;56(suppl_1):1-113. lines for the Management of Patients With Atrial Fibrillation): developed
7. Laine L, Jensen DM. Management of patients with ulcer bleeding. Am J in collaboration with the European Heart Rhythm Association and the
Gastroenterol 2012;107:345-60; quiz 61. Heart Rhythm Society. Circulation 2006;114:e257-354.
8. Barkun AN, Bardou M, Kuipers EJ, et al. International consensus recom- 28. Singer DE, Albers GW, Dalen JE, et al. Antithrombotic therapy in atrial
mendations on the management of patients with nonvariceal upper fibrillation: American College of Chest Physicians Evidence-Based Clin-
gastrointestinal bleeding. Ann Intern Med 2010;152:101-13. ical Practice Guidelines (8th Edition). Chest 2008;133:546S-92S.

www.giejournal.org Volume 83, No. 6 : 2016 GASTROINTESTINAL ENDOSCOPY 1159


Risk stratification in acute upper GI bleeding Robertson et al

29. Gage BF, Waterman AD, Shannon W, et al. Validation of clinical classi- 36. Lee JG, Turnipseed S, Romano PS, et al. Endoscopy-based triage signif-
fication schemes for predicting stroke: results from the National Regis- icantly reduces hospitalization rates and costs of treating upper GI
try of Atrial Fibrillation. JAMA 2001;285:2864-70. bleeding: a randomized controlled trial. Gastrointest Endosc 1999;50:
30. Wang TJ, Massaro JM, Levy D, et al. A risk score for predicting stroke or 755-61.
death in individuals with new-onset atrial fibrillation in the community: 37. Bjorkman DJ, Zaman A, Fennerty MB, et al. Urgent vs. elective endos-
the Framingham Heart Study. JAMA 2003;290:1049-56. copy for acute non-variceal upper-GI bleeding: an effectiveness study.
31. Lim WS, van der Eerden MM, Laing R, et al. Defining community ac- Gastrointest Endosc 2004;60:1-8.
quired pneumonia severity on presentation to hospital: an interna- 38. Lee JG. What is the value of early endoscopy in upper gastrointestinal
tional derivation and validation study. Thorax 2003;58:377-82. bleeding? Nat Clin Pract Gastroenterol Hepatol 2006;3:534-5.
32. Aujesky D, Auble TE, Yealy DM, et al. Prospective comparison of three 39. Tsoi KK, Ma TK, Sung JJ. Endoscopy for upper gastrointestinal bleeding:
validated prediction rules for prognosis in community-acquired pneu- how urgent is it? Nat Rev Gastroenterol Hepatol 2009;6:463-9.
monia. Am J Med 2005;118:384-92. 40. Jung SH, Oh JH, Lee HY, et al. Is the AIMS65 score useful in predicting
33. Mandell LA, Wunderink RG, Anzueto A, et al. Infectious Diseases Soci- outcomes in peptic ulcer bleeding? World J Gastroenterol 2014;20:
ety of America/American Thoracic Society consensus guidelines on the 1846-51.
management of community-acquired pneumonia in adults. Clin Infect 41. Chandra S. AIMS65 score predicts short-term mortality but not the
Dis 2007;44(Suppl 2):S27-72. need for intervention in acute upper GI bleeding. Gastrointest Endosc
34. Lim WS, Baudouin SV, George RC, et al. BTS guidelines for the manage- 2013;78:381-2.
ment of community acquired pneumonia in adults: update 2009. Tho- 42. Dworzynski K, Pollit V, Kelsey A, et al. Management of acute upper
rax 2009;64(Suppl 3):iii1-55. gastrointestinal bleeding: summary of NICE guidance. BMJ 2012;344:
35. Lin HJ, Wang K, Perng CL, et al. Early or delayed endoscopy for patients e3412.
with peptic ulcer bleeding. A prospective randomized study. J Clin 43. Villanueva C, Colomo A, Bosch A, et al. Transfusion strategies for acute
Gastroenterol 1996;22:267-71. upper gastrointestinal bleeding. N Engl J Med 2013;368:11-21.

Read Articles in Press Online Today!


Visit www.giejournal.org

Gastrointestinal Endoscopy now posts in-press articles online in advance of their ap-
pearance in the print edition of the Journal. These articles are available at the Gastroin-
testinal Endoscopy Web site, www.giejournal.org, by clicking on the “Articles in Press”
link, as well as at Elsevier’s ScienceDirect Web site, www.sciencedirect.com. Articles in
Press represent the final edited text of articles that are accepted for publication but not
yet scheduled to appear in the print journal. They are considered officially published as
of the date of Web publication, which means readers can access the information and
authors can cite the research months prior to its availability in print. To cite Articles in
Press, include the journal title, year, and the article’s Digital Object Identifier (DOI),
located in the article footnote. Please visit Gastrointestinal Endoscopy online today to
read Articles in Press and stay current on the latest research in the field of gastrointestinal
endoscopy.

1160 GASTROINTESTINAL ENDOSCOPY Volume 83, No. 6 : 2016 www.giejournal.org

Potrebbero piacerti anche