Original research

Prediction of esophageal varices and variceal

J Investig Med: first published as 10.1136/jim-2015-000047 on 12 February 2016. Downloaded from http://jim.bmj.com/ on 28 April 2019 by guest. Protected by copyright.
hemorrhage in patients with acute upper
gastrointestinal bleeding
Don C Rockey,1 Alan Elliott,2 Thomas Lyles3

▸ Additional material is ABSTRACT

published online only. To In patients with upper gastrointestinal bleeding Significance of this study
view please visit the journal
online (http://dx.doi.org/10.
(UGIB), identifying those with esophageal variceal
1136/jim-2015-000047). hemorrhage prior to endoscopy would be clinically
1 useful. This retrospective study of a large cohort of What is already known about this
Department of Internal
Medicine, Medical University patients with UGIB used logistic regression analyses subject?
of South Carolina, to evaluate the platelet count, aspartate ▸ In patients with upper gastrointestinal
Charleston, South Carolina, aminotransferase (AST) to platelet ratio index (APRI), bleeding (UGIB), identifying those with
USA AST to alanine aminotransferase (ALT) ratio (AAR)
2
Department of Statistical
esophageal variceal hemorrhage prior to
Science, Southern Methodist
and Lok index (all non-invasive blood markers) as endoscopy would be clinically useful.
University, Dallas, Texas, predictors of variceal bleeding in (1) all patients ▸ Non-invasive markers of esophageal varices
USA with UGIB and (2) patients with cirrhosis and UGIB. appear to be helpful in identifying large
3
Digestive Health Associates 2233 patients admitted for UGIB were identified; esophageal varices in stable (out)patients
of Texas, Bedford, Texas, 1034 patients had cirrhosis (46%) and of these,
USA with known cirrhosis.
555 patients (54%) had acute UGIB due to ▸ Non-invasive markers of esophageal varices
Correspondence to esophageal varices. In all patients with UGIB, the have not been examined in the setting of
Dr Don C Rockey, platelet count (cut-off 122,000/mm3), APRI (cut-off acute UGIB.
Department of Internal 5.1), AAR (cut-off 2.8) and Lok index (cut-off 0.9)
Medicine, Medical University
had area under the curve (AUC)s of 0.80 0.82, What are the new findings?
of South Carolina,
96 Jonathan Lucas Street, 0.64, and 0.80, respectively, for predicting the ▸ The platelet count, AST to platelet ratio
Suite 803, MSC 623, presence of varices prior to endoscopy. To predict index, AST to ALT ratio and Lok index
Charleston 29425, SC, USA; varices as the culprit of bleeding, the platelet count accurately predicted the presence of varices
[email protected] prior to endoscopy among the larger group
(cut-off 69,000), APRI (cut-off 2.6), AAR (cut-off
2.5) and Lok Index (0.90) had AUCs of 0.76, 0.77, of all patients with acute UGIB.
Accepted 8 January 2016
Published Online First 0.57 and 0.73, respectively. Finally, in patients with ▸ These same non-invasive measures were
12 February 2016 cirrhosis and UGIB, logistic regression was unable to slightly less accurate in predicting a
identify optimal cut-off values useful for predicting variceal culprit lesion as the cause of
Copyright © 2016 American
Federation for Medical varices as the culprit bleeding lesion for any of the bleeding.
Research non-invasive markers studied. For all patients with ▸ In patients with cirrhosis and UGIB, logistic
UGIB, non-invasive markers appear to differentiate regression was unable to identify optimal
patients with varices from those without varices and cut-off values useful for predicting varices
to identify those with a variceal culprit lesion. as the culprit bleeding lesion for any of the
However, these markers could not distinguish non-invasive markers studied.
between a variceal culprit and other lesions in How might these results change the focus
patients with cirrhosis. of research or clinical practice?
▸ The data suggest that in cirrhotics with
UGIB, endoscopy is required to make a
INTRODUCTION definitive diagnosis of variceal bleeding.
The development of esophageal varices is a
common clinical complication in patients with
cirrhosis; severe bleeding from esophageal
varices has been estimated to occur in approxi- analogues or terlipressin) to reduce portal pres-
mately 30–40% of patients with cirrhosis1 and sure.7–9 Prophylactic antibiotics are also indi-
carries significant morbidity and mortality.2–6 cated in these patients.7 10 Finally, it has been
Since the management of patients with cir- suggested that patients with esophageal varices
rhosis, portal hypertension and esophageal should undergo endoscopy sooner than those
variceal hemorrhage requires specific therapy, it without varices.4 Thus, rapid identification of
is important to differentiate variceal bleeding patients with variceal hemorrhage would
from other causes of upper gastrointestinal greatly enhance the care of this patient
To cite: Rockey DC, bleeding (UGIB). First, patients with variceal population.
Elliott A, Lyles T. J Investig hemorrhage require pharmacological therapy Prediction of the presence of varices prior to
Med 2016;64:745–751. with an agent (ie, intravenous somatostatin endoscopy has previously been studied in stable

Rockey DC, et al. J Investig Med 2016;64:745–751. doi:10.1136/jim-2015-000047 745

 Original research

outpatients, primarily by using non-invasive markers of in patients with any form of UGIB as defined above unless

J Investig Med: first published as 10.1136/jim-2015-000047 on 12 February 2016. Downloaded from http://jim.bmj.com/ on 28 April 2019 by guest. Protected by copyright.
portal hypertension, such as the platelet count, AST-ALT contraindications to endoscopy exist. We have developed a
ratio (AAR), AST-platelet ratio index (APRI), Lok index, standardized approach to management of UGIB including
Forn’s index, functional assays,11 transient elastography12 institution of proton pump inhibitor infusion in all patients
and other novel imaging techniques.13 14 These markers and adjunctive use of octreotide infusion and antibiotics in
have had variable correlation with the presence of varices patients with cirrhosis. Standard operating procedure for
when used in outpatients.11–17 However, they have not practitioners is to use multimodality therapy for all high-
been studied in patients presenting with acute UGIB. risk lesions (ie, those with stigmata of recent bleeding or
Importantly, many of these indices utilize laboratory values those actively bleeding). However, specific management
that are routinely obtained as part of initial assessment and was left to the individual attending physician.
triage in patients with UGIB, making them attractive for Patients with UGIB were identified, and data related to
routine clinical use. the hospital admission were abstracted and entered into the
In this study, we hypothesized that non-invasive markers chronic liver disease database and a Gastrointestinal
of portal hypertension may be useful in determining the Bleeding Healthcare Registry. Data captured included the
presence of varices, and also of varices as etiology of bleed- following: clinical features associated with gastrointestinal
ing, prior to endoscopy. The aims of this study were there- bleeding, patient demographics, mortality and cause of
fore to determine whether readily available and practical death, medical history, packed red blood cell transfusions,
clinical markers might predict the presence of varices prior hospital course data (intensive care unit stay, intubation,
to endoscopy in several different clinical contexts. First, in use of vasoactive agents), laboratory data on admission,
the entire population of all patients with acute UGIB, we vital signs at admission and endoscopic data (types of pro-
sought to determine whether non-invasive markers could cedures performed, endoscopic diagnosis, stigmata of
accurately predict the presence of varices (ie, in cirrhotic recent or active hemorrhage, and endoscopic therapies). A
patients) prior to endoscopy. Further, we analyzed this bleeding lesion in any given case was labeled as the primary
group to determine whether non-invasive markers could diagnosis and the culprit gastrointestinal bleeding lesion. In
accurately predict varices as the culprit bleeding lesion. some patients, more than one diagnosis was present (eg,
Finally, in the specific cohort of patients with cirrhosis and esophagitis or gastritis was identified in addition to a
UGIB, we sought to determine whether non-invasive primary lesion, and considered a secondary lesion) but not
markers could identify varices as the culprit bleeding lesion deemed to be the cause of hemorrhage. Variceal bleeding is
prior to endoscopy. deemed to occur when varices with stigmata of bleeding
are present7 18 when stigmata of variceal bleeding are
METHODS present, varices were deemed to be the culprit lesion. For
This cohort study focused on non-invasive blood markers varices, additional data were collected such as number and
of portal hypertension to predict presence of varices and grade of varices, use of band ligation as hemostasis and
also variceal bleeding as etiology of acute UGIB. The study number of bands placed. Determination of bleed etiology
included patients admitted to Parkland Memorial Hospital and necessary treatment was determined by the attending
(Dallas, Texas, USA), a University of Texas Southwestern physician at the time of the patient’s endoscopic examin-
(UTSW) teaching hospital, from January 1, 2003 through ation. For the purposes of analyses, patients with esopha-
July 1, 2012. Inclusion required documented UGIB. geal, but not gastric varices were considered to have
Patients were identified retrospectively using the institu- variceal hemorrhage.
tion’s chronic liver disease database (2003–2006) and Demographics, laboratory values and physical examin-
UTSW’s Gastrointestinal Bleeding Healthcare Registry ation findings are the initial values on the patient’s admis-
(2007–2012); patient data in the latter registry are col- sion to the hospital. APRI, AAR and Lok index were
lected prospectively, and all patients admitted with UGIB calculated using admission laboratory data, and in accord-
from this institution were included in this cohort. Patients ance with previously published formulas.19–21 Child-Pugh
with incomplete clinical data (laboratory, endoscopy, or scores and model for end stage liver disease (MELD) scores
30-day outcome) were excluded from the current analysis. were calculated using admission laboratory values, physical
We first subdivided the main cohort into patients with examination and clinical presentation. Endoscopic data
cirrhosis and those without cirrhosis. Cirrhosis was defined were gathered at the time of endoscopy, and hospital
as follows: a history consistent with chronic liver disease, course data were reviewed and collected after the
plus the presence of clinical features consistent with cirrho- hospitalization.
sis, including clinical findings of cirrhosis and portal hyper- Patients were excluded from this analysis if they had a
tension (spider angiomata, gynecomastia, splenomegaly, form of gastrointestinal bleeding other than an upper
thrombocytopenia) or a documented complication of gastrointestinal source, age under 18, did not have endos-
chronic liver disease (ie, ascites, varices, hepatic encephal- copy during their hospitalization, or were pregnant. For
opathy), and/or imaging consistent with cirrhosis, and/or patients who had multiple hospitalizations for bleeding
liver histology consistent with cirrhosis. during the study period, rebleeding after 42-days from the
UGIB was defined as reported or witnessed melena, index bleed was considered a new bleed. A readmission
hematemesis, coffee ground emesis or hematochezia (with within 42-days from the index bleed was considered a
a documented upper gastrointestinal tract lesion) in the rebleed and data from that admission were not captured in
setting of at least a four point drop in hematocrit from the study database. The study was approved by the
baseline or lower than normal. It is standard practice at our University of Texas Southwestern Institutional Review
institution to perform esophagogastroduodenoscopy (EGD) Board and met all criteria for good clinical practice.22
746 Rockey DC, et al. J Investig Med 2016;64:745–751. doi:10.1136/jim-2015-000047
 Original research

Descriptive statistics were used to present demographic

J Investig Med: first published as 10.1136/jim-2015-000047 on 12 February 2016. Downloaded from http://jim.bmj.com/ on 28 April 2019 by guest. Protected by copyright.
and clinical characteristics of patients in the study cohort.
The primary outcomes of this study were identification of
the presence of esophageal varices prior to EGD, identifica-
tion of esophageal varices as the culprit lesion in patients
presenting with acute UGIB, and identification of esopha-
geal varices as the culprit lesion in patients with cirrhosis.
We specifically aimed to examine previously reported non-
invasive tests used to assess portal hypertension and varices.

Statistics
Descriptive statistics were used to present demographic and
clinical characteristics of patients in the study cohort. The
primary outcomes of this study was identification of the
presence of esophageal varices prior to EGD, identification
of varices as bleed etiology in patients presenting with
acute UGIB and identification of varices as primary eti-
ology of acute UGIB in patients with cirrhosis. We specific-
ally aimed to examine previously published non-invasive, Figure 1 Flow diagram of the patient cohort. 3018 patients
serum markers of portal hypertension and their correlation were admitted to our institution with acute gastrointestinal
with esophageal varices. Student’s t test or analysis of vari- bleeding during the study period. After excluding patients with
ance was used to compare means by group. χ2 Analysis was lower gastrointestinal bleeding and removing the upper and lower
used to compare categorical data by group. To reduce the 5% (Winsorization), we divided the cohort into those with
effect of outliers in the data set, 5% Winsorization was cirrhosis (1034 patients) and those without cirrhosis (991). We
applied to the data set as a whole in order to adjust then further subdivided the cirrhosis cohort into variceal (555
extreme values and provide more robust estimation for stat- patients) and non-variceal bleeding (479 patients) groups.
istical inference.23 Univariate logistic regression was used to
determine which variables were predictive of the outcome
variables with area under the curve (AUC) and measures of normalized ratio (INR), higher total bilirubin, higher AST,
sensitivity and specificity used to assess the variable’s pre- higher ALT (all p<0.001 for cirrhosis with or without vari-
dictive capabilities. Predictive cut-off values for variables ceal bleeding vs patients without cirrhosis) (table 1). As
were determined using the logistic model to indicate what might be predicted, patients with cirrhosis had substantially
values of each independent variable predicts the outcome different non-invasive markers of portal hypertension, such
with greater than 50% probability. Multiple logistic regres- as APRI, Lok index and AAR than patients without cirrho-
sion was used to determine which combination of variables sis (all p<0.001 for cirrhosis with or without variceal
was predictive of the outcomes. A p value less than 0.05 bleeding vs patients without cirrhosis) (table 1). Notably,
was considered statistically significant. No adjustment of although both groups had similar hematocrit levels at times
the p values was made for multiple tests. Analyses were of admission, the mortality in patients with cirrhosis (both
performed using SAS V.9.3 (SAS Institute, Cary, North groups) had significantly higher mortality than patients
Carolina, USA). without cirrhosis ( p<0.001).
The cirrhosis cohort was then specifically examined to
RESULTS determine differences between cirrhosis patients with an
We identified a total of 2233 unique admissions of patients esophageal variceal culprit lesion and those with a non-
with acute UGIB (figure 1). After 5% Winsorization, the variceal culprit lesion. The two groups had similar causes
cohort included 1034 patients with cirrhosis, and 991 of cirrhosis and similar distribution across the
patients without cirrhosis. Of the patients with cirrhosis, Child-Turcotte-Pugh classification. However, there were
815 (79%) were found to have esophageal varices at the several clinical and laboratory features that appeared to be
time of endoscopy. Of these patients, 260 (25% of the different (table 2); patients with variceal bleeding had
entire cirrhosis cohort) were found to have small or grade lower systolic blood pressure and lower mean arterial pres-
1 varices (our institutional practice grades varices on a sure. Interestingly, those with non-variceal bleeding had
scale from 1 to 4, per the Paquet scale24) that were not the higher MELD and MELD-Na (sodium) scores (table 2).
causative etiology of bleeding. Thus, of the group of 1034 Several of the non-invasive markers had statistically signifi-
patients with cirrhosis, 555 (54%) patients were classified cant differences between the variceal and non-variceal cir-
as having bled from esophageal varices. The remaining 479 rhotics. Patients with variceal bleeding had lower platelet
(46%) patients from this group bled from other etiologies counts, a lower Lok index and a lower AAR. APRI was the
(figure 1, online supplementary table S1). same between the two groups.
We first sought to examine differences between patients
with cirrhosis and those without cirrhosis. Patients with cir- Prediction of the presence of esophageal varices in all
rhosis and acute UGIB were more likely to be male, patients with UGIB
Hispanic in ethnicity, and had a higher in-hospital mortal- Logistic regression analysis was used to determine the pre-
ity rate. They also had several notable laboratory abnormal- dictive ability of non-invasive markers to identify the pres-
ities including lower platelets, higher international ence of esophageal varices in all patients presenting with
Rockey DC, et al. J Investig Med 2016;64:745–751. doi:10.1136/jim-2015-000047 747
 Original research

J Investig Med: first published as 10.1136/jim-2015-000047 on 12 February 2016. Downloaded from http://jim.bmj.com/ on 28 April 2019 by guest. Protected by copyright.
Table 1 Demographic, laboratory and clinical characteristics Table 2 Characteristics of patients with cirrhosis presenting
of all patients presenting with UGIB with UGIB
Patients with cirrhosis Esophageal Non-variceal
(n=1034) variceal etiology
etiology (n=555) (n=479) p value
Esophageal Patients
variceal Non-variceal without Systolic blood pressure 120 (±23) 124 (±22) 0.006
bleeding bleeding cirrhosis
Mean arterial pressure 86 (±16) 89 (±16) 0.011
(n=555) (N=479) (n=991)
Pulse 97 (±20) 94 (±21) 0.144
Age—years (SD) 50 (9) 52 (10) 53 (15) Ascites present on 137 (25%) 86 (18%) 0.01
Gender—female (%) 125 (23) 163 (34) 351 (35) physical examination
Ethnicity Cirrhosis etiology 0.63
Hispanic 305 (55%) 212 (44%) 329 (33%) Alcohol 195 (35%) 162 (39%)
White 164 (30%) 170 (36%) 261 (27%) Hepatitis C 270 (49%) 187 (45%)
African-American 67 (12%) 85 (18%) 369 (37%) Hepatitis B 31 (6%) 21 (5%)
Other 19 (3%) 12 (2%) 32 (3%) NAFLD 22 (4%) 21 (5%)
In-hospital mortality 40 (7.0%) 29 (7.0%) 66 (7%) Cryptogenic 22 (4%) 16 (4%)
WBCs (×109/L) 9 (± 6) 9 (±5) 11 (±6) Other 15 (3%) 7 (2%)
Hematocrit (%) 27 (± 7) 28 (±8) 29 (±8) CTP class 0.41
Platelets (×109/L) 114 (±61) 133 (± 89) 240 (±119) A 187 (34%) 124 (30%)
BUN (mg/dL) 25 (± 17) 25 (± 18) 33 (±16) B 257 (46%) 197 (48%)
Sodium (mmol/L) 135 (± 5) 135 (±5) 136 (±5) C 111 (20%) 93 (22%)
Creatinine (mg/dL) 1 (±0.7) 1.2 (±1.2) 2.1 (±9.0) MELD 14 (±6) 15 (±7) 0.002
INR 1.5 (±0.5) 1.5 (±0.9) 1.3 (±0.9) MELD-Na 17 (±6) 18 (±7) 0.007
Total bilirubin (mg/dL) 2.7 (± 4.4) 2.8 (±4.4) 0.7 (±1.3) Platelets 114 (±61) 133 (±89) <0.001
Albumin (g/dL) 2.9 (± 0.6) 2.9 (± 0.7) n/a APRI 1.8 (± 1.9) 1.7 (±1.9) 0.67
AST 72 (±44) 71 (±44) 54 (±137) Lok index 0.89 (±0.14) 0.84 (±0.25) <0.001
ALT 45 (±33) 39 (±32) 32 (±35) AAR 1.8 (±0.9) 2.1 (±1.1) <0.001
APRI 1.8 (± 1.9) 1.7 (±1.9) 0.5 (±1.6) APRI, AST to platelet ratio index; AAR, AST to ALT ratio; CTP, Child-Turcotte-
Lok Index 0.89 (±0.14) 0.84 (±0.25) 0.5 (±0.3) Pugh; MELD, model for end stage liver disease; MELD-Na, MELD-Na (sodium);
NAFLD, nonalcoholic fatty liver disease; UGIB, upper gastrointestinal bleeding.
AST:ALT ratio 1.8 (±0.9) 2.1 (±1.1) 1.5 (±0.8)
ALT, alanine aminotransferase; AST, aspartate aminotransferase; APRI, AST to
platelet ratio index; BUN, blood urea nitrogen; INR, international normalized (cut-off 69,000), APRI (cut-off 2.6), AAR (cut-off 2.5) and
ratio; n/a, not applicable; UGIB, upper gastrointestinal bleeding; WBC, white
blood cell. Lok Index (0.90) had AUCs of 0.76, 0.77, 0.57 and 0.73,
respectively. For prediction of bleeding from varices, speci-
ficity was marginal for all markers, with the Lok index
UGIB. By performing AUC analyses, we determined having the best specificity at 63%. Overall, the platelet
optimal cut-off values. For all patients presenting with count and Lok index had the best test performances of the
UGIB, the platelet count (cut-off 122,000/mm3), APRI four markers. A cut-off platelet count for the presence of
(cut-off 5.1), AAR (cut-off 2.81) and the Lok index (cut-off varices was determined to be less than or equal to 69,000/
0.86) had AUCs of 0.80 0.83, 0.64, and 0.80, respectively, mm3, while a cut-off Lok index for the prediction of
for predicting the presence of esophageal varices prior to varices was determined to be less than or equal to 0.9.
endoscopy. When we considered simple identification of
the presence of esophageal varices, and associated AUC
values, we created a summary of best fit performance Table 3 Best fit performance of non-invasive markers for
characteristics (table 3). Of the four non-invasive prediction of the presence of esophageal varices in all patients
approaches studied, AAR and APRI had poor specificity. with UGIB
Platelets and the Lok index had similar sensitivities, specifi-
Platelets AST:ALT ratio APRI Lok index
cities, positive predictive value and negative predictive
value. AUC (95% CI) 0.80 0.64 0.83 0.80
Although we determined optimal cut-off values through Cut-off 122 2.81 5.1 0.86
best-fit analysis of the AUC, we explored other cut-off Sensitivity (%) 80 93 96 73
values for each non-invasive marker, with corresponding Specificity (%) 63 15 13 77
unique sensitivities and specificities (table 4). PPV (%) 78 64 64 84
NPV (%) 61 57 64 64
Prediction of esophageal varices as the culprit lesion in Accuracy (%) 74 64 64 75
patients with UGIB LR+ 2.1 1.1 1.1 3.2
We utilized logistic analysis to examine the predictive LR− 0.4 0.5 0.3 0.4
ability of non-invasive markers to predict esophageal AUC, area under the curve; LR+, Positive likelihood ratio; LR−, negative
varices as the bleeding etiology in patients presenting with likelihood ratio; NPV, negative predictive value; PPV, positive predictive value;
UGIB, upper gastrointestinal bleeding.
UGIB (table 5). In this patient subset, the platelet count
748 Rockey DC, et al. J Investig Med 2016;64:745–751. doi:10.1136/jim-2015-000047
 Original research

J Investig Med: first published as 10.1136/jim-2015-000047 on 12 February 2016. Downloaded from http://jim.bmj.com/ on 28 April 2019 by guest. Protected by copyright.
Table 4 Cut-off values of non-invasive markers for Table 6 Performance of non-invasive markers for prediction
prediction of esophageal varices prior to endoscopy in all of esophageal varices as the bleeding etiology in cirrhotics
patients with UGIB with UGIB
Non-invasive marker Cut-off Sensitivity (%) Specificity (%) Platelets AST:ALT ratio APRI Lok index

Platelets 58 95 17 AUC 0.50 0.62 0.52 0.56

80 87 43 Cut-off 461 2.8 10.5 1
142 71 77
Sensitivity 0.2% 22% 0.2% 0%
176 61 88
Specificity (%) 100 88 99 100
AST-ALT ratio 3.7 97 6
1.9 77 40 AUC, area under the curve; UGIB, upper gastrointestinal bleeding.
1.3 55 64
0.9 22 92
APRI 11.4 99 2
7.9 98 6 including history, examination, and laboratory data. While
3.8 88 37 medical history may be known for many patients at the
1.8 73 76 time of admission, a subset of patients present without a
Lok index 0.93 79 63 known medical history. For example, some patients with
0.78 68 85 acute UGIB are not known to have cirrhosis at the time of
0.70 63 90
0.60 56 95
presentation; in a previous study, the diagnosis of cirrhosis
was unknown in 35% of patients at the time of presenta-
APRI, AST to platelet ratio index; UGIB, upper gastrointestinal bleeding.
tion with acute UGIB.27 This is especially important as
in-hospital mortality for cirrhotics with acute UGIB is
greater than for patients without cirrhosis,28 and it is pos-
Prediction of esophageal varices as the culprit lesion in sible that patients with cirrhosis may benefit from early
cirrhotic patients with UGIB endoscopy.26
We next attempted to identify ideal cut-off values for We found that in patients with known cirrhosis, esopha-
detection of bleeding esophageal varices in cirrhotic geal variceal hemorrhage caused bleeding in approximately
patients presenting with acute UGIB. Although there was a half of patients. Thus, the precise etiology of bleeding is
statistically significant difference in the measured non- often unknown until the time of endoscopy. In this study,
invasive markers between patients with cirrhosis with vari- we have shown that in a large combined cohort of patients
ceal bleeding, and patients with cirrhosis and non-variceal presenting to an emergency room with acute UGIB, the
bleeding, logistic regression was unable to clinically differ- platelet count was the best predictor of esophageal varices
entiate variceal from non-variceal bleeding. Overall, the being present on endoscopy. In comparison to platelets, the
examined markers had poor AUCs, poor sensitivities and Lok Index had a similar negative predictive value (64%)
clinically irrelevant cut-offs (table 6). and a higher positive predictive value (84%) for the predic-
tion of the presence of esophageal varices in patients pre-
DISCUSSION senting with acute UGIB.
Acute UGIB is a common clinical problem in the USA, with In patients with cirrhosis, however, non-invasive markers
an annual incidence of 82–146 cases per 100,000 adults25 of portal hypertension were unable to differentiate esopha-
and estimated medical costs of $1 billion.26 Accordingly, geal variceal from non-variceal bleeding prior to endos-
accurate identification of UGIB and appropriate triage of copy. On examination of the cirrhosis cohort (table 2),
care is essential in the management of these patients. A there was a significant difference between the platelet
variety of clinical features assist the clinician in triage, count, APRI and Lok. However, further analysis with logis-
tic regression deemed these markers unable to accurately
identify varices as bleeding etiology prior to endoscopy.
Table 5 Performance of non-invasive markers for prediction While non-invasive markers of portal hypertension have
of esophageal varices as the bleeding etiology in all patients previously been studied in the context of esophageal vari-
with UGIB ceal screening in high-risk patients12 29 and correlation
with fibrosis, they have not been applied to patients pre-
Platelets AST:ALT ratio APRI Lok index
senting with acute UGIB. Further, with the exception of
AUC 0.76 0.57 0.77 0.73 platelet count, the respective non-invasive marker indices
Cut-off 69 2.5 2.6 0.9 examined in this study were formulated using chronic
Sensitivity 91 85 92 70 hepatitis C virus patient cohorts. Thus, this is a novel use
Specificity (%) 24 19 21 63 of previously reported indices.
PPV (%) 76 73 76 83 Although the Lok Index is calculated with AST, ALT, pla-
NPV (%) 49 31 49 44 telets, and INR (all of which can be easily obtained and
Accuracy (%) 72 67 72 68 reported in a time efficient manner), it requires a dedicated
LR+ 1.2 1.0 1.2 1.9 calculation; whereas the platelet count is readily obtainable
LR− 0.4 0.8 0.4 0.5 and easily interpretable. Thus, we speculate that the appli-
APRI, AST to platelet ratio index; AUC, area under the curve; NPV, negative cation of the platelet count in the initial assessment and
predictive value; PPV, positive predictive value; LR+, Positive likelihood ratio; triage of patients with acute UGIB could potentially
LR−, negative likelihood ratio; UGIB, upper gastrointestinal bleeding. augment a clinician’s initial decision-making by rapidly
Rockey DC, et al. J Investig Med 2016;64:745–751. doi:10.1136/jim-2015-000047 749
 Original research

identifying those patients likely to have portal hyperten- content; statistical analysis. TL was involved in the study concept and design;

J Investig Med: first published as 10.1136/jim-2015-000047 on 12 February 2016. Downloaded from http://jim.bmj.com/ on 28 April 2019 by guest. Protected by copyright.
sion, and thus in need of adjunctive octreotide, antibiotics acquisition of data; analysis and interpretation of data; drafting of the
manuscript; critical revision of the manuscript for important intellectual
and potential transfer to facilities that have gastroenter- content.
ology services on call. Unfortunately, while reasonably sen-
Competing interests None declared.
sitive, this clinical measure is not specific.
There are several strengths of this study, notably the size Ethics approval UTSW IRB.
of the overall patient population, and the size of the cirrho- Provenance and peer review Not commissioned; externally peer reviewed.
sis cohort. Furthermore, data were collected over the
course of almost a decade, at a single center, during a time
in which standard of care did not substantially change, (ie,
REFERENCES
the use of proton pump inhibitors, vasoactive agents 1 Jensen DM. Endoscopic screening for varices in cirrhosis: findings,
(octreotide) and antibiotics in cirrhosis presenting with implications, and outcomes. Gastroenterology 2002;122:1620–30.
UGIB). Further, data collection was complete since the 2 Lyles T, Elliott A, Rockey DC. A risk scoring system to predict in-hospital
premise of the study is to utilize routine, easily obtainable mortality in patients with cirrhosis presenting with upper gastrointestinal
bleeding. J Clin Gastroenterol 2014;48:712–20.
laboratory markers that are commonly drawn at the time of 3 Cheung J, Soo I, Bastiampillai R, et al. Urgent vs. non-urgent endoscopy in
an initial assessment. stable acute variceal bleeding. Am J Gastroenterol 2009;104:1125–9.
We recognize limitations of this study. The study was at a 4 Hsu YC, Chung CS, Tseng CH, et al. Delayed endoscopy as a risk factor for
single center, was retrospective, and not all patients diag- in-hospital mortality in cirrhotic patients with acute variceal hemorrhage.
J Gastroenterol Hepatol 2009;24:1294–9.
nosed with cirrhosis had a biopsy proven diagnosis.
5 Ananthakrishnan AN, McGinley EL, Saeian K. Outcomes of weekend
However, practice at our center followed closely currently admissions for upper gastrointestinal hemorrhage: a nationwide analysis.
published guidelines suggesting that the practices followed Clin Gastroenterol Hepatol 2009;7:296–302e1.
are generalizable nationally.7 30 31 Additionally, our popula- 6 Myers RP, Papay KD, Shaheen AA, et al. Relationship between hospital
tion was extremely heterogeneous, suggesting that the find- volume and outcomes of esophageal variceal bleeding in the United States.
Clin Gastroenterol Hepatol 2008;6:789–98.
ings are likely generalizable to a wide range of patients. 7 Garcia-Tsao G, Sanyal AJ, Grace ND, et al. Prevention and management of
With regard to concern about a biopsy diagnosis of cirrho- gastroesophageal varices and variceal hemorrhage in cirrhosis. Hepatology
sis, it should be emphasized that we used commonly 2007;46:922–38.
accepted clinical criteria for the diagnosis of cirrhosis. 8 Azam Z, Hamid S, Jafri W, et al. Short course adjuvant terlipressin in acute
variceal bleeding: a randomized double blind dummy controlled trial.
Interestingly, mortality in our cohort was comparatively
J Hepatol 2012;56:819–24.
low in comparison with other studies,32 33 raising concerns 9 Ioannou GN, Doust J, Rockey DC. Systematic review: terlipressin in acute
about differences in care provided at our institution com- oesophageal variceal haemorrhage. Aliment Pharmacol Ther 2003;17:53–64.
pared to other institutions. However, this should have no 10 Mayorga CA, Rockey DC. Clinical utility of a standardized electronic order set
bearing on the conclusions drawn here since we focused on for the management of acute upper gastrointestinal hemorrhage in patients
with cirrhosis. Clin Gastroenterol Hepatol 2013;11:1342–8.
diagnosis rather than outcome. 11 Lisotti A, Azzaroli F, Buonfiglioli F, et al. Indocyanine green retention test as
In conclusion, by examining an extremely large, diverse a noninvasive marker of portal hypertension and esophageal varices in
and carefully phenotyped cohort of patients with UGIB, compensated liver cirrhosis. Hepatology 2014;59:643–50.
we have shown that patients with cirrhosis frequently bleed 12 Sebastiani G, Tempesta D, Fattovich G, et al. Prediction of oesophageal
varices in hepatic cirrhosis by simple serum non-invasive markers: results of a
from lesions other than esophageal varices. We have also
multicenter, large-scale study. J Hepatol 2010;53:630–8.
demonstrated that patients with cirrhosis have a higher 13 Takuma Y, Nouso K, Morimoto Y, et al. Measurement of spleen stiffness by
mortality rate than those without cirrhosis and that the acoustic radiation force impulse imaging identifies cirrhotic patients with
mortality rate in patients with cirrhosis with esophageal esophageal varices. Gastroenterology 2013;144:92–101.e2.
variceal and non-variceal bleeding was equivalent. This 14 Park Y, Kim SU, Park SY, et al. A novel model to predict esophageal varices
in patients with compensated cirrhosis using acoustic radiation force impulse
underscores that other factors, such as underlying severity elastography. PLoS ONE 2015;10:e0121009.
of liver disease are clinically important, as is the clinical 15 Colecchia A, Montrone L, Scaioli E, et al. Measurement of spleen stiffness to
course of a patient with cirrhosis, especially if they develop evaluate portal hypertension and the presence of esophageal varices in
other poor prognostic signs such as sepsis or multiorgan patients with HCV-related cirrhosis. Gastroenterology 2012;143:646–54.
16 Shi KQ, Fan YC, Pan ZZ, et al. Transient elastography: a meta-analysis of
failure. Finally, while our study indicates that the platelet
diagnostic accuracy in evaluation of portal hypertension in chronic liver
count may be potentially useful in augmenting a clinician’s disease. Liver Int 2013;33:62–71.
ability to quickly identify patients with cirrhosis prior to 17 Berzigotti A, Seijo S, Arena U, et al. Elastography, spleen size, and platelet
endoscopy, we conclude that non-invasive tests are not able count identify portal hypertension in patients with compensated cirrhosis.
to accurately predict whether the cause of bleeding is due Gastroenterology 2013;144:102–111.e1.
18 Sharara AI, Rockey DC. Gastroesophageal variceal hemorrhage. N Engl J Med
to esophageal varices. In this setting, endoscopy is required 2001;345:669–81.
to make a definitive diagnosis. 19 Wai CT, Greenson JK, Fontana RJ, et al. A simple noninvasive index can
predict both significant fibrosis and cirrhosis in patients with chronic hepatitis
C. Hepatology 2003;38:518–26.
Acknowledgements The authors thank Rebecca Chason for her many
20 Giannini E, Risso D, Botta F, et al. Validity and clinical utility of the aspartate
contributions to this research, including maintenance of the Gastrointestinal
aminotransferase-alanine aminotransferase ratio in assessing disease severity
Bleeding Healthcare Registry and data base. The author thank the housestaff,
and prognosis in patients with hepatitis C virus-related chronic liver disease.
fellows, faculty, nurses, and staff of the Parkland Endoscopy unit for their
Arch Intern Med 2003;163:218–24.
diligence in care for patients with gastrointestinal bleeding.
21 Lok AS, Ghany MG, Goodman ZD, et al. Predicting cirrhosis in patients with
Contributors DCR was involved in the study concept and design; acquisition hepatitis C based on standard laboratory tests: results of the HALT-C cohort.
of data; analysis and interpretation of data; drafting of the manuscript; critical Hepatology 2005;42:282–92.
revision of the manuscript for important intellectual content; statistical 22 [No authors listed]. World Medical Association Declaration of Helsinki:
analysis. AE was involved in the analysis and interpretation of data; drafting recommendations guiding physicians in biomedical research involving human
of the manuscript; critical revision of the manuscript for important intellectual subjects. JAMA 1997;277:925–6.

750 Rockey DC, et al. J Investig Med 2016;64:745–751. doi:10.1136/jim-2015-000047

 Original research
23 Kotz S, Balakrishnan N, Read CB, et al. Encyclopedia of statistal sciences. 29 Castéra L, Le Bail B, Roudot-Thoraval F, et al. Early detection in

J Investig Med: first published as 10.1136/jim-2015-000047 on 12 February 2016. Downloaded from http://jim.bmj.com/ on 28 April 2019 by guest. Protected by copyright.
Hoboken, NJ: Wiley-Interscience, 2006. routine clinical practice of cirrhosis and oesophageal varices in chronic
24 Paquet KJ. Prophylactic endoscopic sclerosing treatment of the esophageal hepatitis C: comparison of transient elastography (FibroScan) with
wall in varices—a prospective controlled randomized trial. Endoscopy standard laboratory tests and non-invasive scores. J Hepatol 2009;50:
1982;14:4–5. 59–68.
25 Zhao Y, Encinosa W. Hospitalizations for gastrointestinal bleeding in 1998 30 de Franchis R. Updating consensus in portal hypertension: report of the
and 2006: statistical brief #65. Healthcare cost and Utilization Project Baveno III Consensus Workshop on definitions, methodology and therapeutic
(HCUP) statistical Briefs. Rockville, MD, 2006. strategies in portal hypertension. J Hepatol 2000;33:846–52.
26 Wysocki JD, Srivastav S, Winstead NS. A nationwide analysis of risk factors 31 de Franchis R. Evolving consensus in portal hypertension. Report of the
for mortality and time to endoscopy in upper gastrointestinal haemorrhage. Baveno IV consensus workshop on methodology of diagnosis and therapy in
Aliment Pharmacol Ther 2012;36:30–6. portal hypertension. J Hepatol 2005;43:167–76.
27 del Olmo JA, Peña A, Serra MA, et al. Predictors of morbidity and mortality 32 D’Amico G, De Franchis R., Cooperative Study Group. Upper digestive
after the first episode of upper gastrointestinal bleeding in liver cirrhosis. bleeding in cirrhosis. Post-therapeutic outcome and prognostic indicators.
J Hepatol 2000;32:19–24. Hepatology 2003;38:599–612.
28 Lecleire S, Di Fiore F, Merle V, et al. Acute upper gastrointestinal bleeding in 33 González-González JA, García-Compean D, Vázquez-Elizondo G, et al.
patients with liver cirrhosis and in noncirrhotic patients: epidemiology and Nonvariceal upper gastrointestinal bleeding in patients with liver cirrhosis.
predictive factors of mortality in a prospective multicenter population-based Clinical features, outcomes and predictors of in-hospital mortality. A
study. J Clin Gastroenterol 2005;39:321–7. prospective study. Ann Hepatol 2011;10:287–95.

Rockey DC, et al. J Investig Med 2016;64:745–751. doi:10.1136/jim-2015-000047 751