EVIDENCE-BASED DIAGNOSTICS

History, Physical Examination, Laboratory

Testing, and Emergency Department
Ultrasonography for the Diagnosis of Acute
Cholecystitis
Ashika Jain, MD, Ninfa Mehta, MD, Michael Secko, MD, Joshua Schechter, MD,
Dimitri Papanagnou, MD, Shreya Pandya, MD, and Richard Sinert, DO

ABSTRACT
Background: Acute cholecystitis (AC) is a common differential for patients presenting to the emergency
department (ED) with abdominal pain. The diagnostic accuracy of history, physical examination, and bedside
laboratory tests for AC have not been quantitatively described.

Objectives: We performed a systematic review to determine the utility of history and physical examination (H&P),
laboratory studies, and ultrasonography (US) in diagnosing AC in the ED.

Methods: We searched medical literature from January 1965 to March 2016 in PubMed, Embase, and
SCOPUS using a strategy derived from the following formulation of our clinical question: patientsED
patients suspected of AC; interventionsH&P, laboratory studies, and US ndings commonly used to
diagnose AC; comparatorsurgical pathology or denitive diagnostic radiologic study conrming AC; and
outcomethe operating characteristics of the investigations in diagnosing AC were calculated. Sensitivity,
specicity, and likelihood ratios (LRs) were calculated using Meta-DiSc with a random-effects model (95% CI).
Study quality and risks for bias were assessed using the Quality Assessment Tool for Diagnostic Accuracy
Studies.

Results: Separate PubMed, Embase, and SCOPUS searches retrieved studies for H&P (n = 734), laboratory
ndings (n = 74), and US (n = 492). Three H&P studies met inclusion/exclusion criteria with AC prevalence of
7%64%. Fever had sensitivity ranging from 31% to 62% and specicity from 37% to 74%; positive LR [LR+]
was 0.711.24, and negative LR [LR] was 0.761.49. Jaundice sensitivity ranged from 11% to 14%, and
specicity from 86% to 99%; LR+ was 0.8013.81, and LR was 0.871.03. Murphys sign sensitivity was 62%
(range = 53%71%), and specicity was 96% (range = 95%-97%); LR+ was 15.64 (range = 11.4821.31), and
LR was 0.40 (range = 0.320.50). Right upper quadrant pain had sensitivity ranging from 56% to 93% and
specicity of 0% to 96%; LR+ ranged from 0.92 to 14.02, and LR from 0.46 to 7.86. One laboratory study met
criteria with a 26% prevalence of AC. Elevated bilirubin had a sensitivity of 40% (range = 12%74%) and
specicity of 93% (range = 77%99%); LR+ was 5.80 (range = 1.2526.99), and LR was 0.64 (range = 0.39
1.08). Five US studies with a prevalence of AC of between 10% and 46%. US sensitivity was 86% (range =
78%94%) and specicity was 71% (range = 66%76%); LR+ was 3.23 (range = 1.746.00), and LR was 0.18
(range = 0.100.33).

From the Department of Emergency Medicine, SUNY-Downstate Medical Center (AJ, NM, MS, JS, SP, RS), Brooklyn, NY; and the Department of
Emergency Medicine, Thomas Jefferson University Hospital (DP), Philadelphia, PA.
Received May 27, 2016; revision received October 23, 2016; accepted November 2, 2016.
This study was presented at the Society for Academic Emergency Medicine Annual Meeting, Dallas, TX, May 16, 2014; the Society for Academic
Emergency Medicine Mid-Atlantic Regional Meeting, Washington, DC, February 22, 2014; and the Society for Academic Emergency Medicine
North-East Regional Conference, New Haven, CT, March 26, 2014.
The authors have no relevant nancial information or potential conicts to disclose.
Supervising Editor: Christopher R. Carpenter, MD.
Address for correspondence and reprints: Ninfa Mehta, MD; e-mail: ashikajainmd@gmail.com.
ACADEMIC EMERGENCY MEDICINE 2017;24:281297.

2016 by the Society for Academic Emergency Medicine ISSN 1069-6563

doi: 10.1111/acem.13132 PII ISSN 1069-6563583 281
282 Jain et al. A SYSTEMATIC REVIEW OF ED ULTRASOUND FOR CHOLECYSTITIS DIAGNOSIS
Conclusion: Variable disease prevalence, coupled with limited sample sizes, increases the risk of selection bias.
Individually, none of these investigations reliably rule out AC. Development of a clinical decision rule to include
evaluation of H&P, laboratory data, and US are more likely to achieve a correct diagnosis of AC.
A bdominal pain is one of the most common chief
complaints encountered by emergency depart-
ment (ED) physicians. Many abdominal pain patients
are found to have a benign etiology. However, the pos-
sibility of acute cholecystitis (AC), which accounts for
3%11% of hospital admissions13 with a mortality
rate of 0.8%,4 usually requires a comprehensive diag-
nostic evaluation using the modalities of history and
physical examination (H&P), laboratory results, and
imaging studies. Many studies have previously
described individually the H&P,5,6 laboratory,7,8 and
imaging9,10 consistent with AC. The Tokyo AC guide-
lines by Hirota et al.11 utilized an expert consensus
methodology to integrate multiple diagnostic modalities
to predict AC. Hirota et al.l1 determined a definitive
diagnosis of AC could be obtained if the patient
exhibited at least one of the two local signs of inflam-
mation (Murphys sign or right upper quadrant
[RUQ] mass/pain/tenderness) plus one of the sys-
temic signs of inflammation (fever, elevated C-reactive
protein or elevated while blood cell count [WBC]).
Then if AC was clinically suspected, a definitive diag-
nosis of AC would still depend on a radiology depart-
ment study (ultrasound [US], computerized
tomography [CT], magnetic resonance imaging [MRI],
or hepatobiliary iminodiacetic acid scan/cholescintigra-
phy [HIDA]). Recently an updated version of the
Tokyo AC guidelines (TG13)12 was validated in a sim-
ilar population to the derivation cohort with a sensitiv-
ity of 87.6% and specificity of 77.7%.
Although the TG1312 integrates the diagnostic
modalities for AC, they are derived by expert opinion
based on retrospective multicenter analysis and vali-
dated in the same population as the derivation cohort
may limit the generalizability to other patient popula-
tions. We decided to develop a similar integrated
approach to diagnosing AC but utilizing a systematic
review/meta-analytic approach. The primary objective
of this systematic review is to determine the diagnostic
test accuracy (sensitivity, specificity, and likelihood
ratios [LRs]) of H&P, laboratory data, and imaging
studies to predict AC for ED patients. Specifically, we
were interested in finding elements of the H&P, labo-
ratory data, and imaging studies available to ED physi-
cians at the point of care, which would allow
expedited disposition of patients suspected of AC
without utilizing radiology department resources (US,
CT, MRI, or HIDA). We limited our population to
ED patients suspected of AC and specifically limited
sonography as performed and interpreted by ED physi-
cians. While there are several variants of AC, this
study does not address variants such as emphysema-
tous cholecystitis or acalculous cholecystitis, as they
would require formal radiologic evaluation for diag-
noses and therefore beyond the scope of point-of-care
US (POCUS).
METHODS
Study Design
We conducted a systematic review of studies that
examined the operating test characteristics of the
modalities used by emergency physicians to diagnose
AC. The systematic review was conducting using the
Preferred Reporting Items for Systematic Review and
Meta-analyses (PRISMA) guidelines.13
Search Strategy
The design and manuscript structure of this systematic
review conform to the recommendations from the
Meta-analysis of Observational Studies in Epidemiol-
ogy (MOOSE)14 statement. In conjunction with a
medical librarian, six investigators independently
searched the medical literature from January 1965 to
March 2016 in PubMed, Embase, and SCOPUS for
the search terms diagnosis and cholecystitis. Diagnosis
was searched under MeSH headings diagnosis, diagno-
sis-related groups, delayed diagnosis, computer-assisted
diagnosis, early diagnosis, differential diagnosis,
immunologic tests, ultrasonography, laboratory tech-
niques and procedures, or radiography. Cholecystitis
was searched under MeSH headings blood, diagnosis,
epidemiology, etiology, history, microbiology, pathol-
ogy, physiology, physiopathology, radiography, ultra-
sonography, and cholecystitis. The two searches were
combined and limited by human subjects, adults, and
English language articles. The PubMed, Embase, and
SCOPUS searches were combined for the three sepa-
rate search topics: H&P, laboratory data, and US.
Studies were included if they recruited adult patients
in the ED who had a bedside emergency US. Studies
ACADEMIC EMERGENCY MEDICINE March 2017, Vol. 24, No. 3 www.aemj.org
were included only if the patient had the criterion
standard for final diagnosis, which was predetermined
to be pathology diagnosis or biliary scintography. Nar-
rative reviews, case reports, or studies focused on chil-
dren or therapy were not included.
Criteria for Considering Studies for This
Review
Types of Participants. We included studies that
recruited adult patients presenting to the ED with
abdominal pain or RUQ pain as their chief complaint.
Studies that recruited patients who presented to an
urgent care setting were excluded, as these patients are
substantively different than those who present to the
ED with true, emergent abdominal pain. Patients were
not excluded based on comorbidities.
Types of Index Tests. We included studies that
used H&P findings, laboratory tests, and US as index
tests for the diagnosis of AC. We only included
abdominal sonographic studies if performed and inter-
preted by ED physicians. While many EDs continue
to order formal USs, there is a growing number of
emergency physicians who are US trained and rely
solely on POCUS. Additionally, RUQ US is a core
emergency US application per ACEP15 guidelines. As
a result, many emergency physicians do in fact make
the diagnosis of AC based on POCUS. Some institu-
tions may continue to utilize formal US for a variety
of other reasons. Nevertheless, there is a substantial
amount of evidence that has shown POCUS to be just
as sensitive and specific as formal US for diagnosis of
AC.16 19 For this reason, formal US was excluded
from the analysis.
Types of Reference Standard. We included
studies that used as a reference standard a final diag-
nosis of AC based on pathologic finding of AC at sur-
gery or a positive biliary scintigraphy study.
Data Abstraction
Two or more authors for each index test category inde-
pendently selected articles from the combined
PubMed/Embase search for full text review (H&P =
734; laboratory data = 74; US = 492). Each reviewer
independently selected potentially eligible studies
before both authors agreed on the list of studies for
full text review. Differences in study selection were
resolved by consensus. Having read the methods sec-
tions of the full-text version of the studies potentially
283
eligible for inclusion, each author then applied the sta-
ted inclusion and exclusion criteria to determine
which studies to include in our systematic review. Dif-
ferences were resolved by consensus after discussion
and adjudication.
Data Analysis
Sensitivities, specificities, and LRs were calculated
based on constructed two-by-two tables for each
included study. To compute meta-analysis summary
estimates when more than one study assessed the
same index test, we combined test characteristic data
using a random-effects model with Meta-DiSc soft-
ware.20 Interstudy heterogeneity was assessed for
pooled estimates of sensitivity and specificity using the
DerSimonian-Laird random-effect model.21 Publication
bias was not assessed because of the questionable
validity of this approach in diagnostic meta-analyses.22
Quality Assessment
Two authors (NM, MS) used the Quality Assessment
Tool for Diagnostic Accuracy Studies (QUADAS-2)23
for systematic reviews to evaluate the overall quality of
evidence for the trials included. For the purposes of
this diagnostic systematic review, several considerations
were established a priori to assess the quality of indi-
vidual trials. The ideal patient population would be
those presenting to an ED with abdominal complaints
as mentioned.
The QUADAS-2 method assesses four categories of
study design, including patient selection, index test, ref-
erence standard, and flow and timing. Four domains
were assessed for biases. 1) Patient selectionWere
the patients enrolled at random or consecutively?
Were there inappropriate exclusions? Could the
patients included not be representative of all patients
presenting to the ED with a clinical picture of AC? 2)
Index testWas the history and physical examination
obtained without knowledge of the results of criterion
standard test for AC? Were thresholds for vital signs
pre-determined for the study? Is there concern that
the way the history and physical were obtained would
be different than done in clinical practice? 3) Refer-
ence standardWas the criterion standard test for
AC obtained on every patient in the study? Were the
radiologists blinded to the clinical findings? 4) Flow
and timingCould the order of how the history and
physical and criterion standard test for AC were
obtained and read have introduced bias? Studies
would be considered low risk of bias if all four
284 Jain et al. A SYSTEMATIC REVIEW OF ED ULTRASOUND FOR CHOLECYSTITIS DIAGNOSIS
domains were rated no bias. All included studies used
pathology as the reference standard, but if the refer-
ence standard was not clearly defined that portion of
QUADAS-2 was at high risk for bias. Similarly, if exe-
cution of the index test was not clearly defined, that
portion of QUADAS-2 would also be at high risk for
bias. Each category is accompanied by a set of yes or
no questions and answering no to any question places
that portion of the study design at high risk for bias.
An unweighted Cohens kappa was calculated to mea-
sure agreement. Statistical agreement between these
two reviewers was assessed via a kappa analysis using
SPSS (v21.0). Two of the authors (NM and MS) indi-
vidually rated the QUADAS-2 assessment with a
kappa of 0.87. A meeting was held between the two
QUADAS-2 raters and the third author (RS) who
adjudicated any differences in QUADAS-2 rating.
TestTreatment Threshold
The Pauker and Kassirer decision threshold model
was used to develop a treatment algorithm.24 This
method is based on considering six variables: false-
negative and false-positive proportions, sensitivity,
specificity, risk of a diagnostic test, risk of treatment,
and anticipated benefit of treatment. Estimates of these
variables were abstracted from our systematic review to
derive theoretical test and treatment thresholds for ED
patients with AC diagnosed via bedside emergency
US (Figure 2).
RESULTS
Description of Included Studies
The PubMed, Embase, and SCOPUS search identi-
fied 734 citations for H&P (see Figure 1A). For labo-
ratory studies, PubMed, Embase, and SCOPUS
identified a total of 74 articles (see Figure 1B). For
US, PubMed, Embase, and SCOPUS search identi-
fied a total of 492 articles (see Figure 1C). Reviewing
the bibliographies of the pertinent articles identified
no additional studies. We excluded all studies of
emphysematous and acalculous cholecystitis, which are
present with an indefinable risk factor such as multi-
system trauma, burns, chronic debilitation, total par-
enteral feeding, or immunosuppression, not typical of
usual ED presentation of AC.
We decided to remove all the retrospective
studies57,25,26 from our review. Although, the work
by Lijmer et al.27 has shown that retrospective
compared to prospective diagnostic studies, when cor-
rected for methodologic flaws, do not produce differ-
ent results, our retrospective studies all had significant
flaws. All of our retrospective studies57,25,26 had
issues related to reliability of their retrospectively
abstracted data. Gilbert et al.28 has defined eight crite-
ria: 1) training, 2) case selection, 3) definition of vari-
ables, 4) abstraction forms, 5) meetings, 6)
monitoring, 7) blinding and 8) testing of inter-rater
agreement of retrospective chart reviews to improve
accuracy and minimize inconsistencies in data acquisi-
tion. All of our retrospective studies57,25,26 failed to
document any of these methods to assure unbiased
data collection from their medical records. For these
reasons, all retrospective studies were excluded from
the final data analysis.
As detailed in Figure 1A (H&P, n = 3), Figure 1B
(laboratory studies, n = 1), and Figure 1C (US
(n = 4), a total of nine diagnostic studies were
included in our final analysis, respectively. A full
description of the reviewed studies, including the study
design, subject characteristics, variables assessed, crite-
rion standard, and AC prevalence is included in the
tables for each diagnostic modality: H&P (Table 1A),
laboratory studies (Table 1B), and US (Table 1C).
Prevalence
The combined population from the eight unique
cohorts included in this review was 1,990, of which
297 patients were diagnosed with AC. The weighted
prevalence of AC across all studies was 14.9% with a
range of 7% to 64%. Our reviewed studies using
RUQ pain or suspected AC (seven studies,8,10,2934
n = 657 patients) as their primary inclusion criteria
and a higher prevalence of AC of 31% (range =
10%64%) versus 7% (range = 7%46%) compared
to those with abdominal pain (Eskelinen et al.,35
n = 1,333 patients) as their primary inclusion criteria.
This clearly suggests a selection bias for AC in those
studies using RUQ pain10,29,30,32 or suspected AC
compared to studies including all generalized abdomi-
nal pain8,31,3335 as their inclusion criteria.
H&P
All of the three studies29,30,35 reviewed in the H&P
section used a prospective observational methodology.
Inclusion criteria were not uniform across our
reviewed studies. Acute abdominal pain was the inclu-
sion criteria for Eskelinen et al.,35where Schofield
et al.29 and Bednarz et al.30 included only those
ACADEMIC EMERGENCY MEDICINE March 2017, Vol. 24, No. 3 www.aemj.org 285

Figure 1. Consort diagram. (A) History and physical examination; (B) laboratory studies; (C) ultrasound studies.
286 Jain et al. A SYSTEMATIC REVIEW OF ED ULTRASOUND FOR CHOLECYSTITIS DIAGNOSIS

Figure 1. continued.

Table 1A
Description of Reviewed Studies for H&P

Criterion Prevalence,
Study Study Design Subject Characteristics Variables Assessed Standard % (95% CI)
Schoeld Prospective Inclusion: Temperature > 64 (5473)
et al., 198629 observational RUQ pain 37.5C Gallstones at
study Exclusion: Mass Surgery
Not stated Vomiting
Sample Size: 100
Mean Age: 52 yrs.
Range: 2277 yrs.
Gender: 70% (f)
Bednarz Prospective Inclusion: Temperature > Gallstones at 39 (2850)
et al., 198630 observational Suspected cholecystitis fever surgery (71%)
study Exclusion: Mass Clinical (29%)
Not stated Jaundice
Sample Size: 70 RUQ pain
Mean age: 56 y RUQ tenderness
Range: 1992 y RUQ rebound
Gender: 50% (female)
Eskelinen Prospective Inclusion: Temperature > Gallstones 7 (69)
et al., 199335 observational Acute abdominal pain 37.1C at surgery
study Less than 7 days duration Mass
Exclusion: Jaundice
Not stated RUQ pain
Sample size: 1,333 RUQ tenderness
Mean age: 38 y RUQ rebound
SD: 22.1 y Murphys sign
Gender: 52.3% (female)

H&P = history and physical examination; RUQ = right upper quadrant.

patients with RUQ pain or those suspected of AC.
This difference in inclusion criteria explains the signifi-
cantly higher AC prevalence (64 and 39%) in the lat-
ter two studies compared to Eskelinen et al.35 (7%).
Sample size also varied considerably between Bednarz
et al.30 (n = 70) and Eskelinen et al.35 (n = 1,333).
Ages of included patient ranged from means of 388 to
5630 yrs. All studies showed a female preponderance
of subjects from 50% in Bednarz et al.30 to Schofield
et al.29 with 70%.
ACADEMIC EMERGENCY MEDICINE March 2017, Vol. 24, No. 3 www.aemj.org
Table 1B
Description of Reviewed Studies for Laboratory Studies
Study Study Design Subject Characteristics
Eikman et al., 19758 Prospective Inclusion:
observational study Suspicion of cholecystitis
Exclusion:
Not stated
Sample size: 39
Mean age: 44 y
Range: 1582 y
Gender: 67% (female)
Of the H&P variables in Table 1A, an elevated
temperature (37.138.C) or fever was the most
common AC sign reported in all of reviewed studies.
All of our reviewed studies also contained mass as
an assessed variable. RUQ pain, tenderness, and
rebound were evaluated only by Bednarz et al.30 and
Eskelinen et al.35 Only Bednarz et al.30 tested jaundice
while vomiting was only investigated by Schofield
et al.29 Studies by Schofield et al.29 and Eskelinen
et al.35 used the finding of gallstones at surgery as
their criterion standard for AC, and the study by Bed-
narz et al.30 used both gallstones at surgery (79%) as
well as clinical definition (29%).
Test Characteristics of H&P for AC
From Table 2A, we found that the effects of only two
to three studies per risk factor coupled with the differ-
ences in study populations between generalized
abdominal pain and RUQ pain resulted in such
marked heterogeneity; that pooling of the data was not
adequate. We decided to report only point estimates
and not pooled data for any of our test characteristics.
We grouped the variable fever for the three reviewed
studies together even though the study by Bednarz
et al.30 did not specify a specific temperature and the
other two studies reported very similar cutoffs
of >37.5C29 and >37.1C.35 We found that fever
had very poor test characteristics with sensitivities
between 31%29 and 62%,35 specificities of 37%30 to
74%,29 positive LR (LR+) of 0.7130 to 1.24,35 and
negative LR (LR) of 0.7635 to 1.49.30 Between the
Bednarz et al.30 and Eskelinen et al.35 studies, marked
heterogeneity was noted for the variables RUQ pain,
mass, and tenderness, most likely secondary to biases
documented in our QUADAS-2 analysis. Even with
these biases none of characteristics of RUQ mass,
pain, or tenderness or the other risks such as RUQ
rebound, jaundice, Murphys sign, or vomiting had
287
Prevalence,
Variables Assessed Criterion Standard % (95% CI)
Total bilirubin Gallstones at surgery 26 (1441)
sufficiently low LRs to significantly decrease the
probability of AC.
QUADAS-2 Analysis for H&P for AC
All of our reviewers agreed 100% on the QUADAS-2
scoring for the three H&P studies we reviewed (Table
3). All reviewers found all three studies to have high risks
of bias in reporting H&P test characteristics. The study
by Eskelinen et al.35 suffered from differential verifica-
tion bias. Differential verification bias, also called double
criterion standard bias, as described by Kohn et al.,36
occurs when the results of the index test determine differ-
ent gold standards. Since the study by Eskelinen et al.35
was a study of all patients with abdominal pain, those
patients who tested positive for any of the index tests
commonly associated with AC such as RUQ findings
(mass, pain, tenderness), jaundice, or Murphys sign
were all preferentially tested for the criterion standard for
AC. Those patients without any of these signs of AC
were then tested for other differential diagnoses of acute
abdominal pain, utilizing different criterion standard
tests. Only 10.1% of the study population of Eskelinen
et al.35 had AC, while 30.2% had acute appendicitis
with 40% diagnosed as nonspecific abdominal pain with
remainder having nephrolithiasis, dyspepsia, small
bowel obstruction, and other less common etiologies of
abdominal pain. Differential verification bias in the case
of the study by Eskelinen et al.35 significantly falsely
raised the specificity to greater than 95% for RUQ
(mass, pain, tenderness), jaundice, and Murphys sign,
with a smaller effect on sensitivity with result of inflating
the LR+ (>13) for all these test characteristics.
When we examine the test characteristic of RUQ
pain in Table 2A, we see an excellent example of par-
tial verification bias in the study of Bednarz et al.30
Also as described by Kohn et al.36 partial verification
bias, verification bias, referral, ascertainment, or
workup bias occurs when patients positive for the
288 Jain et al. A SYSTEMATIC REVIEW OF ED ULTRASOUND FOR CHOLECYSTITIS DIAGNOSIS

Table 1C
Description of Reviewed Studies for US

Experience of Criterion Prevalence,

Study Study Design Subject Characteristics Sonographers Standard % (95% CI)
Kendall and Prospective Inclusion: Image acquisition: Surgical 10 (617)
Shimp, 200132 observational RUQ pain Full-time ED physicians pathology
study Epigastric pain 2-h didactics
Jaundice 3 h. supervised
History of stones Image interpretation:
Exclusion: As above
Non-ED visit
Ascites
HIV
Sample size: 109
Mean age: 39 y
Range: 1688 y
Gender: 79% (female)
Rosen et al., 200133 Prospective Inclusion: Image acquisition: Surgical 46 (357)
observational Suspected cholecystitis 52% < 25 previous pathology
study Radiology US RUQ US
Age > 18 y Image interpretation:
Exclusion: As above
No radiology US
Sample size: 116
Mean age: 49 y
Range: not stated
Gender: 72% (female)
Summers et al., 201010 Prospective Inclusion: Image acquisition: Surgical 14 (920)
observational Suspected cholecystitis Variable experience pathology
study Age > 18 y (ED physicians, attendings,
Exclusion: residents, and fellows
No follow-up, pathology Image interpretation:
Presentation 12AM8AM As above
Sample size: 193
Median age: 36 y
Range: 1887 y
Gender: 73% (female)
Noble et al., 201034 Prospective Inclusion: Image acquisition: Surgical 37 (2256)
observational Suspected cholecystitis Not stated pathology
study Age > 18 y Image interpretation:
Exclusion: As above
No RDMS EP on shift
Sample Size: 30
Mean age: not stated
Range: not stated
Gender: not stated

EP = emergency physician; RDMS = registered diagnostic medical sonographer; RUQ = right upper quadrant; US = ultrasound.

index test (RUQ pain) are more likely (studys inclu-
sion criteria) to meet the criterion standard test (hepa-
tobiliary scintigraphy) than patients without RUQ pain
who would not be included in the study by Bednarz
et al.30 This causes the sensitivity to be falsely raised
(93%) while deceptively depressing the specificity (0%).
Laboratory Findings
Only the study by Eikman et al.8 all met our inclusion
criteria for review in the laboratory findings section in
Table 1B. Studies by Brewer et al.31 and Potts et al.7
were both retrospective studies and were excluded
from our review for the already stated reason for fail-
ure to abide by the quality criteria established by Gil-
bert et al.28 In addition, the study by Potts et al.7 only
enrolled patients > 80 years old, significantly limiting
the generalizability of their observations.
Eikman et al.8 included patients with abdominal
pain suspected of AC and were tested by hepatobiliary
scintigraphy. Although the main focus of the study
Eikman et al.8 was hepatobiliary scintigraphy, with suf-
ficient data on AC to permit evaluation of the test
characteristic of total bilirubin. Eikman et al.8 defined
their criterion standard by the presence of gallstones at
surgery, with a AC prevalence of 26%.
Test Characteristics of Laboratory Findings
for AC
From Table 2B, we only report single study test char-
acteristics for total bilirubin from the study by Eikman
ACADEMIC EMERGENCY MEDICINE March 2017, Vol. 24, No. 3 www.aemj.org 289

Table 2A
Test Characteristics of H&P for AC

Sensitivity, Specicity, % Positive Negative

H&P Study % (95% CI) (95% CI) LR (95% CI) LR (95% CI)
Fever or temperature Bednarz et al., 198630 44 (2665) 37 (2353) 0.71 (0.441.14) 1.49 (0.892.50)
> 37.1 or > 37.5C Schoeld et al., 198629 31 (2044) 74 (5687) 1.18 (0.612.30) 0.94 (0.721.21)
Eskelinen et al., 199335 62 (5371) 50 (4753) 1.24 (1.071.44) 0.76 (0.600.96)
RUQ mass Bednarz et al., 198630 19 (638) 72 (5685) 0.66 (0.261.68) 1.13 (0.871.46)
Schoeld et al., 198629 15 (726) 83 (6794) 0.87 (0.342.25) 1.03 (0.861.23)
Eskelinen et al., 199335 16 (1024) 99 (98100) 16.25 (8.1432.44) 0.85 (0.780.92)
RUQ pain Bednarz et al., 198630 93 (7699) 0 (08) 0.92 (0.821.04) 7.86 (0.39157.69)
Eskelinen et al., 199335 56 (4765) 96 (9597) 14.02 (10.1919.28) 0.46 (0.380.56)
RUQ tenderness Bednarz et al., 198630 37 (1958) 58 (4273) 0.89 (0.481.62) 1.08 (0.741.59
Eskelinen et al., 199335 75 (6682) 95 (9496) 15.11 (11.5719.73) 0.26 (0.190.35)
RUQ rebound Bednarz et al., 198630 37 (1958) 58 (4273) 0.89 (0.481.62) 1.08 (0.721.59)
Eskelinen et al., 199335 42 (3450) 53 (5056) 0.89 (0.721.09) 1.10 (0.951.28)
Jaundice Bednarz et al., 198630 11 (229) 86 (7295) 0.80 (0.222.92) 1.03 (0.861.27)
Eskelinen et al., 199329 14 (821) 99 (98100) 13.81(6.7528.25) 0.87 (0.810.94)
Murphys sign Eskelinen et al., 199335 62 (5371) 96 (9597) 15.64 (11.4821.31) 0.40 (0.320.50)
29
Vomiting Schoeld et al., 1986 83 (7191) 56 (3872) 1.86 (1.272.73) 0.31 (0.170.57)

AC = acute cholecystitis; H&P = history and physical examination.

Table 2B
Test Characteristics of Laboratory Tests for AC

Laboratory Test Study Sensitivity, % (95% CI) Specicity, % (95% CI) Positive LR (95% CI) Negative LR (95% CI)
Bilirubin Eikman 19758 40 (1274) 93 (7799) 5.80 (1.2526.99) 0.64 (0.391.08)

AC = acute cholecystitis; LR = likelihood ratio.

Table 2C
Operating Characteristics Bedside Sonography for AC

Risk Factor Sensitivity, % (95% CI) Specicity, % (95% CI) Positive LR (95% CI) Negative LR (95% CI)
33
Rosen 2001 91 (7798) 66 (4980) 2.68(1.734.15) 0.13 (0.040.39)
Kendall 200132 82 (4898) 85 (8089) 4.42 (3.687.98) 0.21 (0.060.75)
Summers 201010 87 (6697) 82 (7488) 4.72 (3.226.91) 0.16 (0.060.46)
Noble 201034 82 (4898) 95(74100) 15.55 (2.26106.88) 0.19 (0.060.68)

AC = acute cholecystitis; LR = likelihood ratio.

et al.8 The presence of an elevated bilirubin not sur-
prisingly increased (LR+ = 5.80) the probability of
AC. Elevated bilirubin was defined as greater than
2.0 mg/100 mL. Total bilirubin was not sufficiently
robust to significantly decrease (LR = 0.64) the prob-
ability of AC.
QUADAS-2 Analysis for Laboratory Findings
for AC
All of our reviewers agreed 100% on the QUADAS-2
scoring for the single laboratory study we reviewed
(Table 3B). All reviewers found the study by Eikman
et al.8 to have high risks of bias secondary to partial
verification bias. Since the study by Eikman et al.8
included only those patients suspected of AC, the
index test (an elevated total bilirubin) we chose to
review was probably used as an inclusion criterion
defining at least a subset of patients suspected of AC.
If the index test increased the probability of patients
chosen to receive the diagnostic test (hepatobiliary
scintigraphy) then the study is at risk for partial verifi-
cation bias. As was the case in the study by Bednarz
290 Jain et al. A SYSTEMATIC REVIEW OF ED ULTRASOUND FOR CHOLECYSTITIS DIAGNOSIS

Table 3A
QUADAS: H&P

Item Schoeld 198629 Bednarz 198630 Eskelinen 199335

1. Was the spectrum of patients described in the Yes Yes Yes
paper and was it chosen adequately?
2. Were selection criteria described clearly? Yes Yes Yes
3. Is the reference standard likely to classify the target condition? Yes Yes Yes
4. Was there an abnormally long time period between the performance Yes, 48 h Yes, unknown No, unknown
of the test under evaluation and the conrmation of the prospect
diagnosis with the reference standard?
5. Did the whole sample, or a random selection of the sample, Yes, whole Yes, whole Yes, whole
receive verication using a reference standard of diagnosis?
6. Did all patients receive the same reference standard Yes Yes Yes
regardless of the index test result?
7. Were the results of the index test incorporated in Yes Yes Yes
the results of the reference standard?
8. Was the execution of the index test described in sufcient Yes Yes Yes
detail to permit replication of the test?
9. Was the execution of the reference standard described in sufcient Yes Yes Yes
detail to permit replication of the test?
10. Were the index test results interpreted blind Yes Yes Yes
to the results of the reference standard?
11. Were the reference standard results interpreted No No No
blind to the results of the index test?
12. Were clinical data available when test results were interpreted? Yes Yes Yes
13. Were uninterpretable/indeterminate/intermediate results reported No, not described Yes No
and included in the results?
14. Were reasons for dropout from the study reported? No, all admitted No No

H&P = history and physical examination; QUADAS = Quality Assessment Tool for Diagnostic Accuracy Studies.

Table 3B
QUADAS: Laboratory Data

Item Eikman 19758

1. Was the spectrum of patients described in the paper Yes
and was it chosen adequately?
2. Were selection criteria described clearly? No, vague description
3. Is the reference standard likely to classify the target condition? Yes
4. Was there an abnormally long time period between the Yes, 72 h
performance of the test under evaluation and the conrmation
of the diagnosis with the reference standard?
5. Did the whole sample, or a random selection of the sample, Yes, whole
receive verication using a reference standard of diagnosis?
6. Did all patients receive the same reference standard regardless of the index test result? Yes
7. Were the results of the index test incorporated in the results of the reference standard? Yes
8. Was the execution of the index test described in sufcient detail to permit replication of the test? No
9. Was the execution of the reference standard described in Yes
sufcient detail to permit replication of the test?
10. Were the index test results interpreted blind to the results of the reference standard? No
11. Were the reference standard results interpreted blind to the results of the index test? Yes
12. Were clinical data available when test results were interpreted? Yes
13. Were uninterpretable/indeterminate/intermediate results reported and included in the results? Yes
14. Were reasons for dropout from the study reported? No

QUADAS = Quality Assessment Tool for Diagnostic Accuracy Studies.

et al.30 in the H&P group who used RUQ pain as
both an index test of AC and a study inclusion crite-
rion.
ED Bedside US
Four studies, by Kendall and Shimp,32 Rosen et al.,33
Summers et al.,10 and Noble et al.,34 met our final
inclusion criteria for review. Blaivas and Adhikari9
was removed from final analysis due to concerns of
potential biases related to data entry as per the require-
ments set forth by Gilbert et al.28 Villar et al.37 used
the same data set as Summers et al.10 and therefore
was excluded from the final reviewed studies. Jang
et al.19 was excluded due to significant overlap of data
point and lack of reproducibility of the data. Studies
by Schlager et al.38 and Jehle et al.39 were both
ACADEMIC EMERGENCY MEDICINE March 2017, Vol. 24, No. 3 www.aemj.org 291

Table 3C
QUADAS: US

Summers Kendall
Item Noble 201034 Rosen 200133 201010 200132
1. Was the spectrum of patients described in the Yes Yes Yes Yes
paper and was it chosen adequately?
2. Were selection criteria described clearly? Yes Yes Yes Yes
3. Is the reference standard likely to classify the target condition? Yes Yes Yes Yes
4. Was there an abnormally long time period between the No No No No
performance of the test under evaluation and the conrmation of the
diagnosis with the reference standard?
5. Did the whole sample, or a random selection of the sample, receive Yes, whole Yes, whole Yes, whole Yes, whole
verication using a reference standard of diagnosis?
6. Did all patients receive the same reference Yes Yes Yes Yes
standard regardless of the index test result?
7. Were the results of the index test incorporated No No No No
in the results of the reference standard?
8. Was the execution of the index test described in Yes Yes Yes Yes
sufcient detail to permit replication of the test?
9. Was the execution of the reference standard described in Yes Yes Yes Yes
sufcient detail to permit replication of the test?
10. Were the index test results interpreted blind to Yes Yes Yes Yes
the results of the reference standard?
11. Were the reference standard results interpreted blind Yes Yes Yes Yes
to the results of the index test?
12. Were clinical data available when test results were interpreted? No No No No
13. Were uninterpretable/indeterminate/intermediate results No No Yes No
reported and included in the results?
14. Were reasons for dropout from the study reported? No Yes Yes Yes

QUADAS = Quality Assessment Tool for Diagnostic Accuracy Studies; US = ultrasound.

removed because they both did not differentiate the
findings of gallstones from AC. All the reviewed US
studies were prospective. All the reviewed studies used
RUQ pain or suspected AC. In addition, Kendall and
Shimp32 obtained RUQ USs in patients with epigas-
tric pains, history of stones, and jaundice. Exclusion
criteria were only defined by Kendall and Shimp32 as
any patient with ascites or HIV+. Sample sizes varied
from 3034 to 19310 subjects, with mean ages ranging
from 3610 to 4933 years. Female gender predominated
in all studies from 79%32 to 72%.33
The experience of the ED sonographers varied
greatly across the studies. Only Kendall and Shimp32
detailed the training of the ED sonographers used for
their study. Rosen et al.33 documented that at least
48% of their ED sonographers in their study had
greater than 25 previous RUQ USs before the study.
The studies by Summers et al.10 and Noble et al.34
failed to document the experience of their ED sonog-
raphers. All studies used the same sonographer to
both acquire and interpret their studies. None of the
studies utilized overreads by more experienced ED
ultrasonographers or radiologists. None of the studies
reported intra- or inter-rater reliability of their ED or
radiology USs.
Test Characteristics of ED Bedside
Sonography for AC
Marked heterogeneity (chi-square p > 0.05, I2 = 75%)
between the test characteristics precluded us from
reporting pooled results. From Table 2C, we noted a
range of sensitivity for AC between (82%91%), speci-
ficity (66%95%), LR+ (2.6815.55), and LR (0.13
0.21). The study by Noble et al.34 provided for a sub-
stantial degree of heterogeneity between test character-
istics for sonography. Although Noble et al.34 used
RUQ pain similar to the other reviewed studies, the
primary hypothesis of this study was not the operating
characteristics of sonography, but was a comparison of
the sonographic Murphy sign (SMS) before and after
analgesia. Since Noble et al.34 only studied SMS and
not the whole range of the sonographic findings (wall
thickening, pericholecystic fluid, sludge, etc.) associated
with AC, it is not surprising their test characteristics
are outliers. A sensitivity analysis of the operating char-
acteristics after removing the Nobel et al.34 still
showed significant heterogeneity for specificity (chi-
square p = 0.20, I2 = 74%; range = 66%84%) and
LR+ (chi-square p = 0.24, I2 = 73%; range = 2.7
5.4). After excluding Nobel et al.,34 heterogeneity for
sensitivity and LR significantly decreased allowing
292 Jain et al. A SYSTEMATIC REVIEW OF ED ULTRASOUND FOR CHOLECYSTITIS DIAGNOSIS

Ttesting threshold = [(Ppos/nd) x (Rrx) + Rt] [(Ppos/nd x Rrx) + (Ppos/d x Brx)] = U/s (4%), MRI (4%),
HIDA (2%)
Ttreatment threshold = [(Pneg/nd) x (Rrx) - Rt] [(Pneg/nd x Rrx) + (Pneg/d x Brx)] = U/s (4%), MRI (52%),
HIDA (74%)

Where assumptions are based upon the summary estimates for probability treatments for cholecystitis*

Ppos/nd = probability of a positive result in patients without disease = 1-speciicity = 1-0.90= 0.10 (HIDA) 0.18
(MRI) 0.19 (U/s)*
Pneg/nd = probability of a negative result in patients without disease = speciicity = 0.90 (HIDA) 0.82 (MRI) 0.81
(U/s)*

Rrx = risk of treatment in patients without disease = 0.168

Rt = risk of diagnostic test = 0.0

Ppos/d = probability of a positive result in patients with disease = sensitivity = 0.94 (HIDA) 0.86 (MRI) 0.82
(U/s)*

Pneg/d = probability of a negative result in patients with disease = 1 sensitivity = 1-0.57= 0.06 (HIDA) 0.14
(MRI) 0.18 (U/s)*

Brx = beneit of treatment in patients with disease = 0.90

* Keiwiet et al

Figure 2. Testtreatment threshold formulas. *Kiewiet et al.40 AC = acute cholecystitis; HIDA = hepatobiliary iminodiacetic acid scan/cho-
lescintigraphy; MRI = magnetic resonance imaging; U/s = ultrasound.

pooling of sensitivity (chi-squared p = 0.67, I2 = 0%;
pooled = 88%; range = 75%95%) and LR (chi-
square p = 0.84, I2 = 0%; pooled = 0.16; range =
0.080.31).
QUADAS-2 Analysis for ED Bedside
Sonography for AC
All of our reviewers agreed 100% on the QUADAS-2
scoring for the four ED US studies we reviewed (Table
3C). The study by Summers et al.10 may also have
been influenced by partial verification bias, in that 189
ED USs were studied but only 125 of these patients
were referred to radiology US. In the study by Sum-
mers et al.10 the results of the index test (ED US) par-
tially determined workup (radiology US; partial
verification bias). Of the 189 ED US examinations, 26
went to the operating room (24 positive for AC) and
163 were discharged to telephone follow up, of whom
23 patients were unable to be contacted. Since we do
not know the prevalence of AC in the patients lost to
follow-up, the potential exists for follow-up bias. Partial
verification bias may have also affected the study by
Noble et al.,34 which only examined SMS, so positive
SMS patients may have been preferentially referred to
radiology US.
In the study by Rosen et al.,33 of the 116 patients
enrolled in the ED US study, 40 (34%) were excluded
from further analysis if the ED US had discordant
results either positive for gallstones and a negative
SMS or vice versa. Since the results of the index test
partially determined the etiology of findings requiring
a continued workup, this study also was exposed to
ACADEMIC EMERGENCY MEDICINE March 2017, Vol. 24, No. 3 www.aemj.org
partial verification bias. Partial verification was not an
issue with the study by and Kendall and Shimp32 as it
appears that all patients in both studies who received
a ED US also received a radiology US.
TestTreatment Threshold Estimates
The testtreatment threshold model we developed was
designed to aid physicians in efficiently and accurately
ruling in or ruling out the diagnosis AC. We built this
model to investigate which element(s) of the H&P, labo-
ratory findings, and ED US had sufficient discrimina-
tory power given the pretest probability of AC to
obviate a formal test by radiology: an official US, MRI,
or HIDA scan. Given that these formal radiology
department tests are not available 24 hours per day, 7
days a week, if element(s) of the H&P, laboratory stud-
ies, or ED US were sufficiently robust to rule in or rule
out AC, patient disposition to either discharge or begin
empiric AC therapy would be facilitated.
In Figure 2, we created three testtreatment thresh-
old models (1, radiology US; 2, MRI; and 3, HIDA)
to diagnose AC. The top half of the Figure 2
describes the variables and calculations used to pro-
duce the test and treatment thresholds depicted in gra-
phic below for each the three formal diagnostic tests
(1, radiology US; 2, MRI; 3, HIDA) for AC. For each
diagnostic modality a separate set of test and treatment
thresholds were defined across a range 0% to 100%.
Based on the work of Pauker and Kassirer24 these
three models utilized the unique operating characteris-
tics of each diagnostic modality while controlling for
the risks of treatment of patients without AC (Rrx),
the risk of the diagnostic test (Rt), and the benefit of
treatment with AC (Brx).
For each diagnostic modality we used individual
operating characteristics coupled with the variables
Rrx, Rt, and Brx to develop unique testtreatment
thresholds. Since the operating characteristics for the
three diagnostic modalities differ, their testtreatment
thresholds also vary.
The test thresholds are depicted as the left-most
open arrow for each diagnostic modality (radiology US
and MRI 4%, nuclear 2%). If the posttest probability
of the index test we choose from our systematic review
(dashed vertical line) falls to the left the testthreshold
than further testing for AC is not warranted and an
alternative diagnosis other than AC should be consid-
ered. The treatment thresholds are represented by the
right-most open arrow for each diagnostic modality (ra-
diology US 46%, MRI 52%, HIDA 74%). When the
293
posttest probability of the index test is to the right of
the treatment threshold, further testing is unnecessary
and treatment should be initiated based only on the
results of the index test. If the posttest probability of
the index test falls between the test and treatment
thresholds, then our analysis recommends continued
testing with that diagnostic modality for AC. The hori-
zontal dashed lines represent the ranges of posttest
probabilities of a positive or negative ED US consis-
tent with AC.
Our three models for the testtreatment thresholds
for diagnosing/treating AC are based on the individ-
ual operating characteristics of radiology US, MRI,
and nuclear that were recently documented in a sys-
tematic review by Kiewiet et al.40 They reviewed 57
studies encompassing 5,859 patients and found HIDA
with the highest sensitivity (94%) and specificity
(90%), followed by MRI sensitivity (86%) and speci-
ficity (82%) and radiology US sensitivity (82%)/speci-
ficity (81%).
The definitive treatment, i.e., operative manage-
ment, is always delayed while obtaining these radiol-
ogy-based studies and delayed further during off-hours
when these tests are not routinely available. While AC
is an inflammatory process, secondary infection can
occur due to cystic duct obstruction and bile stasis,
leading to gangrene, sepsis, and gallbladder perfora-
tion.1,2,57,41,42 The possibility of coexistent bacterial
infection in AC is the justification for starting empiric
antibiotics especially if delays of definitive diagnosis
and surgical management are expected. Empiric antibi-
otic therapy should include activity against the most
frequently associated pathogens: Escherichia coli, Entero-
coccus, and Klebsiella.41 We defined the risks of treat-
ment without disease (Rrx) as the risks of antibiotics
for presumed AC whom after radiology-based tests
rules out AC. Patients without AC but exposed to
antibiotics would suffer all the risks of antibiotics with-
out any of their intended benefits. We will define
(Rrx) as the overall risk of adverse drug reaction is
16.8%.42
We judged the risk of the diagnostic tests (Rt) US,
MRI, and HIDA as zero. Finally, the benefit of treat-
ment (Brx) of patients with AC has never been, nor
ever will be, tested by a randomized double-blinded
placebo controlled methodology; it would be unethical
to study the spontaneous recovery rate of AC without
antibiotics or surgery. Without evidence-based studies
we used a conservative estimate of the benefit from
treating AC (Brx) = 0.90.
294 Jain et al. A SYSTEMATIC REVIEW OF ED ULTRASOUND FOR CHOLECYSTITIS DIAGNOSIS
Of the H&P, laboratory, and US characteristics in
this review, only ED US resulted in a significant
change in LR of positive test ranged from (LR+ =
2.684.72) or a negative US (LR = 0.130.21) after
excluding Noble et al.,34 for reasons explained. We
defined the pretest probability of AC by the weighted
prevalence (31%) across the reviewed studies of using
RUQ as entry criteria. Applying Bayes theorem of a
pretest probability of AC = 31% with LR+ of 2.68
4.72 and LR of 0.130.21 would lead to a range of
posttest probabilities after a positive ED US for AC
(55%68%) or a negative US (6%7%).
In Figure 2, in the case of a negative ED US for AC
the posttest probability would only decrease from 31% to
6%7%; our analysis would recommend further testing
for all three diagnostic modalities (test thresholds: radiol-
ogy US = 4%, MRI = 4%, and HIDA = 2%). Thus, a
negative ED US is not sufficient to adequately rule out
AC and continued testing is required before discharge of
a patient suspected of AC. If the ED US is positive for
AC, the posttest probability of AC would increase from
31% to 55%68% and obviate further testing for the
diagnostic modalities of radiology US (treatment thresh-
old = 46%) and MRI (treatment threshold = 52%). The
nuclear scan model, because of higher sensitivity (90%)
and specificity (94%) compared to the other two modali-
ties, recommends further testing (treatment threshold =
74%) even if the ED US was positive for AC.
DISCUSSION
This systematic review examines the utility of H&P,
laboratory studies, and ED US for the diagnosis of
AC in the ED population. We found eight studies
that met inclusion criteria for H&P (n = 3), for labo-
ratory studies (n = 1), and for ED US (n = 5) with
varying quality based on the degree of bias.
Similar to the findings by the rational clinical exami-
nation of Trowbridge et al.,43 Does This Patient Have
Acute Cholecystitis? we found with an updated search
that no single H&P finding or laboratory test was suffi-
ciently robust to rule out AC. This is not to suggest that
H&P and laboratory findings are of no utility in the
diagnosis of AC, as the presence of these variables is
necessary to define the at-risk population in the first
place. Ross et al.18 compared ED US to other modali-
ties if imaging for cholelithiasis; however, H&P and lab-
oratory were not included in the primary review.
Observational studies of AC patients from Copes
Early Diagnosis of the Acute Abdomen44 in 1921 to the
most recent (2013) update of the Tokyo AC guideli-
nes12 help us form our clinical gestalt of an AC
patient. Unfortunately, these observational studies by
their very nature suffer from verification bias, severely
limiting their potential to find characteristics of
patients suspected of AC, which can substantially rule
out AC. While the index tests for AC (RUQ pain/
mass/tenderness/Murphys sign/fever/WBC) are very
common in AC patients, we only understand their
true clinical utility when we examine the prevalence of
AC in patients without these cardinal signs of AC. In
our study, LR for H&P ranged from 0.26 to 7.86,
and LR for an elevated WBC was 0.64, meaning
that we were not able to use any of these findings to
significantly decrease the posttest probability of AC
and obviate further formal radiology testing.
From our ED US studies reviewed we were able to
find studies with LRs (LR+s of 2.68 to 4.72) and
(LRs of 0.13 to 0.21) significantly robust enough to
produce marked changes from pre- to posttest proba-
bilities of AC. In comparing the performances of radi-
ology to ED US, our reviewed studies found disparate
results. Summers et al.10 found no significant differ-
ences in the operating characteristics of formal radiol-
ogy US (LR+ = 5.7 [range = 3.39.8]; LR = 0.20
[range = 90.080.5]) compared to ED US (LR+ = 4.7
[range = 3.26.9]; LR = 0.16 [range = 0.060.46])
in detecting AC. Yet, the study by Rosen et al.33 only
found a kappa of 0.46 for the AC agreement of ED
and formal radiology US. Kendall and Shimp32 only
directly compared the ED US to formal radiology US
for the detection of SMS and found the ED US (sen-
sitivity = 82%) greater than the formal US (45%).
Some of this heterogeneity in comparing the accuracy
between radiology and ED US can be ascribed to the
variations in the sonographic experience across the
various studies. In addition, none of these studies
included tests of inter-rater or intra-rater reliability of
the ED sonographers.
Using our testtreatment threshold estimates (Fig-
ure 2) with a pretest probability of AC (weighted AC
prevalence) of 31% and a positive ED US for AC an
ED physician could initiate empiric AC antibiotics
and a surgery consultation obtained for admission,
without the need for formal radiology department test-
ing with US or MRI scans. A negative ED US would
only decrease the posttest AC probability to 6%7%,
still just above the testing threshold (US and MRI
4%, HIDA 2%) for the three formal diagnostic modal-
ities. The decision not to go onto further formal AC
ACADEMIC EMERGENCY MEDICINE March 2017, Vol. 24, No. 3 www.aemj.org
testing will be a clinical decision based on your own
clinical judgment of either accepting a miss rate of AC
between 6 and 7% or applying a lower pretest proba-
bility of AC than 31% maybe commensurate with
your clinical experience. A change in the estimate of
the pretest probability of AC to 25% would place a
negative ED US posttest probability of AC (4%) at the
cutoff for the test threshold for both radiology US and
MRI (4%). If the pretest probability is as low as 15%,
then a negative ED US could hit the lower limit of
the test threshold (2%) for even the HIDA scan. On
the other hand, considering that ED US and radiology
US are equally accurate, as described, since the postt-
est probability of an ED US is 6%7% and the test
treatment threshold for all three modalities are 4%,
the difference between 6 and 4% is insignificant
enough to obviate further testing as per clinical judg-
ment, especially in the absence of an elevated biliru-
bin. In this case, perhaps an alternate diagnosis
should be pursued.
With an increasing movement toward early interven-
tion for patients with AC, the utility of ED US for early
diagnosis will aid to avoid unnecessary delay in inter-
vention, as well as reduction of outpatient referrals and
repeated ED visits, which overall is more cost-effective
and improve patient suffering and experience.45
Implications for Future Research
Characteristics of disease have not been studied in the
current era. Lifestyles and food habits have changed in
the past 40 years. All of studies in the H&P and labo-
ratory section were done before 2000. None of the
studies evaluated the discriminatory power of different
historical, physical examination, and laboratory find-
ings in combination. High-quality diagnostic studies of
AC in the future should combine H&P, laboratory
data, and ED US with specific concerns for limiting
bias and directly measuring the efficiencies obtained
by ED POCUS, that is, the decreased time to achieve
diagnosis. The outcome test should not be based on
the index test. Additionally, future studies should
report inter-rater reliability of the ED US interpreta-
tions given that US is a user-dependent modality.
Combining features may lead to better operating char-
acteristics and should be considered in future studies.
LIMITATIONS
Only three databases, PubMed, Embase, and SCOPUS,
were used. Also notable is the paucity of H&P and
295
laboratory studies. Despite a 4:1 female predisposition to
AC, only one study by Irvin et al.3 examined this
parameter. These studies did not include other factors
that predispose patients to AC, i.e., race and diabetes.
When examining many other parameters included in
this study, there are few studies per parameter. For exam-
ple, Singer et al.25 is the only study that met inclusion
criteria and took history of gallstones into account. Fur-
thermore, these studies are over 20 years old and may
no longer represent current lifestyles and risk factors.
The diagnostic tools are associated with limitations as
well. While the risk of diagnostic tests such as US,
MRI, and HIDA are estimated as zero, these tests do
have some quantifiable risk. MRI and HIDA require
patients to leave the ED, which holds some risk as well.
While not a direct or immediate risk, oftentimes imag-
ing studies find results not expected with current investi-
gation purposes; these incidentalomas may lead to
further testing and expose patients to further risk.
Ultrasound is a user-dependent modality. Many
studies that include USs as diagnostic criteria include
USs predominantly performed by emergency physi-
cians who are very experienced and motivated to per-
form ultrasonography. This does not represent actual
practice where bedside US is performed by the practi-
tioner caring for the patient, who may or may not be
a registered sonographer.
CONCLUSION
Right upper quadrant complaints are a frequently
encountered complaint among ED visits. The decision
to pursue and treat these complaints is multifactorial.
The primary objective of this systematic review is to
assess variables of history and physical examination,
laboratory testing, and ultrasound impact on making
the diagnosis of acute cholecystitis. We found that
there is no one parameter that would suggest an
immediate cause to treat and pursue surgical interven-
tion. Individually, the many variations of complaints
as well as clinical findings do not reliably rule out
acute cholecystitis, thereby necessitating a clinical deci-
sion rule to include multiple parameters (history and
physical, laboratory data, and sonographic imaging) to
achieve a correct diagnosis of acute cholecystitis.
References
1. Hastings RS, Powers RD. Abdominal pain in the ED: a
35 year retrospective. Am J Emerg Med 2011;29:7116.
296 Jain et al. A SYSTEMATIC REVIEW OF ED ULTRASOUND FOR CHOLECYSTITIS DIAGNOSIS
2. Miettinen P, Pasanen P, Lahtinen J, Alhava E. Acute abdom-
inal pain in adults. Ann Chir Gynaecol 1996;85:59.
3. Irvin TT. Abdominal pain: a surgical audit of 1190 emer-
gency admissions. Br J Surg 1989;76:11215.
4. Udekwu PO, Sullivan WG. Contemporary experience with
cholecystectomy: establishing benchmarks two decades
after the introduction of laparoscopic cholecystectomy. Am
Surg 2013;79:12537.
5. Adedeji OA, McAdam WA. Murphys sign, acute chole-
cystitis and elderly people. J R Coll Surg Edinb
1996;41:889.
6. Navarro Fernandez JA, Tarraga Lopez PJ, Rodriguez Montes
JA, Lopez Cara MA. Validity of tests performed to diagnose
acute abdominal pain in patients admitted at an emergency
department. Rev Esp Enferm Dig 2009;101:6108.
7. Potts FE 4th, Vukov LF. Utility of fever and leukocytosis
in acute surgical abdomens in octogenarians and beyond.
J Gerontol A Biol Sci Med Sci 1999;54:M558.
8. Eikman EA, Cameron JL, Colman M, Natarajan TK, Dugal
P, Wagner HN Jr. A test for patency of the cystic duct in
acute cholecystitis. Ann Intern Med 1975;82:31822.
9. Blaivas M, Adhikari S. Diagnostic utility of cholescintigra-
phy in emergency department patients with suspected
acute cholecystitis: comparison with bedside RUQ ultra-
sonography. J Emerg Med 2007;33:4752.
10. Summers SM, Scruggs W, Menchine MD, et al. A
prospective evaluation of emergency department bedside
ultrasonography for the detection of acute cholecystitis.
Ann Emerg Med 2010;56:11422.
11. Hirota M, Takada T, Kawarada Y, et al. Diagnostic criteria
and severity assessment of acute cholecystitis: Tokyo
Guidelines. J Hepatobiliary Pancreat Surg 2007;14:7882.
12. Kiriyama S, Takada T, Strasberg SM, et al. TG13 guideli-
nes for diagnosis and severity grading of acute cholangitis
(with videos). J Hepatobiliary Pancreat Sci 2013;20:2434.
13. Shamseer L, Moher D, Clarke M, et al. Preferred report-
ing items for systematic review and meta-analysis protocols
(PRISMA-P) 2015: elaboration and explanation. BMJ
2015;349:g7647.
14. Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis of
observational studies in epidemiology: a proposal for
reporting. Meta-analysis Of Observational Studies in Epi-
demiology (MOOSE) group. JAMA 2000;283:200812.
15. American College of Emergency. P. Emergency ultrasound
guidelines. Ann Emerg Med 2009;53:55070.
16. Miller A, Pepe P, Brockman C, et al. ED Ultrasound in
hepatobiliary disease. J Emerg Med 2006;30:6974.
17. Woo MY, Taylor M, Loubani O, Bowra J, Atkinson P.
My patient has got abdominal pain: identifying biliary
problems. Ultrasound 2014;22:2238.
18. Ross M, Brown M, McLaughlin K, et al. Emergency
physicians in patients with suspected cholelithiasis. Acad
Emerg Med 2011;18:22735.
19. Jang TB, Ruggeri W, Kaji AH. The predictive value of
specific emergency sonographic signs for cholecystitis. J
Med Ultrasound 2013;21:2931.
20. Zamora J, Abraira V, Muriel A, Khan K, Coomarasamy
A. Meta-DiSc: a software for meta-analysis of test accuracy
data. BMC Med Res Methodol 2006;6:31.
21. DerSimonian R, Laird N. Meta-analysis in clinical trials.
Control Clin Trials 1986;7:17788.
22. Deeks JJ, Macaskill P, Irwig L. The performance of tests
of publication bias and other sample size effects in system-
atic reviews of diagnostic test accuracy was assessed. J Clin
Epidemiol 2005;58:88293.
23. Whiting PF, Rutjes AW, Westwood ME, et al. QUADAS-
2: a revised tool for the quality assessment of diagnostic
accuracy studies. Ann Intern Med 2011;155:52936.
24. Pauker SG, Kassirer JP. The threshold approach to clinical
decision making. N Engl J Med 1980;302:110917.
25. Singer AJ, McCracken G, Henry MC, Thode HC Jr,
Cabahug CJ. Correlation among clinical, laboratory, and
hepatobiliary scanning findings in patients with suspected
acute cholecystitis. Ann Emerg Med 1996;28:26772.
26. Staniland JR, Ditchburn J, De Dombal FT. Clinical pre-
sentation of acute abdomen: study of 600 patients. Br
Med J 1972;3:3938.
27. Lijmer JG, Mol BW, Heisterkamp S, et al. Empirical evi-
dence of design-related bias in studies of diagnostic tests.
JAMA 1999;282:10616.
28. Gilbert EH, Lowenstein SR, Koziol-McLain J, Barta DC,
Steiner J. Chart reviews in emergency medicine research:
where are the methods? Ann Emerg Med 1996;27:3058.
29. Schofield PF, Hulton NR, Baildam AD. Is it acute chole-
cystitis? Ann R Coll Surg Engl 1986;68:146.
30. Bednarz GM, Kalff V, Kelly MJ. Hepatobiliary scintigra-
phy. Increasing the accuracy of the preoperative diagnosis
of acute cholecystitis. Med J Aust 1986;145:3168.
31. Brewer BJ, Golden GT, Hitch DC, Rudolf LE, Wangen-
steen SL. Abdominal pain. An analysis of 1,000 consecu-
tive cases in a University Hospital emergency room. Am J
Surg 1976;131:21923.
32. Kendall JL, Shimp RJ. Performance and interpretation of
focused right upper quadrant ultrasound by emergency
physicians. J Emerg Med 2001;21:713.
33. Rosen CL, Brown DF, Chang Y, et al. Ultrasonography
by emergency physicians in patients with suspected chole-
cystitis. Am J Emerg Med 2001;19:326.
34. Noble VE, Liteplo AS, Nelson BP, Thomas SH. The
impact of analgesia on the diagnostic accuracy of the
sonographic Murphys sign. Eur J Emerg Med 2010;17:
803.
35. Eskelinen M, Ikonen J, Lipponen P. Diagnostic
approaches in acute cholecystitis; a prospective study of
1333 patients with acute abdominal pain. Theor Surg
1993;8:1520.
ACADEMIC EMERGENCY MEDICINE March 2017, Vol. 24, No. 3 www.aemj.org
36. Kohn MA, Carpenter CR, Newman TB. Understanding
the direction of bias in studies of diagnostic test accuracy.
Acad Emerg Med 2013;20:1194206.
37. Villar J, Summers SM, Menchine MD, Fox JC, Wang
R. The absence of gallstones on point-of-care ultrasound
rules out acute cholecystitis. J Emerg Med 2015;49:475
80.
38. Schlager D, Lazzareschi G, Whitten D, Sanders AB. A
prospective study of ultrasonography in the ED by emer-
gency physicians. Am J Emerg Med 1994;12:1859.
39. Jehle D, Davis E, Evans T, et al. Emergency department
sonography by emergency physicians. Am J Emerg Med
1989;7:60511.
40. Kiewiet JJ, Leeuwenburgh MM, Bipat S, Bossuyt PM, Sto-
ker J, Boermeester MA. A systematic review and meta-ana-
lysis of diagnostic performance of imaging in acute
cholecystitis. Radiology 2012;264:70820.
297
41. Csendes A, Burdiles P, Maluenda F, Diaz JC, Csendes P,
Mitru N. Simultaneous bacteriologic assessment of bile
from gallbladder and common bile duct in control sub-
jects and patients with gallstones and common duct
stones. Arch Surg 1996;131:38994.
42. Col N, Fanale J, Kronholm P. The role of medication
noncompliance and adverse drug reactions in hospital-
izations of the elderly. Arch Intern Med 1990;150:841
5.
43. Trowbridge RL, Rutkowski NK, Shojania KG. Does this
patient have acute cholecystitis? JAMA 2003;289:806.
44. Silen W. Copes Early Diagnosis of the Acute Abdomen.
New York: Oxford University Press, 2010.
45. Kulvatunyou N, Joseph B, Gries L, et al. A prospective
cohort study of 200 acute care gallbladder surgeries: the
same disease but a different approach. J Trauma Acute
Care Surg 2012;73:103945.