ORIGINAL ARTICLE
Endoscopic Ultrasound Versus CT Scan for Detection
of the Metastases to the Liver
Results of a Prospective Comparative Study
Pankaj Singh, MD,*w Phalguni Mukhopadhyay, MD,z Bankim Bhatt, MD,*
Tushar Patel, MD,y Alex Kiss, PhD,J Rahul Gupta, MD,z Sanjay Bhat, MD,z
and Richard A. Erickson, MD, FACP, FACG#
Background: Computed tomography (CT) scan is a standard test
for the detection of the liver metastases; however, metastases are
often missed on the CT scan.
Objective: To compare the accuracy of the endoscopic ultrasound
(EUS)/endoscopic ultrasound-guided fine needle aspiration (EUS-
FNA) with CT scan for detection of the liver metastases.
Design: Prospective study.
Patients: Subjects with newly diagnosed tumors of the lung,
pancreas, biliary tree, esophagus, stomach, and colon were
enrolled.
Interventions: A CT scan and EUS examination of the liver was
performed. EUS-FNA was performed on noncystic liver lesions.
Results: One hundred thirty-two cases were enrolled. The presence
of liver metastasis was established in 26 cases. The diagnostic
accuracy of EUS/EUS-FNA and CT scan was 98% and 92%,
respectively (P = 0.0578). In comparison to CT scan, EUS detected
significantly higher number of metastatic lesions in the liver (40
vs.19; P = 0.008). CT scan detected lesions in liver that were too
small to be characterized in 8 cases (malignant—3; benign—5). Of
these, EUS-FNA correctly characterized the lesion to be malignant
in 3/3 cases and benign in 4/5 cases. No complications were
observed as a result of EUS-FNA.
Received for publication August 20, 2007; accepted January 8, 2008.
From the Division of *Gastroenterology and Hepatology; zOncology;
zDepartment of Radiology, Central Texas Veterans Health Care
System; #Division of Gastroenterology and Hepatology, Scott and
White Memorial Hospital, Temple; wDepartment of Biostatistics
and Epidemiology School of Rural and Public Health, Texas A&M
University College Station, TX; yDivision of Gastroenterology and
Hepatology, Ohio State University Medical Center, Columbus,
OH; and JDivision of Clinical Epidemiology, Department of
Research Design and Biostatistics, Sunnybrook and Women’s
College Health Sciences Center, Toronto, Ontario.
The authors declare no conflict of interest.
Supported by a grant award from Veterans Affairs and by a grant from
Scott and White Hospital and Texas A&M University.
Institution where work was performed: Central Texas Veterans Health
Care System, Temple, TX.
The trial is registered at clinicaltrials.gov (NCT00290316).
Some of the findings of this study were presented in DDW 2006 and
were published as an abstract: ‘‘Endoscopic ultrasound can
accurately detect liver metastases—results from a prospective
controlled trial. Gastrointest Endosc. 2006;63:AB96’’ and as a
manuscript ‘‘Endoscopic ultrasound as a first test for diagnosis and
staging of lung cancer: a prospective study. Am J Respir Crit Care
Med. 2007;175:345–354.’’
Reprints: Pankaj Singh, MD, Central Texas Veterans Health Care
System, 1901 South First street, Temple, TX 76504 (e-mail:
pankaj.singh@med.va.gov).
Supplemental digital content is available for this article. Direct URL
citations appear in the printed text, and links to the digital files are
provided in the HTML text of this article on the journal’s Web site
(www.jcge.com).
Copyright r 2009 by Lippincott Williams & Wilkins
J Clin Gastroenterol  Volume 43, Number 4, April 2009
Limitations: Endoscopist was not blinded to the findings of the CT
scan.
Conclusions: In comparison with the CT scan, there was trend in
favor of EUS/EUS-FNA for the superior diagnostic accuracy. EUS
was distinctly superior to the CT scan in detecting
the number of metastatic lesions. EUS-FNA was also useful to
identify the nature of lesions that were too small to be characterized
on the CT scan.
Key Words: EUS, EUS-FNA, liver, metastasis, metastases, liver
cancer, liver secondaries, FNA, cancer, oncology
(J Clin Gastroenterol 2009;43:367–373)
T he liver is a common site for metastases for many
different malignancies. The presence of metastasis
in the liver can profoundly affect the management and
prognosis. Thus, radiologic evaluation of the liver is usually
considered essential for the complete staging of primary
tumors of the pancreas, esophagus, colon, gastric, and lung.
Computed tomography (CT) imaging is the most com-
monly performed imaging procedure for evaluating liver
metastases and National Comprehensive Cancer (2005)
guidelines for the management of cancer recommends CT
scan as a standard test for the evaluation of the liver for
metastases.
Recent studies have highlighted the limitations of CT
imaging for the detection of metastases of the liver.1,2
Endoscopic ultrasound (EUS) is a well-established tool for
the diagnosis and/or staging of esophageal, gastric, or
pancreatic cancer and has emerged as an alternative tool for
imaging the liver.3–6 Nguyen et al3 suggested that EUS
might be an adjunct to CT scan for detection of the liver
metastases, as it can detect lesions that are missed on the
CT scan. In their study, occult liver metastases were noted
in 2.4% of 574 patients with suspected gastrointestinal
(GI) or pulmonary malignancies.3 In another retros-
pective study, EUS detected metastatic lesions overlooked
by conventional, cross-sectional imaging studies in 5 of 222
cases (2.3%).4 These studies provide sufficient evidence to
suggest that EUS may be used for the evaluation of the liver
metastases.
We conducted a prospective study to compare the
accuracy of the EUS/EUS-FNA and CT scan for detection
of liver metastases in patients with primary malignancy of
other organs.
METHODS
This study was conducted in Central Texas Veterans
Health Care System, Temple, TX. The Institutional Review
Board of the Central Texas Veterans Health Care System,
367
Singh et al
TX, approved the protocol. Informed consent was obtained
for the participation in the study.
Patients with newly diagnosed pancreatic, biliary,
esophageal, colon, and gastric cancer were prospectively
enrolled for participation in the study from July 2004 to
July 2005; and patients with diagnosis of lung cancer were
prospectively included from March 2004 to July 2005.
Patients with advanced heart or lung disease that precluded
conscious sedation were excluded. Patients were referred
from the Veterans Affairs inpatient service, gastroenterol-
ogy clinic, primary care clinic, surgery clinic, and subspeci-
alty clinics (Oncology, Pulmonary). Patients were evaluated
by means of a history taking, physical examination, blood
count, measurement of electrolytes, and tests of renal and
liver function. All participants underwent CT scan and
EUS examination of the liver. In majority of the cases, EUS
was performed after the CT scan. Endoscopic ultrasound-
guided fine needle aspiration (EUS-FNA) was performed in
subjects with endosonographic evidence of a noncystic liver
lesion. After the procedure patients were referred back to
the clinic from which the patient was identified for further
management and follow-up.
Outcomes
The primary outcome of the study was to compare the
accuracy of EUS/EUS-FNA and CT scan for the detection
of liver metastases. The secondary outcome was to
determine the safety of EUS-FNA. Gold standard for the
diagnosis of metastases included cytopathologic confirma-
tion of malignant cells. If tissue could not be obtained from
liver lesions and there was progression in the size and
number of lesions in follow-up imaging, lesions were
considered to be malignant.
EUS/EUS-FNA Procedure
EUS and EUS-FNA was performed by a single
endoscopist. The curved linear-array echoendoscope (GF-
UCT/P 1140; Olympus America Corp, Melville, NY) was
used for all endosonographic examination and fine needle
aspiration. EUS/EUS-FNA test was called positive if
EUS–FNA-guided cytology showed malignant cells. EUS/
EUS-FNA was called negative if EUS-FNA was negative
for malignant cells or EUS did not show lesion in the liver
that required sampling. The liver was examined at 7.5 MHz
frequencies. EUS-FNA was performed using a 22-gauge
needle (Olympus FNA needle, Olympus America Corp). A
pathologist and the cytotechnologist were present in the
room to provide immediate interpretation as to whether
adequate material was obtained for diagnosis.
CT Scan
All CT examinations were performed with the use of a
helical CT scanner (model PQ-5000, Picker International,
Cleveland). Images were acquired with the use of 10-mm
collimation, a table speed of 6.25 mm/s at 175 mA and
120 kV, and a pitch of 1.50. Images were obtained during a
single breath-holding session when possible. Omnipaque
300 (100 mL) was administered intravenously to the
patients. Injection rate of 3 mL/s and scan delay
of 60 to 70 seconds was used.
Follow-up for Complications
Follow-up consisted of a patient interview by a nurse
or a physician, communication with the primary physician,
collection of additional radiologic test results, and review of
368
J Clin Gastroenterol  Volume 43, Number 4, April 2009
cytopathologic findings. Follow-up data were gathered 24
to 72 hours after the procedure, from clinic visits, collection
of additional radiologic test results, and review of
cytopathologic findings. Patients were followed to the time
of death or censored at the time of last visit to their
physicians.
Statistics
Sample Size Assumptions and Estimates
The study was designed to detect a difference of 35%
between the sensitivity of EUS and that of CT scan with a
power of 0.80 and a 2-sided a level of 0.05. To identify such
a difference, the study required a minimum of 27 patients
with liver metastases. Assuming that liver metastases are
present in about 15% of patients with resectable cancer, a
minimum of 170 patients had to be enrolled.
Statistical Tests
Sensitivity, specificity, positive predictive value, nega-
tive predictive value, and accuracy of the CT scan, EUS,
EUS-FNA, and EUS/EUS-FNA as a combined test were
calculated and compared. EUS/EUS-FNA was defined
positive when diagnosis of malignancy could be established
cytologically. EUS/EUS-FNA was defined as negative
when lesions were not visualized on EUS and in cases
when lesions were visualized, the EUS-guided FNA yielded
nonmalignant cells. The results were analyzed on a per-
lesion and per-patient basis. McNemar tests were used to
assess the differences between accuracies of tests. Nonpara-
metric tests (Wilcoxon signed-rank test) were used to
compare the number of lesions among EUS and other
tests. A Spearman correlation coefficient was run to assess
the correlation between the CT scan and EUS for the
multiplicity of the lesions. If one lesion in a subject was
cytologically established to be malignant, then all other
lesions with similar imaging characteristics were also
considered to be malignant. Lesions were defined as false
positive when the cytology was negative for malignancy and
there was no progression on follow-up imaging. False-
negative lesion was defined as lesion that was read as benign
lesion due to negative cytology or due to characteristics of
the lesion; however, on follow-up the lesion increased in
size. A ‘‘P’’ value of less than 0.05 was considered to
indicate statistical significance. All analyses were carried
out using SAS Version 9.1 (SAS Institute, Cary, NC).
RESULTS
One hundred thirty-two patients with newly diagnosed
pancreatic, biliary, esophageal, colon, gastric, or lung
cancer were prospectively enrolled in the study. Mean age
of the subjects was 67 years (SD ± 9.7; range: 45 to 86).
EUS and CT scan were performed in 132 and 131 cases,
respectively. Intravenous contrast for CT scan was used in
85% (112) of the cases. Twenty-six subjects were identified
with metastases of the liver. One patient was excluded from
the analysis as pancreatic mass was diagnosed as non-
Hodgkin lymphoma (Table 1).
EUS/EUS-FNA
EUS was performed in 132 cases of which 35 were
identified with definite liver lesions. EUS-FNA was
performed in 32 cases. EUS-FNA was not performed in 3
cases with EUS findings characteristics of cystic lesions in
the liver. FNA established the tissue diagnosis of metastases
r 2009 Lippincott Williams & Wilkins
J Clin Gastroenterol  Volume 43, Number 4, April 2009 ELIS Versus CT Scan for Detection of Liver Metastases

In 1 patient, a lesion was not visualized and, therefore,

TABLE 1. Demographics FNA was not performed. In another patient, lesion was
N easily visualized and EUS-FNA was performed. As the
No. cases with primary tumors 131 onsite-cytopathology showed malignant cells, subsequent
Mean age (y) 67 passes were not made. Subsequently, final diagnosis of
Sex (male) 128 cellblock showed the cells to be normal liver cells. Follow-
Metastases 26 up imaging showed the lesions to progressively increase in
Tumor types with metastases the size and number strongly suggesting that lesions were
Lung 14/115 malignant (Table 2).
Colon 7/7 CT scan missed metastases in 2 cases. CT missed
Esophagus 1/1 1 case with multiple lesions (2 lesions >1 cm, 8 lesions
Pancreas 2/5
<1 cm; 7 lesions in the right lobe, 3 lesions in the left lobe)
Biliary 2/3
that were detected on the EUS/EUS-FNA. In another case
missed by CT scan (also missed on EUS), new lesions in the
liver were detected on follow-up (repeat) CT imaging at 4
months.
to the liver in 24 cases. Sixteen lesions were hypoechoic, 7
lesions were hyperechoic, and 1 lesion was isoechoic Follow-up Imaging
(Fig. 1). Endosonographic findings based on echogenicity Sixty-nine percent (91/132) cases had follow-up
were not predictive of malignant lesions (P = 0.2). Of 24 imaging (CT scan or magnetic resonance imaging) of the
cases with liver metastases, EUS showed numerous lesions liver. There were 6 cases in whom follow-up tests showed
(Z6) in 11 cases. The median number of needle passes to lesions that were not detected in the previous tests (EUS
establish the cytologic diagnosis of malignancy was 2.1 and CT scan). One subject had metastases detected within 3
(SD ± 0.73; range: 1 to 4). months (missed on EUS and CT scan). This subject was
included as false negative in the analysis. Of remaining
Comparisons of EUS/EUS-FNA With 5 cases, lesions highly suspicious for metastases were
CT Scan for the Accuracy detected in 3 cases. The time interval for the detection of
The diagnostic accuracy of EUS/EUS-FNA and CT metastases was 10 months (CT at 6 mo was normal), 10
scan was 98% and 92%, respectively (P = 0.0578). EUS/ months (CT at 6 mo was normal), and 11 months (CT at
EUS-FNA missed 2 patients with metastases to the liver. 7 mo was normal). As follow-up CT scan at 6 months/7

FIGURE 1. A, EUS image showing hypoechoic metastatic lesions. B, EUS-FNA of the hypoechoic metastatic lesion in the liver.
C, EUS image showing isoechoic metastatic lesion in the left lobe of the liver. D, EUS image showing hyperechoic metastatic lesions in
the left lobe of liver. EUS indicates endoscopic ultrasound; EUS-FNA, endoscopic ultrasound-guided fine needle aspiration.

r 2009 Lippincott Williams & Wilkins 369

Singh et al J Clin Gastroenterol  Volume 43, Number 4, April 2009

TABLE 2. Comparison of Sensitivity, Specificity, Positive Predictive Value, Negative Predictive Value, and Accuracy of the CT Scan, EUS
and EUS-FNA for the Diagnosis of Liver Tumors
Sensitivity (%) Specificity (%) PPV (%) NPV (%) Accuracy (%)
CT scan 96 (91-98) 92 (85-96) 73 (64-80) 99 (95-100) 92 (86-96)
EUS 100 (96-100) 96 (91-99) 87 (80-92) 100 (96-100) 97 (92-99)
EUS/EUS-FNA 92 (86-96) 99 (95-100) 96 (91-99) 98 (94-100) 98 (93-99)
CT indicates computed tomography; EUS endoscopic ultrasound; EUS-FNA, endoscopic ultrasound-guided fine needle aspiration; NPV, negative
predictive value; PPV, positive predictive value.

months did not show lesions in these 3 cases, it was
assumed that these patients developed metastases later.
Two of the five cases had very small (2 to 3 mm)
indeterminate lesions detected at 6 and 15 months. These
patients are currently under follow-up. Follow-up radi-
ologic imaging of the liver is not available in 42 cases.
Reason for not having follow-up imaging are as follows: 7
of the 42 cases had cytologically proven liver metastases by
EUS-FNA; and 23 subjects died within 3 to 4 months.
There were 12 cases with survival more than 5 months and
did not have a follow-up imaging.
Comparison of EUS With CT Scan for
Number of Lesions
Of 26 patients with malignant lesions in the liver, 13
subjects had numerous (>6) lesions in the liver. The range
of numerous lesions was from 6 to 20. Eleven subjects had
numerous lesions on EUS and 11 subjects had numerous
lesions on CT scan. EUS and CT scan were found to be
significantly positively correlated (P =0.0002) for numer-
ous lesions. Patients with numerous lesions were excluded
when comparing the EUS and CT scan.
Total Lesions (Benign and Malignant)
Compared with CT scan, EUS detected significantly
higher number of total lesions (51 vs. 27; P = 0.005).
Stratifying the lesions to left and right lobes of the liver
showed that EUS detected significantly higher number
of lesions than CT scan in the left lobe (32 vs. 12;
P = 0.002). There was no significant difference between
EUS and CT for the number of lesions in right lobe
(19 vs. 15; P = 0.79) (Table 3).
Content 1, which demonstrates EUS of the liver showing
three hypoechoic lesions, http://links.lww.com/A867)].
Interobserver Variation for the Identification
of the Lesions
Radiologist with an expertise in abdominal imaging
performed the blinded review of CT and EUS photographs
at a remote time. There was a high degree of reliability
between the 2 raters (endoscopist and radiologist). A
radiologist was in agreement with endoscopist for the
identification of liver lesions in 96% of cases.
Safety of the EUS/EUS-FNA
Follow-up of patients is categorized as immediate (24
to 72 h after the procedure), intermediate (4 to 14 d after the
procedure), and long term (15 d onwards). Immediate
follow-up was available in 68% (90 of the 132) cases.
Intermediate follow-up was available in 37 of the 42 cases
with no immediate follow-up. Five cases in whom neither
short nor intermediate follow-up was available, had long-
term follow-up. Overall, 116/132 (88%) had long-term
follow-up. Six of the 132 cases died within a period of 1
month due to advanced metastatic disease and comorbid-
ities (median survival 18 d, range: 4 to 25).
Thirty-two patients underwent EUS-FNA. In total, 66
needle passes were made in 34 lesions. No complications
were observed as a result of EUS-FNA of the liver.
DISCUSSION
This prospective controlled study showed that EUS/
EUS-FNA is a safe and an accurate test for the detection

Malignant Lesions
For malignant lesions, EUS detected significantly TABLE 3. Comparison of EUS and CT Scan for Detection of
higher number of lesions than the CT scan (40 vs. Hepatic Lesions
19; P = 0.008). EUS detected more lesions than CT scan CT Scan EUS P
in left lobe of the liver (25 vs. 8; P = 0.005). EUS detected
more lesions in the right lobe as compared with the CT scan Lesions (benign)
(15 vs. 11; P = 0.59); however, the difference was not Both lobes 8 11 1.0
Left lobe 4 7 0.50
statistically significant (Table 3; Figs. 2, 3). Right lobe 4 4 1.0
Lesions (malignant)
Both lobes 19 40 0.008
Small Lesions on CT Scan Left lobe 8 25 0.005
CT scan detected lesions in liver that were too small to Right lobe 11 15 0.59
be characterized (<1 cm) in 8 patients of which 3 were Lesions (benign+malignant)
malignant and 5 were benign. EUS visualized these small Both lobes 27 51 0.005
lesions in 7 (88%) cases. Of these, EUS-FNA correctly Left lobe 12 32 0.002
characterized the lesion to be malignant in 3 cases and benign Right lobe 15 19 0.79
in 4 cases. In one case with benign lesion, the lesion was not CT indicates computed tomography; EUS endoscopic ultrasound.
identified by EUS [Fig. 2, (see Video, Supplement Digital

370 r 2009 Lippincott Williams & Wilkins

J Clin Gastroenterol  Volume 43, Number 4, April 2009 ELIS Versus CT Scan for Detection of Liver Metastases

FIGURE 2. A, CT scan with contrast showing 1 lesion in the right lobe of the liver that was too small to be characterized.
B, Transcutaneous ultrasound showed 2 lesions. C, Magnetic resonance imaging showed 2 lesions, a small single enhancing nodule in
the right lobe of the liver measuring 1.2 cm and another small nodule (not shown in the figure) in the left lobe measuring 6 mm. D,
EUS image showing one of the 3 hypoechoic lesions detected during endosonographic examination of the left and right lobe of the
liver. EUS-FNA of the lesion confirmed the lesion to be malignant. CT indicates computed tomography; EUS, endoscopic ultrasound;
EUS-FNA, endoscopic ultrasound-guided fine needle aspiration.

of the metastases of the liver. In comparison to CT scan,
there was a trend in favor of EUS/EUS-FNA for the higher
diagnostic accuracy. EUS detected significantly higher
number of malignant lesions in the liver as compared with
CT scan. EUS-FNA was highly accurate (88%) in
establishing the exact nature of the liver lesions that were
too small to be characterized on the CT scan.
Dewitt et al5 in a large retrospective study showed the
sensitivity of the EUS to be in the range of 82% to 94%.
EUS/EUS-FNA detected 17 malignant hepatic lesions in
patients with a previously normal CT scan. Prasad et al4
showed cytologically positive malignant lesions in 2.3% (5
of the 222 patients) who underwent EUS examination.
Nguyen et al3 found occult liver metastases in 2.4% of 574
patients with suspected GI or lung malignancy. The
findings of this prospective study are in agreement with
the hypothesis that the accuracy of the EUS is superior to
the CT scan. The reason for the high accuracy was as a
result of high sensitivity of the EUS for the visualization of
the hepatic lesions and high specificity of the EUS-FNA for
establishing the cytologic nature of the lesions. Though the
sensitivity of the EUS and CT was similar, EUS was
significantly superior to the CT scan in detection of the
number of malignant lesions. The ability to detect more
lesions has potential implications on the management as
patients with single lesion are potential candidates for liver
resection in some GI cancers (eg, colon).6
Contrary to popular belief, lesions in the right lobe of
the liver could be visualized and successfully sampled
during EUS examination (Fig. 3). In comparison with CT
scan, the ability of the EUS for the detection of the lesions
in the left lobe was significantly higher (25 vs. 8, P = 0.005)
than in the right lobe (15 vs. 11, P = 0.59).
The plausible reason seems to be the relative ease in the
complete examination of the left lobe as there is large
endosonographic window of anterior gastric wall and
antrum. The right lobe is examined from the duodenum,
which is technically difficult because of small endosono-
graphic window (duodenal bulb and second portion of the
duodenum) and is possibly further compounded by
the limited depth of penetration. Particularly, lesions
in subdiaphragmatic region of right lobe of liver (right
anterior superior and right posterior superior anatomic
segments) may not be accessible to EUS. In addition,
acoustic shadows as a result of the gallstones poses
significant problem in complete examination of the right
lobe of liver (see Video, Supplemental Digital Content 2,
which demonstrates EUS showing the limitation of

r 2009 Lippincott Williams & Wilkins 371

Singh et al
FIGURE 3. Illustration showing EUS-FNA of the lesion in the right
lobe of the liver. EUS-FNA indicates endoscopic ultrasound-
guided fine needle aspiration.
complete liver examination becuase of presence of gall-
stones, http://links.lww.com/A868).
Focal hepatic lesions that are smaller than 1 cm in size
are not readily characterizable on the CT scan. In a patient
without a primary cancer, small liver lesions are often
discovered incidentally and have a high likelihood of being
benign (cysts, hemangiomas, focal nodular hyperplasia).
These lesions are usually evaluated by serial follow-up
imaging tests.7 However, in patients with cancer, the
probability of smaller lesions being metastatic lesion is
higher and, therefore, determination of the nature of lesion
is critical for the management. Lawrence et al8 showed the
prevalence of small hepatic lesions of 12.7% in a large
cohort of 2978 patients with cancer. Twelve percent of these
small lesions showed interval growth and were, therefore,
considered to be malignant and 8% patients were stable at
follow-up of less than 6 months and were considered
indeterminate. Eberhardt et al9 assessed the role of
sonography in evaluating the small indeterminate liver
lesions detected on the CT scan in patients with cancer.
Sonography detected and characterized 48% of the 124
indeterminate lesions that were evident on the CT scan.
Sonographic detection was significantly dependent on the
size of the lesions and the body habitus. Follow-up studies
supported the sonographic diagnoses in 70% of the cases.
In this study, CT scan detected lesions that were too small
to be characterized in 6% cases. EUS/EUS-FNA was
successful in visualizing and establishing the nature of these
lesions in 88% cases in a single setting. The high success
rate of EUS as compared with transabdominal ultra-
sound seems to be due to much higher resolution; the
success rate of EUS examination is not dependent on the
body habitus and importantly the nature of lesions can
be determined cytologically rather than merely by imaging
characteristics.
372
J Clin Gastroenterol  Volume 43, Number 4, April 2009
A distinct advantage of EUS is that EUS examination
and EUS-FNA can be performed simultaneously and,
therefore, confirmation of malignancy can usually be
accomplished in a single procedure, whereas conventional
imaging studies typically require 2 sessions to accomplish
these tasks, one for detection and another for FNA.
Endosonographically, it is difficult to differentiate benign
from malignant liver lesion as malignant lesions can be
hyperechoic, hypoechoic, or isoechoic. Therefore, we
recommend that once a lesion is identified, FNA should
be performed regardless of echogenic features. Tissue
diagnosis plays a critical role particularly when therapies
involve major surgical procedures. This study showed that
the positive predictive value of EUS is 87%. As EUS-FNA
diagnostic criteria were based on cytologic diagnosis, it had
a high accuracy of 100%.
Though a large retrospective survey of 167 cases
showed a complication rate of 4%, which included
bleeding, abdominal pain, fever, and 1 death10; this study
showed that EUS-FNA of the liver is a safe procedure. Our
findings regarding the safety of EUS-FNA correlate with
those of Dewitt et al5; however, because of the smaller
number of patients with metastases in our study,
a low frequency of significant complications could have
been missed.
There are some limitations to this study. First, as EUS
examination was performed after the CT scan in majority
of the cases, the endoscopist was not blinded to the findings
of the CT scan. Second, in an individual patient with liver
metastases all of the lesions were not sampled. We decided
to call the lesions as malignant when cytopathologic
assessment of one of the similar looking lesions was
cytologically positive for malignancy. It would have been
unethical to subject patients to additional procedural time
and risk, however small, to try and sample all the lesions for
malignancy when the diagnosis was established from the
initial sample of one lesion. This is in concordance with
standard practice as in any radiologic imaging modalities
the number of lesions is decided by the identification of
the lesions that are similar in appearance. Third, there
was no priori definition of ‘‘metastatic lesion.’’ As multiple
radiologists read the CT scan as opposed to single
endosonographer, there is a possibility of interobserver
variability in how one defines a metastatic lesion as it
appears on CT scan. Another possible limitation of
the study is that helical CT scans were performed with
10-mm collimation, which may have lowered the detection
rate of the number of lesions in comparison with EUS.
Though it seems logical that with CT imaging with
thinner collimation, the actual number of smaller lesions
may be higher, a recent study did not support this
hypothesis.11 Another possible downside of thinner
collimation would be that a greater number of lesions that
are ‘‘too small to be characterized’’ would be detected,
which would require additional follow-up diagnostic
studies. The high resolution of EUS and ability to perform
EUS-FNA in a single setting is a distinct advantage
to characterize such lesions. Also, CT-FNA of the small
(<1 cm) lesions is difficult, whereas EUS-FNA is feasible
in as small as 4-mm lesion. Another possible limitation
of the study is that surgical specimen was not used as the
gold standard. We do not consider this as a limitation
because in practice majority of the patients with liver
metastases do not undergo liver resection. We used
follow-up imaging tests to determine if there were
r 2009 Lippincott Williams & Wilkins
J Clin Gastroenterol  Volume 43, Number 4, April 2009
metastatic lesions that were missed by both EUS and
CT scan.
In conclusion, EUS examination of the liver should be
considered an integral aspect of the staging of the
esophageal, gastric, pancreatic, biliary, and lung cancer.
As EUS detects significantly more lesions than CT,
it should be considered for the preoperative evaluation of
tumors before consideration of liver resection of metastatic
lesions. In patients with cancer, EUS/EUS-FNA should be
considered as the next test after CT scan to identify the
nature of lesions that are too small to be characterized on
the CT scan.
REFERENCES
1. Glover C, Douse P, Kane P, et al. Accuracy of investigations
for asymptomatic colorectal liver metastases. Dis Colon
Rectum. 2002;45:476–484.
2. Stark DD, Wittenberg J, Butch RJ, et al. Hepatic metastases:
randomized, controlled comparison of detection with MR
imaging and CT. Radiology. 1987;165:399–406.
3. Nguyen P, Feng JC, Chang KJ. Endoscopic ultrasound (EUS)
and EUS-guided fine-needle aspiration (FNA) of liver lesions.
Gastrointest Endosc. 1999;50:357–361.
r 2009 Lippincott Williams & Wilkins
ELIS Versus CT Scan for Detection of Liver Metastases
4. Prasad P, Schmulewitz N, Patel A, et al. Detection of occult
liver metastases during EUS for staging of malignancies.
Gastrointest Endosc. 2004;59:49–53.
5. DeWitt J, LeBlanc J, McHenry L, et al. Endoscopic
ultrasound-guided fine needle aspiration cytology of solid liver
lesions: a large single-center experience. Am J Gastroenterol.
2003;98:1976–1981.
6. Petrelli NJ, Nambisan RN, Herrera L, et al. Hepatic resection
for isolated metastasis from colorectal carcinoma. Am J Surg.
1985;149:205–209.
7. Jones EC, Chezmar JL, Nelson RC, et al. The frequency and
significance of small hepatic lesions (<15 mm) detected by CT.
AJR. 1992;158:535–539.
8. Lawrence H, Schwartz MD, Eric J, et al. Prevalence and
importance of small hepatic lesions found at CT in patients
with cancer. Radiology. 1999;210:71–74.
9. Eberhardt SC, Choi PH, Bach AM, et al. Utility of sonography
for small hepatic lesions found on computed tomography in
patients with cancer. J Ultrasound Med. 2003;22:335–343.
10. ten Berge J, Hoffman BJ, Hawes RH, et al. EUS-guided fine
needle aspiration of the liver: indications, yield, and safety
based on an international survey of 167 cases. Gastrointest
Endosc. 2002;55:859–862.
11. Haider MA, Amitai MM, Rappaport DC, et al. Multi-detector
row helical CT in preoperative assessment of small (< or
1.5 cm) liver metastases: is thinner collimation better? Radi-
ology. 2002;225:137–142.
373