Shandra Bipat, MSc
Maarten S. van Leeuwen,
MD, PhD
Emile F. I. Comans, MD,
PhD
Milan E. J. Pijl, MD, PhD
Patrick M. M. Bossuyt, PhD
Aeilko H. Zwinderman, PhD
Jaap Stoker, MD, PhD
Published online before print
10.1148/radiol.2371042060
Radiology 2005; 237:123–131
Abbreviations:
CI
confidence interval
FDG
fluorine 18
fluorodeoxyglucose
SPIO
superparamagnetic iron
oxide
1
From the Department of Radiology
(S.B., J.S.) and Department of Epide-
miology and Biostatistics (P.M.M.B.,
A.H.Z.), Academic Medical Center,
University of Amsterdam, Meiberg-
dreef 9, 1105 AZ Amsterdam, the
Netherlands; Department of Radiology,
University Medical Center Utrecht, Utre-
cht, the Netherlands (M.S.v.L.); Depart-
ments of Nuclear Medicine and PET Re-
search, VU University Medical Center,
Amsterdam, the Netherlands (E.F.I.C.);
and Department of Radiology, Martini
Hospital, Groningen, the Netherlands
(M.E.J.P.). Received December 6,
2004; revision requested January 10,
2005; revision received January 24; ac-
cepted February 1. Supported by a
grant from the Dutch Order of Medi-
cal Specialists. Address correspon-
dence to S.B. (e-mail: s.bipat@amc.uva
.nl).
Authors stated no financial relation-
ship to disclose.
Author contributions:
Guarantor of integrity of entire study,
S.B.; study concepts/study design or
data acquisition or data analysis/in-
terpretation, all authors; manuscript
drafting or manuscript revision for im-
portant intellectual content, all au-
thors; approval of final version of sub-
mitted manuscript, all authors; litera-
ture research, all authors; statistical
analysis, S.B., P.M.M.B., A.H.Z., J.S.;
manuscript editing, all authors
?
RSNA, 2005
Colorectal Liver Metastases:
CT, MR Imaging, and PET for
Diagnosis—Meta-analysis
1
PURPOSE: To perform a meta-analysis to obtain sensitivity estimates of computed
tomography (CT), magnetic resonance (MR) imaging, and fluorine 18 fluorodeoxy-
glucose (FDG) positron emission tomography (PET) for detection of colorectal liver
metastases on per-patient and per-lesion bases.
MATERIALS AND METHODS:MEDLINE, EMBASE, Web of Science, and CANCERLIT
databases and Cochrane Database of Systematic Reviews were searched for relevant
original articles published from January 1990 to December 2003. Criteria for inclu-
sion of articles were as follows: Articles were reported in the English, German, or
French language; CT, MR imaging, or FDG PET was performed to identify and
characterize colorectal liver metastases; histopathologic analysis (surgery, biopsy, or
autopsy), intraoperative observation (manual palpatation, intraoperative ultra-
sonography [US]), and/or follow-up US was the reference standard; and data were
sufficient for calculation of true-positive or false-negative values. A random-effects
linear regression model was used to obtain sensitivity estimates in assessment of liver
metastases.
RESULTS: Of 165 identified relevant articles, 61 fulfilled all inclusion criteria. Sen-
sitivity estimates on a per-patient basis for nonhelical CT, helical CT, 1.5-T MR
imaging, and FDG PET were 60.2%, 64.7%, 75.8%, and 94.6%, respectively; FDG
PET was the most accurate modality. On a per-lesion basis, sensitivity estimates for
nonhelical CT, helical CT, 1.0-T MR imaging, 1.5-T MR imaging, and FDG PET were
52.3%, 63.8%, 66.1%, 64.4%, and 75.9%, respectively; nonhelical CT had lowest
sensitivity. Estimates of gadolinium-enhanced MR imaging and superparamagnetic
iron oxide (SPIO)–enhanced MR imaging were significantly better, compared with
nonenhanced MR imaging (P
.019 and P .001, respectively) and with helical CT
with 45 g of iodine or less (P
.02 and P  .001, respectively). For lesions of 1 cm
or larger, SPIO-enhanced MR imaging was the most accurate modality (P  .001).
CONCLUSION: FDG PET had significantly higher sensitivity on a per-patient basis,
compared with that of the other modalities, but not on a per-lesion basis. Sensitivity
estimates for MR imaging with contrast agent were significantly superior to those for
helical CT with 45 g of iodine or less.
?
RSNA, 2005
Colorectal cancer is the second leading cause of cancer-related deaths in the United States.
According to the National Program of Cancer Registries, 146 940 new patients received a
diagnosis of the disease in 2004, with an estimated 56 730 deaths due to colorectal cancer
in that year. Liver metastasis is a common consequence of colorectal carcinoma; up to 70%
of patients with colorectal cancer eventually develop liver metastases. In 30%–40% of
those patients, the metastases are still confined to the liver at the time of detection, and
only a limited number of patients with colorectal metastases confined to the liver are
surgical candidates because of the larger size of the lesions, the broad distribution of the
lesions, or the difficulty in assessing the tumors or because the volume of the remaining
liver is inadequate (1–5).
Preoperative selection of patients with colorectal metastases who are most likely to
benefit from surgery is necessary and challenging. The armamentarium for imaging-based
Evidence-based Practice
123
R
a
d
i
o
l
o
g
y

preoperative selection comprises trans-
abdominal ultrasonography (US), com-
puted tomography (CT), fluorine 18 flu-
orodeoxyglucose (FDG) positron emis-
sion tomography (PET), and magnetic
resonance (MR) imaging (6–12). During
the past 10 years, improvements in these
imaging modalities were either intro-
duced or great progress has been made in
their application (6,8,13–16). Although
extensive research has been performed in
regard to the diagnostic performance of
CT, MR imaging, and FDG PET for the
detection of colorectal liver metastases,
the optimal imaging staging strategy has
not been defined.
Kinkel et al (17) performed a meta-
analysis to compare current noninvasive
imaging methods such as US, CT, MR
imaging, and FDG PET for the detection
of hepatic metastases from colorectal,
gastric, and esophageal cancers. Treat-
ment approaches for liver metastases
from various cancerous origins (pancre-
atic or colorectal cancer), however, are
different, and, therefore, the importance
of certain findings with respect to these
origins differs.
Because of its noninvasive character,
low cost, and widespread availability, US
is a valuable screening tool for the imag-
ing of liver metastases. US, however, has
two relative disadvantages: US is more
operator independent than are CT and
MR imaging, and parts of the liver re-
main nonvisible in certain patients at US.
In daily practice, though, US is highly
efficient in helping to distinguish be-
tween two groups of patients with liver
metastases: the group of patients with
diffuse metastases who are no longer eli-
gible for curative treatment and the
group with no metastases or a very lim-
ited number of them. The patients in the
latter group require CT, MR imaging, or
FDG PET for the selection of appropriate
therapeutic approaches. Thus, the aim of
our study was to perform a meta-analysis
to obtain the estimates of sensitivity of
CT, MR imaging, and FDG PET for the
detection of colorectal liver metastases
on a per-patient and a per-lesion basis.
MATERIALS AND METHODS
Literature Search
A comprehensive computer literature
search (18) of abstracts about studies in
human subjects was performed to iden-
tify articles about the diagnostic perfor-
mance of CT, MR imaging, and FDG PET
for the detection of liver metastases in
patients with colorectal cancer compared
with the diagnostic performance of intra-
operative US, surgery, follow-up US, and
histopathologic analysis as the reference
standard. The MEDLINE and EMBASE da-
tabases, from January 1990 to December
2003, were used with the following key-
words: (“Colorectal Neoplasms” [MeSH])
AND (“Liver neoplasms” [MeSH]) AND
(“Laparoscopy” [MeSH] OR “Tomogra-
phy, Emission-Computed” [MeSH] OR
“magnetic resonance imaging” [MeSH]
OR “Tomography, X-Ray Computed”
[MeSH] AND (sensitivity and specificity
[MeSH] OR sensitivity [WORD] OR speci-
ficity [WORD] OR false negative [WORD]
OR false positive [WORD] OR diagnosis
[MeSH] OR diagnostic use [MeSH] OR de-
tection [WORD] OR accuracy [WORD]).
Other databases, such as CINAHL and
SUMSEARCH, were also checked for rele-
vant articles with the following key-
words: Colorectal Neoplasm [MeSH]
AND (Liver Neoplasms [MeSH] OR Neo-
plasm Metastasis [MeSH]). The databases
of Web of Science and CANCERLIT and
the Cochrane Database of Systematic Re-
view were checked with the following
words: Colorectal cancer AND (liver me-
tastases OR hepatic metastasis). Review
articles, letters, comments, case reports,
unpublished articles, and articles that did
not include raw data were not selected.
The list of articles was supplemented
with extensive cross-checking of the ref-
erence lists of all retrieved articles.
Selection of Studies
Four observers independently checked
all retrieved articles for inclusion criteria.
One observer (S.B.) checked all articles.
Three observers checked a subset of arti-
cles: One observer (M.S.v.L.) checked
studies that predominantly focused on
evaluation of CT, another (M.E.J.P.)
checked studies that predominantly fo-
cused on evaluation of MR imaging, and
another (E.F.I.C.) checked studies that
predominantly focused on evaluation of
FDG PET. Disagreements were resolved in
consensus. The inclusion criteria were as
follows: (a) Articles were reported in the
English, German, or French language. (b)
CT, MR imaging, or FDG PET was used to
identify and characterize colorectal liver
metastases. (c) Histopathologic analysis
(performed at surgery, biopsy, and au-
topsy), intraoperative observation (eg,
manual palpation, intraoperative US)
and/or follow-up US were used as the ref-
erence standard. (d) For per-patient or
per-lesion statistics, sufficient data were
presented to calculate the true-positive
and false-negative values for imaging
techniques. (e) When data or subsets of
data were presented in more than one
article, the article with the most details or
the most recent article was chosen. Stud-
ies were excluded if results of different
imaging modalities were presented in
combination and could not be differen-
tiated for performance assessment of
tests on an individual modality.
Data Extraction
The same observers independently ex-
tracted relevant data about study (design)
characteristics and examination results,
which will be discussed later, from each
article by using a standardized form. One
observer (S.B.) extracted data from all ar-
ticles. Three observers extracted data
from a subset of articles: One observer
extracted only data from studies that pre-
dominantly focused on evaluation of CT
(M.S.v.L.), another extracted only data
from studies that predominantly focused
on evaluation of MR imaging (M.E.J.P.),
and still another extracted only data
from studies that predominantly focused
on evaluation of FDG PET (E.F.I.C.). Ob-
servers were not blinded with regard to
information about the authors, the au-
thors’ affiliation, or the journal name,
since this has been shown to be unnec-
essary (19). To resolve disagreement be-
tween reviewers, a fifth reviewer (J.S.) as-
sessed all discrepant items, and the ma-
jority opinion was used for analysis.
Study design characteristics.—The
QUADAS quality assessment tool was
used to extract relevant study design
characteristics of each study. This tool
and the definitions of the characteristics
are fully described elsewhere (20).
Other study characteristics.—In addi-
tion, the following characteristics were
recorded: (a) year of publication; (b) sam-
ple size (number of patients with colorec-
tal liver metastases); (c) description of
study population, which included dis-
ease severity (tumor stage), age, and
male-female distribution; (d) description
of interpretation of diagnostic tests,
which included the reporting of the char-
acterization of lesions as benign versus
malignant or the detailed subcharacter-
ization of lesions as cysts, hemangiomas,
or metastases and the confidence rating
used for identification of lesions; (e) de-
scription of reference tests, which in-
cluded intraoperative findings (at inspec-
tion and/or palpation), intraoperative US
features (probe frequency), pathologic
features (staining, lamination, thickness
of slices), or follow-up characteristics (in-
terval between examinations, frequen-
124


Radiology


October 2005 Bipat et al
R
a
d
i
o
l
o
g
y