Lymph nodes measuring larger than or equal to 4 mm, especially those located ante- rior to the mid portion of the aorta, should raise a suspicion of metastases.
CT Detection Lymph Node Patients with Testicular
of Retroperitoneal Metastases in Clinical Stage I
Nonseminomatous
Germ Cell Cancer: Size and Distribution
Assessment Criteria
of
OBJECTIVE.
Susan Hmlton1 HarryW. Herr2 Jerrold B. Teitcher1 Cohn B. Begg3 Ronald A. Castehlmno1
Patients with nonseminomatous germ cell cancer of the testis with no eviof metastatic disease after orchiectomy may be managed with either retroperitoneal node dissection or surveillance. The present retrospective study was undertaken to de-
dence lymph termine
with
the accuracy newly diagnosed
tis when
smaller
CT-revealed than
are
size
posterior
were
METHODS.
by three
The
sizes
nodes
are applied are more
and
sites
the basis of the size of the largest
CT
Abdominal
observers
who of all
were
lymph
CT scan was judged
Each
10 mm)
than
lymph
node metastases germ cell cancer
in patients of the tes-
and to test the hypothesis
likely
to correlate
with
that
metastases
nodes.
AND
reviewed
recorded.
(smaller
retroperitoneal
lymph
phadenectomy. were
criteria
anterior
MATERIALS tients
of CT for revealing retroperitoneal lymph clinical stage 1 testicular nonseminomatous
nodes
as positive
measured
lymph
scans
unaware
before
obtained
of the
measuring
or negative
surgery
results
of retroperitoneal
larger
than
or equal
for retroperitoneal
node at the expected
in 70 palymto 4 mm
metastasis
metastatic
on
site. Diameters
6, 8, and 10 mm were successively applied to each case as the criteria for a positive scan. Using a criterion of 10 mm or larger for metastases, we calculated a sensitivity
of4,
RESULTS.
of
37% and a specificity of 100%; with a 4-mm criterion, the sensitivity was 93% and the specificity was 58%. Receiver operating characteristic curves comparing the accuracy of CT for revealing similar-sized
lymph
nificantly
(p
=
nodes
located
CONCLUSION.
10 mm to as low as 7% using Lymph
ificity.
nodes
urvival tous
based
Received October 2, 1996; accepted March 3, 1997. 1
Department
of Radiology,
Cancer Center, l2l5York Address correspondence 2 Department
of Urology, Memorial
Center, NewYork, 3Department Sloan-Kettering
ate
Memorial
Sloan-Kettering
Cancer
NY 10021.
of Epidemiology Cancer
Center,
and Biostatistics,
Memorial
New York, NY 10021.
Roentgen
the
Ray Society
AJR:169, August 1997
in-
cisplatin-
use
in human
chorionic
serum
radiograph,
tumor
findings
of cispl-
gonadotropin
findings
on
an abdominal CT scan) was retroperitoneal lymph node dissection (RPLND). Surgical removal
of
all
metastatic
microscopic
disease
and generally
macroscopic produced
[1]. The high salvage rate therapy for patients with relapsing
with chemostage 1 dis-
ease
RPLND
who
the aorta
differed
sig-
did not undergo
initial
a corresponding
to 4 mm,
a suspicion
(including
decrease
especially
those
of
in spec-
located
ante-
of metastases.
about
infertility
the morbidity of RPLND due to injury of the sympa-
thetic nerves controlling ejaculation and potential complications of major abdominal surgery) [3], led to surveillance tive
I disease.
protocols
for patients
to RPLND The
as an alterna-
with
drawback
of
clinical
this
stage
approach
is
that an abdominal
markers after on a preopera-
and negative
with
and the concern
of the testis of
Before
negative
tive chest
to a line bisecting
mm,
or equal raise
markedly
cures
036i-803X/97/1692-521
© American
decline
orchiectomy,
nodal AJR1997;169:521-525
have
chemotherapy.
than
should
nonseminoma-
cell cancer
introduction
and ct-fetoprotein
NY 10021.
Sloan-Kettering
the
of4
atm, the standard treatment for patients with clinical stage I disease (based on an appropri-
after revision
Ave., NewYork, to S. Hilton.
with
for
rates germ
larger
of the aorta,
(NSGCT) creased
or posterior
a size criterion
measuring
rior to the mid portion
S
anterior
when the same criteria were applied to lymph nodes in both regions. False-negative rates were decreased from 63% using a size criterion
.04)
[2],
CT scan using conventional size criteria for lymph node enlargement will miss a significant percentage (false-negative rates of 22-44%) of patients with retroperitoneal metastases [4-6]. Before beginning this thors
observed
association cated
anteriorly
the aorta) site
in
the
what between
study,
appeared visible
(anterior
at the expected retroperitoneum
two of our au-
to be a frequent lymph nodes lo-
to the mid portion primary and
of
metastatic subsequent
521
Hilton
RPLND
for
positive
part of the decided nodes mary
current
to
test
located metastatic
metastatic
the site
the accuracy
peritoneal
we
hypothesis
anteriorly
that
correlate
with
study
was
to
criterion
axis
diameter
is used to define
ings and to determine lymph nodes anterior
deter-
retrowhen a in short-
positive
whether to a line
find-
CT-revealed bisecting
aorta at the expected
primary
metastatic
are
posterior
lymph
likely
than
to correlate
size
with
enrolled
were 70
the
site nodes
reviewed
with
a minimum
and Methods
The
initial
consecutive
study
population
patients
with clinical
treated
at our institution
tween
October
pre-RPLND
with
1988
and
abdominal
for retrospective
consisted
stage
primary
April
1995,
These
76
1 NSGCT RPLND
CT scans
review.
of
for
were
bewhich
patients
of the total of 255 patients with clinical evaluated at our institution during the period of the study. The CT scans of six patients were judged by consensus to be technically suboptimal. Our study analyzed the images of the remaining 70 patients. 1 NSGCT
stage
Pathology 62
slides
patients
outside
of the primary
institutions
were
Microscopic
specimens and
no vascular
vascular invasion
of
at
by our pathol-
slides from orchiectomies
analysis
showed
tumors
orchiectomy
reviewed
ogy department along with patients who had undergone center.
testis
who had undergone
of the
the
eight at our
CT
scans
interpreters
were
in academic
unaware
of the
whether
of
present
institu-
had
retroperitoneal
results
per-lymph
of
node
histopathologic
RPLND 43%)
review
of
basis).
The
vals.
For
case.
each
maximum
size
(short-axis
of retroperitoneal or larger
diameter)
lymph
at stipulated and
nodes sites
nodes
also
ceiver
renal
groups
were
considered
true-positive
for right-sided groups
tumors:
for left-sided
left paraaortic, were
considered
sites
the primary
of
metastatic
recorded the size of the largest lymph node at the primary site of spread in each paTo obtain
overall
accuracies
of CT results,
the
three interpreters’ data were pooled; for each patient, a single value for the largest lymph node at the primary metastatic site was obtained by averaging the values given by the three interpreters. Various size criteria (l0 mm, 8 mm, 6 mm. and 4 mm) for a CT scan positive for metastases were then applied to each
case.
At each
size
criterion,
each
scan
was
For
curve
that interpreter’s
measurement
curve
of the three
was
interpreters
re-
false-positive
size on CT
for each
observer
ratio
was
using
a standard
A re-
then
con-
such
that
the calculated
scans. The
for
a given
area under
calculated.
the
These
re-
curves were then corn-
ceiver operating characteristic pared
each in-
node size for each patient.
and
node
statistical
test
for
matched
receiver operating characteristic data [10]. To assess
iliac
sites. Each interpreter tient.
lymph
tumors,
and left common
CT interpreters
at each size criterion
characteristic
for each
between
size for each
each point on the curve represented
bifur-
classified
the primary
of results node
pooled).
of CT
using lymph
operating
structed
by the interpreter as anterior or posterior with respect to a line bisecting the aorta (Fig. I). On the basis ofpreviously published results [8, 9]. the paracaval, interaortocaval. and right common iliac node
a
and specific-
of the three (not
the accuracy
of maximum
the common
were
by each separately
was calculated
4 mm
the aortic
recorded analyzed
terpreter,
number
at the
through
For each site above lymph
the
measuring
beginning
extending
measured
and
patient
inter-
recorded
than
ity of CT were then calculated for each size criterion.
was
interpreter
were
rather
sensitivity
10-mm
each
metastases basis
using
at 10-mm
node
whether
defined
size
lymph
nodes
separate
criteria
(4,
located
more
application
6,
8,
or lO
anteriorly
of these
mm) to
(as compared
posteriorly) within the expected metastatic site would yield more accurate results. two receiver operating characteristic curves were constructed, one for lymph nodes classified as anterior and another with
for three
lymph
nodes
observers’
ceiver operating to the calculated tio
for
a given
classified averaged
as posterior, data.
characteristic
true-positive lymph
node
Each curve
point
using
the
on the re-
corresponded
and false-positive
ra-
size:
the
areas
under
curves were calculated and compared.
in 42 patients
in 27 patients:
a negative
node
(on a per-case
obtained
slices
(i.e., on
a per-lymph
orchiectomy
invasion
metastases,
for metastases than
retropentoneal
or absent
for
one
patient this information was unavailable at the time of this study. Thirty patients had a positive RPLND (retroperitoneal metastatic disease), and 40 patients
rather
basis). Surgical pathology results were designated as positive or negative for each patient on the basis
indepen-
radiologist
or negative
basis
were
node
a subset
were
of 1 1-20 years of experience
readers
interaortocaval,
composed
three
a per-patient
To assess the reproducibility observers, the maximum lymph
spread
available
by
scans
corded a.s positive
pa-
1 30
program.
RPLND. Fifty-seven CT scans had been obtained at outside institutions. All patients received oral contrast material: 66 (94%) received IV contrast material. In most of the cases (n = 61 ), images
cation,
Materials
CT
abdominal The
approximately
in the surveillance
abdominal
hilar vessels iliac regions.
metastasis.
study.
this
dently
tions.
of CT for revealing
size
of
tients
reading
at RPLND.
node metastasis of less than 10 mm
of similar
pri-
positively
the period The
lymph
lymph
more
As
therefore
at the expected
the presence of metastases The purpose of this mine
disease.
investigation,
et aI.
(prevalence on
specimens
the
of
basis
of
obtained
ivc
aorta
anterior posterior
at
RPLND. Forty-eight percent (20/42) of the patients with vascular invasion of the testicular tumor had positive RPLND, and 33% (9/27) of the patients
without
vascular
invasion
had
positive
RPLND (p = .24, not significant). At our institution, we require RPLND (rather than a surveillance program) as the primary treatment for patients with vascular invasion by the primary tumor. This management policy is based on the resuIts of prior studies, which indicate that the presence
of vascular
invasion
on histologic
exami-
of primary tumors predicts occult metastatic disease [7], although this prediction was not nation
true in our subgroup
of patients.
out vascular invasion
involving
are given program
522
the choice for initial
of RPLND treatment
All
patients
the primary
withtumor
or a surveillance
at our center.
During
Fig. 1.-Drawing
shows classification
scheme for anterior-posterior
lymph node site.
AJR:169, August
1997
CT
of Lymph
Node
Metastases
in Testicular
Cancer
Results The results lustrated Pairwise
ceiver observer
accurate
curve,
Fig.
ues for the other (Fig.
4).
These
for the that one
significantly
observer
(bottom
.01): the comparison beI and 2 remained significant correction
to account
comparisons
[I I]. The
comparisons
were
comparisons
interobserver lymph
Fig. 4) was
a Bonferroni
multiple
10] of the recurves
=
observers
the
I and ii-
showed
another
than 4) (p
after
[
interpreters
(top curve.
more
even
tests
characteristic
radiologist
tween
in Table
2 and 3.
significance
operating
three
for
are reported
in Figures
show
variability
p val-
.24 and .24 significant
.
in measurement
nodes.
The terior
area
estimates
versus
posterior
for the two location
curves, lymph
annodes
at the primary metastatic site, were 0.79 0.51 (Fig. 5): the areas were significantly ferent
(p
and dif-
ing left paraaortic
studies size
node
lymph nodes at retroperitoneal have been interpreted
indicate
criteria
that
as evidence
metastasis
the
use
of
of retroperito-
in this group
a size
scheme
of patients
of 3 mm
criterion
sensitivity,
on
the
developed
terpretation
basis using
rate.
mor in retroperitoneal
et al. [6] found
led to a sensitivity
ity of 67%, 37%
and
Stomper and
lymph
node
had left testicular
germ cell cancer.
that a criterion
of 5
further
improved
lymph
reader
for
in-
static
and applying
a lo-
the
they
threshold
nodes
tu-
at a given
is significantly the expected
with
affected
primary
meta-
site. Application of a 3-mm threshold in predicted metastatic site and a 10-mm
In
addition, of metastatic
diameter
by correlation
resulted
the predicted
outside
in a sensitivity
ity of 52%.
The
use
metastatic
site
of 91 % and a specific-
of smaller
size
criteria
as a
of 7 1% and a specific-
as compared
with
a specificity
of
a a sensitivity
98%
using
of
15 mm.
0
et al. [5] reported a sensitivity of 88% of 44% using a criterion of 5
a specificity
mm for positive of 58% accuracy
et al.
[4]
of CT could
and
fourth-generation
tive
rate
of
33%
no improvement the size
CT findings
and a specificity
Fernandez
cently,
Patient
nor-
of a measurement
of the CT scans
gistic regression model. showed that the probability
Lien
as showing
a single
decreases false-negative findings on CT scans, with an expected increase in false-positive mm
section.
lymph
node dissection. Using conventional 10-mm size criteria,
ing
Previous
neal
metastatic disease at surgery. This patient, with germ cell cancer of left testis, had metastatic disease involv-
3.-CT scan of patient who had lymph nodes measuring up to 9 mm on the scan (arrow) and no evidence of metastatic disease at retroperitoneal lymph node disFig.
mal findings. (Reprinted with permission from [14])
Discussion smaller
Fig. 2-CT scan of patient with 5- to 6-mm lymph node (arrow) in left paraaortic region located anteriorly and
this scan would
.04).
=
.
of
criteria Leibovitch
using
investigated
in this
1994 prior
15 mm.
whether
be improved scanners;
over
and a sensitivity
of 76%
the
C
the
using
third-
false
nega-
study
indicated
results,
although
U0 F-
used were not specified. Reet al. [ 12] reported that us-
0
a
0 C 0.0
0.2
0.4
0.6
0.8
1.0
FPF
Fig. 4.-Graph
shows receiver
operating
characteristic
analyses of three observers
of CT scans. Note that
each point on curves represents y-axis (true-positive fraction, TPF) and x-axis (false-positive fraction, FPF) for given lymph node diameter as revealed with CT. Significant difference exists in accuracies between observer 1 (bottom curve) and observer 2 (top curve). Asterisk = observer 1, plus sign = observer 2, circle = observer 3.
AJR:169, August 1997
523
Hilton
et al.
negative,
the
question
arose
error of measurement
Q
would
render
reader
to reader.
the
small
the
structures
not reproducible
The
significant
operating
comparing agreement
0
the results
receiver
as to whether
of such
from
difference
characteristic
in
curves
two of the readers confirms lack of of exact measurements from one to another.
reader
Thus,
rather
than
relying
Co
0
on specific
solely
U-
should
F-
the
a,
size criteria, the reader suspicious any lymph nodes at
consider
primary
would
C
preset
metastatic
10 mm)
CT criteria, of this
form technique CT scans, most outside
0 C
0.0
0.2
0.4
0.6
0.8
1.0
Fig. 5.-Graph shows receiver operating characteristic analyses of anterior and posterior location lymph nodes (at primary metastatic site). Note that each point on curves represents y-axis (TPF) and x-axis (FPF) for given lymph node diameter on CT. Asterisk = anterior size, plus sign = posterior size.
institutions.
would
tives
been study when
that a CT scan
in this group
addressed validates small size
is posi-
of patients
has
only recently; the present improvement in sensitivity criteria are used and reveals
metastatic
disease
a given nodes
lieved given
in
patient.
retroperitoneum
the
We
did
size
not
record
ing.
Also,
ably
distinguished
tice, with particular attention to lymph located anteriorly, should result in better
Our results show CT identification node metastases
that false-negative of retroperitoneal in this patient
can be decreased
from
was
for nonoperative
The purpose to correlate
CT scans
of the present investigation the size of lymph nodes on
at the expected
sites
with
logic
results
selec-
surveillance.
positive
primary
or
metastatic
negative
of RPLND
for
metastatic
dis-
ease on a per-patient (rather than per-lymph node) basis. No attempt was made to correlate
individual
lymph
nodes
at specific
sites
nodes examination.
idenThe
on CT scans with specific tified at histopathologic
lymph
specimens
are sent
from
thologist tiple
RPLND
as en bloc
lymph
nodes,
site-specific
vidual
containing
a method
that
histopathologic
lymph
lymph nodes
nodes. measured
to the pa-
packets
the
mul-
precludes
analysis
In fact,
mdi-
of
individual
on CT scans
may
not
actually
represent
the identical
that
positive
or negative at histopathorather, these results show
are
logic
examination;
the effect accuracy
524
lymph
a size
nodes
of varying the size criteria on the of CT in predicting the presence of
nodes
from
primary
site
blood
using 4
mm
imagbe relivessels.
rates for lymph population
to as low
of
visualized
metastatic
cannot
small
63%
criterion
Importantly,
histopatho-
lymph
in standard
terion of 10 mm or larger ing
power
a size
cri-
as 7%
us-
or
ofhistologically
lymph nodes that are more
to determine
normal
lymph
the
nodes.
from
That his-
comparisons
of clinical
stage
remain a problem, and other reports. diagnosed
testicular
being
considered
as
stage
1 , size
rently ment
accepted 10-mm for lymph nodes
criteria
of
the small
criteria
other
than newly
ticular
diagnosed
nonseminomatous
tients would
lymphadenopathy to any
population
clinical
stage
germ
cell
cancer
I tespa-
be inappropriate.
In this study, in which gradations in measured size of relatively small lymph nodes were used to classify findings as positive or
number
the
pected,
the result
criteria
is a decrease
when
a 4-mm
surveillance
NSGCT
NS-
than
in this with who are clinical the
cur-
short-axis measurecould be used to
of patients disease
with
these
in-
As
ex-
smaller
in specificity is used.
offered
occult
treated
surveillance.
of using
be
at
smaller
criterion will
in the
1 testicular
possibly
with
for
oper-
be largely
as evidenced For patients
newly
appropriately
study
receiver should
curves
is well known.
present
and of our
unaffected by these selection biases. False-negative results of CT scans
study
fur-
lymph
the validity
and the estimated
characteristic
ter on CT scans size
targeted might
of sensitivity
However,
metastatic
to apply
For ex-
vessels.
retroperitoneal
Thus,
indicate.
exclusion of patients who RPLND could have affected
of our estimates
GCT
size
blood
as at all
A smaller,
[13].
at the anteri-
CT
increase.
specificity
evaluation
rebe
identified be lower
of the retroperitoneum accuracy in distinguishing
the validity
ating
in dis-
been performed of IV contrast material,
minimize
in the
IV
had
tologically normal retroperitoneal lymph nodes may measure at least I cm in short-axis diame-
used
for
accuracy
our results
injection
The selective did not receive
larger.
orly located should, on the basis of the ceiver operating characteristic analysis, viewed as especially suspicious. We did not attempt
of uni-
rates
actually
case
might
nodes
of CT
the field
tion of patients
with
specificity
lymph
scanners
resolution
and
small
is lack
mistakenly
every
the
the degree of interobserver variability that can occur when these size criteria are applied. Use of these revised size criteria in clinical pracnodes
of
than
if
field of view ther improve
used
aorta.
lymph nodes from blood the number of false-posi-
could
criteria
we bereliably,
of view
those
ofthe
Injection
vessels
nodes)
smaller than 4 mm, because these could not be measured spatial
study
be to diminish
(blood
lymph ample,
for deciding
especially
in obtaining the abdominal of which were obtained at
tinguishing small vessels; therefore,
FPF
tive for metastasis
that
contrast material could not be determined. The expected effect of a suboptimal CT technique
threshold
those
(i.e., less than
to the mid portion
anterior
A limitation
cJ
even
unenlarged
by usual
located C
site,
be considered
size
to 58% Given
at some
that
centers
to patients with stage 1 disease, however, the percentage of patients who have an unnecessary RPLND (false-positive results of CT)
as a result
of use
of these
smaller
size
AJR:169, August 1997
CT criteria mm
will size
still
be much
lower
(42%
for adenopathy)
criterion
at a 4than
of Lymph
4. Fernandez McLeod
the
as an option.
phy
JA, Foster
Rowland RG, Einhom toneal lymphadenectomy cancer:
the Indiana
RS, Bihrle R,
LH. The role of retroperiin clinical stage B testis
University
experience.
J Urol
1995:153:85-89
2. Sogani
PC, Whitmore
WF,
Herr
HW,
et al. Or-
chiectomy
alone in the treatment of clinical stage I nonseminomatous germ cell tumor of the testis.
J Cliii
1984:2:267-270 JP, Thornhill JA,
RG, Bihrle
for clinical modification
tion.
PC, Fung
J Urn!
AJR:169, August
Foster
RS,
Rowland
R. Retroperitoneal lymphadenectomy stage A testis cancer (1965 to 1989): of technique and impact 1993:149:237-243
1997
MB,
on ejacula-
JW,
Foley
JP,
imaging
Cancer
Colon with
I nonseminomatous
E,
9. Donohue JP, Zachary tion of nodal
third germ
tis cancer. 10. Metz
1994;44:548-552 CY, Socinski Richie
MA,
JR Detection
Jochelson
between
metastases
JM. Maynard
BR. Distribuin nonseminomatous tes-
J Uro! 1982:128:315-320
CE,
proach
metastases in early-stage testicular cancer: analysis
Wang
for testing ROC
PL,
Kronman
HB.
the significance
curves
A new
ap-
of differences
for correlated
data.
In: De-
II.
F, ed. Information prncessing in medical imaging. The Hague: Nijhoff. 1984:432-445 Altman DO. Practical statistics for medical ,r-
12.
search. London: Chapman and Hall, 1991:210-212 Leibovitch I, Foster RS, Kopecky KK, Donohue
conick
of retrononsemiof different
CT criteria. AiR 1987:149:1187-1190 6. Lien HH, Stenwig AE, Ous 5, Fossa SD. Influence of different criteria for abnormal lymph node size on reliability of computed tomography in patients with non-seminomatous testicular tumor. Ada Ra-
JP. Improved accuracy of computerized tomography based clinical staging in low stage nonseminomatous germ cell cancer using size criteria of
dio!Diagn
retroperitoneal
(Stock/i)
1986:27:199-203
7. Moul JW, McCarthy WF, Fernandez EB, Sesterhenn IA. Percentage of embryonal carcinoma and
of vascular invasion
Oncol
3. Donohue
stage
Urn!ogy
Garnick
peritoneal nomatous
1. Donohue JP, Thornhill
Moul
in Testicular
generation computed axial tomogra-
in clinical
Stomper MS.
References
EB,
cell tumors. 5.
Metastases
DO. Retropentoneal
and fourth
70% who would have received unnecessary RPLND at centers that do not offer surveillance
Node
predicts
clinical stage I nonseminomatous cer. Cancer Res 1994:54:362-364 8. Biswamay R, Hajdu tion of retroperitoneal ticular
germinal
pathological
stage
testicular
in
can-
SI, Whitmore WF, Jr. Distribulymph node metastases in tes-
tumors.
Cancer
1974:33:340-348
lymph
nodes.
J
Urn!
1995:154:
1759-1763
I 3. Begg CB. Biases
in the assessment
of diagnostic
tests. Stat Med 1987:6:41 1-424 14. Hilton S. Schwartz LH. Radiologic genitourinary neoplasms. In: Raghavan HI, Leibel SA, Lange P. eds. Principles tice ofgenitourinarv oncology. pincott-Raven, 1997:41
imaging
of
D, Scher and
Philadelphia:
prac-
Lip-
525