Abdominal Jose F. Botet, MD Carlos Urmacher,

#{149} Charles J. Lightdale, MD #{149} Ann MD #{149} Murray F. Brennan, MD

Preoperative Comparison Fifty

with

patients

proved by means went preoperative scopic

(US);

in 42 of

the patients, dynamic CT of the chest and abdomen was also performed. All results were compared with the findings at pathologic examination of resected specimens. In staging the depth of tumor growth, endoscopic US was significantly more accurate (46 of 50 tumors [92%1) than CT (25 of 42 tumors [60%J) (P < .0003). In staging regional lymph nodes, it was more accurate (44 of 50 patients [88%]) than CT (31 of 42 patients [74%]), but this was not statistically significant. In staging distant metastases,

rate

however,

CT was

(38 of 42 patients

doscopic

US

more

[90%J)

accu-

than

(35 of 50 patients

en-

[70%])

(P < .016). The highest concordance with surgical and pathologic findings in overall stage (36 of 42 tumors [86%]) occurred with the combined use

of CT

was

and

endoscopic

US,

significantly

more

accurate

I

From

the

#{149} Hans

of Medical

of the esophagus is a relatively uncommon neoplasm in the United States, and approximately 10,000 new cases were expected to be diagnosed in 1990 (1). ARCINOMA

This carcinoma

now

accounts

for

about 1 % of all malignancies and 2% of all cancer-related deaths. A cumulative literature review by Earlam and Cunha-Melo in 1980, including 83,783 surgical

cases,

reported

an

overall

5-year survival rate of 4% (2). In the United States, most patients have advanced disease at the time of initial diagnosis (3). The prognosis after surgical resection is closely related to pathologic stage (4). This

prospective

study

was

de-

signed to compare the accuracy of endoscopic ultrasonography (US) in preoperative clinical staging of esophageal cancer with that of dynamic computed tomography (CT) with use of pathologic examination of resected specimens as a standard.

Study

181:419-425

Departments

C

PATIENTS

Computed tomography (CT), preoperative #{149} Endoscopy, 718.12981 #{149} Esophagus, neoplasms, 71.321 #{149} Lymphatic system, neoplasms, 996.8323 #{149} Ultrasound (US), comparative studies #{149} Ultrasound (US), tissue characterization, 718.321 1991;

PhD

than

terms:

Radiology

G. Zauber,

which

use of CT alone (27 of 42 tumors [64%]) (P < .008).

Index

Gastrointestinal

Staging ofEsophageal ofEndoscopic US and

esophageal cancer of biopsy understaging with endo-

ultrasonography

and

Imaging

(J.F.B.), Medicine, Gastroenterology Service (C.J.L., HG.), Epidemiology and Biostatistics (A.G.Z.), Pathology (CU.), and Surgery (M.F.B.), Memorial Sloan-Kettering Cancer Center and Cornell University Medical College, 1275 York Aye, New York, NY 10021. From the 1988 RSNA scientific assembly. Received March 6, 1991; revision requested April 24; revision received June 18; accepted June 19. Address reprint requests toJ.F.B. 0 RSNA, 1991

See also the article by Botet et al (pp 426-432) and the editorial by Baker and Kopecky (pp 342-343) in this issue.

AND

METHODS

Group

Radiology

Gerdes,

MD

Cancer: CT’

Dynamic

screening chemistry

hematologic tests.

tests,

and

The study group consisted and 14 female patients with

blood

of 36 male a mean age of

60.3 years (median age, 60 years). They had 12 epidermoid carcinomas and 38 adenocarcinomas. Location of the tumor in

the esophagus was divided into upper (if the tumor was 10-19 cm from the incisors), middle lower

One

(20-29 cm from the incisors), and (more than 30 cm from the incisors).

tumor

(2%)

was

in the

upper

third,

eight tumors (16%) were in the middle third, and 41 tumors (82%) were in the lower third of the esophagus. All imaging findings were compared

with those obtained at surgery logic evaluation of the resected

and pathospecimens

with regard to depth of penetration of the tumor through the esophageal wall, regional lymph node metastasis, and distant metastasis, according to the 1988 staging classification of the American Joint Committee on Cancer (AJCC) (Tables 1, 2) (5).

Endoscopic

US

We used a commercially available US endoscope with a 7.5-MHz radial transducer (GF-UM2; Olympus, Lake Success, NY) and one with a switchable 7.5- and 12-MHz transducer (GF-UM3, Olympus).

The scanning

plane

of these

instruments

1988, 50 consecutive patients who underwent surgery for epidermoid carcinoma or adenocarcinoma of the esophagus enrolled in a prospective study for preoperative evaluation and staging of their disease. All patients were selected with the understanding that surgery was planned for either paffiation or attempt at cure.

was orthogonal to the axis of the insertion tube, and the field of view was 360#{176}. The images were obtained in real time with a scanning speed of 10 revolutions per second. A balloon with a radius of 1.5 cm, which could be filled with water for better acoustical contact, was placed over the transducer head. The maximum resolution with the 12-MHz transducer was less than

Evaluation

0.5

Between

December

and staging

1986 and December

included

endos-

copy,

endoscopic US, and, in 42 of the 50 patients, dynamic CT of the chest and abdomen. Eight patients had previously Un-

dergone

non-contrast

material-enhanced

CT performed elsewhere. Their scans were not included in our study because we considered it essential to have standard, high-quality CT scans for comparison. All endoscopic US scans were obtamed prior to dynamic CT. All CT scans

were

interpreted

by radiologists

rable experience who results of endoscopic also included physical

mm,

and

with

the

7.5-MHz

transducer,

1 .0 mm. The scanning radius was a maximum of 7 cm at 7.5 MHz and was reduced to 3 cm at 12 MHz. The instrument was introduced by an endoscopist with a method similar to the method used with conventional endoscope and was advanced to the gastric antrum. Sonographic evaluation was performed as the instrument was withdrawn. Anatomic land-

a

of compa-

were blinded to the US. Clinical staging examination,

Abbreviations:

mittee

on Cancer,

AJCC

=

American

CI = confidence

Joint Corninterval. 419

marks

(6) were

guide

the

used

by the radiologist

positioning

Images were vals but were

of the

not obtained chosen during

to document

amination

to

Table 1 AJCC Staging

endoscope. at fixed interreal-time cx-

the maximum

US.

Five

layers

of alternating

ies,

the

first

Primary

two

layers

to the

balloon

Regionallymph

Distant

(adventitia)

through

the

as T3 (Fig 3a), and

fifth

TX TO

Primary tumor cannot No evidence of primary

Tis

Carcinoma

Ti T2 T3 T4

Tumor Tumor Tumor Tumor

NX NO Ni

Regional lymph nodes cannot be assessed No regional lymph node metastasis Regional lymph node metastasis

MX

Presence

in situ

invades invades invades invades

lamina propria or submucosa muscularis propria adventitia adjacent structures

metastasis/M of distant

metastasis

cannot

be

assessed MO Mi *

In the cervical

distant.

In the

all others,

esophagus,

thoracic

including

regional

esophagus,

lymph and

nodes

are the supraclavicular

lymph nodes are mediastinal

regional

supraclavicular

No distant metastasis Distant metastasis

celiac

lymph

nodes,

lymph

nodes;

and perigastric

all others

lymph

are

nodes;

are distant.

2

3

layer

invasion

be assessed tumor

nodes/N*

method,

from the upper esophagus to the cardia. Alterations in thickness of individual layers are easily and reproducibly identified. Esophageal cancer was diagnosed as a hypoechoic disruption of the layers. Involvement of any or all of the three inner layers (mucosa and submucosa) was interpreted as TI (Fig 2a), invasion of the fourth layer (muscularis propria) as T2 invasion

Criterion

tumor/T

hyperechoic

correspond

US with the water-filled is approximately 3 mm

2b),

Stage

layers can usually be imOn the basis of in vitro stud-

interface and mucosa, the third layer corresponds to the submucosa, the fourth layer to the muscularis propria, and the fifth layer to the adventitia/serosa (7) (Fig 1). The normal thickness of the esophageal wall, as observed by means of endoscopic

(Fig

Cancer

cx-

tent of disease. Individual layers of the gastrointestinal wall can be visualized with endoscopic and hypoechoic aged in vivo.

of Esophageal

Tissue/Symbol

of 6 5

Table

2

Stage

Grouping

in the TNM

System

Tumor

Node

Metastasis

0

Tis

NO

MO

I

Ti

hA

T2 13 Ti T2 T3 T4

NO NO NO Ni Ni Ni AnyN

AnyT

AnyN

MO MO MO MO MO MO MO Ml

Stage

IIB III

IV

(a) and US scan (b) depict five-layered internal structure normal esophagus. I = mucosa (hyperechoic), 2 = deep mucosa (hypoechoic), (hyperechoic), 4 = muscularis propria (hypoechoic), 5 = adventitial interface and 6 = transducer.

1.

Diagram

b.

a. Figure

(a) Endosonogram shows Ti tumor (t) that dosonogram shows T2 tumor (f) that invades mucosa, row). (c) CT scan of Ti-T2 tumor (t) shows circumferential face.

420

b.

a. Figure

2.

#{149} Radiology

invades

deep

of the

3

wall

of the

= submucosa

(hyperechoic),

C.

the

mucosa, thickening

mucosa

and

and

deep

submucosa

of esophageal

mucosa.

Notice

the

intact

but does

not extend

through

wall

no definite

extension

but

submucosa

(arrow).

the muscularis beyond

(b)

propria

adventitial

November

En-

(arinter-

1991

adjacent

structures

or organs

as T4 (Fig

4a).

The lymphatic drainage areas routinely investigated are the peritumoral area, paratracheal area, aorticopulmonary window, subcarinal area, crural area, celiac axis, splenic vein, portal vein, and gastrohepatic

ligament

nodes were (Ni) if they

(Fig

from surrounding echoic, or if they

characteristics mary cancer The AJCC

a. Figure 3. (a) Endosonograph shows T3 tumor but does not involve any adjacent structure. A

to the

disease

(Fig

diameter. does not

right

b. (T) that

extends

beyond

adventitial

interface

tissue, showed

celiac

7a).

round, or sonographic

lobe

because

axis

as distant

heft

hepatic

The

scanned of distant

hepatic

imaged

lymph

hypo-

similar to those of the pri(Fig 6a, 6b). classifies lymph node me-

tastases

routinely evidence

5). Regional

considered positive for cancer were sharply demarcated

(MI) lobe

was

with endoscopic US for metastasis. Most of the cannot of the

be limited

satisfactorily penetration

LA = left atrium, N = node 1.9 cm in to extension beyond adventitia, but tumor A = aorta, LA = left atrium.

(b) CT scan of T3 tumor. Arrow invade surrounding structures.

= aorta,

points

Tmcsal

I.,.

uodss

TrUcSS-bS’SScMaI

Mds$

uodss

usds

ClUE

b.

a. Figure

shows

4.

(a) Endosonograph T4 tumor (T) that

shows

invades

T4 tumor

the aorta

that

(a) Endosonograph

dosonograph

of malignant

Volume

wall

(arrow).

(b) CT scan

b. 6.

esophageal

Figure aortic

(a).

a. Figure

invades

node

181

1.2 cm

#{149} Number

of benign

node

7 mm

nodes

in diameter

(N)

7 mm

axis nodn

in diameter. (arrow). Note that

5. Lymphatic drainage of the esophagus. Endoscopic US can enable evaluation of all nodes depicted. Jut. = internal.

C.

Note that they they are round

are elongated and hypoechoic

and hyperechoic (at 12 MHz).

(at 12 MHz). (b) En(c) CT scan shows para-

in diameter.

2

Radiology

#{149} 421

of high-frequency

ultrasound.

metastases

be evaluated

cannot

Pulmonary with

en-

doscopic US. In nine tumors, no difference in staging resulted from examination with both 7.5and 12-MHz transducers. The 12-MHz transducer did provide increased detail of the esophageal wall but did not result in any change in evaluation N and M was best assessed

MHz

transducer

depth

of field.

Dynamic

because

of

with the 7.5of the greater

of Chest

CT

Upper

of T. Staging

and

Abdomen

This technique was performed proximately 150-200 mL of 60%

with apiodinated

contrast material given intravenously with a commercially available injector (Mark V; Medrad, Pittsburgh) at a rate of 1.0 mL/sec for the first 60 seconds and then at 0.7 mL/ sec

for

the

remainder

all 42 patients

CT, scans

of the

who

were

underwent dynamic by means of a

International,

Ohio)

with

tions

10-mm-thick

from

the

(GE 9800: or 1200

Highland margin

to the thoracic inlet. These received 200 mL of diluted

GE SX:

Heights,

contiguous

inferior

sec-

of the

liver

patients also diatrizoate

meglumine (Gastrografin; Bristol-Myers Squibb, Princeton, NJ), which was administered orally, to minimize contrast inter-

face

artifacts.

tention

We chose

or change

patients during The thickness wall is variable

the

not

to use gas dis-

position

of these

or after dynamic CT. of the normal esophageal because of the distensibil-

ity of its lumen. We used 5 mm as the upper limits of normal; any increase beyond this was considered abnormal. The thickness of individual layers of the esophageal

wall by

cannot means

be determined of CT.

The

with

following

(Fig

cent

3b); and structures

T4,

wall greater of the outer

than 15 margin

tumor invasion of adjasuch as the trachea, aortic

pericardium, or vertebral body (Fig 4b). Lymph nodes greater than 10 mm in diameter were considered abnormal, a widely used standard (Fig 6c). The drain-

age

areas

examined

were

identical

those examined at endoscopic determined according to the

Note

relationship

metastatic

US. AJCC

to M was classifi-

cation (eg, metastasis to any organ, including lung, or nonregional lymph nodes such as mesenteric or celiac axis nodes [Fig

racoabdominal patients with esophagus agectomy. pect in the sected, identified

underwent Lymph diagnostic

for

Pathologic

Procedures lesions of the upunderwent right

#{149} Radiology

Of the 38 30 underwent

of a combined

(arrow) arteries

9 mm

in diameter

in the

celiac

(b) CT scan of possible

(ART).

attempted.

whereas eight lesions of the distal

nodes

transhiatal identified studies were

esophas susre-

nodes modali-

not

in the were ob-

The resected specimens were examined with special attention to the depth of invasion of the esophageal wall as well as to evidence of invasion into adjacent struc-

tho-

Figure operative ELIS

8.

=

Diagnostic staging

algorithm

of

endoscopic

esophageal

for the precancer.

US.

tures and organs. All peritumoral and resected lymph nodes were examined. Their diameters ranged from 3 mm to 2.6 cm.

Statistical

dent on CT scans in two patients biopsy samples were not obtained

Methods

in whom at sur-

gery.

The

percentage of cases in which endoscopic US findings were in agreement or concordant with the pathologic classification after surgery is given for T, N, and stage in Tables 3-5. A 95% confidence interval (CI) about the percentage of concordant findings is also given, as are the percentage of CT findings concordant with pathologic examination and the concomitant 95% CI. The McNemar test (8,9) for matched was used significant centage

ings

were

pairs with to assess

continuity whether

correction a statistically

difference existed in the of cases in which pathologic

concordant

Surgical

422

aggressively

(HEP)

Examination

findings in the both endoscopic determination

by means

hepatic

node

confirmation.

findings comparisons

exploration

was

samples of metastasis or upper abdomen

dant with statistical

and laparotomy. distal lesions,

metastatic

and

as were other suspected by use of the imaging

ties. Biopsy mediastinum tamed

approach, localized

7b]).

with

splenic

Diagnosis

US and pathologic

thoracotomy patients

to

lesions.

doscopic in which

The 12 patients with per or middle esophagus

of Mi

certainty criteria

therefore were used to determine the T component. It was not possible to differentiate TI from T2 disease. TI and T2 disease was thickening of the wall greater than 5 mm and less than 15 mm (Fig 2c); T3, thickening of the mm with irregularity

axis.

b. (a) Endosonograph

7.

In

obtained

commercially available unit Medical Systems, Milwaukee;

Picker

injection.

a. Figure

with the

findings

percentage findings at dynamic were

42 patients US and of whether

These on

who underwent CT. Also, in our endoscopic US

respect to concordance of TI and T2 were

combined

CT

follow-up

because

between confirmed

cannot

enable

TI and lung

dif-

T2. Clinical

metastases

2 test (8) was also used a significant difference ex-

are

at pathologic test gave

examination. similar results

test for the binomial distribution (P = .5); only the McNemar given

in the

The as the

results

text.

at en-

and CT differ with for T, the categories ferentiation

findings McNemar testing

perfind-

McNemar whether

isted in direction, either overstaging or understaging, in the patients in whom findings at endoscopic US were discordant with findings at pathologic examination and whether such direction existed in findings at dynamic CT compared with

exact

of cases were concorCT. based

The to test

cvi-

RESULTS Tumor

Growth

In 46 of 50 patients with esophageal cancer (92%), endoscopic US findings were concordant with pathologic findings for T (95% CI = 84%, 100%) (Table 3). When Ti and T2 classifications were combined, findings at endoscopic US were concordant with findings at pathologic examination for

November

1991

Invasion

in Esophageal

Findings

Dynamic

CT

.

versus

Cancer. Concordance of with Pathologic Oassification

Findings

Endoscopic Pathologic

Classification of Tumor

TO

Ti

T2

2*

j

Ti T2 T3

Note.-Numbers

are

of tumors.

number

Findings

were

concordant

in 46 of 50 tumors

t Findings

were

concordant

in 25 of42

US and dynamic

scopic

Table

.

T4

1 19*

(92%). tumors(60%).

Endoscopic Dynamic

US 13

2*

T4 *

TO

1 1 23

Ti

CT

T2

T3

14

it it 1

2 1 16t 10

7t

1 I

1

.

P

.0003 in 42 patients