Michiel W. M. van den Brekel, MD #{149} Herbert V. Stel, MD, PhD Jos J. P. Nauta, MSc #{149} Isa#{227}kvan der Waal, DD, PhD #{149} Jaap Valk, Chris J. L M. Meyer, MD, PhD #{149} Gordon B. Snow, MD, PhD

Cervical Assessment To estimate ent radiologic

Lymph Node of Radioloic

the

accuracy of differcriteria used to detect cervical lymph node metastasis in patients with head and neck carcinoma, seven different characteristics of 2,7i9 lymph nodes in 7i neck dissection specimens from 55 patients were assessed. Three lymph node diameters, their location, their number, the presence of a tumor, and the amount of necrosis and fatty metaplasia were recorded. The minimal diameter in the axial plane was found to be the most accurate size criterion for predicting lymph node metastasis. A minimal axial diameter of 10 mm was determined to be the most effective size criterion. The size criterion for lymph nodes in the subdigastric region was i mm larger (ii mm). Groups of three or more borderline nodes were proved to increase the sensitivity but did not significantly decrease the spedficity. Radiologically detectable necrosis (3 mm or larger) was found only in tumorous nodes and was present in 74% of the positive neck dissection specimens. Shape was not a valuable criterion for the radiologic assessment of the cervical lymph node status. Index

terms: Lymphatic system, CT, 276.121 1 Lymphatic system. MR studies, 276.1214 #{149} Neck CT, 28.1211 #{149} Head and neck neoplasms, 262.373, 271.373, 276.373, 28.314 #{149} Neck, MR studies, 28.1214 Radiology

1990;

177:379-384

I From the Departments of Otorhinolaryngology and Head and Neck Surgery (M.W.M.v.d.B., J.J.P.N., G.B.S.), Radiology (J.A.C., J.V.), Pathology (H.V.S., C.J.L.M.M.), and Oral Pathology (I.v.d.W.), Free University Hospital. P0 Box 7057, 1007 MB Amsterdam,

The Netherlands.

Supported

by grant

IKA 88-

19 from the Queen Wilhelmina Fund. Received April 2, 1990; revision requested May 24; revision received June 18; accepted June 27. Address reprint requests to M.W.M.v.d.B. C RSNA, 1990

#{149} Jonas

MD,

A. Castelijns, PhD #{149}

of the status of the cervical lymph nodes in patients with primary squamous cell carcinoma in the head and neck region is difficult. The overall error rate (falsepositive cases + false-negative cases! total number of cases X 100) in assessing the presence or absence of cervical lymph node metastasis by palpation ranges from 20% to 28% (1,2); therefore, the treatment of patients with a clinical stage NO neck is controversial. If these patients are at great risk of having occult cervical lymph node metastasis, most head and neck surgeons advocate elective treatment of the neck (3,4). However, of these

elective

neck

dis-

sections prove to be tumor-free at histopathologic examination (3,4). On the other hand, up to 32% (5) of patients with clinically negative nodes who are not treated develop lymph node metastasis in the neck, depending on the site of the primary tumor. In most studies, the use of computed tomography (CT) compares favorably to palpation in assessing the status

.

of the

PhD

#{149}

Metastasis: Criteria’

SSESSMENT

75%-77%

MD,

lymph

nodes;

the

overall

error rate ranges from 7.5% to 19% (6-9). To our knowledge, only one publication cites an overall error rate of 28% with use of CT (10). Many different radiologic criteria are being used to assess the presence or absence of metastasis in cervical lymph nodes (6-i3). These criteria involve the maximal axial diameter (6-13), the irregular enhancement due to tumor necrosis (6-il), the shape (8,i 1,13), and the grouping of nodes (7-9,11). The differences in criteria used account in part for the difference in reported overall error rates. Most authors defined their criteria by retrospective radiologichistopathologic correlation. To our knowledge, different radiologic criteria were compared in only one study (8). These varying, and often

inadequate, criteria flated overall error Retrospective

may rate.

comparison

cause

an inof CT

and magnetic resonance (MR) images with histopathologic findings of neck dissection specimens has many disadvantages.

For

instance,

metastat-

ic and reactively enlarged nodes found in the specimen cannot always be identified accurately on the images. Whether or not these nodes can be identified at CT or MR depends on the skill of the radiologist and the information provided by the pathologist about the exact location and size of the nodes. Information about the size of the node is rarely supplied. Volume averaging effects, patient compliance, and selected techniques (eg, section thickness, contrast material administration, and dosage) may influence the measurements obtamed at CT and MR and make them not only less reproducible, but also more difficult to compare with those of other studies. Furthermore, the longitudinal diameter cannot be accurately measured on axial images. To avoid the previously mentioned disadvantages and to prospectively evaluate all morphologic characteristics of metastatic and reactively enlarged lymph nodes, these characteristics were studied in 2,719 lymph nodes from 71 neck dissection specimens. This enabled us to define the validity of different radiologic criteria such as size, shape, and grouping of lymph nodes. Regional variations in lymph node size and the prevalence of tumor necrosis and adipose metaplasia were also assessed. Knowing the maximal accuracy and the negative and positive predictive values of radiologic techniques may help to select patients who should be treated.

Abbreviation:

ROC

=

receiver

operating

characteristic.

379

sx.min.

#{149}x.max.

,.

long.

.

--. --

,

100 70

-

--

#{149}0

,

--

/

I

,

sx.msx.

-

I,

#{149}0 C C C

..

sx.mln.

.

long.

40 - - 30

50 20

S

10

IS

20

25

diameter

30

35

40

(mm)

1

2

3

4

I 00

Figure

1. Percentage of nodes malignant versus lymph node diameter. Note that the minimal axial diameter curve has the steepest course. With all three measurements, all nodes are malignant beyond a certain diameter (see also Table 1). In this and subsequent figures, ax. mm. minimal axial diameter, ax. max. = maximal axial diameter, long. = longitudinal diameter.

MATERIALS Within

AND

a period

dissection

of

from

18 had ryngeal cally

patients

proved

71

neck

and

upby

(a structure

cell

imen

was

ethanol

neck

Minimal

specimens

show lymph

Table 2 Percentage

in

a solution

absorption diagnost

as fat (14). U-M (Philips

Shelton,

Conn)

tion. After

ble

all

nodes

and,

vein. same

if

A sox-ray

expo-

The

radiograph is especially

palpable

and/or

dissected

obusecorrelavisi-

from

the

Note-One tive, 33 were

eter

jugular

perpendicular

dinal

sected

were

lymph

nodes.

diameter

380

diameters

diameter

corresponds

Radiology

#{149}

80

90

100

(%)

the relation

between

the sensitivity

and

specificity

of

and

the

measured

The

of Nodes

That

Are Malignant (mm)

11-15

16-20

21-25

26

1 (20/1,895) 1 (23/1,935) 2 (44/2,411)

6 (28/440) 8 (49/615) 23 (58/257)

16 (38/240) 33 (43/130) 75 (27/36)

29 (26/91) 60 (12/20) 100 (8/8)

47 (14/30) 82 (9/il) 100 (4/4)

78 (18/23) 100 (8/8) 100 (3/3)

longitufor

minimal

to the

are percentages. of that size.

Numbers

in parentheses

and Specificity

widest

all

(mm)

are the number

of malignant

nodes/total

of Size Criteria

dis-

axial diam-

of the

eter. the

The internal

ameters

Node

Sensitivity

Per

Specimen

Specificity

Sensitivity

Specificity

69.4 54.2 47.9 41.7 29.2 25.7

91.9 96.8 98.6 99.3 99.7 99.9

100 97 89 89 79 66

12 21 58 73 85 94

72.9 59.0 42.4 30.6 27.1 20.1 13.9

85.9 93.3 97.3 98.9 99.5 99.6 99.8

100 97 89 82 71 58 47

0 9 24 48 73 79 88

72.9 56.3 49.3 27.8 15.3 10.4

85.0 92.5 95.0 98.8 99.6 99.9

100 97 97 74 53 32

3 18 24 55 88 94

axial

axial

i6 20 Longitudinal i0 i3 15 20 25 30

specimen, measured, and cut into 2-4mm-thick slices in the axial plane for microscopic examination. The minimal and maximal axial (perpendicular to the course of the internal vein)

70

6-10

15

MammoSystems,

programmed

were

to

of the to the

nodes,

A Philips Medical

with

fixation,

lymph

and size relation jugular has the

sure control was used. tamed with this method ful for radiologic-histopathologic

Minimal 6 8 9 10 i1 i2 Maximal 8 lO l2 i4

of 96% obtained

other

the internal of 96% ethanol

60

5

Sensitivity

Diameter

dis-

the spec-

was

the exact location nodes and their

resected, lution

After

opacity),

gland,

show

Per

muscle

a radiograph

submandibular

50

specificity

were

a high

immersed

axial

Note-Numbers number of nodes

dissecNone of

sternocleidomastoid

and

curves

Percentage

Maximalaxial

55 comprehen-

for 36 hours. with

Versus

Longitudinal

photographed (Polaroid and slides) postoperatively. Next, the specimen was nailed to a board and fixed in a mixture of of the

(2) ROC

Diameter

the patients had previously undergone a neck dissection on the affected side or received preoperative radiation therapy or

section

40 -

bilateral

Overall,

4% formaldehyde

30

I 00

Size

origin. Thirtyunilateral neck

dissection

2, 3.

Table 1 Diameter

10 had lahad cytologi-

and 16 supraomohyoid specimens were examined.

chemotherapy. All neck

20

(%)

the different size criteria per node. Note that the minimal axial diameter is the most valid criterion because the area under the curve is the largest. The high specificity (x axis) is caused by the enormous number of true-negative cases for all values (see also Table 2). (3) ROC curves show the relation between the sensitivity and specificity of the size criteria per neck dissection specimen. The minimal axial diameter is again the most valid criterion. If the criterion of groups of three or more borderline nodes is added, the sensitivity of the minimal axial diameter criterion increases slightly near the y axis (high specificity). See also Tables 2 and 3.

carcinoma,

16 underwent

dissections.

specificity

-

10

#{149} 1011121314

B

3.

Figures

neck

of a squamous

of unknown underwent

dissections

oral

carcinoma, and two

metastasis

carcinoma nine patients

sive tion

had

pharyngeal carcinoma,

7

2.

55 patients

with squamous cell carcinoma of the per aerodigestive tract were examined the same investigator (M.W.M.v.d.B.). Twenty-five

5

METHODS

10 months,

specimens

1

hundred negative.

node

forty-four

in the axial to the

of a node

plane

maximal

longitudinal jugular

nodes

were

If all less

that is diamparallels

axial

diameter vein.

were

than

three

di-

5 mm,

positive.

the

2,575

were

negative;

size

was

not

further

38 specimens

were

specified.

posi-

The

shape of the node was defined as the maximal axial diameter divided by the minimal axial diameter. To measure the possible shrinkage of

November

1990

Table 3 Percentage

Sensitivity

and Specificity

of Some

Modified

Minimal

Axial

Minimal is

Diameter

Sensitivity

(mm)

8 9 ?:il 12 S

Three

Specificity

on more

borderline

nodes

Specificity

92 89 82 71 61

(1 or 2 mm smaller

than

Criterion

axial

diameter)

41

Sensitivity

Specificity

Sensitivity

Specificity

97 92 92 87 84

21 58 73 85 94

95 95 87 87 82

30 58 76 94 100

36 70 82 94 iOO

the minimal

per Side

31

Sensitivity

21 52 67 85 94

Subdigastnic nodes 1 mm larger. Minimal axial diameter combined I Subdigastnic nodes 1 mm larger

Axial Diameter

Criteria

2t

97 92 89 79 74

10

Diameter

grouped

in tumor

drainage

region.

t

benign to the

and malignant fixation procedure,

with necrosis (3 mm). and three or more grouped

lymph nodes due three metastatic

and three benign lymph nodes sected from random specimens measured

before was

and

after

measurements

The location

and the sia inside

of axial

diameters to the

of all nodes

examination,

nodes,

diame-

was indicated

number

size

of adipose recorded.

foci,

metapla-

operating characteristic These curves represent

is represented

the ROC curve; better the test.

Sensitivity

by

the

of differcharted as

(ROC) the rela-

area under

the

larger

the

area,

the

177

combined.

(validity),

(trueX 100.

as follows:

Number

#{149}

2

Lymph

Nodes

The mean number of nodes med from the 55 comprehensive neck dissections was 45 (range, 74),

while

were neck the

an

average

exam20-

of 16 nodes

found in the 16 supraomohyoid dissections (range, 7-41). Out 2,719

lymph

nodes

examined,

of 144

nodes in 38 specimens from 34 patients contained metastatic squamous cell carcinoma. The number of metastatic nodes in the positive comprehensive neck dissections ranged from one to 23 (mean, 3.8). Only one of the i6 supraomohyoid neck dissections contained metastatic nodes (two nodes were positive). Node

Size

Table 1 and Figure 1 show the number and percentage of malignant lymph nodes in relation to longitudinal and maximal and minimal axial The

smooth

ure 1 were obtained logistic regression

curves

in Fig-

by applying model to the

the data

(15).

The longitudinal diameter varied between 3 and 60 mm. Figure 1 shows that only nodes with longitudinal diameters measuring 35 mm or larger (found in only 10 of the 144 malignant nodes) correspond to 100% malignancy.

was estimated

number of true-positive cases/(number of true-positive + false-negative cases) X 100; specificity, true-negatives/(true-negatives + false-positives) X 100; positive predictive value, true-positives/(true-positives + false-positives) X 100; negative predictive value, true-negatives/(truenegatives + false-negatives) X 100; prevalence, (true-positives + false-negatives)/

Volume

of Positive

diameters.

and specificity calculated and

tion between the sensitivity and the specificity: Lines that follow the highest course near the y axis have the best sensitivity at a high specificity. The overall accuracy

Number

Lymph

of

of necrotic

Analysis

The sensitivity ent criteria were

receiver curves.

the

the

size of areas nodes were

Statistical

necrosis

RESULTS

on the photograph and radiograph of the specimen. All macroscopically negative nodes were removed for microscopy. Two or more representative slices were made of each macroscopically positive node for microscopic study, and the presence or absence of gross necrosis was recorded. All macroscopically positive and negative nodes were examined microscopically by at least two investigators. The number of nodes removed from the specimen for mimetastatic

X 100; and accuracy + true-negatives)/total

fixation.

of 20 nodes were compared ters in the specimen.

croscopic

and

also

diameter and the longitudinal diameter of 15 nodes on both radiographs. To assess volume averaging at CT, the on-

screen

nodes

were disand were

assessed by obtaining a radiograph of three specimens before and after fixation, with use of exactly the same technique, and measuring one axial Shrinkage

total positives

borderline

There

is a large

range

of

overlap in size between positive and negative nodes. In this respect, the use of the maximal axial diameter (343 mm) was not significantly more accurate (Fig 2). At a maximal axial diameter of 14-15 mm, which is often used as a size criterion, only 45% of the nodes contained tumor (Fig 1). The minimal axial diameter (2-30

mm) proved to be the most accurate diameter to use in predicting tumorpositive nodes (Fig 2). All 34 lymph nodes with a minimal axial diameter larger than 12 mm were metastatic. Although metastatic lymph nodes with a minimal axial diameter smaller than 10 mm made up 58% of all malignant nodes and were found in 30 specimens, only four specimens contained exclusively metastatic nodes of this size. Table 2 and Figure 2 show the sensitivity and specificity for several cutoff points of these size criteria per lymph node. Note that the specificity remains high for all criteria (greater than 85%) because of the large number of (true) negative nodes. Because clinicians are concerned about whether a side contains metastasis or not, the sensitivity and specificity of these criteria were calculated per neck dissection as well. As illustrated in Tables 2 and 3 and Figure 3, the minimal axial diameter proved to be the most valid size criterion in predicting a tumor-positive side. Lymph node shrinkage by fixation could not be measured with use of the previously described tests. Comparing lymph node size at CT and in the specimen, we found that in 20 nodes, axial lymph node diameters were 1 1% smaller to 22% larger (mean, 6.1% larger) in the specimen than at CT. Lymph

Node

Shape

The shape criterion itself (maximal axial diameter divided by minimal axial diameter) or the minimal or maximal axial diameter combined with a roundish shape were all less valid than the minimal axial diameter criterion alone (results not shown). The ROC curves of combinations

of the

axial

diameter

minimal

criterion

axial

or maximal

together Radiology

with 381

#{149}

Figure

4.

(a) Contrast-enhanced

CT scan

(Philips 350 tomoscan; Philips Medical Systems) with a 6-mm section thickness was obtained at man with a T2NO supraglottic carcinoma. Two metastatic subdigastnic nodes lying against each other minimal axial diameter, thick arrows indicate maximal axial diameter) show several areas of irregular necrotic tumor foci. c common carotid artery. j internal jugular vein, in = sternocleidomastoid Gadolinium diethylenetniaminepentaacetic acid-enhanced Ti-weighted gradient recalled echo MR Systems, Milwaukee]) obtained with a 3-mm section thickness at the same level shows small necrotic nodes. Thin arrows = minimal axial diameter, thick arrows = maximal axial diameter, c = common cain = stennocleidomastoid muscle, s submandibular gland. (c) Histopathologic section through the of necrosis (X) that were also seen at CT and MR imaging. L = preexistent lymphatic tissue, T = squa-

the subdigastnic level in a 58-year-old on the right side (thin arrows indicate enhancement caused by small (3-mm) muscle, s = submandibular gland. (b) image (0.6 1; Technicane [GE Medical areas (low signal intensity) inside the notid artery, j = internal jugular vein, metastatic nodes shows the small areas mous cell carcinoma.

a roundish shape criterion for borderline nodes (1 or 2 mm smaller than the minimal axial diameter)

were

below

the

curve

for

the

mini-

mal axial diameter (results not shown). Thus, any combination shape with either the minimal

maximal

axial

diameter

valid a criterion diameter alone.

Lymph Region

Node

Negative

average

Size and

subdigastric gion had

axial

of 1-2

subdigastric

region

was

maximal

axial

Per

nodes

in the

submandibular diameters that

mm

as axial

Variations

larger

mm,

diameter

were minimal nodes

the

was

in

larger regions mm), tivity

4-5

mm

.

F

increased was a higher course of the ROC curve obtained. When even

slightly

Radiology

used for these diameter + 2 loss of sensi..

Groups

.

diameters were (minimal axial an unacceptable resulted.

mean

larger than metastatic nodes in other regions. The longitudinal diameter of both malignant and reactive lymph nodes was 3-6 mm larger for nodes in the jugular chain than for nodes in other regions. The use of a larger minimum axial diameter (minimal axial diameter + i mm) for lymph nodes in the subdigastric and submandibular region did not change the course of the ROC curve for acceptable (85% or greater) values of specificity. Only when the minimal axial diameter criterion for the subdigastric region (Table 3) was

382

a.

an

nodes

submandibular and

‘;!

re-

than

The mean of metastatic

and 2-3

not

minimal

positive

and

in other regions. axial diameter

the

was

as the

of or

Figure 5. (a) Contrast-enhanced CT scan (6-mm section thickness) obtained at the submandibular level in a 69-year-old man with a T3NO oropharyngeal carcinoma on the right side. The node behind the submandibular gland on the right side (thin arrows indicate minimal axial diameter, thick arrows indicate maximal axial diameter) shows irregular enhancement by adipose metaplasia inside the node. c common carotid artery, j = internal jugular vein, s = submandibular gland. (b) Histopathologic section of the reactively enlarged submandibular node shows the central adipose metaplasia (F). No tumor was found in the neck dissection specimen of this patient. L lymphatic tissue.

of Borderline

Nodes

The presence of groups of two or more borderline lymph nodes (minima! axial diameter of 1-2 mm smaller) in the first or second lymph node drainage region of the primary tumor (as described by Lindberg [16]) cornbined with the minimal axial diameter criterion resulted in an ROC

curve

similar

to the

curve

obtained

with the diameter criterion alone (not shown). Groups of three or more borderline nodes combined with the minimal axial diameter criterion high

increased specificity

the (Table

sensitivity 3, Fig

at a 3).

;

b.

Lymph

Node

Tumor

Necrosis

Areas of tumor necrosis, cystic tumor growth, or extensive tumor keratinization larger than 3 mm were visualized with both contrast materialenhanced CT and magnetic nance (MR) imaging (Fig

reso4) (i7).

These areas were seen in 46 (32%) of the 144 metastatic lymph nodes in 28 neck dissection specimens. Necrotic November

1990

nal

I

and

sagittal

MR

images

will

not

add significant information for cervical lymph node staging. Friedman et a! (i8) found that lymph node size can be up to 24% (mean, 15%) larger in the specimen than the measurement obtained at CT. We found that lymph node diameters in the specimen are 1 1% smaller to 22% larger (mean, 6.1% larger) than those measured at CT. This effect is probably caused by volume averaging effects or oblique scanning planes. We have also determined that the fixation procedure does not shrink lymph nodes. Our thus have to be diminished

size criteria by an av-

erage of 6.1% (we suggest using 10% to minimize sensitivity loss), dependa. Figure

ing

C.

6.

(a) Contrast-enhanced

CT scan (6-mm section thickness) obtained at the submandibular level in a patient with a T3N1 supraglottic carcinoma on the left side. Note the horseshoe-shaped submandibular node (thin arrows indicate minimal axial diameter, thick arrows indicate maximal axial diameter) with the hilar structure clearly visible. s submandibular gland, t pharyngeal lumen. (b) Radiograph of the left neck dissection specimen. The submandibular node with adipose metaplasia is indicated with two arrows (maximal axial diameter). The internal jugular vein (j) is radiolucent, as it contained no blood. The positive subdigastnic nodes (N) are in front of and behind the internal jugular vein. m = undissected part of sternocleidomastoid muscle, z reactively enlarged nodes, o omohyoid muscle, p = parotid gland underpole, s submandibular gland. (c) Histopathologic section of the reactively enlarged submandibular node shows the widened hilar structure filled with fat (F). L lymphatic tissue.

on

some

niques

1-3 mm were not reliably by current imaging techbut were present in another

27 lymph neck small found seen

nodes

dissection necrotic in large in nodes

vi-

in six additional specimens. These areas were not only nodes but were often with a minimal axial

diameter smaller than i cm as well. The sensitivity of the criterion of tumor necrosis, cystic tumor growth, or tumor keratinization in areas larger than 3 mm was 32% per node, with a specificity of 100%. Per side, the prevalence of necrotic areas larger than 3 mm was 39%, while the sensitivity and specificity of this criterion were 74% and 100%, respectively. of a combination

Table

with

modified

several

3 shows the of this criterion

minimal

Node

Adipose

177

Number

#{149}

2

dissection

classified

used.

that,

in

specimens,

re-

as true-positive.

In contrast (8,ii,i3),

to some

we

found

authors that

lymph

node

shape combined with any axial diameter did not increase the accuracy of the minimal axial diameter criterion. may

be

caused

by

the

fact

that

shape on section (depending on the section plane). In 15 nodes from 13 specimens, this hilar fat measured 3 mm or more and was visualized at CT (Fig 5). The fat was situated centrally in 20 nodes, and these central foci

the maximal axial diameter combined with a roundish shape (greater than

were

among the different regions in the neck, and, consequently, different criteria for lymph nodes in the subdi-

3 mm

nodes

these

or larger

of four

four

in only

specimens.

specimens

areas

of

contained

metastatic nodes. Central mm or larger may simulate (Fig 6) by causing central

uated

four

Two

in lymph

no

foci of 3 necrosis hypoatten-

nodes

at CT.

literature,

the

gastric

size

criterion

used

tamed

from

the

using assess study,

was

size

criteria.

all authors

the maximal metastasis. the maximal

not

found

(7-9,ii), may in-

To our

recommend

axial diameter to However, in our axial diameter

to be as valid

as the

minimal axial diameter in predicting tumor-positive nodes (Table 2; Figs 2, 3). The most useful criterion for the minimal axial diameter is between 10

and was non

12 mm. Because the

longitudinal

diameter

found to be a less valid size criteas well (Table 2; Figs 2, 3), coro-

in the

per

Consequently, these nodes

However, nodes are perilaryngeal,

specimen

the may

in our

size differ

study.

criteria for from the ones

study.

Groups

of borderline

area of lymphatic mary tumor (16) metastasis, grouping (7-9,1

in other

or posterior regions, never contributed to the of the sensitivity and

specificity

in our

and specificity 3). Results ob-

acceptable. metastatic

situated

peritracheal, these nodes calculations

are

found

the minimal axial for lymph nodes region, optimal

modifications

seldom

knowledge,

We have

that by increasing diameter by 1 mm in the subdigastric

The and fluence

areas

(11,12).

axial

shape (8,ii,i3), number site (11,12) of the node(s)

to the (shape diameter). shape vary

or submandibular

being

values for sensitivity were obtained (Table

DISCUSSION In the

0.8 mm) in fact corresponds minimal axial diameter maximal/minimal axial Lymph node size and

effect

Areas of adipose metaplasia larger than 1 mm (minimal diameter) were found in 123 lymph nodes, especially in the submandibular region. However, this fat was part of the hilar structures in 103 of these lymph nodes, giving the node a horseshoe

Volume

neck

areas were less because isolated

Metaplasia

thickness

to realize

for cervical lymph node metastasis varies between 8 and 30 mm (6-13).

diameter criteria. It is clear that necrosis increases the sensitivity of all modified minimal axial diameter criteria.

Lymph

section

actively enlarged nodes were larger than the actual metastatic nodes. These reactively enlarged nodes may have caused these specimens to be

This foci of sualized

the

It is important

and as an 1).

Recently,

has been questioned show that groups

nodes

drainage

in an

of the

pri-

are suggestive of many authors use additional criterion

however, (13). of three

its value Our data or more

Radiology

383

#{149}

borderline (minimal axial diameter i or 2 mm) lymph nodes are suggestive of lymph node metastasis. Adding this criterion increased the sensitivity of the size criterion but did not significantly influence the specificity (Table 3, Fig 3). However, the number of patients meeting this criterion was too small to make definite conclusions. Groups of two borderline lymph nodes, however, proved to be a rather aspecific criterion. Irregular contrast enhancement in the nodes, which can be caused by tumor necrosis, cystic tumor growth, or (avascular) keratinization, is used as a 100% specific criterion (6-li). As expected, this criterion was the most specific criterion in our study as well. To our knowledge, the sensitivity of necrosis has never been defined. For necrosis measuring 3 mm or larger, we found a sensitivity of 74%. These small foci can be detected reliably with use of conventional contrast-enhanced CT or MR imaging with a section thickness of 3-6 mm (Fig 6) (17). Because smaller necrotic foci are more common, the sensitivity of this criterion may be increased by making thinner sections or by increasing the contrast. However, spontaneous lymph node necrosis, other malignancies, cysts, abscesses, or adipose metaplasia may also simulate necrosis radiologically. In our series, only centrally located adipose metaplasia imitated necrosis at contrast-enhanced CT (Fig 5). Central adipose metaplasia of 3 mm or larger was present in four nodes of four specimens. This feature decreases the specificity of the tumor necrosis criterion at CT (hypoattenuated area) from 100% to 94%; thus it me still a reliable criterion. In MR imaging, fat tissue has a characteristic signal intensity, which decreases the probability of false-positive findings. To interpret the criteria used in this study, it is important to realize that all percentages were calculated for neck dissections with a high prevalence of metastasis (54%) and may not apply to patients with a lower prevalence of lymph node metastasis. -

384

Radiology

#{149}

To select the most appropriate criteria for cervical lymph node imaging, one must decide whether a specific or a sensitive test is to be used. To reliably select patients who do not need elective neck treatment, criteria with a high negative predictive value should be chosen. To obtain this high negative predictive value, the number of false-negative results should be as low as possible. In conclusion, we propose using the following radiologic criteria for assessing cervical metastasis in patients with a primary squamous cell carcinoma in the head and neck (Table 3): 1 Nodes with a minimal axial diameter of 1 1 mm or more in the subdigastric region and 10 mm or more in other lymph node-bearing regions should be considered metastatic. 2. Groups of three or more borderline lymph nodes with a minimal axial diameter of 9 or 10 mm in the subdigastric region and of 8 or 9 mm in other lymph node drainage regions of the tumor should be considered metastatic. 3. At CT, all nodes that show irregular enhancement and that are surrounded by a rim of enhanced tumor or lymph node tissue should be considered metastatic. One must beware of hypoattenuated areas that are continuous with the lymph node border. Together, these criteria had a sensitivity of 87% and a specificity of 94% per neck dissection specimen in our patient group (Table 3). With a prevalence of 54%, the negative predictive value is 86% and the positive predictive value is 94%. The overall error rate when using these criteria is 9.9% (seven of 71). U

4.

EL.

for his

technical

We support.

thank

5.

6.

7.

9.

2.

Sako K, Pradier RN, Marchetta FC, Pickren JW. Fallibility of palpation in the diagnosis of metastases to cervical nodes. Surg Cynecol Obstet 1964; 118:989-990. Byers RM, Wolf PF, Ballantyne AJ. Rationale for elective modified neck dissection. Head Neck 1988; 10:160-167.

Neck

3.

False nodes.

for subclini-

metastasis.

Arch Otolar-

Surg 1989; 115:981-984. HI, Lindberg RD, Fletcher CH. Cancer of the oral cavity: is elective neck dissection beneficial? Am Sung 1970; 120:505-508. Friedman M, Shelton VK, Mafee M, Bellity P. Crybauskas V. Skolnik E. Metastatic neck disease: evaluation by computed tomography. Arch Otolaryngol Head Neck Sung 1984; 110:443-447. Stevens MH, Harnsberger R, Mancuso

Davis

RK, Johnson

LP, Parkin

JL.

tomography of cervical lymph staging and management of head

cancer.

Arch

Otolaryngol

Head

Neck Sung 1985; 111:735-739. Close LC, Merkel M, Vuitch MF, Reisch J, Schaefer SD. Computed tomographic evaluation of regional lymph node involvement in cancer of the oral cavity and oropharynx. Head Neck 1989; 11:309-317.

Mancuso W. AJR

AA, Macen

D, Rice D, Hanafee

CT of cervical lymph 1981; 136:381-385.

node

cancer.

10.

Feinmesser R, Freeman JL, Nojek AM, Birt BD. Metastatic neck disease: a clinical/nadiographic/pathologic correlative study. Arch Otolaryngol Head Neck Surg 1987;

11.

Som PM. Lymph nodes of the neck. Radiology 1987; 165:593-600. Bruneton JN, Roux P. Caramella E, Dcmard F, Vallicioni J, Chauvel P. Ear,

113:1307-1310.

12.

nose, and throat nosis of metastasis nodes. Radiology

13.

Crasl

cancer: ultrasound to cervical lymph

diag-

1984; 152:771-773. Neuwirth-Riedl K, Critzmann

MCh,

N, Schurawitzki

H, Braun

sonomorphologischer

0.

Wertigkeit

knitenien

bei der

identifikation

region#{227}rer metastasen von plattenepithelkarzinomen des HNO-ber-

14.

15.

16,

17.

CB. neck

or surgery

node

nodes: and neck 8.

BC, Knivit JS, Applebaum

Head Neck RH, Barkley

Computed

References Ali 5, Tiwani RM, Snow tive and false negative

yngol Jesse

AA,

W. W. de Jong

1.

JM, Levin Radiotherapy

cal cervical

,

Acknowledgment:

Chow

posiHead

eiches. HNO 1989; 37:333-337. Croote AD, Oosterhuis JW, Molenaar WM, Vermey A, van Osnabrugge-Bondon C, Arnold Ky. Radiographic imaging of lymph nodes in lymph node dissection specimens. Armitage in medical

Lab Invest 1985; 52:326-329. P, Berry C. Statistical methods research. 2nd ed. Oxford:

Blackwell Scientific Publications; 1987; chap 12. Lindberg R. Distribution of cervical lymph node metastases from squamous cell carcinoma of the upper respiratory and digestive tracts. Cancer 1972; 29:14461449. van

den

Brekel

MWM,

HV, et al.

Detection

of metastatic

cervical

imaging: comparison techniques. J Comput

1985; 8:78-82.

Castelijns

JA, Stel

and characterization adenopathy

by MR

of different MR Assist Tomogr 1990;

14:581-589.

18.

Friedman igl TL, Dew for elective ryngoscope,

M, Mafee MF, Pacella BL, StrorLL, Toniumi DM. Rationale neck dissection i990; 100:54-59.

in 1990.

November

La-

1990