Morphological characteristics and distribution patterns of epithelial cells in the cytological diagnosis of gastric cancer.

Abstract

The morphological features of epithelial cells in gastric biopsy touch smears from 296 cases (248 benign, 48 malignant) were retrospectively reviewed. By subjective criteria no single purely morphological characteristic could be identified that was consistently capable of separating benign from malignant specimens. The value of cell distribution patterns in diagnosis was therefore objectively analysed in the 231 adequate specimens (190 benign, 41 malignant). A score of over 20% for single epithelial cells was found to have a sensitivity of 95.1% and specificity of 99.5% in the diagnosis of carcinoma. Quantitative analysis of cell distribution pattern is a simple test, which does not require special equipment, yet can be helpful in the interpretation of gastric cytology.

Full text

Page 1

J Cliii Pathol 1982;35:585-590
Morphological characteristics and distribution patterns
of epithelial cells in the cytological diagnosis of gastric
cancer
JENNIFER A YOUNG, HELENA E HUGHES, DJ HOLE
From the Department of Cytology, The Royal Infirmary, Glasgow and the West of Scotland Cancer
Surveillance Unit, Ruchill Hospital, Glasgow
SUMMARY
cases (248 benign, 48 malignant) were retrospectively reviewed. By subjective criteria no single purely
morphological characteristic could be identified that was consistently capable of separating benign
from malignant specimens. The value of cell distribution patterns in diagnosis was therefore objec-
tively analysed in the 231 adequate specimens (190 benign, 41 malignant). A score of over 200% for
single epithelial cells was found to have a sensitivity of95- 1
ofcarcinoma. Quantitative analysis of cell distribution pattern is a simple test, which does not require
special equipment, yet can be helpful in the interpretation of gastric cytology.
The morphological features of epithelial cells in gastric biopsy touch smears from 296
and specificity of99 5%in the diagnosis
The examination ofendoscopically collected cytology
specimens
cedure during the investigation of patients with
gastric disease. Brushing is the most widely used
sampling technique,1-4 direct vision lavage5 6 and
biopsy touch smear cytology47-9 are additional
methods. However, accurate cytological diagnosis
can sometimes be difficult, irrespective of the sam-
pling technique. Although in very occasional cases,
equivocal cytological findings may be the reflection of
severe epithelial dysplasia,1011 the great majority of
difficulties do not arise for this reason. Most problems
in the interpretation of gastric cytology are due to
two far more common causes.
cellular changes are seen in cytological specimens
from benign conditions, such as gastritis and peptic
ulcer,'213 which on histology show only inflam-
matory or degenerative changes not amounting to
severe dysplasia. Secondly, cells from gastric adeno-
carcinomas display a wide variety of morphological
appearances.'4 A cytological specimen can appear
puzzling for either of these reasons, yet correct
interpretation is essential for patient management.
In order to try to identify a simple criterion, which
could be used as an objective aid in the interpretation
of gastric cytology specimens, a retrospective study
of the characteristics of benign atypical and malig-
nant epithelial cells was carried out. From this
is now an established diagnostic pro-
Firstly, marked
Accepted for publication 21 October 1981
subjective study, the single most valuable feature
appeared to be an alteration in intercellular relation-
ships. The diagnostic accuracy of cell distribution
patterns was therefore examined, as this was the
component of intercellular relationship which could
be analysed objectively.
Material and methods
The specimens studied were those collected during a
previous trial of gastroduodenal cytology.4 From a
total of 329 patients, 303 had specimens collected
from the stomach. The original subjective cytology
reports had classified the specimens as normal;
showing benign atypia of mild, moderate or severe
degree; or malignant, with small or large numbers of
malignant cells. "Suspicious" or inconclusive reports
were not issued, as a definitive diagnosis was one of
the requirements of the trial protocol. The final
diagnosis was benign in 248 cases and malignant in
55. The benign diagnoses were substantiated by a
period of follow-up ranging from two to four years
and the malignant cases were all verified by histology.
Six ofthe tumours were non-epithelial or of uncertain
primary origin and these were excluded from the
present study. The remaining malignant tumours
were all primary adenocarcinomas. There was one
case with false-negative cytology, due to sampling
error, among these and this was also deleted. A total
of296 cases (248 benign and 48 malignant) was there-
585
group.bmj.com on July 15, 2011 - Published by jcp.bmj.com Downloaded from

Page 2

586
fore available for retrospective study.
fn the original trial4 both brush and biopsy touch
smear cytology specimens were examined. Cellular
detail was better in the touch smears than the brush-
ings, due to the smaller quantity of contaminating
debris. The "best" biopsy touch smear specimen was
therefore selected from each case. The incidence and
severity of the cellular characteristics listed in Table I
were subjectively assessed for each of the four
categories: mild, moderate and severe benign atypia
and malignancy. Specimens containing small and
large numbers of malignant cells were considered
together. From this subjective review, it appeared
that alteration in cell distribution pattern was a
feature, which could possibly be used as an objective
discriminating factor between benign and malignant
specimens.
To test the validity of this hypothesis, the cell
distribution pattern on the 296 slides was examined
"blind," without knowledge of the previous cyto-
logical classification. In each of the 200 microscopic
fields (at x 400 magnification), it was recorded (i)
whether the epithelial cell at the top right hand
border of the field occurred as a discrete single cell
with cytoplasm; (ii) in a small group of two to five
cells; (iii) in a large group of six or more cells or (iv)
was a bare nucleus without cytoplasm. It was neces-
sary to look at bare nuclei as a separate category, as
these may represent either undifferentiated malignant
cells, or may occur as artefacts due to cytolysis in
normal specimens.
STATISTICAL METHODS
The frequency distribution for each type of cell
distribution pattern was constructed for all benign
Young, Hughes, Hole
cases,
cases. These statistical distributions could not be
adequately represented by a "normal" curve and,
consequently, all statistical tests performed were non-
parametric. Comparison was made of the median
percentage value between all benign and malignant
cases, and between severe benign atypia and malig-
nant cases. These calculations were made for each
cell distribution pattern separately. The statistical
procedure used was the Mann-Whitney test and the
probability levels quoted were based on a two-sided
test of significance.
severe benign atypia cases and malignant
Results
MORPHOLOGICAL CHARACTERISTICS OF
EPITHELIAL CELLS
The subjective assessment of the incidence and
severity of cellular characteristics in the four cat-
egories of mild, moderate and severe benign atypia
and malignancy are given in Table 1. This is an
expanded version of the description of cellular
characteristics included in the original report of the
gastric cytology trial.4
The characteristics fall into four groups. In group I
are features which occurred on a progressive scale
through mild, moderate and severe benign atypia.
The degree to which they were found in malignant
cells was generally greater than in atypia, but in
occasional malignant cases was less marked than in
some examples of benign atypia. The features in
group 2 were never seen in mild atypia, but occurred
to a variable extent in both moderate and severe
atypia and malignancy. The characteristics in group 3
were never seen in benign specimens. If present they
Table 1
Incidence and severity of cellular characteristics in benign atypical and malignant gastric cytology speciments
Characteristic
Classification
Benign atypia
Malignant
Mild
Moderate
Severe
Group I
Variation in nuclear/cytoplasmic ratio
Anisocytosis
Anisonucleosis
Abnormal nucleoli
Variation in nuclear chromasia, granularity or
clumping
Group 2
Nuclear pleomorphism
Abnormal coarse cytoplasmic vacuolation
Mitotic figures, normal
Group 3
Mitotic figures, abnormal
Irregular multinucleation
Fine "foamy" cytoplasmic vacuolation
"Opaque glass" nuclei
Group 4
Abnormal intercellular relationships (polarity and
distribution pattern)
-+
+
+- A
++
-t -- t
+
++
--- + +--
++
+
+
+-
+ + + +
++
+-4-I
+
+
-4-
-
-
-
-
+ -
-
-+
-
t-
-
+ ±
-
+
-*4
-
-
-
-
---
-
-
-
-
r
-
-+ +
+ + + +
group.bmj.com on July 15, 2011 - Published by jcp.bmj.comDownloaded from

Page 3

Morphology and distributioni of epithelial cells in gastric cancer
were indicative of malignancy, but only occurred in
cells from a small proportion of carcinomas and
therefore could not be used as a consistent basis for
diagnosis. Altered intercellular relationships were
placed in group 4 as the only characteristic, which
was always present in malignant specimens and
always present to a more marked degree than in
severe benignatypia.Intercellular
however, consist of two closely associated features,
polarity and cell distribution pattern. Polarity is the
way cells relate to one another within a given group
and assessment is essentially a subjective judgement,
which is dependent on previous experience and, there-
fore, not suitable for objective examination. Cell
distribution pattern is the way cells relate to one
another on the slide as a whole and this appeared to
be a factor which could be objectively analysed by
simple means.
relationships,
CELL DISTRIBUTION PATTERN
The material was adequate for the assessment of cell
distribution in 231 cases (190 benign, 41 malignant)
and inadequate in 65. The definition of an adequate
specimen was the presence ofwell preserved epithelial
cells in 200 microscopic (x 400) fields. The overall
range of values found for single cells, small and large
groups and bare nuclei, in the 231 specimens is shown
in Table 2. The original cytological classification of
these 231 specimens was then checked. Sixteen had
been classified as normal, 76 displayed mild benign
atypia, 68 moderate atypia, 30 severe atypia and 41
had been diagnosed as malignant. In the 30 cases
with severe benign atypia the final definitive diag-
nosis
was gastric ulcer
gastritis in 12 and gastric erosions in two. One patient
was found to have a hyperplastic "polyp" associated
with chronic gastritis and one an abnormal gastric
fold, also associated with chronic gastritis. Table 3
compares the median values for the percentage of
each type of cell distribution in all 190 benign cases,
with the values in the 41 malignant cases. Table 4
gives the median percentages for each of the four
benign classes (normal, mild, moderate and severe
in 14 patients, chronic
Table 2
touch smears
Overall cell distributioni pattern in 231 biopsy
Single cells
2-5 cells
>-6 cells
Bare nuclei
0 -64-5
3-0-44-5
2-0-79-0
1-5-76-5
Table 3
distribution in biopsy smears by original cytological
diagnosis
Median percentage ofeach type of cell
No of
cases
Median values ('%)
Single
cells
2-5
cells
>6
cells
Bare
nuclei
Benign
Malignant
190
41
6-0
38-0
15-5
13-0
43-5
19 5
28 5
27-5
benign atypia) compared with malignancy.
There was found to be a significant difference in
the median values of the distribution of both single
cells (p < 0 001) and large groups (p < 0-01), not
only between
benign and malignant specimens
overall, but also between severe benign atypia and
malignancy, the distinction which causes most prob-
lems during routine reporting. In both instances the
difference was greater for the distribution of single
cells than for large groups. The frequency distribution
of single cells in all the 190 benign cases in com-
parison with the 41 malignant cases is shown in Fig. I
and the frequency distribution in the 30 cases of
severe benign atypia in comparison with malignancy
is shown in Fig. 2. The data for the region where the
values for benign and malignant overlap is shown in
greater numerical detail in Table 5. It can now be
seen that all benign cases except one had 20% or less
single cells and that all except two malignant cases
had more than 20% single cells. If a score of more
than 20% single cells is therefore considered as an
indication of malignancy, the diagnostic sensitivity
of this observation is 95-1 %, the specificity 99-5 0O
and the efficiency 98 7 %.
Twenty-nine of the 30 cases displaying severe
Table 4
Median percentages ofeach cell distribution in biopsy touch smears by cytological classification
Classification
No of cases
Median values (%)
Single cells
2-5 cells
>6 cells
Bare nuclei
Normal
Mild benign atypia
Moderate benign atypia
Severe benign atypia
Malignant
16
76
68
30
41
3-5
5 5
7-0
10-0
38-0
15-0
11 5
16-6
23-0
13-0
43-5
38-0
43-5
46-0
19 5
33 5
28 5
27-5
27 5
27-5
Differences between all benign and malignant specimens
Difference between severe benign atypical and
malignant specimens
p < 0-001
NS
p < 0 01
NS
p < 0-001
NS
p < 0-01
NS
587
group.bmj.com on July 15, 2011 - Published by jcp.bmj.com Downloaded from

Page 4

Younig, Hughes, Hole
Benign cases
(n=190)
Malignant cases
(n =41)
10
20
30
40
'la single cell occurrence
Frequency distributioni ofsingle cells in benign
and malignant cases.
Fig.
1
Severe benign atypia
(n=30)
Malignant cases
(n =41)
0
10
20
'/o single cell occurrence
30
40
50
Fig. 2
cases with severe atypia and in malignant cases.
Frequency distributionojsingle cells in benign
Table 5
malignant cases
Percentage of single cells in benign and
Percentage of
single cells
All benign cases
(n = 190)
Severe benign
atypia (n = 30)
All malignant
cases (n =41)
<15-0
15-1-17-5
17-6-20-0
20-1-22-5
22-6-25-0
--25-0
180
25
1
0
1
2
5
3
6
1
0
0
1
3
1
0
0
32
atypia were correctly diagnosed as benign by assess-
ment of single cell distribution pattern. The one so-
called false-positive case with a slightly raised single
cell score of 215 %0 was from this group. The patient
had a gastric ulcer, which in the original endoscopic
biopsies had shown dysplasia, although the final
definitive diagnosis, after follow-up, was benign.
Two malignant cases had false-negative scores, one
had 15% single cells and the other 18%. Both these
specimens
were from
tumours and the slides contained large numbers of
undifferentiated malignant cells, which appeared as
bare nuclei in the cell distribution counts. The bare
nuclei scores for these two cases were 56.50% and
3150%.
In
terms
presented
obviously malignant on microscopic examination
due to extreme chromatin abnormalities.
very poorly
differentiated
practical
a diagnostic problem
neither would have
as both looked
Discussion
Gastric epithelial cells exhibit great morphological
variety in their response to disease. The cells from
two carcinomas are never quite alike14 and benign
specimens also show a great range of appearances,
particularly in the cells arising from areas of atrophic
gastritis or the regenerating edges of gastric ulcers.
Cells from one carcinoma may display a particular
morphological abnormality to such a degree that the
diagnosis of malignancy is obvious, yet the same
abnormality may be almost completely absent from
the cells of another tumour, which in turn may
exhibit a quite different characteristic of malignancy.
The majority of carcinomas display gross stigmata of
malignancy
phology occurs on a progressive scale through mild,
moderate and severe benign atypia to malignancy.
However,
the
degree
characteristic develops shows such variation that no
single purely morphological feature was identified
which could be used as a constant discriminating
factor between benign and malignant specimens. This
is in keeping with the findings of Kalnins et al15 based
on discriminant function and multiple regression
analysis by computer of variable morphological
factors in cells arising from gastric carcinomas.
A group I characteristic may occasionally be less
pronounced in malignancy than in severe benign
atypia. For example, anisocytosis and anisonucleosis
are less pronounced in the small cell anaplastic type
of cellular picture associated with Lauren's diffuse
carcinoma,'617 than in groups of cells arising from
the regenerating edges of occasional benign ulcers.
The cellular changes in group 2, though not seen in
mild atypia, occur to such a variable degree both in
benign and malignant lesions that they
cnd the "overall disturbance" of mor-
to which each abnormal
are un-
40
30
20
10
CO
v 0
:3
0
Lt
30,
1
20
10
0
40
30
20
10
310
u
ci
4,
20-
10-
0J
"N'K\
I.
V
i
\\\\
--\
\ -,
588
group.bmj.com on July 15, 2011 - Published by jcp.bmj.comDownloaded from

Page 5

Morphology and distribution of epithelial cells in gastric cancer
reliable. Nuclear pleomorphism is a serious abnor-
mality and if present to more than a moderate degree
is indicative of malignancy, but occasional carci-
nomas occur that are surprisingly monomorphic.18
Coarse cytoplasmic vacuolation may be seen in
intestinal metaplasia and conversely some poorly
differentiated carcinomas show no evidence ofmucin
secretion, even with special staining techniques, in
cytological preparations. Mitotic figures are seen in
both benign and malignant specimens and with
routine light microscopy it is sometimes difficult to
distinguish normal from abnormal forms. They are a
known cause of false-positive reports.19
The third group consists of characteristics which
only occur in malignant cells and are not found in
severe benign atypia. However, unfortunately they
can only be seen in the cells from a minority of
carcinomas. Abnormal mitotic figures are indicative
of malignancy, but are seldom sufficiently bizarre to
be identified with certainty. Irregular multinucleation
has not been observed in benign columnar cells, but
occurs only infrequently in malignant cells. There are
also two unusual varieties of malignant cells, which
display specific morphological abnormalities
encountered in cells showing benign atypical changes.
The first is a large cell with abundant finely vacuo-
lated "foamy" cytoplasm and a prominent
membrane. The second type of cell has a peculiar
"thick" nucleus, which looks like "opaque glass" and
appears to rise above the surrounding cytoplasm.
Altered intercellular relationship (polarity and cell
distribution pattern) is the only characteristic, which
appears
carcinomas,
irrespective
variety of the malignant cells. It is also the only
feature which is always more pronounced in malig-
nancy than in severe benign atypia. Alteration of
polarity is a constant feature in groups ofmalignant
cells, but judgement of this depends on experience
and
occasional specimens malignant cells may occur
almost exclusively as discrete, single cells and groups
of malignant cells may be of small size or few in
number. In these cases loss of polarity may be
inconspicuous, even to an experienced observer. The
distribution pattern of the cells on the slide as a
whole, however, is easy to measure and requires no
special equipment.
The objective assessment of the cell distribution
pattern in the 231 specimens examined confirmed the
subjective impression
contained greater numbers of single cells and fewer
groups of cells than benign specimens. Furthermore,
this difference remained significant even when severe
atypia and malignancy were compared. The sen-
sitivity
(95 1%) and
not
cell
to be displayed by
all
the
gastric adeno-
morphological
of
itis impossible
to quantify. Moreover,
in
that malignant specimens
specificity (99*5%O)
of the
observation that more than 20% single cells indicates
malignancy are sufficiently high for this measure-
ment to be clinically useful when assessing specimens
with equivocal cellular changes. Two specimens,
which contained large numbers of undifferentiated
(bare nuclei) malignant
results. The nuclei of both these tumours, however,
showed extreme chromatin abnormalities and neither
would have presented any difficulty on routine
screening. The nuclear
structure must
always be carefully scrutinised in specimens with a
high count of bare nuclei, so that undifferentiated
malignant cells are separated from bare nuclei due
to cytolysis.
The specimens from the 30 cases with severe benign
atypia were, as expected, from patients with chronic
atrophic gastritis or from the regenerating edges of
gastric ulcers. In one case there was an associated
hyperplastic polyp and in another an abnormal
gastric fold. We agree with Husain et al20 that this is
a group of patients who require special identification
and follow-up. The cases in the present series were
followed for a period ranging from two to four years
and none has developed carcinoma. However Morson
et al"l stress that it is epithelial dysplasia which is
actually the marker in increased risk of malignant
change. It is therefore patients with this particular
lesion who must be singled out for especially thorough
follow-up. In the present study, histology of en-
doscopic biopsies showed epithelial dysplasia in only
one patient with cytological severe benign atypia.
It was this patient (with a gastric ulcer) who had the
raised single cell score (21-5%o) among the benign
cases. This probably represented the equivocal state
of the epithelium at the time of endoscopy. Morson
et all" state that it is important to realise that the
degree of risk associated with epithelial dysplasia is
not yet established. They consider that even severe
dysplasia, on
surgical intervention, though they stress close follow-
up is necessary. The patient who had dysplasia in the
present series remains well and the gastric ulcer has
healed without problems. Further studies of cell
distribution pattern in patients with histologically
proven dysplasia are required to see if this test can
"pick out" true epithelial dysplasia from the cyto-
logical atypias and thus help to identify the "high
risk" patient, who requires vigilant observation.
The altered cell distribution seen in malignant
specimens is due to the reduced mutual adhesiveness
of malignant cells first noted by Coman in 1944,21 in
connection with squamous cell carcinomas. Study of
the ultrastructure and surface properties of tumour
cells has subsequently greatly expanded this obser-
vation and shown that a wide variety of factors
determine cell to cell adhesion.
cells, gave false-negative
therefore
its own,
is not an indication for
589
group.bmj.com on July 15, 2011 - Published by jcp.bmj.com Downloaded from