ANATOMIC
PATHOLOGY
Original
N u c l e a r
a n d
D N A
N u c l e a r
a n d
a n d
C o n t e n t ,
S i z e
C i r r h o t i c
M a l i g n a n t
Article
P r o l i f e r a t i o n
D e t e r m i n a t i o n
L i v e r ,
a n d
P r i m a r y
H e p a t i c
i n
a n d
i n
I n d e x ,
N o r m a l
B e n i g n
M e t a s t a t i c
T u m o r s
CARINE DEPREZ, MD,* DANIEL VANGANSBEKE, MD,f ROLAND FASTREZ, M D 4
JEAN-LAMBERT PASTEELS, MD,§ ALAIN VERHEST, MD,1
AND ROBERT KISS, PHD§
The distribution values of ploidy, of the proliferation index, of the percentages of diploid cells/case, of nuclear size, and DNA histogram type
are described in a series of 92 liver samples from 87 patients. The 92
samples include normal (31 cases) and cirrhotic (14 cases) tissues, benign tumors (7 cases), well-differentiated hepatocellular carcinomas
(HCCs, 13 cases), moderately and poorly differentiated HCCs (8
cases), and colorectal glandular metastatic tissues (19 cases). The samples are from either fine-needle aspiration biopsy (FNAB) or histologic
imprint smears (HIS). Nuclear assessments were computed on Feulgen-stained nuclei by means of a cell image processor. The results show
that the mean DNA index value, the mean nuclear area (NA) value, and
the mean percentages of diploid cells per sample are significantly different in the three benign groups under study (normal and cirrhotic
tissues and benign tumors) as compared with the mean parameter values from the three neoplastic liver groups (well-differentiated and
poorly differentiated HCCs and colorectal metastases). None of these
three parameters, however, makes it possible to discriminate clearly
As reported by Stillwagen and colleagues,1 the overall incidence of primary liver carcinomas in the United States ranges
from 0.21-0.57%, of which more than 70-90% are hepatocellu-
between these six histopathologic groups at the individual case level.
The mean proliferation index values were significantly lower in the
normal tissues and the benign tumors than in the four other histopathologic groups. Recognizing six DNA histogram types, ie, diploid, hyperdiploid, triploid, hypertriploid, tetraploid, and polymorphic, we observed that all the benign samples (the normal and cirrhotic tissues and
the benign tumors) exhibited a diploid and/or tetraploid DNA histogram pattern, whereas the neoplastic samples exhibited the six DNA
histogram patterns. The combination of the five computerized parameters into a cytologic score (CS) ranging from 5 to 15 permits clear-cut
discrimination between nonneoplastic and neoplastic liver cases. The
specificity and sensitivity, and the positive and negative predictive values relating to this score were as high as 100%, 95%, 100%, and 96%,
respectively (Key words: Liver tumors; Ploidy; Proliferation index; Nuclear size; Feulgen staining; Cytophotometry) Am J Clin Pathol
1993;99:558-565
lar cancers (HCCs). Worldwide incidence ranges between 0.4
and 6.9%.' These authors also report that the survival rates of
unresectable hepatocellular cancers have generally been poor,
with a 5% response rate to treatment and a medium survival
time of 2.3 months. In addition, Ferrucci estimated that in
1989 nearly 150,000 new cases of colorectal carcinomas were
diagnosed in the United States. Of these, approximately 60,000
From the Departments of*Pathology and ^.Surgery, Hospital Brag2
mann, the Departments of } Radiology and ^Pathology, Hospitalpatients
Er- would ultimately prove to have liver metastases.
Because the successful treatment of hepatocellular carciasme, and the ^Laboratory of Histology, Faculty of Medicine, Free
University of Brussels, Brussels, Belgium.
nomas by surgical resection relies on early diagnosis,3 fine-needle aspiration biopsy (FNAB) is an increasingly popular means
Supported by grants awarded by the Fonds de la Recherche Scienti-of evaluating liver masses.4"7 Accordingly, Noguchi and cofique Medicale (FRSM, Belgium).
workers4 found that for a reliable diagnosis of hepatic maligReceived January 6, 1992; revised manuscript accepted for publicanancy, the percutaneous aspiration biopsy of the liver with a
tion April 13, 1992.
fine
needle is safe and useful. A correct cytodiagnosis of HCCs
Dr. Kiss is a Research Associate at the Fonds National de la Reis not always possible, however, because available cytologic decherche Scientifique (FNRS, Belgium).
Address reprint requests to Dr. Kiss: Laboratory of Histology, Fac- scriptions of HCC4 in aspirated materials obtained by FNAB
were too sketchy. These authors and others also argue that
ulty of Medicine, Free University of Brussels, 808 route de Lennik,
1070 Brussels, Belgium.
generally the cells in well-differentiated types of HCC cannot
558
DEPREZ ET AL.
DNA Ploidy Assessments in Liver Tumors
TABLE 1. HISTOLOGIC CRITERIA FOR HEPATOCELLULAR
CARCINOMA CLASSIFICATION PROPOSED
BY FRIAS-HIDVEGI"
Well-differentiated
Microtrabecular and macrotrabecular
Cobblestone
Fibrolamellar
Moderately well-differentiated
Acinar or adenoid
Clear cell
Poorly differentiated
Giant cell
Spindle cell
Solid
Sclerotic
559
Switzerland) after hydrolysis in 200 mL of 6N HC1 at 22 °C.
The Feulgen reaction was used because it permits selective and
quantitative (stoechiometric) DNA staining.9,10
Histopathologic and Cytopathologic Diagnoses
be as easily identified as tumor cells because they show little
cellular atypism.4,5,8 Cohen and co-workers5 state that the distinction between HCCs and the cells of "normal" livers is often
more difficult than that between an HCC and a metastatic
neoplasm.
Our purpose is to describe the distribution values of ploidy,
the proliferation index, the percentages of diploid cells/case,
nuclear size, and DNA histogram type in a series of 92 samples.
The series comprise 46 FNAB samples and 46 histologic imprint smears (HISs) from 87 patients. Nuclear assessments
were computed on Feulgen-stained nuclei by means of a cell
image processor. We set up a cytologic score (CS) ranging from
5 to 15, taking thefiveparameters into account. We report the
specificity, the sensitivity and the positive and negative predictive values.
The histopathologic and cytopathologic diagnoses were
based on the classification proposed by Frias-Hidvegi." The
histologic and cytologic criteria corresponded essentially to architectural features. These criteria took into account the presence of microtrabeculae, macrotrabeculae, and endothelial
cells; the cohesive pattern of the cellular sheets; the thickness of
the aspirated cell clusters; the overlapping of the cells inside the
clusters; the acinar formation, and the percentage of naked
nuclei (Table 1). The two moderately differentiated HCCs were
combined arbitrarily with the six poorly differentiated HCCs,
and all were placed in the poorly differentiated histopathologic
group.
Our series contained 31 normal samples from 26 patients.
These 31 samples corresponded to 10 FNAB specimens and 21
DIPLOID ( I )
»,
HYPERDIPLOID ( IV )
MATERIALS AND METHODS
Histologic and Cytologic Samples
The samples came from 21 patients who had undergone surgery for a pathologic condition other than liver disease ("normal" tissues), from 26 patients who had undergone surgery for
clinical purposes, and from 40 who had undergone FNAB for
diagnostic purposes. In addition, FNAB procedure was performed preoperatively on 6 of the 40 patients who underwent
diagnostic surgery. All the FNABs were performed under echographic guidance with a 22-gauge needle. With respect to each
of the surgically resected specimens, part of each was processed
using standard histologic techniques. In addition, two 30-mm3
tissue slices were smeared onto two slides to produce HISs,
which were immediately fixed for 20 minutes in a fixative of
ethanol 96° (70 vol), 4% neutral formol (25 vol), and acetic
acid, (5 vol). The FNAB samples were also smeared onto glass
slides. In each case some of the FNABs were processed routinely and stained by the Diff-Quik method (American Scientific Products, Columbia, MD). Others were immediately fixed
in a solution of ethanol 96° (70 vol), 4% neutral formol (25
vol), and acetic acid, (5 vol) as described above. The time-lapse
from smearing to fixation was similar for the HISs and FNAB
samples-10 to 20 seconds. The HISs and FNAB specimens
were stored at 4 °C until they were stained using the Feulgen
reaction according to the procedure described previously.9,10
Briefly, all of the cell preparations from a given case were
stained together for 1 hour in the Feulgen reagent (Fluka, CH,
C DNA CONTENT
FIG. 1. Six typical DNA histogram patterns were encountered in 90 of
92 liver samples analyzed. These six patterns correspond to a diploid
(type I), a triploid (type II), a tetraploid (type III), a hyperdiploid (type
IV), a hypertriploid (type V), and a polymorphic (type VI) pattern. The
breakdown of DNA histogram type in relation to histopathologic grading is given in Table 2.
Vol. 99 • No. 5
560
ANATOMIC PATHOLOGY
Original Article
TABLE 2. BREAKDOWN OF DNA HISTOGRAM TYPE AS OPPOSED TO HISTOPATHOLOGIC GRADING
DNA Histogram
Type
Histopathologic
Group
Normal tissue
Cirrhosis
Benign tumor
Well-differentiated HCC
Poorly differentiated HCC*
Colorectal metastasisf
/
//
III
IV
V
VI
25/31
(81%)
13/14
(93%)
7/7
(100%)
2/13
(15%)
—
—
6/31
(19%)
—
—
—
1/14
(7%)
—
—
—
—
4/13
(32%)
1/8
(13%)
3/19
(16%)
2/13
(15%)
2/8
(25%)
2/19
(11%)
1/19
(5%)
—
—
2/13
(15%)
1/8
(13%)
4/19
(21%)
1/13
(8%)
1/8
(13%)
3/19
(16%)
2/13
(15%)
2/8
(25%)
4/19
(21%)
pattern.
HCC = hepatocellular carcinoma.
* One poorly differentiated hepatocarcinoma exhibited a hyperhexaploid DNA histogram t One colorectal hepatic metastasis exhibited a hypcrtetraploid DNA histogram type.
HISs. We obtained both an FNAB sample and an HIS from 5
of 26 patients. The 14 cirrhotic samples (14 patients) were from
4 FNAB specimens and 10 HISs. The 7 benign tumors (nodular focal hyperplasia, 7 patients) were from 2 FNAB samples
and 5 HISs. The 13 well-differentiated HCCs (13 patients) were
from 8 FNAB specimens and 5 HISs, whereas the 8 moderately
(2 patients) and poorly differentiated HCCs (6 patients) were
from 4 FNAB specimens and 4 HISs. Lastly, the 19 metastatic
glandular samples (19 patients) were from 17 FNAB specimens
and 2 HISs. The 19 patients had clinically documented colorectal carcinomas.
Statistical Analyses
The results are presented as individual assessments performed on each sample and their mean values ± standard error
of the mean (white dots ± bars). The statistical comparisons of
data were performed using the Fisher F test (one-way variance
analysis). Paired data were compared using the nonparametric
Kendall rank correlation test.
RESULTS
Characterization of the DNA Histogram Type and the
Ploidy Value Distribution
Image Analysis
Cell image analysis was performed using a SAMBA 2005
microscope image processor (Alcatel-TITN, Grenoble,
France). The nuclear parameters, ie, size, DNA content, and
integrated optical density (IOD), were computed on 200 epithelial nuclei per sample according to the procedure reported previously.9-10
The nuclear area corresponds to the area of nuclear profile,
ie, the number of pixels (0.16 tim2/pi\e\) occupied by a nucleus. It was computed in a high-resolution mode, ie, at 100X
magnification (numeric aperture: 1.30). The IOD parameter
was measured on a 540-nm wavelength and computed at each
pixel on 256 densitometric levels. The IOD calculates the
amount of absorbent material, ie, the nuclear DNA content in
arbitrary units, and makes it possible to assess the DNA histogram of each tumor. We determined the DNA index from the
IOD, the value of which is 1.00 in the case of a normal G0-G1
diploid population. This value corresponds to 2,100 arbitrary
integrated optic density units. Normal liver nuclei in the G0G1 phase of the cell cycle were used as an external standard.
We calculated the proliferation index as previously described.10
The proliferation index represents the percentage of cells outside the major peak (regardless of its ploidy level) and the related peaks (peaks standing at 50% of the modal value and/or
double the modal value).
We recognized 6 DNA histogram types in 90 of 92 liver
samples under study. These 6 DNA histogram types corresponded to diploid (type I), hyperdiploid (type IV), triploid
(type II), hypertriploid (type V), tetraploid (type III), and polymorphic (type VI) patterns, as illustrated in Figure 1. The DNA
histogram types from the two remaining liver samples corresponded to a hypertetraploid and a hyperhexaploid pattern (Table 2). The breakdown of the 6 DNA histogram types according
to the 6 histopathologic groups is given in Table 2. All 31 normal cases exhibited either a diploid or a tetraploid DNA histogram pattern. All except one cirrhotic tissue sample exhibited a
diploid DNA histogram, the exception being a hyperdiploid
one (Table 2). All the benign tumors, ie, nodular focal hyperplasias, exhibited diploid type I DNA histograms. In contrast to
this relatively homogeneous distribution of normal through cirrhotic to benign tumor tissues according to either diploid or
tetraploid DNA histogram patterns, the neoplastic liver samples (the well-differentiated or poorly differentiated primary
cancers and colorectal metastases) had a wide and heterogeneous distribution with respect to their DNA histogram patterns. The three neoplastic groups of liver tissue exhibited the 6
DNA histogram types in relatively similar proportions (Table
2). No typical DNA histogram pattern emerged from any neoplastic histopathologic group. Furthermore, it should be emphasized that 8 of 40 (20%) of the neoplastic samples (3 welldifferentiated HCCs, 1 poorly differentiated HCC, and 4
A.J.C.P.-May 1993
561
DEPREZ ET AL.
DNA Ploidy Assessments in Liver Tumors
O
2.62.5-
^
2.4-
3
2.2-
^
t
2.1 "
2.0 1.9-
Q
Z
-
1.51.4-
<
1 3
-I
' "
1.1 -
.
O
•
•
$
OO
8 <J) 1
881
8t«»
NORMAL
TISSUES
0
CIRRHOTIC
TISSUES
1o e n y o i o
«•
0
o«BENIGN
•"•
§
[
o
o
o
8°
•
o
•
•
0
o
v
WELLPOORLYDIFFERENTIATED
PRIMARY CANCERS
TUMORS
METASTATIC
CANCERS
SAMPLES
8I«°
0 to
L
9
•
8o«
Bo *
OO
cm
8
LIVER
80
•
O
A
90
0
8
i.o -
100
•
o
O
O
•8 1.8co
*~* 1.7"
X 1.6 -
FIG. 2. DNA index versus histopathologic grouping, (top) The DNA index
was calculated from the integrated
optic density, which was measured
by means of a cell image processor on
200 Feulgen-stained nuclei per case.
The black and open dots represent individual DNA index assessments
from histopathologic imprint smears
andfineneedle aspiration biopsy respectively. The open dots ± bars rep- W
resent the mean DNA index values ±
their standard error. Percentage of o
diploid cells/case versus histopatho- Z
logic grouping (bottom).
o
0
9
o»»
70
ga60
o
5
OO
%
u. w
—
o
Hi r 30
<
20
Z
UJ 10 O
QC
UJ
Q.
o*
ooo
9o
O
•o
O
•O*
NORMAL
TISSUES
CIRRHOTIC
TISSUES
BENIGN
TUMORS
B
LIVER
colorectal metastases) exhibited either a diploid or a tetraploid
DNA histogram pattern that might have been falsely considered to be specific to the benign liver tissue groups (Table 2).
Figure 2 illustrates the distribution of the DNA index values
in the six histopathologic groups.The mean DNA index values
of the three nonneoplastic groups (the normal and cirrhotic
tissues and the benign tumors) were similar (P > 0.05) to each
other (Fig. 2). The same feature was observed when the mean
O
O
8
^> 8o.
8
WELLPOORLYDIFFERENTIATED
PRIMARY CANCERS
§
METASTATIC
CANCERS
SAMPLES
DNA index values from the three neoplastic groups were taken
into account. In contrast, the three mean DNA index values of
the neoplastic groups were significantly distinct from the three
mean DNA index values of the nonneoplastic groups (P < 0.01
to P < 0.001), as shown in Figure 2.
Figure 2 also reveals that 5 of 31 (16%) normal tissues were
aneuploid (DNA index > 1.25), as was also the case for 1 of 14
(7%) cirrhotic tissues. All the benign tumors appeared to be
Vol. 99 • No. 5
562
ANATOMIC PATHOLOGY
Original Article
12
u 11
o>
OO
QOo
o
0)
Q.
X
UJ
Q
Z
$
O
o
0
z
o
o
•
•oo
o»«a
<
rr
UJ
oo
OO
O
O*
o»
O
M
000««o*
O
rr
o.
O
O
t
o
OO
oo
NORMAL
TISSUES
O
CIRRHOTIC
TISSUES
••Q««
BENIGN
TUMORS
WELL-
POORLY-
DIFFERENTIATED
PRIMARY CANCERS
METASTATIC
CANCERS
FIG. 3. Proliferation index versus histopathologic grouping (top). Nuclear
area measurements versus histopathologic grouping (bottom).
LIVER SAMPLES
5200 4800 (0
8
REA
V 4400
R
4000
O
OOO
3600
<
Q. 3200
<
LU
_i 2800
O
3
Z 2400
2000
B
cm—
fifeo.
NORMAL
TISSUES
$
*
£•_
CIRRHOTIC
TISSUES
BENIGN
TUMORS
WELL-
POORLY-
DIFFERENTIATED
PRIMARY CANCERS
METASTATIC
CANCERS
LIVER SAMPLES
euploid (DNA index < 1.25). In the case of the three neoplastic
groups, 3 of 13 (23%) well-differentiated HCCs, 2 of 8 (25%)
poorly differentiated HCCs, and 1 of 19 (5%) colorectal metastases were also euploid (DNA index < 1.25). Lastly, Figure 2
shows that sample origin (FNAB rather than HIS) did not influence the results. The FNAB- and HIS-related DNA index values are mixed (P > 0.05/K.endall test).
As illustrated in Figure 2, we calculated the percentage of
diploid cells in each liver case under study to go deeper into the
characterization of the liver sample ploidy level. All samples of
the three nonneoplastic groups of liver tissues contained more
than 65% diploid cells. In sharp contrast, most but not all of the
neoplastic samples contained less than 30% diploid cells.
Determination of the Proliferation Index
Figure 3 shows that the proliferation index values varied
widely in 4 of 6 histopathologic groups under study. In the
A.J.C.P. -May 1993
X
DEPREZ ET AL.
DNA Ploidy Assessments in Liver Tumors
cirrhotic group, the proliferation index values ranged from
0.1% to > 10%. The same feature was observed with respect to
the groups of well-differentiated and poorly differentiated
HCCs, in which the mean proliferation index values were not
statistically distinct (P > 0.05) from those of the cirrhotic
group. The mean proliferation index value of the colorectal
metastatic samples was statistically higher than that of the welldifferentiated HCCs (P < 0.05) and that of the cirrhotic group
(P < 0.01), whereas it was not statistically distinct from that of
the poorly differentiated HCC group (P > 0.05).
All the 31 normal samples showed a proliferation index
value of less than 3%, with a mean proliferation index value of
less than 1 %. The same feature was observed with respect to the
benign tumor group. As reported previously about the DNA
index assessments, sample origin (FNAB or HIS) did not influence the results (P > 0.05/Kendall test).
563
TABLE 3. DETERMINATION OF THE CUTOFF VALUES
RELATING TO THE COMPUTERIZED PARAMETERS
ENABLING A CYTOLOGIC SCORE TO BE SET UP
Subscore*
1
2
3
NA
%2C
DI
PI
DHT
>65
<3.0 I + III
<3,500
<1.25
3,501-4,000 1.25-1.50 50-64 3.1-7.0 II + IV + V
<50
>4,000
>7.0 VI
>1.50
NA = nuclear area: DI = DNA index: %2C = percentage of diploid cells/case: PI = proliferation index: DHT = DNA histogram type (illustrated in Fig. 1).
* The cutoff values selected for each parameter so enabling a value ranging from 1 to 3 to be
associated with each parameter were determined according to the value distribution illustrated in Figures 2-6 and in Table 2. In any given liver case the sum of each subscore relating
to the five parameters leads to a cytologic score ranging from 5 to 15. The cytologic score
distribution in the six histopathologic groups under study is shown in Figure 6.
variables described above, ie, the DNA index, the proliferation
index, the nuclear area, the percentage of diploid cells/case,
and the DNA histogram type. The CS thus ranges from 5 to 15.
The mean nuclear area values in the three nonneoplastic
The cut-off values were chosen according to the data distribugroups were similar (P > 0.05), with individual values ranging
tion illustrated in Figs. 2, 3 and Table 2. This subscoring of
from 2000 to 3600 pixels (Fig. 3). The three mean nuclear area
parameter value therefore simply corresponds to an assessment
values in the nonneoplastic groups were significantly distinct
of degree relating to each parameter. Figure 4 illustrates the CS
from those of the three neoplastic groups (P < 0.05 to P
value distribution in the six liver histopathologic groups under
< 0.01). The three mean nuclear area values in the neoplastic
study. It appears that none of 52 nonneoplastic samples obgroups were similar (P > 0.05) within the groups, with individtained a score >8. In contrast, we observed no neoplastic-reual nuclear area measurements ranging from 2,000 to >5,000
lated CS value <7. The CS of 7 therefore represents a frontier
pixels.
between nonneoplastic and neoplastic liver cases. Three of 52
nonneoplastic liver cases obtained a score of 7, compared with
2
of 40 neoplastic samples that obtained it. Thus a CS = 7 in the
Setting up a Cytologic Score
malignant groups might be considered as a false-negative diagnosis. Referring to this CS = 7 value, the sensitivity ([trueIn Table 3 we report the cut-off values that we chose to set up
five subscores ranging from 1 to 3 and corresponding to the five positives]/[true-positives + false-negatives]) of the CS index is
Determination of the Nuclear Size
1
••
14 -
LLI
DC
O
O
C/>
13 -
••
•
•••••••
12 -
••••
••
——
11 -
•••
•••
•
••
•
•
•
•••
FIG. 4. CS versus histopathologic
10 grouping. The CS was set up from _ l
five subscores relating to thefivepa- <
9rameters under study, ie, DNA index, percentage of diploid cells/case, o
proliferation index, nuclear area, and o 8DNA histogram type. Each subscore o
ranges from 1 to 3, as explained in
7- • •
Table 3. The CS therefore ranges >from 5 to 15.
o 6- • • • • • •
5-
ittttttnt
NORMAL
TISSUES
•
•
•
•
••••
•———•
•••••••
CIRRHOTIC
TISSUES
BENIGN
TUMORS
LIVER
Vol. 99 • No. 5
POORLYWELLDIFFERENTIATED
PRIMARY CANCERS
SAMPLES
METASTATIC
CANCERS
564
ANATOMIC PATHOLOGY
Original Article
therefore 95% (38/40 malignant cases), whereas its specificity
([true-negatives]/[true-negatives + false-positives]) is 100%
(52/52 cases). The positive predictive value ([true-positives]/
[true-positives + false-positives]) of this CS is equal to 100%
(38/38 cases), whereas its negative predictive value ([truenegatives]/[true-negatives + false-negatives]) is 96% (52/54
cases). The statistical level of significance associated with the
CS-index-related discrimination between benign (normal, cirrhotic, and benign tumor tissues) than neoplastic liver samples
(well-differentiated to poorly differentiated HCCs and colorectal metastases) is as high as P < 0.0001 (Kendall test).
DISCUSSION
Fine-needle aspiration biopsy is an increasingly popular
means of evaluating liver masses.4,5'7 Cohen and co-workers5
report that FNAB offers advantages over core-needle biopsy,
including safety and accuracy, because in six combined series
totalling 7,941 patients, only two complications were reported
after liver FNAB. Both were related to intra-abdominal hemorrhage controlled by means of medical treatment. These authors
also report that accuracy in the diagnosis of HCCs by liver
FNAB has been reported to vary between 73% and 95% with a
mean of approximately 85%; this difference is related in part to
the lack of key cytologic criteria.5 Accordingly, Noguchi and
co-workers4 state that making a correct diagnosis of malignancy from the cytologic features of individual cells is often
difficult in the case of well-differentiated hepatocarcinoma because the cells resemble normal hepatocytes. In contrast, in
moderately differentiated and poorly differentiated types of hepatocarcinoma a correct diagnosis of malignancy is easily
made from the features of individual cells, but there is little or
no cytologic evidence of their hepatic origin.4 Recently, Cohen
and co-workers5 reviewed a series of FNAB specimens of the
liver to identify useful cytologic criteria to distinguish hepatocellular carcinomas from nonneoplastic livers. They examined
ten cytologic features in their study, ie, high cellularity, acinar
pattern, trabecular pattern, hyperchromasia, pleomorphism, irregularly granular chromatin, uniformly prominent nucleoli,
multiple nucleoli, an increased nuclear/cytoplasmic ratio, and
atypical naked hepatocytic nuclei. With the use of a stepwise
logistic regression analysis, three features—increased nuclear/
cytoplasmic ratio, trabecular pattern, and atypical naked hepatocytic nuclei—were identified as predictive of hepatocarcinomas. When these three criteria are used, the sensitivity in the
diagnosis of HCCs by FNAB is 100%, and specificity is 87%.5
The authors argue that the relatively lower (87%) specificity of
the three key cytologic criteria is likely to be related to the
continued difficulty in distinguishing nonneoplastic liver conditions from well-differentiated HCCs.5 The same group of authors has already showed that the cytologic features that distinguish HCCs from metastatic neoplasms are polygonal cells
separated by sinusoidal capillaries and bile.6 Ryoo and Bushmann12 report that the diagnosis of liver cirrhosis depends on
assessingfibrosisand architectural alterations of the liver. In a
needle biopsy specimen the connective tissue often is sampled
inadequately, however, a weakness that leads to an uncertain
diagnosis.12
We report the use of the digital cell image analysis of liver
Feulgen-stained nuclei to assess the nuclear area, the DNA index, the DNA histogram type, the percentage of diploid cells in
the sample analyzed, and the proliferation index in FNAB and/
or HIS of the liver. We observed that the material sampling, ie,
FNAB rather than HIS did not influence the biologic interpretation of data. Such a feature is partly corroborated by a previous work carried out on breast cancers,13 in which the
method of cell sampling, although significantly influencing the
results, did not alter the general conclusions regarding evolution of the morphonuclear features. The present results further
show that the mean DNA index value, the mean nuclear area
value, and the mean percentages of diploid cells per sample are
significantly different in the three benign groups under study
(normal, cirrhotic, and benign tumor tissues) as compared with
the mean parameter values from the three neoplastic liver
groups (well-differentiated HCCs, poorly differentiated HCCs,
and colorectal metastases). The results also show, however,
that none of these three parameters makes it possible to discriminate clearly between these six histopathologic groups at individual case level. By way of an example, several neoplastic liver
samples showed DNA indices which were not distinct from
those of normal cases. This was true with respect to all the
euploid cases. Such a feature also was observed while considering the nuclear area, in which several neoplastic tissues have
nuclei whose mean size values were as small as those in nonneoplastic liver cases. In contrast, the percentage of diploid
cells clearly distinguishes most nonneoplastic cases from neoplastic ones.
To go deeper into the characterization of the nuclear DNA
content distribution in the 92 liver cases under study, we also
characterized their DNA histogram type. We estimate that the
determination of the DNA index alone is too approximate to
characterize the nuclear DNA content of a biologic sample. For
example, a nonproliferating case that contains the same proportion of diploid and tetraploid cell nuclei will exhibit a DNA
index value equal to 1.50—the same result that would be obtained from a pure nonproliferating triploid tumor. Considering the DNA index values, it appears that some normal or
nonneoplastic cases are aneuploid, ie, the DNA index is higher
than 1.25. The DNA histogram typing reveals that all these
apparently aneuploid cases are actually composed of diploid
and tetraploid cell nuclei populations. Some neoplastic cases,
however, exhibited an apparently normal DNA histogram pattern, ie, diploid and/or tetraploid.
Usingflowcytometry, Fujimoto and colleagues recently studied the prognostic value of the DNA index alone in 203 cases of
resected hepatocellular carcinomas. DNA aneuploidy detected
in 50% of patients correlated significantly with tumor size and
the presence of vascular invasion or intrahepatic metastasis.
Patients with DNA aneuploid tumors had a significantly worse
prognosis than those with DNA diploid tumors. Lastly, among
the DNA aneuploid patients, the survival times were significantly shorter for the patients with a low DNA index (DNA
index < 1.50) than for those with a high DNA index (DNA
index > 1.50).14
CONCLUSION
By combining the five above-mentioned parameters into a
CS, we obtained a CS index = 7, which clearly discriminates
between benign and malignant cases. Below this value all the
cases were benign, while up to this value all the cases were
malignant. When the CS is equal to 7 for a given patient, the
case must be considered "suspect," and we think that a new
AJ.C.P.-May 1993
DEPREZ ET AL.
DNA Ploidy Assessments in Liver Tumors
FNAB must be performed on such a patient. The score proposed here does not make it possible to discriminate between
well-differentiated and poorly differentiated HCCs, or between
primitive and secondary lesions. In contrast, we are able to
discriminate clearly between nonneoplastic and neoplastic
liver samples, and to be more precise, between FNAB specimens taken from well-differentiated HCCs and from normal
livers. The specificity and sensitivity of the current CSs are 100
and 95%, respectively. Because we studied a relatively few tumors in some groups, however, their observed values were variable. Therefore, to be sure that an adequate sample is tested,
each week we enter new cases into our data bank.
In conclusion, none of the five computerized parameters
under study (the DNA index, the proliferation index, the DNA
histogram type, the percentage of diploid cells/case, and the
nuclear area) makes it possible to discriminate clearly between
nonneoplastic and neoplastic liver FNAB specimens. In sharp
contrast, the combination of these five parameters into a CS
ranging from 5 to 15 permits such discrimination. We are now
investigating whether an inverse relationship might exist between the increasing value of the CS and the clinical outcome
of the patient.
4.
5.
6.
7.
8.
9.
10.
Acknowledgments. The authors thank Dr. Isabelle Salmon for help- 11.
ing us in the grading of the cytologic samples and Dr. Claude Depardieu for providing us with many cytologic samples.
12.
REFERENCES
13.
1. Stillwagen GB, Order SE, Guse C, et al. Prognostic factors in unresectable hepatocellular cancer: Radiation Therapy Oncology
Group study 83-01. Int J Rad Oncol Biol Phys 1991;20:65-71.
2. Ferrucci JT. Liver tumor imaging. Cancer 1991;67:1189—1195.
3. Zaman SN, Johnson PJ, Williams R. Silent cirrhosis in patients
14.
565
with hepatocellular carcinoma: Implications for screening in
high-incidence and low-incidence areas. Cancer 1990;65:16071610.
Noguchi S, Yamamoto R, Tatsuta M, et al. Cell features and patterns infine-needleaspirates of hepatocellular carcinoma.
Cancer 1986;58:321-328.
Cohen MB, Haber MM, Holly EA, et al. Cytologic criteria to distinguish hepatocellular carcinoma from nonneoplastic liver. Am J
Clin Pathol 1991;95:125-130.
Bottles K, Cohen MB, Holly EA, et al. A stepwise logistic regression analysis of hepatocellular carcinoma: An aspiration biopsy
study. Cancer 1988;62:558-563.
Kung ITM, Chan SK, Fung K.H. Fine-needle aspiration in hepatocellular carcinoma: Combined cytologic and histologic approach. Cancer 1991;67:673-680.
Zeppa P, Zabatta A, Fulciniti F, et al. The role of morphometry in
the cytology of well-differentiated hepatocarcinoma and cirrhosis with atypia. Anal Quant Cytol Histol 1988;10:343-348.
Larsimont D, Kiss R, de Launoit Y, Melamed MR. Characterization of the morphonuclear features and DNA ploidy of typical
and atypical carcinoids and small cell carcinomas of the lung.
Am J Clin Pathol 1990;94:378-383.
Petein M, Michel P, van Velthoven R, et al. Morphonuclear relationship between prostatic intraepithelial neoplasia (PIN) and
cancers as assessed by digital cell image analysis. Am J Clin
Pathol 1991;96:628-634.
Frias-Hidvegi D. Guides to Clinical Aspiration Biopsy: Liver and
Pancreas. New York: Igahu-Shoin, 1988, p 78.
Ryoo JW, Bushmann RJ. Morphometry of liver parenchyma in
needle biopsy specimens from patients with alcoholic liver disease: Preliminary variables for the diagnosis and prognosis of
cirrhosis. Mod Pathol 1989;2:382-389.
Salmon I, Coibion M, Larsimont D, et al. Comparison of fine
needle aspirates of breast cancers to imprint smears by means of
digital cell image analysis. Anal Quant Cytol Histol
1991;13:193-200.
Fujimoto J, Okamoto E, Yamanaka N, et al. Flow cytometric
DNA analysis of hepatocellular carcinoma. Cancer 1991 ;67:
939-944.
Vol. 99 • No. 5