A New Prognostic System for Hepatocellular Carcinoma:

A Retrospective Study of 435 Patients

THE CANCER

OF THE

LIVER ITALIAN PROGRAM (CLIP) INVESTIGATORS*

The clinical outcome of cirrhotic patients with hepatocellular carcinoma (HCC) depends both on the residual liver
function and tumor characteristics. However, the relative
prognostic weight of these variables is not well defined. The
aims of this study were to verify the value of known
prognostic factors and to devise a prognostic index more
sensitive than the commonly used Okuda stage. A retrospective analysis of the cases of HCC diagnosed at 16 Italian
institutions from 1990 to 1992 was performed. Overall
survival was the only end point used in the analysis. The
Cox model, stratified by locoregional treatment, was used
for multivariate analyses. The final model was derived from
10 randomly chosen training samples, and the prognostic
validity of the Cancer of the Liver Italian Program (CLIP)
score was assessed on the corresponding testing samples.
Four hundred thirty-five cases of HCC were collected. As of
January 1997, 313 patients (72%) were deceased. Overall
median survival was 20 months. At multivariate analysis,
independent predictive factors of survival were Child-Pugh
stage, tumor morphology, a-fetoprotein (AFP), and portal
vein thrombosis. A simple scoring system (CLIP score) was
thus produced, assigning linear scores (0/1/2) to the covariates. Compared with Okuda stage, the CLIP score, structured as a six-category tool, has a greater discriminant
ability, revealing a class of patients with an impressively
more favorable prognosis and another class with a relatively
shorter life expectancy. The CLIP score is a new prognostic
system that accounts for both liver function and tumor
characteristics. It is easy to calculate and appears to give
more precise information than the Okuda stage. (HEPATOLOGY 1998;28:751-755.)

diseases with a possibly very unfavorable quoad vitam prognosis.

Nevertheless, the relative prognostic weight of variables
measuring residual liver function and of those describing
tumor characteristics is not well defined. Thus, in clinical
practice, it is frequently difficult to make reliable estimates of
life expectancy in cirrhotic patients with HCC.
This may induce unfavorable effects. Namely, when there is
an underestimation of life expectancy, patients may be denied
chances of locoregional or systemic treatment that could
positively affect their outcome. On the other hand, when
there is an overestimation of life expectancy, patients may be
exposed to treatments that will result in no benefit, and their
prognosis may even be worsened by treatment toxicity.
An attempt of combining in a single staging system both
variables measuring liver function and a measure of tumor
extension has been successfully made by Okuda.1 The Okuda
staging system is based on serum levels of albumin and total
bilirubin, the presence of ascites, and on a gross evaluation of
whether or not the tumor has replaced more than 50% of liver
parenchyma.
The Cancer of the Liver Italian Program (CLIP) was settled
in 1993, with the aim of performing prospective, therapeutic
trials and studies on diagnostic and prognostic assessment of
HCC. The first prognostic study of the project was named
CLIP-03; it is a retrospective analysis on consecutive HCC
cases observed in the CLIP Institutions between 1990 and
1992. The aims of the study were to verify the value of known
prognostic factors and to explore whether it is possible to
produce a prognostic index more sensitive than Okudas,
accounting for both liver function and tumor characteristics.

Hepatocellular carcinoma (HCC) mainly arises in patients

already affected with chronic hepatitis or cirrhosis. To reliably
predict the outcomes of patients with HCC, both liver
function and tumor-related variables should be accounted for,
because both cirrhosis and cancer may be considered severe

PATIENTS AND METHODS

Abbreviations: HCC, hepatocellular carcinoma; CLIP, Cancer of the Liver Italian

Program; AFP, a-fetoprotein.
*See the Appendix for authors and their institutions.
Received October 21, 1997; accepted April 28, 1998.
Presented in part at the Digestive Disease Week, Washington, DC, May 10-16, 1997,
and published in abstract form in Gastroenterology 1997;112:A1252.
Supported in part by CNR-PF ACRO.
Address reprint requests to: Prof. Ciro Gallo, Cattedra di Metodologia Epidemiologica
Clinica, Seconda Universita` di Napoli, via L.Armanni 5, 80138 Napoli, Italy. Fax:
39-081-5666013/7702938.
Copyright r 1998 by the American Association for the Study of Liver Diseases.
0270-9139/98/2803-0022$3.00/0

Data Collection and Processing. All cases diagnosed between January

1, 1990, and December 31, 1992, were eligible for the study. No
minimum diagnostic criterium was imposed. The following variables were used for the analysis: age and sex of the patient, date and
modality of HCC diagnosis, date of death or of last information on
vital status, presence of cirrhosis (diagnosed with previous liver
biopsy or unequivocal laboratory and imaging signs), Child-Pugh
stage2 and its constitutive variables (albumin, bilirubin, prothrombin time, ascites, encephalopathy), tumor type, tumor extension,
Okuda stage, portal vein thrombosis, a-fetoprotein (AFP) level
(ng/dL) at diagnosis, evidence of distant metastases at diagnosis,
locoregional treatment, and systemic treatment.
A quality control of data was made by verifying consistency (e.g.,
claimed categories of established staging systems were matched with
reported values of constituting variables) and possible input errors
(all data entered were checked by two operators). When inconsistencies were found, they were discussed with the single investigator.
Statistical Methods. Overall survival was the only end-point used in
the analysis. It was defined as the time elapsed from the date of

751

752

GALLO ET AL.

HEPATOLOGY September 1998

diagnosis and either the date of death or the date of last follow-up
information. Univariate survival curves were estimated using the
Kaplan-Meier method3 and compared by means of the log rank test.4
The observed/expected death ratio for each category is reported. For
each variable, missing data were not used in the analysis if they
accounted for less than 10% of cases.
A stratified Cox proportional hazard regression model5 was used
for multivariate analyses, using Child-Pugh stage, tumor type,
Okuda stage, portal vein thrombosis, and AFP level as covariates.
Locoregional therapy (yes/no) was the stratification factor. Treatment choice strongly depends on the prognostic assessment of
patients, and a retrospective comparison of treated and untreated
patients is accordingly biased. Furthermore, locoregional treatment
may be considered as a proxy for the severity of the disease, and,
stratifying by it, we assumed a different baseline survival for treated
and untreated subjects. Cases with missing values for one or more
variables in the model were excluded from multivariate analysis.
An internal cross-validation procedure was applied, splitting the
whole sample in a training (67%) and in a testing sample (33%). The
process was repeated 10 times; thus, we had 10 training samples and
10 corresponding testing samples. The best-fitting model was
estimated for each training sample. During the analysis, we noticed
an impaired prognostic ability of the Okuda categories caused by the
strict correlation (P , .0001) of the Okuda with the Child-Pugh
stage. Therefore, we replaced the Okuda stage with its single
constitutive variables. A composite variable (tumor morphology)
was further defined by combining together tumor type and tumor
extension. Median value and interquartile range of the 10 estimates
of the relative risks are reported for each covariate retained in the
final model.
A simple scoring system (CLIP score) was derived assigning
linear scores (0/1/2) to the covariates of the final model. Prognostic
validity of the CLIP score was assessed on the 10 testing samples and
compared with the Okuda and Child-Pugh stages. Median values
and interquartile ranges of overall survival for each prognostic
category are presented.
All analyses were performed with the BMDP statistical software.6
RESULTS

A total of 435 case-report forms were collected from 16

institutions. Patients were prevalently males (male/female
ratio: 3.3). The median age was 64 years. The number of cases
recorded per year progressively increased from 1990 to 1992.
About 80% of cases had a cytological or histological confirmed diagnosis. Almost all patients had underlying cirrhosis
(Table 1).
TABLE 1. Description of Cases
Variable

Year of diagnosis
1990
1991
1992
Age (yr)
Median
Range
Sex
Male
Female
Modality of diagnosis
Cytological/histological
Imaging 1 AFP . 400
Imaging 1 AFP , 400 or unknown
Cirrhosis
Absent
Present

n (%)

118 (27.1)
128 (29.4)
189 (43.4)
64
19-89
333 (76.6)
102 (24.4)
338 (77.7)
48 (11.0)
49 (11.3)
14 (3.2)
421 (96.8)

TABLE 2. Univariate Analysis

Variable

Child-Pugh stage (unknown 5 8)

A
B
C
AFP (ng/dL) (unknown 5 13)
#10
11-400
$400
Tumor type (unknown 5 8)
Uninodular
Multinodular
Massive
Tumor extension (unknown 5 15)
#50%
.50%
Okuda stage
I
II
III
Unknown
Portal vein thrombosis (unknown 5 28)
No
Yes
Distant metastases
Absent
Present
Locoregional treatment (unknown 5 6)
No
Yes
Surgery
PEI
TACE
More than 1
Systemic treatment (unknown 5 7)
No
Yes

O/E Death

166
192
69

0.69
1.12
2.40

129
193
100

0.78
0.93
1.88

204
181
42

0.71
1.28
5.26

331
89

0.82
3.54

161
175
46
53

0.63
1.41
4.06
0.79

361
46

0.92
2.42

421
14

0.98
1.93

182
247
12
138
75
22

2.35
0.68
0.48
0.62
0.87
0.66

348
80

1.02
1.00

,.0001

,.0001

,.0001

,.0001
,.0001

,.0001

.03

,.0001

.89

Abbreviations: PEI, percutaneous ethanol injection; TACE, transarterial

chemoembolization; O/E, observed/expected.

As of January 31, 1997, 313 patients (72%) were deceased.

Overall median survival was 20 months. Estimated survival
rates were 61%, 46%, 32%, 18%, and 11% at 1, 2, 3, 4, and 5
years, respectively. All the liver function and tumor variables
were significantly associated with survival at univariate
analysis (Table 2).
As detailed in the Statistical Methods section, the final
multivariate model was derived from 10 training samples. In
Table 3, median values and interquartile ranges of the 10
relative-risk estimates are reported for the four covariates of
the model (Child-Pugh stage, tumor morphology, AFP, and
portal vein thrombosis). Tumor morphology was obtained by
combining tumor type and tumor extension.
The CLIP scoring system was finally derived assigning a
linear score (0/1/2) to the four covariates (Table 4), and its
prognostic validity was assessed on the 10 testing samples.
Median survivals of patients belonging to the different CLIP
categories in the 10 testing samples are reported in Table 5, in
which scores 5 and 6 were grouped because of the low
number of cases. Overall, the discriminant ability of the CLIP
categories is consistent in the 10 different samples: the lower
the CLIP score, the higher the median survival.
Finally, the prognostic ability of the CLIP score was

HEPATOLOGY Vol. 28, No. 3, 1998

GALLO ET AL.

TABLE 3. Multivariate Analysis: Median Values and Interquartile Ranges

of the Relative Risks Estimated From the 10 Training Samples

TABLE 5. Median Survival Estimates (months) in the 10 Testing Samples

According to the CLIP Score Categories

RR (training samples)
Covariates

Median Value

Child-Pugh stage
A
B
C
Tumor morphology
Uninodular and #50%
Multinodular and #50%
Massive or .50%
Portal vein thrombosis
No
Yes
AFP (ng/dL)
,400
$400

IQ Range

1
1.72
3.92

1.66-1.92
3.51-4.21

1
1.74
3.18

1.65-1.96
2.78-3.79

1
1.58

1.47-1.97

1
1.79

1.43-1.74

Abbreviations: RR, relative risk; IQ, interquartile.

compared with Okuda and Child-Pugh stages. In Table 6,

median values and interquartile ranges of the median survival
for the three prognostic systems in the 10 testing samples are
reported. The CLIP score performs better than both ChildPugh (quite obviously because the latter is a component of
the former one) and Okuda stage, allowing a better characterization of the extreme categories and of patients with the
intermediate Okuda stage. Observed median values of the
percentage distribution of the three prognostic systems in the
10 testing samples are also reported in Table 6.
DISCUSSION

In this article, we propose a new scoring system (CLIP

score) that accounts for both liver function and tumor
characteristics useful in prognostic assessment of patients
with HCC. The CLIP score is easy to calculate and is based on
variables that are routinely assessed during clinical examination, biochemical staging, and ultrasound study of the liver.
As with other staging systems, it may be used in the clinical
decision-making process in consideration of the high variability of survival typical of HCC patients, and to stratify
randomization in therapeutic, clinical trials.
Compared with the Okuda stage, the CLIP score has a
higher number of categories and a greater discriminant
ability, revealing a subgroup of patients with an impressively
more favorable prognosis who could be candidates to more
aggressive therapeutic strategies, provided that their efficacy
has been clearly demonstrated. Other treatments of unproven
efficacy should be administered only within clinical trials.
The CLIP score also better defines the prognostic heterogeneTABLE 4. CLIP Scoring System
Scores
Variables

Child-Pugh stage
A
B
C
Tumor moruninodular and
multinodular and massive or extenphology
extension #50%
extension #50%
sion .50%
AFP (ng/dL)
,400
$400
Portal vein
no
yes
thrombosis

753

CLIP Score
Testing
Sample

516

1 (n 5 114)
2 (n 5 110)
3 (n 5 110)
4 (n 5 120)
5 (n 5 107)
6 (n 5 105)
7 (n 5 112)
8 (n 5 123)
9 (n 5 127)
10 (n 5 128)

*
38
44
42
38
*
50
37
43
28

38
38
35
34
25
30
28
38
21
26

19
14
22
20
15
19
14
14
12
18

3
4
5
5
4
5
4
5
5
4

2
2
3
4
6
3
1
3
1
2

1
1
1
1
2
1
0
1
1
1

*Median survival not reached (survival estimate greater than 50%).

ity of Okuda stage 2. Furthermore, among the patients with a

worse prognosis, the CLIP score can sort out a subset of
patients with a median survival long enough to be considered
for clinical trials of palliative antineoplastic treatment.
Because the Okuda stage largely contains the variables
constituting the Child-Pugh stage, whose prognostic role is
well known,7-9 it seems likely that the better performance of
the CLIP score as compared with the Okuda stage is a result
of the inclusion of tumor type, portal vein thrombosis, and
AFP level. Overall, these variables can be regarded as indices
of tumor burden. Indeed, behind the estimation of the
percentage of liver parenchyma replaced by the tumor (that is
an approximate and highly subjective measure), the difference between uninodular and multinodular tumors, which
may be partly regarded as a critical variable to define
eligibility for surgery and percutaneous ethanol injection,
was already recognized to be an independent prognostic
factor for unresectable patients.8,10,11
Portal vein thrombosis, which is a relatively rare condition,
is prevalently found in unresectable cases and is usually
considered as a contraindication to transarterial chemoembolization.12 Its independent prognostic value has been recently
demonstrated in an unselected series of Western patients13
TABLE 6. Median Values and Interquartile Ranges of the Median
Survivals (months) Observed on the 10 Testing Samples for the
Three Prognostic Systems
Prognostic System
(median %)

CLIP score
0 (13.7)
1 (29.4)
2 (29.8)
3 (13.5)
4 (8.2)
5 1 6 (8.0)
Okuda stage
I (43.0)
II (44.7)
III (14.4)
Child-Pugh stage
A (36.3)
B (46.0)
C (16.2)

Median

Interquartile
Range

42.5
32.0
16.5
4.5
2.5
1.0

37.5-*
25.5-38
14-19.5
4-5
1.5-3.5
1-1

32.5
12.0
1.5

30-35.5
11-16.5
1-3

29
19
3.5

25.5-29.5
16-20.5
2-5

*Upper quartile not measurable.

Median percentage of patients observed in the 10 testing samples.

754

GALLO ET AL.

and is likely to represent direct tumor infiltration of the vessel

rather than an effect of portal hypertension.14 Only recently
have the computed tomography and Doppler ultrasound
characteristics of benign and malignant portal vein thrombosis been described.15,16 Thus, we cannot comment on the
ultrasonographic characteristics of portal vein thrombosis in
our retrospective series. However, studies using Fine Needle
Aspiration Biopsy of the portal thrombus showed that the
majority (76%-85%) of portal thromboses in patients with
HCC are neoplastic.17,18 The strong prognostic value of portal
thrombosis in the present series is consistent with these
findings.
The prognostic significance of AFP level is a controversial
issue, mainly because it can be tested as either a continuous
or categorical variable. When AFP was tested as a continuous
variable, conflicting results were reported, positive in one
case7 and negative in two.19,20 On the contrary, testing AFP as
a categorical variable introduces the problem of cut off
choice. In this study, we first used two predefined cut offs (10
and 400 ng/dL) to identify three groups of patients: those
with normal serum levels, those with moderately increased
AFP serum levels, and those with tumors producing very
high AFP serum levels. At univariate analysis, no prognostic
difference was observed between the first two subgroups
(,10 and 10-400 AFP level), while the latter (.400) had a
worse prognosis (observed/expected death: 0.78, 0.93, and
1.88 for the three subgroups, respectively). Therefore, the
cut off of 10 ng/dL was not used in the subsequent analysis.
This choice is consistent with the results found both in
resected21 and in transarterial chemoembolization-treated
patients.22,23 On the contrary, very high AFP serum levels
(.400 ng/dL) may be an indirect measure of tumor burden.
This is consistent with the finding of a direct relationship
between AFP levels and tumor size, and with the observation
of a decreasing rate of patients with high AFP levels as earlier
diagnosis of HCC was made possible by improvement of
diagnostic techniques.24 The multivariate analysis of our
study shows that AFP levels over 400 still have a slight
independent prognostic value, even when tumor morphology
is accounted for, thus suggesting that AFP could be a marker
of a biologically more aggressive phenotype, although we
cannot rule out that some patients were understaged.
To our knowledge, this is the largest unselected series
produced in the Western countries in which survival analysis
has been performed. In fact, another Italian series of similar
size was collected, but only descriptive analyses were performed, either on relationships between tumor morphology
and other variables25 or on relationships of presentation
pattern with age at diagnosis.26 In addition, our series was
accrued in a short, recent period (1990-1992), while accrual
times of other unselected published series range from 619 to
13 years27 and start in the 1970s1,19,20,24,27 or in the 1980s.13,25,28
This is a very important point, because both diagnostic and
therapeutic procedures for HCC have changed notably over
recent years. On the other hand, possible shortcomings of our
series are a result of the number of centers involved and the
consequent heterogeneity in staging and follow-up procedures. This, however, should enhance the generalizability of
our findings.
In our study on Italian patients, which was not different
from other Italian series, we observed a median survival
longer than that reported in the majority of published series
of non-European patients. Possible explanations are differ-

HEPATOLOGY September 1998

ences in race (e.g., Asian patients) or etiology of HCC (e.g.,

American patients). More likely, the increased number of
early diagnoses as a result of regular follow-up procedures in
cirrhotic patients and of the higher diagnostic yield of the
recent imaging techniques (as compared with older patients
series) could induce a false indication of prolonged survival
(i.e., the so-called lead-time bias). Finally, as in all retrospective studies, a selection bias could have occurred, with a
reduced prevalence of patients with a worse prognosis.
However, in our study, this possible bias should be less
important than in other series because of the short and recent
period of recruitment.
All scoring systems arise as a compromise between simplicity and discriminant ability. We think that, compared with the
Okuda staging system, the CLIP score increases predictive
efficacy while remaining simple. Although we tried to reduce
the overfitting bias by means of the internal cross-validation,
a further independent external validation of the CLIP score is
needed. When follow-up data will be mature, the CLIP score
will be validated on the database of the prospective CLIP-01
trial, comparing tamoxifen with no antineoplastic systemic
treatment, whose accrual was closed in January 1997.
APPENDIX

Participating Clinicians: Giuseppe Manghisi,1 Silvana Elba,1

Ascanio Mossa,1 Antonio Giorgio,2 Vincenza Aloisio,2 Anna
Perrotta,2 Berardino Tardio,3 Carlo Del Naja,3 Eugenio Caturelli,3 Maria Calandra,4 Luigi Castellano,4 Ilario De Sio,4
Gaetano Capuano,5 Domenico Pomponi,5 Fabiana Castiglione,6 Pasqualina Cocchia,6 Fabio Farinati,7 Michela Rinaldi,7 Luigi Elio Adinolfi,8 Enrico Ragone,8 Martina Felder,9
Laura Zancanella,9 Giuseppe Pasquale,10 Maria Stanzione,10
Giampiero Marone,11 Valentina DAngelo,11 Giovanni Battista
Gaeta,12 Giancarlo Giolitto,12 Bruno Lamborgese,13 Luigi
Manzione,13 Michele Russo,14 Raffaele Colurcio,14 Enzo Veltri,15 Francesco Izzo.16
Advisory Committee: Gabriele Budillon,5 Nicola Caporaso,4
Camillo Del Vecchio Blanco,4 Bruno Galanti,14 Giuseppe
Giusti,12 Gabriele Mazzacca,6 Silvio Monfardini,20 Felice
Piccinino,10 Giuseppe Ruggiero.8
Data Coordinating Center: Ciro Gallo (biostatistician),18,21
Francesco Perrone (data coordinator), 20,21 Sabino De
Placido,17,21 Giuliana Canzanella.20
Writing Committee (who assume all responsibility for the
content of the manuscript): Gaetano Capuano,5 Bruno
Daniele,19 Giovanni Battista Gaeta,12 Ciro Gallo,18,21 Francesco Perrone,20,21 Sandro Pignata.19
Institutions: 1Divisioni di Gastroenterologia e Radiologia
IRCCS De Bellis, Castellana Grotte, Bari; 2Ospedale Cotugno,
Napoli; 3Ospedale Casa Sollievo della Sofferenza, S. Giovanni
Rotondo, Foggia; 4Dipartimento di Internistica Clinica F.
Magrassi, Seconda Universita` di Napoli (SUN), Napoli;
5Cattedra di Gastroenterologia II, FED, Napoli; 6Cattedra di
Gastroenterologia I, Universita` Federico II (FED), Napoli;
7Cattedra Malattie Apparato Digerente, Policlinico Universitario, Padova; 8Istituto di Terapia Medica II, SUN, Napoli;
9Divisione Gastroenterologia, Ospedale di Bolzano; 10Istituto
Malattie Tropicali e Subtropicali, SUN, Napoli; 11Divisione
Gastroenterologia, Ospedale Ascalesi, Napoli; 12Istituto Malattie Infettive, SUN, Napoli; 13Ospedale S.Carlo, Potenza;
14Istituto Malattie Infettive-Servizio AIDS, SUN, Napoli; 15Divisione Oncologia, Ospedale S. Maria Goretti, Latina; 16Divisione di Oncologia Chirurgica C, Istituto Nazionale Tumori

HEPATOLOGY Vol. 28, No. 3, 1998

(INT), Napoli; 17Cattedra di Oncologia Medica, FED, Napoli;

18Cattedra di Metodologia Epidemiologica Clinica, SUN,
Napoli; 19Divisione di Oncologia Medica B, INT, Napoli;
20Ufficio Sperimentazioni Cliniche Controllate, Direzione
Scientifica, INT, Napoli; 21Centro Elaborazione Dati Clinici
del Mezzogiorno-CNR ACRO, Napoli; Italy.
REFERENCES
1. Okuda K, Ohtsuki T, Obata H, Tominatsu M, Okazaki N, Hasegawa H,
Nakajima Y, et al. Natural history of hepatocellular carcinoma and
prognosis in relation to treatment. Cancer 1985;56:918-928.
2. Pugh RNH, Murray-Lyon IM, Dawson J-L, Pietroni MC, Williams R.
Transection of the oesophagus for bleeding oesophageal varices. Br J
Surg 1973;60:646-649.
3. Kaplan EL, Meier P. Non parametric estimation from incomplete
observation. J Am Stat Assoc 1958;53:457-481.
4. Mantel N. Evaluation of survival data and two new rank order statistics
arising in its consideration. Cancer Chem Rep 1966;50:163-170.
5. Cox DR. Regression models and life tables. J R Stat Soc 1972;B34:187220.
6. Dixon WJ. BMDP Statistical Software, Los Angeles, CA: University of
California Press, 1992.
7. Nomura F, Ohnishi K, Honda M, Satomura Y, Nakai T, Okuda K. Clinical
features of hepatocellular carcinoma in the elderly: a study of 91 patients
older than 70 years. Br J Cancer 1994;70:690-693.
8. Mondazzi L, Bottelli R, Brambilla G, Rampoldi A, Rezakovic I, Zavaglia
C, Alberti A, et al. Transarterial oily chemoembolization for the treatment
of hepatocellular carcinoma: a multivariate analysis of prognostic factors.
HEPATOLOGY 1994;19:1115-1123.
9. Lencioni R, Bartolozzi C, Caramella D, Paolicchi A, Carrai M, Maltinti G,
Capria A, et al. Treatment of small hepatocellular carcinoma with
percutaneous ethanol injection. Cancer 1995;76:1737-1746.
10. Yamashita Y, Takahashi M, Koga Y, Saito R, Nanakawa S, Hatanaka Y,
Sato N, et al. Prognostic factor in the treatment of hepatocellular
carcinoma with transcatheter arterial embolization and arterial infusion.
Cancer 1991;67:385-391.
11. Taniguchi K, Nakata K, KatoY, SatoY, Hamasaki K, Tsuruta S, Nagataki S.
Treatment of hepatocellular carcinoma with transcatheter arterial embolization. Cancer 1994;73:1341-1345.
12. Yamada R, Sato M, Kawabata M, Nakatsuka H, Nakamura K, Takashima
S. Hepatic artery embolization in 120 patients with unresectable
hepatoma. Radiology 1983;148:397-401.
13. Stuart KE, Anand AJ, Jenkins RL. Hepatocellular carcinoma in the
United States. Prognostic features, treatment outcome, and survival.
Cancer 1996;77:2217-2222.
14. Okuda K, Ohnishi K, Kimura K, Matsutani S, Sumida M, Goto N, Musha
H, et al. Incidence of portal vein thrombosis in liver cirrhosis. An
angiographic study in 708 patients. Gastroenterology 1985;89:279-285.

GALLO ET AL.

755

15. Tublin ME, Dodd GD III, Baron RL. Benign and malignant portal vein
thrombosis: differentiation by CT characteristics. Am J Roentgenol
1997;168:719-723.
16. Dodd GD III, Memel DS, Baron RL, Eichner L, Santiguida LA. Portal vein
thrombosis in patients with cirrhosis: does sonographic detection of
intrathrombus flow allow differentiation of benign and malignant
thrombus? Am J Roentgenol 1995;165:573-577.
17. Cedrone A, Rapaccini GL, Pompili M, Aliotta A, Trombino C, De Luca F,
Caturelli E, et al. Portal vein thrombosis complicating hepatocellular
carcinoma. Value of ultrasound-guided fine-needle biopsy of the thrombus in the therapeutic management. Liver 1996;16:94-98.
18. De Sio I, Castellano L, Calandra M, Romano M, Persico M, Del Vecchio
Blanco C. Ultrasound-guided fine needle aspiration biopsy of portal vein
thrombosis in liver cirrhosis: results in 15 patients. J Gastroenterol
Hepatol 1995;10:662-665.
19. Chlebowski RT, Tong M, Weissman J, Block JB, Ramming KP, Weiner JM,
Bateman JR, et al. Hepatocellular carcinoma. Diagnostic and prognostic
features in North American patients. Cancer 1984;53:2701-2706.
20. Attali P, ProdHomme S, Pelletier G, Papoz L, Ink O, Buffet C, Etienne JP.
Prognostic factor in patients with hepatocellular carcinoma. Cancer
1987;59:2108-2111.
21. Vauthey JN, Klimstra D, Franceschi D, Tao Y, Fortner J, Blumgart L,
Brennan M. Factors affecting long-term outcome after hepatic resection
for hepatocellular carcinoma. Am J Surg 1995;169:28-35.
22. Akashi Y, Koreeda C, Enomoto S, Uchiyama S, Mizuno T, Shiozaki Y,
Sameshima Y, et al. Prognosis of unresectable hepatocellular carcinoma:
an evaluation based on multivariate analysis of 90 cases. HEPATOLOGY
1991;14:262-268.
23. Shijo H, Okazaki M, Higashihara H, Koganemaru F, Okumura M.
Hepatocellular carcinoma: a multivariate analysis of prognostic features
in patients treated with hepatic arterial embolization. Am J Gastroenterol
1992;87:1154-1159.
24. Nomura F, Ohnishi K, Tanabe Y. Clinical features and prognostic of
hepatocellular carcinoma with reference to serum alpha-fetoprotein
levels. Cancer 1989;64:1700-1707.
25. Trevisani F, Caraceni P, Bernardi M, DIntino PE, Arienti V, Amorati P,
Stefanini GF, et al. Gross pathologic types of hepatocellular carcinoma in
Italian patients. Cancer 1993;72:1557-1563.
26. Trevisani F, DIntino PE, Grazi GL, Caraceni P, Gasbarrini A, Colantoni
A, Stefanini GF, et al. Clinical and pathologic features of hepatocellular
carcinoma in young and older Italian patients. Cancer 1996;77:22232232.
27. Nagasue N, Yukaya H, Hamada T, Hirose S, Kanashima R, Inokuchi K.
The natural history of hepatocellular carcinoma. Cancer 1984;54:14611465.
28. Yamasaki T, Kurokawa F, Kato A, Irie K, Yutoku K, Terai S, Matsuzaki Y,
et al. Clinicopathologic features of early hepatocellular carcinoma.
Hepatogastroenterology 1996;43:926-931.