Edmondson grade predicts survival of patients with primary clear cell carcinoma of liver after curative resection: A retrospective study with long-term follow-up

Wei XU, and Yilei MAO

Department of Liver Surgery, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences and Peking Union北京协和医院肝脏外科毛一雷

Medical College, Beijing, China

Asia-Pacific Journal of Clinical Oncology 2016

Abstract

Aim: Primary clear cell carcinoma of liver (PCCCL) is a specific and rare subtype of primary hepatocellular

carcinoma (HCC). We performed a retrospective study with long-term follow-up to investigate predictive

factors and prognosis of intrahepatic recurrences of PCCCL after radical resection.

Methods: We retrospectively analyzed records of 38 patients with PCCCL who were diagnosed at Peking

Union Medical College Hospital between January 1989 and September 2010, with a long-term follow up to

January 2015, to determine their clinical characteristics and postoperative survival. The data were compared

with 400 patients received radical hepatectomy for common type hepatocellular carcinoma (CHCC) during

the study period.

Results: PCCCL tumors were smaller than those of CHCC (P<0.001) and the incidence of vascular invasion

of tumors in PCCCL group was significantly lower than that in CHCC (P = 0.029). The 1-, 3-, and 5-year

overall survival (OS) for PCCCL patients were 94.6%, 67.3%, and 58.5%, respectively; 1-, 3-, and 5-year

disease-free survival (DFS) were 89.2%, 54.1%, and 48.6%, respectively. Both OS and DFS were significantly

better for PCCCL patients than for CHCC (P = 0.039 and 0.044). Cox modeling showed high Edmondson

grade to be the only independent predictive factor for survival of PCCCL patients, which were different from

those of CHCC.

Conclusions: PCCCL is a less malignant subtype of HCC than CHCC, patients with PCCCL likely have

later intrahepatic recurrences and a better prognosis. Edmondson grade predicts survival of patients with

PCCCL after curative resection; those with higher Edmondson grades may require more careful follow-up

and aggressive post-hepatectomy therapy.

Key words: clear cell carcinoma, hepatectomy, prognosis, recurrence, risk factor

INTRODUCTION

Hepatocellular carcinoma (HCC) ranks fifth in cancer incidence

and third in cancer mortality worldwide,1 and

Correspondence: Yilei MAO MD PhD, Department of Liver

Surgery, Peking Union Medical College Hospital, Chinese

Academy of Medical Sciences and Peking Union Medical

College, 1# Shuai-Fu-Yuan, Beijing, 100730, China.

Email: pumch-liver@hotmail.com

Conflicts of interest: none

Accepted for publication 13 March 2016.

includes various subtypes according to histological pattern.

Primary clear cell carcinoma of the liver (PCCCL) is

a specific and rare subtype of primary HCC; its incidence

among HCC is reportedly 2.2–6.7%.2 It is pathologically

characterized by a large proportion of tumor cells with

cytoplasm clear to hematoxylin and eosin staining,which

has been attributed to accumulation of glycogens or lipid

and changes or defects of metabolic pathways.When the

proportion of clear cells is larger than 50%, PCCCL is

generally diagnosed.3

Previous studies have shown PCCCL to have some

different clinical and pathological features from CHCC,

C _

2016 John Wiley & Sons Australia, Ltd

2 WXu et al.

including female prevalence, a higher rate of HCV infection,

capsule formation, smaller tumor size, or insufficient

development of the arterial tumor vessels.4 However,

these findings are still somewhat disputed due to the

limited cases. The prognosis of PCCCL patients is also

controversial and the prognostic factors influencing its recurrence

and survival have not been clarified.

Determining the predictive risk factors that affect intrahepatic

recurrence after surgery and prognosis is clinically

important, as it could facilitate appropriate management

during patient follow-up. However, the knowledge

still remains limited. More detailed information regarding

the clinicopathologic features and outcome of PCCCL

patients is needed to facilitate the establishment of

therapeutic strategies for these patients. This retrospective

study was designed to characterize PCCCL patients,

to explore the prognosis, and to investigate factors that

affect recurrence and survival of PCCCL.

METHODS

Between January 1989 and September 2010, 992 patients

with primary HCC underwent radical resection at

the Peking Union Medical College Hospital (PUMCH),

including 38 patients in whom PCCCL was confirmed

pathologically. Participants in our study included 38 patients

with PCCCL and 400 patients with CHCC, who

were randomly selected from 954 cases of primary HCC.

All the recruited patients gave written informed consent.

The study protocol was approved by the Ethics

Committee of PUMCH. Radical resection was defined as

complete macroscopic removal of the tumor without exposure

of tumor cells on the cut surface. Pathologic diagnoses

were confirmed by two experienced pathologists.

According to diagnostic criteria generally accepted by

pathologists in China, PCCCL was diagnosed when clear

cells accounted for more than 50% of the tumor.

Patients’ preoperative data, including age, sex, family

history, serum hepatitis B virus (HBV) surface antigen

(HBsAg), hepatitis C virus (HCV) antibody, and serum

alfa-fetoprotein (AFP) were collected, and histopathologic

information regarding tumor number and size, tumor

location, tumor capsule, vascular invasion, and cirrhotic

change in background liver were recorded. Liver

function was assessed by Child-Pugh score system. Tumors

were graded by the Edmondson grading system,

which was first described by Edmondson and Steiner in

1954 and become one of the most widely used means of

grading the pathologic features of HCC.5 This grading

system relies mainly on cytoplasm morphology (quantity,

granularity, acidophilia) and nuclear characteristics (size,

hyperchromasia).6

All patients were followed-up regularly in the outpatient

department and monitored prospectively for recurrence

by a standard protocol that included serum AFP

level, ultrasound, contrast computed tomography (CT),

and magnetic resonance imaging (MRI). Patients were

followed-up every 3 months during the first postoperative

year and at least every 6 months afterward. Abdominal

CT or MRI scans were performed every 6 months.

Recurrence was diagnosed on the basis of typical imaging

appearance in CT or MRI. Positron emission tomography

(PET) scan was routinely done on patients when

new doubtful lesions were detected by CT or MRI. Median

survival and cumulative 1-, 3- and 5-year survival

rates were calculated. Overall survival (OS) was defined

as the interval between surgery and death or the last date

of follow-up. Disease-free survival (DFS) was calculated

from the date of resection to the date when tumor recurrence

was diagnosed; if recurrence was not diagnosed at

the time of study, the cases were censored on the date of

death or the last date of follow-up.

Clinical and pathological factors were compared using

either Fisher’s exact test or Pearson’s χ2-test, as appropriate.

Survival rates were calculated using the Kaplan–

Meier method. COX-regression analysis with backward

elimination using the entire variable was performed to

identify independent risk factors with hazard ratio (HR)

and 95% confidence interval (CI). P < 0.05 was considered

statistically significant. Data analysis was performed

using SPSS 19.0 software (IBM Corp., Armonk,

NY, USA).

RESULTS

Patient characteristics

Of the 992 patients who underwent radical hepatectomies,

38 (3.83%) had pathologically diagnosed PCCCL.

Clinical and pathological characteristics of PCCCL

and CHCC patients are shown in Table 1. Most patients

in either group had HBV infection. No significant differences

in clinical features were found. The PCCCL tumors

were smaller (P < 0.001) and the incidence of vascular invasion

of tumors in PCCCL group was significantly lower

than that of CHCC (P = 0.029).

Surgical procedures

All PCCCL patients underwent surgeries for HCC, including

single-segmentectomy (n = 3), bi-segmentectomy

or double segmentectomy (n = 28), right anterior

sectorectomy (n = 2), right posterior sectorectomy

C _

2016 John Wiley & Sons Australia, Ltd Asia-Pac J Clin Oncol 2016

Edmondon grade predits survival of PCCCL patients after curative resection 3

Table 1 Comparison of clinicopathologic characteristics between PCCCL and CHCC patients

Characteristic PCCCL (n = 38) CHCC (n = 400) P value†

Age (years) 58.32 ± 11.08 55.74 ± 11.53 0.188

Gender (M/F) 31/7 336/64 0.649

HBsAg status (P/N) 29/9 295/105 0.848

HCV antibody (P/N) 5/33 32/368 0.352

ALT (U/L) 53.37 ± 57.43 48.75 ± 42.62 0.538

GGT (U/L) 82.92 ± 84.84 93.05 ± 116.44 0.601

AFP (ng/mL) 1140.84 ± 4598.20 5211.57 ± 22747.24 0.272

CA19-9 (U/mL) 39.73 ± 45.20 42.92 ± 77.57 0.803

Tumor size (cm) 3.80 ± 2.65 5.70 ± 3.78<0.001< span="">

Tumor location (R/L/B) 30/4/4 303/68/29 0.470

Multiple tumors (Y/N) 10/28 69/331 0.184

Tumor capsule (Y/N) 24/14 220/180 0.394

Vascular invasion (Y/N) 1/37 65/335 0.029

Edmondson grade (III–IV/I–II) 6/32 109/291 0.175

Cirrhosis (Y/N) 29/9 297/103 0.848

Child-Pugh classification (C or B/A) 2/36 19/381 0.702

M/F, Male/Female; P/N, Positive/Negative; R/L/B, Right/Left/Both; Y/N, Yes/No.

(n = 1), right anterior sectorectomy and segmentectomy

(n = 3), left lateral sectorectomy (n = 1). Simultaneously,

one patient underwent removal of portal vein tumor

thrombus and nine patients underwent cholecystectomy.

Lymph node dissection was performed in one patient,

in whom metastatic lymph nodes were suspected

on preoperative MRI, while proved to be chronic inflammation

by pathology after the surgery; and 21 patients

underwent inflow vascular occlusion using Pringle’s maneuver

of clamp/unclamp cycles of 20/5 min.Median surgical

times did not significantly differ (PCCCL patients:

180 min [range: 90–450 min]; controls: 220 min [range:

60–600 min, P = 0.537]), nor did median blood loss (PCCCL

patients: 200 mL [range: 50–2000 mL]; controls:

300 mL [range: 30–15 000 mL, P = 0.992]). There were

no perioperative deaths in either group.

Follow-up and patient prognosis

The median follow-up time for PCCCL patients after

surgeries was 60 months (range: 2 months to 10.9 years).

During the follow-up, 9 (23.7%) patients experienced

intrahepatic recurrence, 19 (50.0%) patients were still

alive, 14 (36.8%) patients died of cancer-related causes,

2 (5.3%) patients died of unclear causes and 3 (7.9%)

patients were unconnected for various reasons.

The overall 1-, 3- and 5-year OS for patients with primary

HCC were 86.2%, 58.4% and 41.9%, respectively.

Univariate analysis indicated that the 1-, 3- and 5-year OS

of PCCCL patients were significantly better than those of

CHCC patients (94.6%, 67.3% and 58.5% vs. 85.3%,

57.4% and 40.4%, respectively; P = 0.039; Figure 1B).

The overall 1-, 3- and 5-year DFS for patients with primary

HCC were 79.6%, 48.9% and 32.8%, respectively.

Univariate analysis indicated that the 1-, 3- and 5-year

DFS for PCCCL patients were significantly better than

those of the 400 CHCC patients treated by curative resection

in our institute during the same period (89.2%,

54.1% and 48.6% vs. 77.2%, 48.3% and 31.2%, respectively;

P = 0.044; Figure 1D). Interestingly, multivariate

analysis did not identify PCCCL as an independent protective

risk factor for either OS or DFS of patients with

primary HCC (P > 0.05).

Furthermore, we evaluated the risk factors for OS in

PCCCL patients. Cirrhotic change in background liver

and Edmondson grade were assessed as the prognostic

factors for OS by univariate analysis (P = 0.004;

Figure 1A). Multivariate analysis also showed that higher

Edmondson grade was the only independent risk factor

for PCCCL patients in terms of OS (P = 0.035, HR =

3.59 [1.10–11.74]), whereas Edmondson grade was not

an independent risk factor for poor overall survival of

CHCC patients (Table 2).

We also evaluated the risk factors for intrahepatic recurrence

in PCCCL patients. Univariate analysis indicated

that cirrhotic change in background liver and Edmondson

grade were prognostic factors for DFS rates in

PCCCL (P = 0.002; Figure 1C). However, in multivariate

analysis, only Edmondson grade remained as an independent

factor in DFS (P = 0.028, HR = 3.22 [1.13–9.17]),

which were also different from those of CHCC (Table 3).

Asia-Pac J Clin Oncol 2016 C _ 2016 John Wiley & Sons Australia, Ltd

4 WXu et al.

Figure 1 Comparison of Kaplan–Meier curves in overall survival between Edmondson grade I–II and III–IV groups (A); PCCCL

and CHCC groups (B); and in disease-free survival between Edmondson grade I–II and III–IV groups (C); PCCCL and CHCC

groups (D).

DISCUSSION

PCCCL is a specific and rare subtype of primary HCC.

Pathologic diagnose criteria for PCCCL were different

and we applied the strictest one when the proportion

of clear cells is greater than 50% (Figure 2).2 The incidence

of PCCL among primary HCC is reportedly 2.2–

6.7%. In this study, only 3.83% of 992 patients with

primary HCC who had received radical hepatectomies

were pathologically confirmed as having PCCCL.The notable

clinical features in previous studies included female

prevalence, high rate of HCV infection, high incidence

of capsule formation and propensity for cirrhotic change

in background liver.7 In our series, no significant differences

were found between PCCCL and CHCC regarding

these clinical features. Both tumor types were prone

to occur in patients with HBV infection, mostly on the

basis of liver cirrhosis. In previous studies, patients with

PCCCL had poorer liver function, possibly due to excess

fat storage.8 However, we did not find the PCCCL and

CHCC groups to significant differ in liver function as

assessed by Child-Pugh scores. Our study demonstrated

C _

2016 John Wiley & Sons Australia, Ltd Asia-Pac J Clin Oncol 2016

Edmondon grade predits survival of PCCCL patients after curative resection 5

Table 2 Comparison of prognostic factors in terms of overall survival between PCCCL and CHCC patients

PCCCL CHCC

Univariate analysis Multivariate analysis Univariate analysis Multivariate analysis

Characteristic Median

overall

survival

(month)

P-value HR (95%

CI)

P-value Median

overall

survival

(month)

P-value HR (95% CI) P-value

Age (years)†

(>/_58.5) 44.5/49.9 0.342 37.1/38.3 0.849

Gender (M/F) 47.5/45.4 0.472 37.2/39.4 0.469

HBsAg status (P/N) 46.9/51.7 0.825 37.4/38.1 0.475 1.47 (1.04–2.08) 0.031

HCV antibody (P/N) 47.0/47.4 0.891 32.3/38.0 0.224 – 0.073

ALT (>/_40 U/L) 47.1/47.7 0.735 36.8/38.1 0.447 – 0.071

GGT (>/_67 U/L) 46.8/47.6 0.893 31.7/41.7<0.001 1.45 (1.10–1.90) 0.008

AFP (>/_20 ng/ml) 45.1/49.1 0.312 33.7/43.0<0.001 1.39 (1.05–1.84) 0.020

CA19-9 (>/_37 U/mL) 47.1/47.4 0.873 32.3/39.6<0.001 1.35 (1.03–1.78) 0.030

Tumor size (>/_5 cm) 47.2/47.4 0.536 29.0/45.7<0.001 2.18 (1.66–2.88)<0.001< font="">

Multiple tumors (Y/N) 37.0/50.6 0.182 — 0.075 34.0/38.2 0.096

Tumor capsule (Y/N) 47.8/47.1 0.425 40.3/34.2 0.004 0.75 (0.57–0.98) 0.033

Vascular invasion (Y/N) 28.0/47.9 0.157 20.7/40.7<0.001 2.44 (1.78–3.35)<0.001< font="">

Edmondson grade

(III–IV/I–II)

31.6/49.8 0.004 3.59

(1.10–11.74)

0.035 31.5/39.9 0.002

Cirrhosis (Y/N) 44.5/56.0 0.036 – 0.087 36.2/41.3 0.015

Child-Pugh classification

(C or B/A)

49.0/47.3 0.970 35.8/37.7 0.638

†Patients’ age was divided by the median age; M/F,Male/Female; P/N, Positive/Negative; Y/N, Yes/No.

Asia-Pac J Clin Oncol 2016 C _ 2016 John Wiley & Sons Australia, Ltd

6 WXu et al.

Table 3 Comparison of prognostic factors in terms of disease-free survival between PCCCL and CHCC patients

PCCCL CHCC

Univariate analysis Multivariate analysis Univariate analysis Multivariate analysis

Characteristic Median

overall

survival

(month)

P-value HR (95%

CI)

P-value Median

overall

survival

(month)

P-value HR (95% CI) P-value

Age (years)??

(>/_58.5) 38.8/45.1 0.397 32.6/33.9 0.564

Gender (M/F) 42.8/39.0 0.484 32.8/34.9 0.600

HBsAg status (P/N) 39.8/48.9 0.429 32.4/35.1 0.114 1.77 (1.26–2.52) 0.001

HCV antibody (P/N) 47.0/41.3 0.542 26.5/33.7 0.092 1.90 (1.18–3.08) 0.009

ALT (>/_40 U/L) 41.4/42.6 0.820 31.3/34.5 0.136

GGT (>/_67 U/L) 40.1/42.9 0.802 27.3/37.2<0.001 – 0.082

AFP (>/_20 ng/ml) 38.9/44.7 0.537 28.7/39.2<0.001 1.43 (1.11–1.86) 0.006

CA19-9 (>/_37 U/mL) 40.9/42.5 0.879 26.6/35.6<0.001 1.38 (1.06–1.79) 0.018

Tumor size (>/_5 cm) 45.6/40.9 0.276 25.0/40.2<0.001 2.23 (1.71–2.89)<0.001< font="">

Multiple tumors (Y/N) 37.0/43.7 0.580 29.4/33.9 0.058

Tumor capsule (Y/N) 44.2/40.7 0.732 35.9/29.6 0.004 0.74 (0.57–0.95) 0.019

Vascular invasion (Y/N) 28.0/42.4 0.360 19.8/35.6<0.001 1.65 (1.21–2.25) 0.002

Edmondson grade

(III–IV/I–II)

19.8/46.4 0.002 3.22

(1.13–9.17)

0.028 27.1/35.4 0.006

Cirrhosis (Y/N) 37.6/56.0 0.009 7.32

(0.96–56.06)

0.055 31.8/36.9 0.011

Child-Pugh classification

(C or B/A)

33.0/42.5 0.167 35.3/33.0 0.944

†Patients’ age was divided by the median age; M/F,Male/Female; P/N, Positive/Negative; Y/N, Yes/No.

C _

2016 John Wiley & Sons Australia, Ltd Asia-Pac J Clin Oncol 2016

Edmondon grade predits survival of PCCCL patients after curative resection 7

Figure 2 Pathological appearances of PCCCL lesions, the mass

is mainly composed of clear cells in Edmondson grade II and

shows a pseudocapsule (A, HE × 40; B, HE × 100).

that PCCCL tumors tended to be in smaller size and less

involved with vascular invasion when they are pathologically

diagnosed, which were in accordance with findings

in the research of Li et al.9

The prognosis of patients with PCCCL is controversial.

Although some researchers claim that their prognosis

is similar to that of CHCC or perhaps even worse,10

more studies have reported PCCCL to have a better prognosis

than CHCC.3,4,7,11–13 Our study confirmed their results

and showed significantly higher 1-, 3- and 5-year

survival rates in PCCCL patients in univariate analysis.

Interestingly, multivariate analysis did not identify the

rare histopathologic type of PCCCL as a significant prognostic

factor. The better OS of PCCCL than of CHCC

patients after radical resection may relate to differences

in these tumors’ pathological characteristics: the former

tend to be smaller and to exhibit less vascular invasion at

the time of diagnosis.We postulate that cell proliferation

and vascular invasion are inhibited in the PCCCL variant

of HCC. However, the possible prognostic superiority associated

with the different biological behavior of PCCCL

requires further confirmation.

Selection of optimal management of PCCCL patients

after radical resection requires clarification of risk factors

for OS. In this study, the only independent risk

factor we identified for PCCCL was Edmondson grade,

which was widely used to assess tumor differentiation of

HCC, whereas risk factors for CHCC patients included

tumor size and vascular invasion (Table 2).14 The difference

in independent risk factors between the PCCCL

and CHCC groups may be attributable to the particular

pathological characteristics and biological behavior

of PCCCL. The presence of clear cells and fatty change

indicate that PCCCL is a less malignant form of primary

HCC and results in these tumors being classified as welldifferentiated

HCC. Thus, PCCCL patients with low Edmondson

grades are likely to have longer OS. Our findings

indicate that tumor size and vascular invasion do not

significantly impact OS of PCCCL patients. PCCCL characteristically

presents with small nodules without vascular

invasion. Although these characteristics may result

in longer survival times for PCCCL than CHCC, tumor

differentiation independently determines OS of patients

with PCCCL.

Prognostic risk factors for PCCCL patients in our serious

differed from those in other investigations. Tumor

size and vascular invasion are once regarded as independent

risk factors of PCCCL in some previous studies, just

like those findings in patients with CHCC (Table 2).15

This may result from enlarged diagnostic criteria used in

their research. When authors confirmed PCCCL patients

by the criteria that their tumor specimen contains more

than 30% clear cells, they may risk mixing the characteristics

of PCCCL and CHCC. Some researchers suggest

that preoperative liver function is an independent risk

factor for OS in PCCCL patients. 3 This inconsistency is

probably related to the varying etiology of liver cirrhosis,

whereas HCV infection is a main cause of cirrhotic

change in background liver in their subjects.

A previous study suggested that HCC prognosis was

significantly improved with increasing proportion of

clear cells. They classified PCCCL into groups according

to whether the clear cell count was 30%, 50% or even

70% of all cells and found that the group with >70%

clear cells had significantly longer survival. 9 Liu et al.

even claimed that PCCCL showed a significantly better

prognosis than CHCC only when the proportion of clear

cells was larger than 75%.7 We dismissed the proportion

Asia-Pac J Clin Oncol 2016 C _ 2016 John Wiley & Sons Australia, Ltd

8 WXu et al.

of clear cells as a risk factor of survival in our series, for

the proportion of clear cells less than 50% were not generally

accepted by pathologists to diagnose PCCCL and

tumors with clear cells ranging from 90% to 100% are

extremely rare.16 We can hardly account the proportion

of clear cells accurately. Further, PCCCL may present in a

focal pattern, a diffuse pattern, or even a mixed pattern,

which complicates precise accounts of the proportion of

clear cells in tumor (Figure 2).

Clarifying risk factors that predict PCCCL recurrence

is also important, as the main cause for the dismal outcome

of primary HCC is the high incidence of intrahepatic

recurrence.17 In our series, PCCCL patients who did

not develop recurrence lived significantly longer than did

the recurrence group. Multivariate analysis showed that

only Edmondson grade, rather than tumor size or vascular

invasion, was an independent predictor of DFS,which

were different from that of CHCC (Table 3).

Cirrhosis is a significant risk factor for recurrence after

resection for CHCC.14,17 Similar findings were discovered

in previous studies in PCCCL patients; recurrence after

resection, especially later than 12 months, has been suggested

to be only influenced by the host status such as

cirrhosis but not by any initial tumor factors.17 However,

we did not find any significant differences regarding cirrhosis

between CHCC and PCCCL patients, and cirrhosis

showed no influence on patients’ survival in multivariate

analysis. Edmondson grade is regarded as an indicator

of recurrence in PCCCL patients. As recurrence shortens

survival, the overlap among factors that predict recurrence

and survival correspond to correlations between

DFS and OS.

PCCCL patients were rare.Our study provided 38 PCCCL

patients who confirmed with strict diagnostic criteria

in pathology, and suggest that PCCCL is a less malignant

subtype of HCC than CHCC with smaller tumor

size and less vascular invasion rate and indicate that PCCCL

patients may have a better prognosis than CHCC

patients, and revealed different risk factors for recurrence

in PCCCL than in CHCC. Edmondson grade, assessed on

the basis of tumor differentiation, may provide a prognostic

diagnosis for PCCCL. The PCCCL patients with

higher Edmondson grades may require more aggressive

therapeutic strategies after their hepatectomies.

This study is subject to the limitations inherent to retrospective

studies. Because it represents the experience

of a single tertiary referral center, it may not be valid to

generalize our findings. Our results do not indicate that

PCCCL has a better prognosis that CHCC of the same

size and with the same vascular invasion profile: a truly

matched study was not possible with the limited number

of subjects in this study. Further, the possibility that tumor

cell proliferation and vascular invasion are inhibited

in the clear cell variant of HCC requires further investigation.

Although the limitations of our study include a

smaller sample size than other published reports, we applied

stricter diagnostic criteria than did other series. A

larger, multicenter study of patients from a larger geographic

region would provide more conclusive results.