Hepatectomy for hepatocellular carcinoma larger than 10 cm: preoperative risk stratification to prevent futile surgery

Abstract

Objectives

Appropriate patient selection is important to achieving good outcomes and obviating futile surgery in patients with huge (≥10 cm) hepatocellular carcinoma (HCC). The aim of this study was to identify independent predictors of futile outcomes, defined as death within 3 months of surgery or within 1 year from early recurrence following hepatectomy for huge HCC.

Methods

The outcomes of 149 patients with huge HCCs who underwent resection during 1995–2012 were analysed. Multivariate logistic regression analysis was performed to identify preoperative independent predictors of futility.

Results

Independent predictors of 3-month mortality (18.1%) were: total bilirubin level <34 μmol/l [P = 0.0443; odds ratio (OR) 16.470]; platelet count of <150 000 cells/ml (P = 0.0098; OR 5.039), and the presence of portal vein tumour thrombosis (P = 0.0041; OR 5.138). The last of these was the sole independent predictor of 1-year recurrence-related mortality (17.2%). Rates of recurrence-related mortality at 3 months and 1 year were, respectively, 6.3% and 7.1% in patients with Barcelona Clinic Liver Cancer (BCLC) stage A disease, 12.5% and 14% in patients with BCLC stage B disease, and 37.8% (P = 0.0002) and 75% (P = 0.0002) in patients with BCLC stage C disease.

Conclusions

According to the present data, among patients submitted to hepatectomy for huge HCC, those with a high bilirubin level, low platelet count and portal vein thrombosis are at higher risk for futile surgery. The presence of portal vein tumour thrombosis should be regarded as a relative contraindication to surgery.

Introduction

According to tumour size, hepatocellular carcinoma (HCC) is defined as very early stage (<2 cm),¹ early stage (<3 cm),¹ large (<5 cm),^2,3 or huge (≥10 cm).⁴ The last designation carries a poor prognosis because it is commonly associated with macroscopic vascular invasion, satellite nodules, and distant intra- or extrahepatic metastasis.⁵ Among the many currently employed classifications,⁶ the Barcelona Clinic Liver Cancer (BCLC) classification is the most reliable system for predicting outcomes, at least in patients with HCC in Western countries.¹ This system is endorsed by the European Association for the Study of the Liver (EASL) and the American Association of the Study of Liver Disease (AASLD). According to the BCLC classification, surgical resection for huge HCC is indicated only in patients who are classified as having stage A disease (i.e. with a single tumour of any size), provided that they show the absence of vascular invasion and extrahepatic disease upon imaging, have liver function of Child–Pugh class A status, do not show portal hypertension and are in good general health. Nevertheless, many centres still claim that resection provides the greatest survival benefit for selected patients with so-called intermediate and advanced (BCLC stages B and C) huge HCC (Table1).^5,7–13 In practice, the decision to proceed with resection for huge HCC is based on two considerations: firstly, the operative risk associated with surgery, and secondly, the chance for longterm survival beyond that which might be achieved with standard non-surgical treatments [i.e. a median survival of 41 months with transarterial chemoembolization (TACE)¹⁴ in patients with a BCLC stage B tumour or 12 months with sorafenib for patients with a BCLC stage C tumour].^15,16 Accordingly, the present authors consider liver resection for huge HCC to be futile when it is followed by either death within 3 months of surgery or recurrence within 1 year of resection (early recurrence-related mortality). Most reports note pre-, intra- and postoperative factors as predictors of prognosis, although intra- and postoperative data are of course not available to aid patients in deciding whether to accept or decline this high-risk surgery when liver resection is initially considered. This situation was the impetus for the present study, which analyses only data available before surgery to identify independent predictors of 3-month mortality and early recurrence-related mortality.

Table 1.

Reported series (≥100 cases) of patients submitted to liver resection for huge (≥10 cm) hepatocellular carcinoma (HCC)

Study,	Cases, n	Tumour size, mma	Cirrhosis, %	BCLC stages A/B/C, %b	Mortality rate, %	Morbidity rate, %	5-year OS rate, %	5-year RFS rate, %
Eastern series
Poon et al. (2002)11	120	138 ± 30	26.7%	NA/NA/16.7%	5%d	35%	27.5%	9.5%
Hong Kong
Yeh et al.(2003)13	211	139 ± 34	29.9%	NA/NA/55.8%	4.3%i	16.1%	16.7%	12.7%
Taiwan
Chen et al. (2006)8	634	141 ± 26	86.3%	NA/NA/21.9%	2.2%f	NA	18.2%	NA
China
Lee et al. (2007)9	100	133 ± 30	NA	42%/32%/26%	0.2%i	NA	31%	20%
South Korea
Pandey et al. (2007)10	166	130 (100–240)	48.2%	NA/NA/61.4%	3%d	NA	28.6%	NA
Singapore
Ariizumi et al. (2012)7	177	130 (10–24)	19.8%	30.5%/30%/39.5%	5.6%i	NA	42%	NA
Japan
Yang et al. (2009) 3	258	132 ± 41	66.1%	NA/NA/47.7%	0.78%e	10.9%c	33%	6%
China
Western series
Pawlik et al. (2005)5	300	NA	NA	22.3%/61.4%/16.3%	5%f	NA	26.9%	NA
USA
Schrager et al. (2012)12	130	142 ± 38	39.8%	NA/NA/52.4%	6.9%d	21.5%c	18.8%	11.5%
USA
Present series (2014)	149	126 ± 31	15.4%	32.2%/37.6%/30.2%	5.4%f/14.8%g/18.1%h	43.6%	42%	24.8%
France

^aMean ± SD or median (range).

^bAs none of the studies categorized patients with huge HCC according to the BCLC classification, the present authors have classified these patients according to the data provided in the respective articles (i.e. stage A includes single huge tumours; stage B includes huge HCC associated with intrahepatic metastases; stage C includes any tumour with a portal vein tumour thrombus).

^cMajor complications only.

^dDefinitions of mortality;

^din-hospital or 30-day mortality;

^ein-hospital mortality;

^f30-day mortality;

^g60-day mortality;

^h90-day mortality;

ⁱdata not available.

BCLC, Barcelona Clinic Liver Cancer; OS, overall survival; NA, not available; RFS, recurrence-free survival; SD, standard deviation.

Materials and methods

Study population

From 1995 to 2012, 820 patients with HCC underwent hepatic resection in two Western hepatobiliary units that use the same criteria for performing resection: 419 patients underwent surgical resection carried out by any of five senior surgeons at Henri Mondor Hospital (Creteil, France), and 401 patients underwent surgical resection performed by any of three senior surgeons at Paul Brousse Hospital (Villejuif, France). All procedures were performed by two senior surgeons working in pairs. The study population included 149 consecutive patients (18.2%) with huge HCC (i.e. tumour diameter of ≥10 cm) and excellent performance status with neither lymph node nor distant metastases. Data for additional surgical patients with contralateral metastases (17 patients), extrahepatic metastases (seven patients), or tumour thrombi in the main hepatic veins or in the vena cava (11 patients) were excluded from the analysis. Data on demographics, comorbidities, liver diseases, tumour characteristics, mortality, morbidity and survival outcomes were retrieved from the prospective database of each centre. This retrospective study was approved by the institutional review boards of both centres and conformed to the ethical guidelines of the 1975 Declaration of Helsinki.

Preoperative assessment

Total bilirubin, aspartate aminotransferase (AST), alanine aminotransferase (ALT) and α-fetoprotein (AFP) levels and platelet count were reported. Preoperative values for albumin and prothrombin time were not recorded. Kidney function was classified as normal (creatinine within the normal range) or abnormal. Patients were transfused before surgery as necessary to achieve a haemoglobin level within the normal range. Need for preoperative transfusion was not recorded in the source databases.

The Charlson Comorbidity Index (CCI) was used to classify patients according to their comorbid conditions, as previously described.¹⁷ Additional points were assigned for age-adjusted CCI, as follows: 1 point for patients aged 50–59 years; 2 points for patients aged 60–69 years, and 3 points for patients aged 70–79 years. Chronological age per se was not a contraindication for surgery.

The diagnosis of HCC was based on imaging findings and/or tumour biopsy, as recommended.^1,18 In each patient, the tumour was staged using contrast-enhanced thoracic, abdominal and pelvic computed tomography (CT) scans, magnetic resonance imaging (MRI), Doppler ultrasound (US) of the liver, and bone scintigraphy. Preoperative assessment included the determination of the number of nodules, the presence of satellite nodules, the maximum tumour size and the presence of portal vein tumour thrombosis. The last condition was diagnosed when the vein was enlarged¹⁹ and displayed a wash-in and wash-out pattern in MRI^20,21 and/or CT²² and/or contrast-enhanced Doppler US.^22–24 Whenever a tumour biopsy was performed, non-tumour liver tissue was also obtained to determine the status of the underlying parenchyma.

Patients were stratified according to the BCLC staging system (Fig.1). Briefly, stage A included only single tumours. Stage B included tumours that were associated with intrahepatic metastases. Stage C included any tumour with a portal vein tumour thrombus observed upon imaging.

Figure 1.

Preoperative computed tomography scans showing (a) a huge Barcelona Clinic Liver Cancer (BCLC) stage A hepatocellular carcinoma (HCC), (b) a BCLC stage B HCC (the white arrow indicates intrahepatic metastases), and (c) a BCLC stage C HCC (the white star indicates tumour portal vein thrombosis)

Study design

The aim of this study was to identify independent preoperative predictors of futile surgery, which is defined as surgery followed by death within 3 months from any cause or death within 1 year of surgery as a result of early tumour recurrence. Multivariable logistic regression analysis was therefore used to identify predictors among 13 preoperative factors (see Statistical analysis) potentially contributing to 3-month mortality and/or early recurrence-related mortality.

Selection criteria for liver resection

Patients were selected for surgery if they fulfilled these criteria: (i) an absence of prohibitive comorbidities; (ii) Child–Pugh class A liver function, and (iii) planning of complete macroscopic resection combined with a sufficient future remnant liver volume upon preoperative CT scan volumetry. Portal vein embolization (PVE) was performed when the expected remnant liver represented <40% of the total non-tumour liver volume.²⁵ Preoperative TACE was never applied. Portal hypertension was defined as oesophageal varices detected upon endoscopy or splenomegaly, with a platelet count of <100 000 cells/ml.²⁶ Portal hypertension was not considered an absolute contraindication for hepatectomy and was discussed on a case-by-case basis, provided that the patient had no history of encephalopathy, ascites or variceal rupture. The present group's experience in this setting was recently reported.²⁶

Liver resection

All resections were performed using a right subcostal open approach.²⁶ Major hepatectomy was defined as the resection of three or more segments according to the Couinaud classification.²⁷

Three types of vascular control were employed in this series. (i) In cases of hemiliver clamping, no clamping was applied to the future remnant liver and thus clamping time was considered to be nil; similarly, clamping time was considered to be nil when only the portal vein was clamped for portal deobstruction, which maintained the blood flow within the hepatic artery. (ii) The intermittent Pringle's manoeuvre consisted of alternating cycles of liver pedicle clamping (15 min in non-cirrhotic livers and 10 min in cirrhotic livers) and clamp release (5 min), repeated until the end of the hepatectomy; the total duration of clamping was recorded. (iii) Standard total vascular exclusion of the liver was performed for tumours in proximity to the hepatocaval confluence.

Operative mortality, morbidity and follow-up

Postoperative mortality was defined as death within 3 months of surgery or at any time during hospitalization for surgery. Complications were classified according to the Dindo–Clavien system of classification.²⁸ The durations of stays in the intensive care unit (ICU) and in hospital were recorded. Patients were defined as showing postoperative liver failure if they fulfilled the 50–50 criteria on day 5 or at any time thereafter.²⁹ Massive ascites was defined as an abdominal drain output of <500 ml/day for <3 days. Renal insufficiency was defined as a serum creatinine level of <150 μmol/l. Quality of life could not be assessed in this retrospective study.

Postoperative follow-up included quarterly clinical examinations, liver function tests, measurement of AFP levels, Doppler US, and either CT or MRI. No patient was lost from follow-up. Site(s) and date of recurrence, if any, were recorded.

Pathological assessment

All specimens were reviewed by a single pathologist from each centre (MS, JC), who were blinded to the postoperative and longterm outcomes of the patients. Tumour characteristics retrieved included: the number and size of lesions; the presence of macroscopic and microscopic vascular invasion, and the presence of satellite nodules.

Cost analysis

Costs were estimated based on reimbursements made to public hospitals by the health care insurance system. The Groupe Homogène de Malade [GHM; derived from the Diagnosis-Related Group (DRG)] cost in euros was determined by the hospitals and the Agence Technique de l'Information sur l'Hospitalisation (ATIH). The GHM-based reimbursements were calculated according to cost : charge ratios, using national median charges and discharges for specific GHMs. For GHM coding, a severity-refined GHM system was defined based on the associated diagnosis, and major complications were considered as modifiers for various reimbursements. In the current context, the GHM code 07C09 was used for hepatectomies for malignant tumours. The cost of the stay for this specific group was calculated according to the relevant diagnosis and associated complications.

The cost depends not only on the medical diagnosis, but also on additional financial supplements calculated according to the severity of the treated condition(s), as well as the number of days spent in the ICU.

Statistical analysis

Continuous variables are presented as the mean ± standard deviation (SD) or as the median (range). Categorical variables are presented as numbers and percentages. The chi-squared or Kruskal–Wallis test was performed to compare clinical data, as well as intra- and postoperative outcomes among patients grouped according to BCLC classification. Survival curves were computed using the Kaplan–Meier method and P-values for comparison were computed with the log-rank test. As the aim of this study was to identify independent preoperative predictors of death within 3 months from any cause or death within 1 year of surgery as a result of early tumour recurrence, Cox multivariate analysis was not used in a survival analysis.

Multivariate logistic regression analysis was performed for 13 variables of potential prognostic importance to identify independent predictors of 3-month mortality and early recurrence-related mortality. These variables included sex, age-adjusted CCI, liver disease aetiology, status of the underlying liver parenchyma, need for PVE, total bilirubin levels, platelet counts, and the levels of AST, ALT and AFP, as well as tumour size, number of nodules, and the presence of portal vein tumour thrombosis. The selection of variables was based on a univariate screen with a liberal P-value of ≤0.1; these variables were then considered for multivariate analysis. A P-value of <0.05 was considered to indicate statistical significance. Statistical analysis was performed using StatView Version 5.0 (SAS Institute, Inc., Cary, NC, USA).

Results

Baseline characteristics of the patients

During the study period, 149 patients underwent liver resection for huge HCC (Table2). A total of 117 male (78.5%) and 32 female (21.5%) patients with a mean ± SD age of 58 ± 15 years (range: 16–89 years; median: 60 years) were included in this study. Thirty-nine patients (26.2%) presented with positive markers for hepatitis B viral infection, and nine patients (6%) presented with positive markers for hepatitis C viral infection. Alcohol consumption and metabolic syndrome were responsible for 6.7% and 4%, respectively, of liver disease in this group. The median AFP level was 31.5 ng/ml (range: 1–1 140 000 ng/ml). The mean ± SD age-adjusted CCI was 1.9 ± 1.3, and 33 patients (22.2%) exhibited an age-adjusted CCI of ≥3. Upon preoperative imaging, the mean ± SD size of the largest nodule and the mean ± SD number of nodules were 126.0 ± 30.9 mm and 1.7 ± 2.8 nodules, respectively. Macroscopic portal vein tumour thrombosis was diagnosed in 45 patients (30.2%). Ten (6.7%) patients required preoperative PVE in preparation for major hepatectomy.

Table 2.

Baseline characteristics of 149 patients submitted to liver resection for huge (≥10 cm) hepatocellular carcinoma

	Total cohort (n = 149)	BCLC class A (n = 48, 32.2%)	BCLC class B (n = 56, 37.6%)	BCLC class C (n = 45, 30.2%)	P-valuea
Patient characteristics
Age, years, mean ± SD	58 ± 15	58 ± 16	59 ± 16.5	57 ± 12	0.522
Male/female, n (%)	117 (78.5%)/32 (21.5%)	36 (75%)/12 (25%)	48 (85.7%)/8 (14.3%)	33 (73.3%)/12 (26.7%)	0.248
CCI
Weighted index, mean ± SD	0.26 ± 0.49	0.28 ± 0.5	0.19 ± 0.4	0.33 ± 0.5	0.353
Age-adjusted index, mean ± SD	1.90 ± 1.3	1.83 ± 1.1	1.77 ± 1.1	1.33 ± 0.2	0.136
Aetiology of liver disease, n (%)
HBV infection	39 (26.2%)	14 (29.2%)	10 (17.9%)	15 (33.3%)	0.302
HCV infection	9 (6%)	2 (4.2%)	3 (5.4%)	4 (8.9%)
Alcohol	10 (6.7%)	4 (8.3%)	3 (5.4%)	3 (6.7%)
Metabolic syndrome	6 (4%)	0	5 (8.9%)	1 (2.2%)
Haemochromatosis	1 (0.7%)	0	1 (1.8%)	0
No liver disease	84 (56.4%)	28 (58.3%)	34 (60.7%)	22 (48.9%)
Status of underlying liver, n (%)
Normal	48 (32.2%)	17 (35.4%)	20 (35.7%)	11 (24.4%)
Chronic hepatitis or fibrosis	78 (52.3%)	25 (52.1%)	26 (46.4%)	27 (60%)
Cirrhosis	23 (15.4%)	6 (12.5%)	10 (17.9%)	7 (15.6%)
Portal vein embolization	10 (6.7%)	3 (6.3%)	4 (7.1%)	3 (6.7%)	0.984
Tumour characteristics
Macroscopic vascular invasion at diagnosis	45 (30.2%)	0	0	45 (100%)	0.0001
Tumour size at diagnosis, mm, mean ± SD	126 ± 30.9	136 ± 37.3	143 ± 36.4	164 ± 48.3	0.0082
Tumour number at diagnosis, mean ± SD	1.7 ± 2.8	1	2.4 ± 4.4	1.5 ± 1.1	0.0002
α-fetoprotein, ng/ml, median (range)	31.5 (1–1 140 000)	14.5 (1–608 100)	28 (1–150 000)	500 (1–1 140 000)	0.01
Preoperative blood tests
Total bilirubin, μmol/l, mean ± SD	16.78 ± 34.84	12.78 ± 7.13	20.63 ± 56.97	16.41 ± 7.97	0.013
AST, IU/l, mean ± SD	83.57 ± 93.30	59.40 ± 80.50	73.55 ± 92.94	119.40 ± 97.05	<0.0001
ALT, IU/l, mean ± SD	53.70 ± 62.40	47.73 ± 78.15	47.33 ± 36.19	66.91 ± 66.56	0.007
Platelet count per ml, mean ± SD	294 000 ± 250.2	267 500 ± 102.5	335 500 ± 392.3	276 500 ± 130.9	0.916

^aChi-squared or Kruskal–Wallis test as appropriate.

ALT, alanine aminotransferase; AST, aspartate aminotransferase; BCLC, Barcelona Clinic Liver Cancer; CCI, Charlson Comorbidity Index; HBV, hepatitis B virus; HCV, hepatitis C virus; SD, standard deviation.

Tumours were classified as BCLC stage A in 48 (32.2%) patients, as BCLC stage B in 56 (37.6%) patients, and as BCLC stage C in 45 (30.2%) patients.

Surgical procedures

Major resections accounted for 123 (82.6%) of the 149 resections (Table3). The mean ± SD number of resected segments was 3.8 ± 1.3 segments (median: 4.0 segments). Some form of vascular clamping involving ischaemia in the remaining liver was required in 117 (78.5%) patients, whereas 32 (21.5%) patients did not require clamping. Intermittent hepatic pedicle clamping was applied in 73 (49%) patients, and total vascular exclusion of the liver was required in 44 (29.5%) patients. The mean ± SD total duration of ischaemia was 35.23 ± 24.11 min. Sixty-six (44.3%) patients received a transfusion of a mean ± SD of 2.2 ± 4.4 units of blood. The mean ± SD duration of surgery was 241 ± 83.8 min.

Table 3.

Intra- and postoperative data for 149 patients submitted to liver resection for huge (≥10 cm) hepatocellular carcinoma

	Total cohort (n = 149)	BCLC class A (n = 48, 32.2%)	BCLC class B (n = 56, 37.6%)	BCLC class C (n = 45, 30.2%)	P-valuea
Intraoperative data
Type of resection, n (%)
Right hepatectomy	42 (28.2%)	11 (22.9%)	16 (28.6%)	15 (33.3%)	0.154
Extended right hepatectomy	40 (26.8%)	14 (29.2%)	12 (21.4%)	14 (31.1%)
Left hepatectomy	18 (12.1%)	5 (10.4%)	6 (10.7%)	7 (15.6%)
Extended left hepatectomy	12 (8.1%)	1 (2.1%)	5 (8.9%)	6 (13.3%)
Central hepatectomy	4 (2.7%)	2 (4.2%)	1 (1.8%)	1 (2.2%)
Left lobectomy	12 (8.1%)	5 (10.4%)	7 (12.5%)	0
Segmentectomy	21 (14.1%)	10 (20.8%)	9 (16.1%)	2 (4.4%)
Number of segments resected, mean ± SD	3.8 ± 1.3	3.6 ± 2.0	3.6 ± 1.7	4.2 ± 1.1	0.0749
Major hepatectomy (≥3 segments), n (%)	123 (82.6%)	33 (68.8%)	40 (71.4%)	43 (95.6%)	0.0027
Vascular control, n (%)					0.0302
Total vascular exclusion	44 (29.5%)	15 (31.3%)	10 (17.9%)	19 (42.2%)	0.0271
Hepatic pedicle clamping	73 (49%)	22 (45.8%)	29 (51.8%)	22 (48.9%)	0.8325
No clamping	32 (21.5%)	11 (22.9%)	17 (30.4%)	4 (8.9%)	0.0317
RBCs transfused, units, mean ± SD	2.20 ± 4.40	2.08 ± 3.41	2.14 ± 5.83	2.53 ± 3.17	0.038
RBC transfusion, n (%)	66 (44.3%)	21 (43.8%)	17 (30.4%)	28 (62.2%)	0.0059
Total ischaemia time, min, mean ± SD	35.23 ± 24.11	37.98 ± 25.34	28.45 ± 20.09	40.14 ± 25.77	0.066
Operation time, min, mean ± SD	241.0 ± 83.8	250.9 ± 85.5	239.0 ± 83.9	232.0 ± 86.0	0.593
Postoperative data, n (%)
1-month mortality	8 (5.4%)	3 (6.3%)	3 (5.4%)	2 (4.4%)	0.9282
2-month mortality	22 (14.8%)	3 (6.3%)	6 (10.7%)	13 (28.9%)	0.0049
3-month mortality	27 (18.1%)	3 (6.3%)	7 (12.5%)	17 (37.8%)	0.0002
Morbidity	65 (43.6%)	21 (43.8%)	21 (37.5%)	23 (51.1%)	0.391
Severity of complications (Dindo–Clavien classification), n (%)
Grade I/II	31 (20.8%)	14 (29.2%)	8 (14.3%)	9 (20%)	0.571
Grade III/IV	17 (11.4%)	5 (10.4%)	7 (12.5%)	5 (11.1%)
ICU stay, days, mean ± SD	7.0 ± 8.5	5.7 ± 5.2	7.6 ± 9.3	8.0 ± 10.0	0.965
Hospital stay, days, mean ± SD	15.0 ± 9.3	12.8 ± 7.7	13.7 ± 8.2	19.2 ± 11.2	0.0004
Surgical margins R1, n (%)	40 (26.8%)	16 (33.3%)	10 (17.9%)	14 (31.1%)	0.153
Satellite nodules, n (%)	80 (53.7%)	0	49 (87.5%)	31 (68.9%)	<0.0001
Microscopic vascular invasion, n (%)	101 (67.8%)	19 (39.6%)	37 (66.1%)	45 (100%)	<0.0001
Capsule tumour, n (%)	112 (75.2%)	37 (77.1%)	45 (80.4%)	30 (66.7%)	0.267
Capsule effraction if present, n (%)	81 (72.3%)	19 (39.6%)	35 (62.5%)	27 (60%)	0.0012

^aChi-squared or Kruskal–Wallis test as appropriate.

BCLC, Barcelona Clinic Liver Cancer; ICU, intensive care unit; R1, positive margin resection; RBC, red blood cells; SD, standard deviation.

Mortality and morbidity

Three-month mortality

Mortality at 1 month post-surgery was 5.4% (eight of 149 patients) (Tables5). An assessment of mortality at 3 months post-surgery found that 27 patients (18.1%) had died within a mean ± SD of 42.2 ± 22.5 days after surgery (median: 44 days; range: 2–90 days). Mortality rates at 3 months were 6.3%, 12.5% and 37.8% in patients with tumours of BCLC stages A, B and C, respectively (P = 0.0002).

Table 5.

Univariate and multivariate analyses of preoperative predictors of early recurrence-related mortality in 149 patients submitted to liver resection for huge (≥10 cm) hepatocellular carcinoma

	Early mortality		Univariate analysis P-value	Multivariate analysis
	Yes n = 21	No n = 101		P-value	OR (95% CI)
Preoperative factors
Patient factors
Male gender, n (%)	16 (76.2%)	79 (78.2%)	0.839
CCI weight-index ≥3, mean ± SD	0.31 ± 0.70	0.24 ± 0.43	0.599
CCI age-adjusted ≥3, mean ± SD	1.94 ± 1.40	1.96 ± 1.25	0.944
Portal vein embolization	1 (4.8%)	6 (5.9%)	0.833
Aetiology of liver disease, n (%)
HBV infection	5 (23.8%)	24 (23.8%)	0.451
HCV infection	3 (14.3%)	4 (4%)
Alcohol	1 (4.8%)	8 (7.9%)
Metabolic syndrome	0	5 (5%)
Haemochromatosis	0	1 (1%)
Normal liver	12 (57.1%)	59 (58.4%)
Preoperative blood tests
Total bilirubin <34 μmol/l, n (%)	0	1 (1%)	0.647
Platelet count <150 000/ml, n (%)	2 (9.5%)	9 (8.9%)	0.929
AST, IU/l, mean ± SD	88 ± 72	69 ± 82	0.33
ALT, IU/l, mean ± SD	59 ± 77	51 ± 61	0.594
α-fetoprotein, ng/ml, mean ± SD	70 027 ± 154 598	18 338 ± 76 446	0.025	0.192
Tumour factors
Maximum tumour size, mm, mean ± SD	153.6 ± 50.5	140.8 ± 37.1	0.181
Multiple tumours, n (%)	7 (33.3%)	17 (16.8%)	0.084	0.159
Macroscopic vascular invasion, n (%)	12 (57.1%)	16 (15.8%)	<0.0001	0.0005	6.5 (2.3–18.6)
Tumour number, mean ± SD	1.9 ± 1.5	1.7 ± 3.2	0.828
Status of underlying liver parenchyma, n (%)
Normal	6 (28.6%)	35 (34.7%)	0.876
Chronic hepatitis or fibrosis	4 (19%)	54 (53.5%)
Cirrhotic liver	11 (52.4%)	12 (11.9%)

95% CI, 95% confidence interval; ALT, alanine aminotransferase; AST, aspartate aminotransferase; CCI, Charlson Comorbidity Index; HBV, hepatitis B virus; HCV, hepatitis C virus; OR, odds ratio; SD, standard deviation.

Univariate analysis revealed six preoperative factors that were related to 3-month mortality with P-values of <0.1: total bilirubin of <34 μmol/l (P = 0.0028); a platelet count of <150 000 cells/ml (P = 0.0037); AST (P = 0.0022); AFP (P = 0.0618); tumour size (P = 0.0062), and portal vein tumour thrombus (P < 0.0001) upon preoperative imaging (Table4). Multivariate analysis identified three independent predictors of 3-month mortality: total bilirubin <34 μmol/l [P = 0.0443; odds ratio (OR) 16.470, 95% confidence interval (CI) 1.073–252.753]; a platelet count of <150 000 cells/ml (P = 0.0098; OR 5.039, 95% CI 1.477–17.193), and portal vein tumour thrombosis (P = 0.0041; OR 5.138, 95% CI 1.678–15.728).

Table 4.

Univariate and multivariate analysis of preoperative predictors of 3-month mortality in 149 patients submitted to liver resection for huge (≥10 cm) hepatocellular carcinoma

	3-month mortality		Univariate analysis P-value	Multivariate analysis
	Yes n = 27 (18.1%)	No n = 122 (81.9%)		P-value	OR (95% CI)
Preoperative factors
Patient factors
Male gender, n (%)	22 (81.5%)	95 (77.9%)	0.6791
CCI weight-index ≥3, mean ± SD	0.26 ± 0.53	0.23 ± 0.46	0.7674
CCI age-adjusted ≥3, mean ± SD	1.519 ± 1.252	1.615 ± 1.153	0.6998
Portal vein embolization, n (%)	3 (11.1%)	7 (5.7%)	0.3126
Aetiology of liver disease, n (%)
HBV infection	10 (37%)	29 (23.8%)	0.7524
HCV infection	2 (7.4%)	7 (5.7%)
Alcohol	1 (3.7%)	9 (7.4%)
Metabolic syndrome	1 (3.7%)	5 (4.1%)
Haemochromatosis	0	1 (0.8%)
None	13 (48.1%)	71 (58.2%)
Preoperative blood tests
Total bilirubin <34 μmol/l, n (%)	3 (11.1%)	1 (0.8%)	0.0028	0.0443	16.470 (1.073–252.753)
Platelet count <150 000/ml, n (%)	8 (29.6%)	11 (9%)	0.0037	0.0098	5.039 (1.477–17.193)
AST, IU/l, mean ± SD	134.6 ± 128.0	72.3 ± 80.2	0.0022	0.8030
ALT, IU/l, mean ± SD	59.8 ± 58.3	52.4 ± 63.4	0.5911
α-fetoprotein, ng/ml, mean ± SD	83 257 ± 2 502 497	27 537 ± 96 306	0.0618	0.4593
Tumour factors
Maximum tumour size, mean ± SD	167.2 ± 46.0	142.9 ± 39.8	0.0062	0.2184
Single versus multiple tumours, n (%)	22 (81.5%)/5 (18.5%)	100 (82%)/22 (18%)	0.9527
Macroscopic vascular invasion, n (%)	17 (63%)	28 (23%)	<0.0001	0.0041	5.138 (1.678–15.728)
Tumour number, mean ± SD	1.5 ± 1.6	1.7  ± 3.0	0.6927
Status of underlying liver parenchyma, n (%)
Normal	7 (26%)	41 (33.6%)	0.2385
Chronic hepatitis or fibrosis	13 (4.1%)	65 (53.2%)
Cirrhotic liver	7 (26%)	16 (13.1%)

95% CI, 95% confidence interval; ALT, alanine aminotransferase; AST, aspartate aminotransferase; CCI, Charlson Comorbidity Index; HBV, hepatitis B virus; HCV, hepatitis C virus; SD, standard deviation.

Early recurrence-related mortality

Among the 122 patients (45, 49 and 28 patients with tumours classified as representing BCLC stage A, B and C disease, respectively) who survived at least 3 months after surgery, 79 (64.8%) patients experienced recurrence. Overall, 28 patients exhibited early recurrence (i.e. within 1 year of surgery). Recurrence was intrahepatic, extrahepatic and both intra- and extrahepatic in 10, nine and nine patients, respectively. Death from recurrence occurred in 21 of the 122 patients (rate of early recurrence-related mortality: 17.2%), within a mean time of 6.6 ± 2.4 months (range: 3.0–11.5 months). Rates of early recurrence-related mortality were 7.1% (n = 3), 14% (n = 6), and 75% (n = 12) in patients with tumours of BCLC stage A, B and C disease, respectively (P = 0.0002).

A univariate analysis identified three risk factors for early recurrence-related mortality: serum AFP level (P = 0.025); the presence of multiple tumours (P = 0.084), and portal vein tumour thrombosis (P < 0.0001) (Table5). According to the multivariate analysis, portal vein tumour thrombosis remained the sole independent risk factor for early recurrence-related mortality (P = 0.0005; OR 6.5, 95% CI 2.3–18.6).

Morbidity

A total of 84 complications were recorded in 65 patients (morbidity rate: 43.6%). According to the Dindo–Clavien system of classification, these complications were minor (33 Grade I or II complications) in 31 patients (20.8%), major (20 Grade III or IV complications) in 17 patients (11.4%), and resulted in death (Grade V) in 17 (11.4%) patients (with a total of 31 complications). Morbidity rates were similar for the three BCLC stages (P = 0.391).

For the entire series, the mean ± SD durations of stays in the ICU and hospital were 7.0 ± 8.5 days and 15.0 ± 9.3 days, respectively. The durations of stays in the ICU and in hospital (13.8 ± 14.1 days and 20.7 ± 11.9 days, respectively) in the group of patients who died within 3 months of surgery were significantly longer (P < 0.0001 and P = 0.0012, respectively) than those in patients who survived beyond this period (5.7 ± 6.0 days and 13.9 ± 8.4 days, respectively). The durations of ICU stay were similar among the three BCLC stages (P = 0.965). Mean ± SD hospital length of stay differed significantly amongst patients with BCLC stage A (12.8 ± 7.7 days), B (13.7 ± 8.2 days) and C (19.2 ± 11.2 days) disease (P = 0.0004).

Longterm results

Survival rates in the whole cohort

After a median follow-up of 17.5 months, 57 patients (25, 24 and eight patients with tumours representing BCLC stage A, B and C disease, respectively) were living and 90 patients had died. Rates of overall survival in the whole cohort at 1, 3 and 5 years were 73%, 50% and 42%, respectively (Fig.2). Rates of recurrence-free survival in the whole cohort at 1, 3 and 5 years were 51.1%, 36.9% and 24.8%, respectively (Fig.2). Rates of overall survival at 1, 3 and 5 years were 80.8%, 63.4% and 54.9%, respectively, in patients with BCLC stage A HCC, (median overall survival: 81.7 months), 71.1%, 54% and 30.9%, respectively, in patients with BCLC stage B HCC (median overall survival: 38.9 months), and 30.8%, 20.5% and 15.4%, respectively, in patients with BCLC stage C HCC (median overall survival: 4.8 months) (P < 0.0001) (Fig. S1, online). Rates of recurrence-free survival at 1, 3 and 5 years were 66%, 50% and 34.6%, respectively, in patients with BCLC stage A HCC, 52%, 42% and 25.2% respectively, in those with BCLC stage B HCC, and 32.5%, 11.6% and 11.6%, respectively, in those with BCLC stage C HCC (P = 0.0016).

Figure 2.

Overall and recurrence-free survival in 149 patients submitted to resection for huge (≥10 cm) hepatocellular carcinoma

Overall survival in patients who survived beyond the first year

A total of 84 patients (56.4% of the entire cohort, including 36, 36 and 12 patients classified as presenting HCCs of BCLC stages A, B and C, respectively) survived more than 1 year after surgery. The 1-, 3- and 5-year overall survival rates in this subgroup of patients were 100%, 74.5% and 57%, respectively, with median overall survival of 84 months. Rates of overall survival at 1, 3 and 5 years were 100%, 78% and 67%, respectively, in patients with BCLC stage A HCC, 100%, 73% and 42%, respectively, in patients with BCLC stage B HCC, and 30.8%, 66.5% and 50%, respectively, in patients with BCLC stage C HCC (P = 0.52).

Cost analysis

The mean ± SD cost of hospitalization for surgery was €17 918 ± 10 173. When stratified according to BCLC stage, mean costs of hospitalization were similar across the three groups (€17 121 ± 8573, €17 783 ± 11 332, and €18 938 ± 10 359 in patients with disease of BCLC stages A, B and C, respectively (P = 0.451). Patients who died within 3 months of surgery incurred significantly higher total charges (€121 806 ± 499 579; median: €27 174; range: €10 551–2 670 000) compared with the 122 patients who survived beyond 3 months (€37 390 ± 239 421; median: €12 501; range: €9251€–2 670 000) (P < 0.0001).

Discussion

This is the first study to analyse outcome variables in the setting of huge HCC and to use only those variables available before surgery. The present data show that good longterm survival outcomes (57% at 5 years, with median survival of 84 months) can be achieved with surgery in patients with huge HCC if they survive the first year after resection.

As clinicians await a consensus definition of futility, physiologic futility remains the most strictly defined type of futility³⁰ and refers to treatment that should not be offered³¹ because it cannot be expected to provide any survival benefit to the patient. On this basis, the present group has defined the outcome of surgery for huge HCC as futile if death occurs within 1 year of resection. This timeframe was subdivided into two periods reflecting the ‘real-life’ situation, based on, respectively, 3-month mortality and early recurrence-related mortality (within 1 year of resection) for survivors of the postoperative period. The present analyses identified high levels of total bilirubin, a low platelet count, and portal vein tumour thrombosis as independent predictors of 3-month operative mortality, whereas the last condition was the sole independent predictor of early recurrence-related mortality.

Although independent predictors of operative mortality have been previously reported (for a review, see32), the existing series concentrate on pre-, peri- and postoperative prognostic criteria. In practice, only preoperative data are available to aid a given patient in making the decision on whether to accept or decline surgery.³³ Therefore, these preoperative data were analysed in the present work to achieve the present study's objective (i.e. to identify predictors of futile surgery). The 3-month mortality rate observed in the present series (18.1%) is considerably higher than rates reported in other surgical series of huge HCC, which range from 0% to 6.9% (Table1). By contrast with the majority of these series, the present study analysed operative mortality at 3 months rather than at 1 month following surgery, as the latter time-point underestimates the true risk of liver surgery by at least 50%.^34–36 Similarly, the operative mortality rate recorded in the present series was 3.3-fold higher when measured at 3 months (18.1%) than at 1 month (5.4%). To refine the indications for surgery and to compare the present data with those in most reported series of patients with huge HCC (Table1), the present analysis has described patients in terms of not only their tumour size, but also their BCLC classification. In patients in the present series with BCLC stage A HCC (32% of patients), 3-month mortality was 6.2% and early recurrence-related mortality was 7.1% (three of 42 survivors); the actuarial survival rate was 54.9% at 5 years. These results, which are concordant with those of other series, confirm that patients with BCLC stage A HCC are optimal candidates for surgery, regardless of tumour size, even when the tumour exceeds 10 cm in diameter.³⁷ Until recently, patients with BCLC stage B HCC were treated with TACE, which achieves a significant increase in survival compared with best supportive care (median survival: 20 months versus 16 months).³⁸ Even more recently, in a randomized study, Yin et al.¹⁴ showed that patients with BCLC stage B HCC treated with surgical resection exhibited a median survival of 41 months, which is significantly longer than the median survival of 14 months reported for those treated with TACE. Multivariate analysis identified type of treatment as an independent predictor of survival, with a 2.3-fold higher likelihood of mortality observed for patients treated with TACE. The present results for the subset of patients with BCLC stage B HCC (operative mortality: 12.5%; early recurrence-related mortality: 14%; 5-year survival: 42%, with a median overall survival of 38.9 months) are concordant with those of Yin et al.,¹⁴ and thus this group will continue to propose surgery for this subset of patients with huge HCC.

Sorafenib is recommended by the current EASL and AASLD guidelines as the primary standard of care for advanced HCC (BCLC stage C). In this context, the median lengths of survival achieved in two randomized trials^15,16 comparing the outcomes of sorafenib treatment with those of placebo care were 10.7 months versus 7.9 months (P < 0.001) (Llovet et al.¹⁵) and 6.5 months versus 4.2 months (P = 0.014) (Cheng et al.¹⁶). Despite the lack of robust evidence, the present authors continue to offer surgery, as do others, to selected patients with BCLC stage C tumours. In the present study, 3-month mortality, early recurrence-related mortality and 5-year survival rates were 37.8%, 75% and 15.4%, respectively, in patients with BCLC stage C HCC. Although surgery can be expected to show better results than treatment using sorafenib, the results achieved in ‘non-ideal’ candidates for surgery are far from the ‘ideal results’ targeted by the BCLC treatment guidelines.^39–41 Many modifications to the perioperative management of these patients might be implemented to improve the current outcomes of surgery (see below). Considering the identified predictors of 3-month mortality, the present group is now reluctant to perform surgery in patients with huge HCC who exhibit increased bilirubin levels (<34 μmol/l) and/or a low platelet count (<150 000 cells/ml). For patients with low platelet counts, the reported success of neoadjuvant splenic artery embolization⁴² or even splenectomy⁴³ to increase the preoperative platelet count needs to be confirmed by further studies. Additional studies are also needed to assess the use of eltrombopag, which is efficient as a preparation for invasive procedures in patients with cirrhosis and a low platelet count.⁴⁴ Although beyond the scope of the present study, it is worth mentioning other measures which should be evaluated, including: (i) an objective measure of the patient's nutritional status and pre- and postoperative optimization;^45–48 (ii) thorough evaluation of portal hypertension and liver functional reserves before resection, including measurement of the hepatic venous pressure gradient⁴⁹ and tests to determine the indocyanine clearance rate;⁵⁰ (iii) more liberal application of PVE, even when the future remnant liver volume seems sufficient, as a test of the regenerative potential of the future remnant liver and, more importantly, so that the regeneration of the future remnant liver precedes hepatectomy by several weeks, and (iv) the implementation of conditioning manoeuvres that are either systemic or that target the remnant liver in order to condition the liver against intraoperative ischaemia–reperfusion injury subsequent to vascular clamping.^51,52 In low-risk patients with portal vein tumour thrombosis, which is a surrogate for intra- or extrahepatic metastatic disease, this group will continue to offer surgery as this remains the best palliative treatment currently available.⁵³

Oncological results require to be improved because surgery alone achieves low rates of longterm disease-free survival (24.8% at 5 years). A multimodal management strategy involving surgery might be explored, possibly including neoadjuvant chemotherapy, such as with licartine, which has been successfully used in the setting of liver transplantation for HCC,⁵⁴ an intraoperative anterior approach,^2,55 and adjuvant TACE of the remaining liver.^56,57 For huge HCC, the present authors remain reluctant to apply preoperative TACE, even using a supra-selective approach, as the size of the tumour has been found to be predictive of rupture and even mortality following this procedure.^58–60

There is a cost associated with death from surgery because most deaths do not occur during or immediately after the procedure. The present data show that patients who died within 3 months of surgery incurred significantly higher costs compared with patients who survived beyond this period.

The shortcomings of the present study include its retrospective nature. Additionally, a treatment selection bias that favours surgery over other treatments cannot be ruled out because the patients in this series were referred to two separate tertiary surgical units. However, this bias remains limited as all potential treatments were available at both centres and all cases were discussed and managed after multidisciplinary meetings during the study period. Another limitation of the study includes the large range in the OR for bilirubin in the multivariate analysis used to identify independent predictors of 3-month mortality. This might be explained by the sample's inclusion of 27 patients who died in the postoperative course and the fact that only four patients in this series were found to have a high bilirubin level according to a bilirubin cut-off of 34 μmol/l. However, the presence of hepatic vein tumour thrombus was not considered at all in this study because hepatic vein tumour thrombosis, like portal vein tumour thrombosis, is common in patients with large HCC.⁶¹ The present authors expect further reports on the impact of hepatic vein tumour thrombus on longterm outcomes in surgically resected patients with huge HCC. The present study did not assess quality of life following hepatectomy. Such a measure might highlight those patients in whom resection provided a survival benefit but sacrificed quality of life. Further investigation should examine this important issue. A recent prospective study showed not only that any physical impairments present after surgery for HCC had returned to baseline at 6 months, but also that, postoperatively, psychological quality of life improved relative to the preoperative state.⁶²

Conclusions

In conclusion, this study shows that acceptable survival rates may be achieved in patients with huge HCC. The use of predictors of futile outcomes that are available before the decision to proceed with this type of high-risk surgery may be of value to improve patient counselling and patient selection, and may help to negotiate the gap between technical resectability and the clinical operability of huge HCCs. Patients with huge HCCs classified as BCLC stage A are good candidates for surgery. For patients with huge HCCs classified as BCLC stage B, liver resection should be discussed on a case-by-case basis in patients who exhibit either a low platelet count or a high bilirubin level. Although hepatectomies performed for huge BCLC stage C tumours can be performed with reasonable longterm survival rates if patients survive beyond the first year (survival rate of 50% at 5 years), high rates of postoperative mortality indicate significant risk. Even if the presence of portal vein tumour thrombosis per se should not be regarded as an absolute contraindication to surgery, this study suggests it should be considered as a major negative prognostic factor in patients with huge HCCs.

Conflicts of interest

None declared.

Supporting Information

Additional Supporting Information may be found in the online version of this article: