- Email: [email protected]
Anatomic versus nonanatomic resection in cirrhotic patients with early hepatocellular carcinoma
Anatomic versus nonanatomic resection in cirrhotic patients with early hepatocellular carcinoma Alessandro Cucchetti, MD,a Guo-Liang Qiao, MD,b Matteo Cescon, MD, PhD,a Jun Li, MD,b Yong Xia, MD,b Giorgio Ercolani, MD,a Feng Shen, MD,b and Antonio Daniele Pinna, MD, PhD,a Bologna, Italy, and Shanghai, China
Background. Whether anatomic resection (AR) for hepatocellular carcinoma (HCC) can really confer a survival advantage over non-AR (NAR), especially for cirrhotic patients, remains unclear. Methods. Prospectively collected data of 543 cirrhotic patients in Child–Pugh class A submitted to AR (n = 228) versus NAR (n = 315) for early HCC in an Eastern (n = 269) and a Western (n = 274) surgical unit, were reviewed. To control for confounding variable distributions, a 1-to-1 propensity score match was applied to compare AR and NAR outcomes (n = 298). Results. The 5-year recurrence-free and overall survivals of the 543 patients were 32.3% and 60.0%, respectively, without differences between the 2 centers (P = .635 and .479, respectively). AR conferred better overall and recurrence-free survival than NAR (P = .009 and .041, respectively), but NAR patients suffered from significantly worse hepatic dysfunction. After 1-to-1 match, AR (n = 149) and NAR (n = 149) patients had similar covariate distributions. In this matched sample, AR still conferred better recurrence-free survival over NAR (P = .044) but the beneficial effect of AR was limited to the reduction of early recurrence (<2 years) of poorly differentiated tumors and of tumors with microvascular invasion (P < .05), resulting in better overall survival (P = .018). Conclusion. In cirrhotic patients, AR for early HCC can lead to a lower early recurrence rate in tumors with unfavorable tumor features, whereas NAR will not worsen the recurrence rate in well/moderately differentiated tumors or in the absence of microvascular invasion. (Surgery 2014;155:512-21.) From the Liver and Multiorgan Transplant Unit,a S. Orsola Hospital, Alma Mater Studiorum – University of Bologna, Bologna, Italy; and the Eastern Hepatobiliary Surgery Hospital,b Shanghai, China
HEPATOCELLULAR CARCINOMA (HCC) is 1 of the 5 most common malignancies worldwide and the third most common cause of cancer related mortality1; although more common in East Asia, the incidence of HCC is increasing in the Western world.2 Liver resection is widely accepted as a safe treatment with a low operative mortality as the result of advances in surgical techniques and perioperative management.3,4 Unfortunately, the high incidence of recurrence remains the major challenge in obtaining long-term results. Most recurrences occur in the liver as the consequence of subclinical metastases, originating from the primary tumor
Supported in part by a grant from Regione Emilia Romagna. Accepted for publication October 8, 2013. Reprint requests: Alessandro Cucchetti, MD, Policlinico Sant’ Orsola-Malpighi, University of Bologna, Via Massarenti 9, 40138 Bologna, Italy. E-mail: [email protected]. 0039-6060/$ - see front matter Ó 2014 Mosby, Inc. All rights reserved. http://dx.doi.org/10.1016/j.surg.2013.10.009
512 SURGERY
growth through microscopic vascular invasion and peripheral spread, and are considered the most important factors associated with poor prognosis.4-6 On this basis, the systematic removal of a hepatic segment, confined by tumor-bearing portal tributaries, namely anatomic resection (AR), was suggested because it should be more effective for eradication of the intrahepatic metastases of HCC. On the contrary, most surgeons prefer to leave a greater portion of parenchyma of this functional unit, such as in non-AR (NAR), focusing on the preservation of a $1-cm tumor-free margin to reduce postoperative liver failure in patients with cirrhosis. It remains unclear whether AR can really confer a survival advantage over NAR. Some authors have described better long-term outcomes after AR compared with NAR, whereas others have not been able to demonstrate these benefits, as outlined by a recent meta-analysis.7 Discrepancies have probably to be considered a consequence of the intrinsic relatively low level of evidence of available literature represented by observational
Surgery Volume 155, Number 3
retrospective studies only. Meta-regression analysis on this topic suggested that both overall survival and disease-free survival after AR seem to be superior to NAR because the worse liver function reserve in the NAR group significantly affects prognosis.7 In particular, comparative data available from pertinent literature showed that patients submitted to NAR were characterized by a significantly higher prevalence of cirrhosis and more advanced hepatic dysfunction compared with patients in the AR group, and that such differences are able to modify postoperative results.7 Although a randomized, controlled trial (RCT) comparing operative approaches would be ideal, a retrospective analysis using a propensity score matching8 patients groups to reduce bias was used herein to better determine the impact of operative approach on recurrence-free and overall survival for cirrhotic patient with HCC undergoing resection. METHODS Patient selection was accomplished through 3 levels of inclusion criteria. First, all patients submitted to portosystemic shunts before or at the same time as hepatic resection, or treated as an emergency, or submitted to preoperative portal vein embolization were excluded from the analysis. Second, we considered eligible for the present study only those patients in whom no evidence of extrahepatic metastasis was present at the time of surgery, and at pathologic examination did not present tumor invasion into a major branch of the portal or hepatic veins, direct invasion of adjacent organs, or spread to the lymph nodes of the hepatic hilum. In addition, we retained only those cases in which a tumor-free margin of $1 cm was confirmed at pathologic examination; consequently, no tumor enucleations were included in the present study and all resections considered in the present analysis were curative resections at histology. Applying these criteria, between February 2001 and August 2010, 508 cirrhotic patients underwent a first curative resection at the Eastern Hepatobiliary Surgery Hospital of Shanghai, and between January 1997 and November 2011, 388 cirrhotic patients underwent the same procedure at the Department of Surgery and Transplantation of the University of Bologna: The policies of the 2 centers regarding indications for hepatic resection have been published elsewhere.4,9 Diagnosis of cirrhosis was confirmed on histologic specimens. To reduce potential confounding nomenclature of AR and NAR, a third level of inclusion criteria was adopted. Patients
Cucchetti et al 513
with large tumors were excluded from the study that was limited to patients with a single nodule of HCC #5 cm or no more than 3 nodules none >3 cm at histologic examination. In addition, the study was also limited to patients belonging to Child–Pugh class A. The final study population thus consisted of 543 cirrhotic patients, in Child– Pugh class A, resected for early HCC: A total of 269 patients came from the Eastern surgical center and 274 from the Western surgical center. All patients underwent intraoperative hepatic ultrasonography and were deemed to have resectable tumors at the time of surgery. AR was defined as the complete removal of $1 Couinaud segment containing the tumor together with the related portal vein branch and the corresponding hepatic territory. The appropriate segment margins were identified with the discoloration of the parenchyma after ligation of the corresponding arterial and portal venous branches and with intraoperative ultrasonography assistance when necessary. NAR was defined as the resection of the tumor with a margin of $1 cm without regard to segmental, sectional, or lobar anatomy. Pathologic and histologic evaluations of the resected specimens were carried out for all cases. The resected tumor, with its surrounding liver, was examined both microscopically and macroscopically for its histopathologic features. The maximal diameter of the tumor was taken as the tumor size. Curative resection was defined as complete macroscopic and microscopic removal of the tumor. Tumor differentiation and microscopic vascular invasion in the resected tumor were also determined.10-12 After discharge, all patients were observed periodically at follow-up to exclude possible recurrence of HCC: Biochemical liver function tests, serum alpha-fetoprotein level measurement and ultrasonography were performed 3 and 6 months after discharge and then according to an annual or semiannual surveillance program. When any recurrence was suspected, a computed tomography or magnetic resonance imaging was performed for confirmation. Recurrent lesions were managed aggressively by a multimodal approach, which included re-resection, transarterial chemoembolization, percutaneous radiofrequency ablation, and percutaneous ethanol injection. The treatment was decided by the pattern of recurrence, liver functional reserve, and the general condition of the patient at the time of recurrence. For selected patients with transplantable recurrence, salvage liver transplantation was also adopted. Since the end of 2008, Sorafenib (Nexavar; Bayer,
514 Cucchetti et al
Leverkusen, Germany) therapy was also adopted, either alone or in combination with percutaneous approaches for treatment of recurrence.13,14 Follow-up data were collected until December 31, 2011. The study was approved by the appropriate institutional review committees and met the guidelines of their responsible governmental agencies. Statistical analysis. Categorical variables were reported as number of cases and prevalence, and differences between subgroups were compared using the Fisher exact test Chi-square analysis. Continuous variables were initially explored for their normal distribution using the KolmogorovSmirnov test. Because normal distribution could not be confirmed for most variables, all were expressed as median and range and differences between subgroups were explored by the MannWhitney test. All analyses were 2-tailed. Overall survival was computed from the day of operation until the most recent follow-up or until patient death; recurrence-free survival was computed from the day of operation until the most recent followup visit or until clinical evidence of tumor recurrence. Patients transplanted for postoperative liver failure or tumor recurrence were censored on the day before transplantation. In particular, at the end of follow-up, in the whole study group of 543 patients, salvage transplantation was performed in 18 cases (3.3%; 11 of them underwent NAR). Survival estimates were obtained with the KaplanMeier method and compared using the log-rank test. To overcome biases owing to the different distribution of covariates among patients submitted to AR and those submitted to NAR, a 1-to-1 match was created using propensity score analysis: The propensity score represents the probability of each individual patient being assigned to a particular condition in a study given a set of known covariates.8 Propensity scores are used to reduce selection bias by equating groups on the basis of these covariates and are used to adjust for selection bias in observational studies through matching. A multivariate logistic regression was built to predict the probability of each individual patient being submitted to AR or NAR (predictive values) on the basis of covariates that showed a different distribution in the 2 groups and taking into account histologic variables that are known to be able to affect postoperative survival or recurrence. The predictive values were then used to obtain a 1-to1 match by using the nearest neighbor matching method: Matching to 5 decimal points was initially performed, followed by 4-, 3-, 2-, and 1-point matching, and cases whose propensity score
Surgery March 2014 deviated >0.10 were considered unmatched. Thus, patients for whom the propensity score was not matchable were excluded from further analysis. The Cox proportional hazard model was finally applied in the matched sample to identify independent predictors of tumor recurrence and overall survival. Effect size was also calculated for each covariate. Effect size is a measure that is independent of the sample size and can give a more robust estimation of a difference in means or proportions: Valuesj0.5j indicate considerable differences. An effect size value of
Cucchetti et al 515
Surgery Volume 155, Number 3
Table I. Baseline characteristics of the cirrhotic patients, belonging to Child–Pugh class A, submitted to surgery for early hepatocellular carcinoma in relationship with anatomic resection (AR) and nonanatomic resection (NAR) Variable Eastern Surgical Centre Western Surgical Centre Age (y) Male gender HCV positive serology HBsAg positive serology Serum albumin (g/dL) Platelet count (3103/mm3) Serum bilirubin (mg/dL) ALT (IU) INR MELD score Histologic solitary tumor Histologic size of largest tumor (cm) Histologic tumor size <2 cm Histologic tumor size >3 cm Tumor grading G3–G4 Presence of MVI
All patients (n = 543) 269 274 58 449 204 291 4.0 118 0.85 53 1.07 8 501 3.0 104 230 366 310
(49.5%) (50.5%) (48–66) (82.7%) (37.6%) (53.6%) (3.6–4.3) (86–162) (0.64–1.16) (33–84) (0.98–1.16) (7–9) (92.3%) (2.0–4.0) (19.2%) (42.4%) (67.4%) (57.1%)
AR (n = 228) 128 100 57 192 73 136 4.1 120 0.84 55 1.05 7 216 3.0 48 105 149 124
(56.1%) (43.9%) (48–65) (84.2%) (32.0%) (59.6%) (3.7–4.3) (87–167) (0.64–1.12) (34–85) (0.90–1.13) (7–8) (94.7%) (2.0–4.0) (21.1%) (46.1%) (65.4%) (54.4%)
NAR (n = 315) 141 174 59 257 131 155 3.9 115 0.85 53 1.09 8 285 3.0 56 125 217 186
(44.8%) (55.2%) (49–67) (81.6%) (41.6%) (49.2%) (3.5–4.2) (84–149) (0.64–1.18) (33–84) (1.00–1.17) (7–9) (90.5%) (2.0–4.0) (17.8%) (39.7%) (68.9%) (59.0%)
P value
Effect size
.009 — .115 .491 .025 .019 .035 .179 .521 .983 .001 .001 .074 .409 .377 .159 .405 .293
0.226 — 0.134 0.059 0.193 0.202 0.182 0.163 0.078 0.009 0.298 0.283 0.154 0.070 0.076 0.121 0.072 0.090
Continuous variables are reported as median values and interquartile ranges (25th and 75th percentiles) because normal distribution cannot be confirmed for most variables. Comparisons were performed with Mann-Whitney U test for continuous variables and Fisher exact test for categorical variables. For continuous variables, the effect size was measured after log10 transformation. The following variables, significantly different between the 2 groups and/or known to be able to affect postoperative survival or recurrence, were selected for propensity score calculation: Age, HCV serology, HBsAg serology, albumin, MELD score (INR was not included because it was already present in the MELD formula), solitary tumor, tumor size, tumor grading and MVI. ALT, Alanine aminotransferase; HBsAg, hepatitis B surface antigen; HCV, hepatitis C virus; INR, International Normalized Ratio; MELD, Model for End-Stage Liver Disease; MVI, microvascular invasion.
belonged to Child–Pugh class A, NAR patients showed lower serum albumin levels (P = .035), higher International Normalized Ratio values (P = .001), and worse Model for End-stage Liver Disease scores (P = .001) than AR patients. Histologic characteristics were similar. The 1-, 3-, and 5-year recurrence-free rates were 83.4%, 57.0%, and 37.2%, respectively, after AR, and 76.8%, 49.2%, and 28.4%, respectively, after NAR (P = .041). The 1-, 3-, and 5-year overall survival rates were 93.7%, 82.2%, and 65.1%, respectively, after AR, and 93.1%, 68.4%, and 56.1%, respectively, after NAR (P = .009). Analysis of the matched sample. After the 1-to-1 propensity score match, 298 patients were selected for comparison: 149 were submitted to AR and an equal number to NAR; details regarding propensity score calculation can be found in Supplementary Table II. As can be observed in Table II, clinical and histologic variables had very similar distributions as confirmed by the effect size of <0.1 in all cases. In this matched sample, the 30- and 90-day postoperative mortality rates were 0.3% and 1.3%, respectively. The 1-, 3-, and 5-year recurrence-free survivals were 78.5%,
52.9%, and 34.6%, respectively, and the corresponding overall survivals were 93.1%, 74.9% and 59.2% (Fig 1). As can be noted, recurrence-free survival changed its slope from the second year onward. In particular, the annual incidence rate of recurrence within 2 years of surgery (early recurrence) was 24.0% (95% confidence interval [CI],19.5–28.6) and the annual incidence of recurrence 2 years after surgery (late recurrence) was 19.6% (95% CI, 15.0–24.1). In this matched sample, patients submitted to AR still showed an overall better recurrence-free survival than NAR patients (P = .044). Results from the Cox regression are reported in Table III and showed that when tumor recurrence was stratified into early and late recurrence, the type of surgery significantly affected early recurrence (P = .015) together with tumor diameter (P = .016) and tumor grading (P = .006), but late recurrence was not significantly affected by AR (ExpB = 1.032; 95% CI, 0.644–1.652; P = .896) or the other available variables (full data not reported). These data were confirmed when patients were stratified on the basis of tumor invasiveness to further investigate how type of surgery could influence early
516 Cucchetti et al
Surgery March 2014
Table II. Baseline characteristics of the matched cohort in relationship with anatomic resection (AR) and nonanatomic resection (NAR) Variable Age, y (range) Male gender HCV-positive serology HBsAg-positive serology Serum albumin (g/dL) Platelet count (3103/mm3) Serum bilirubin, mg/dL (range) ALT, IU/L (range) INR MELD score Histologic solitary tumor Histologic size of largest tumor (cm) Tumor grading G3–G4 Presence of MVI
All patients (n = 298) 56 245 107 169 4.0 124 0.84 55 1.07 8 286 3.0 196 153
(48–66) (82.2%) (35.9%) (56.7%) (3.6–4.3) (91–165) (0.63–1.12) (35–85) (0.97–1.14) (7–9) (96.0%) (2.0–4.1) (65.8%) (51.3%)
AR (n = 149) 57 124 54 84 4.0 124 0.84 58 1.06 8 143 3.0 98 77
(48–66) (83.2%) (36.2%) (56.4%) (3.7–4.3) (91–160) (0.64–1.11) (36–89) (0.97–1.14) (7–9) (96.0%) (2.0–4.0) (65.8%) (51.7%)
NAR (n = 149) 56 121 53 85 4.0 124 0.82 54 1.07 8 143 3.0 98 76
(47–66) (81.3%) (35.6%) (57.0%) (3.6–4.4) (93–169) (0.62–1.15) (34–81) (0.97–1.15) (7–8) (96.0%) (2.0–4.1) (65.8%) (51.0%)
P value
Effect size
.507 .560 .860 .868 .784 .892 .909 .399 .996 .702 1.000 .870 1.000 .903
0.034 0.076 0.016 0.015 0.041 0.074 0.022 0.078 0.033 0.011 0.000 0.085 0.000 0.015
Continuous variables are reported as median values and interquartile range (25th and 75th percentiles). Comparisons were performed with Wilcoxon signed-rank test for continuous variables and McNemar for categorical variables. For continuous variables, effect size was measured after log10 transformation. Effect size < 0.1 indicates very small differences. ALT, Alanine aminotransferase; HBsAg, hepatitis B surface antigen; HCV, hepatitis C virus; INR, International Normalized Ratio; MELD, Model for End-Stage Liver Disease; MVI, microvascular invasion.
Fig 1. Overall (upper plot) and recurrence-free (lower plot) survivals of the matched cohort of 298 cirrhotic patients. Recurrence-free survival changes its slope from year 2 after surgery onward, dividing early from late recurrences. Linear interpolation for early recurrence: Constant = 0.977; b1 = 0.016; r2 = 0.984. For late recurrence: constant = 0.801; b1 = 0.008; r2 = 0.988.
and late recurrence. As can be noted from Figure 2 and Table IV, AR lost its prognostic significance in improving the recurrence-free survival in G1–G2 tumors and in the absence of microvascular invasion (MVI). In these 2 categories neither the overall recurrence rate nor the early recurrence rate were affected by the type of operative procedure (P > .05 in all cases). The benefit obtainable with
AR was narrowed to G3–G4 tumors and in tumors with MVI: In these 2 categories, AR provided better early recurrence rates (P = .003 and .017, respectively) even if late recurrence rates remained substantially unaffected (P > .05 in both cases). Similar findings were observed for tumor size (Fig 2 and Table IV). The 1-, 3-, and 5-year patient survivals after AR were 93.1%, 83.3%, and 65.8%, respectively, and after NAR were 93.1%, 66.8%, and 52.9% (P = .018). In particular, the overall survival was similar, between AR and NAR patients, in G1–G2 tumors (P = .388), in tumors without MVI (P = .197) and in tumors <2 cm (P = .113), whereas AR was related to better overall survival in G3–G4 tumors (P = .014); when MVI was present (P = .016) and in tumor sized $2 cm (P = .043). Cox regression results on overall survival are provided in Supplementary Table III. Regarding data from the 245 patients excluded from the match, 79 patients were submitted to AR and 166 to NAR. Patients submitted to AR had a median Model for End-stage Liver Disease score of 7 (range, 6–8) and a tumor diameter of 3.3 (range, 2.0–4.0) whereas those submitted to NAR had a median Model for End-stage Liver Disease of 8 (range, 7–10) and a tumor diameter of 2.8 (range, 2.0–3.7; P = .001 and .041, respectively), confirming that outliers were excluded from the match. DISCUSSION Surgery is the most important therapeutic approach for patients with HCC, offering a
Cucchetti et al 517
Surgery Volume 155, Number 3
Table III. Uni- and multivariate Cox regression model on early recurrence-free survival in the matched cohort of 298 cirrhotic patients Univariate regression model
Multivariate regression model
Variable
Exp(B)
95% CI
P
Exp(B)
95% CI
P
Western (vs Eastern) Surgical Centre Non-anatomic resection (vs Anatomic) Age (per year increase) Male (vs female) HCV positive (vs negative) HBsAg positive (vs negative) Serum albumin (per g/dL increase) Platelet count (3103/mm3 increase) Serum bilirubin (per mg/dL increase) ALT (per IU/L increase) INR (per unit increase) Solitary tumor (vs multiple) Size of largest tumor (per cm increase) Tumor grading G3–G4 (vs G1–G2) Presence of MVI (vs absent)
1.013 1.629 0.992 0.876 0.900 1.396 1.078 1.001 1.951 0.997 0.961 0.515 1.253 1.875 1.291
0.689–1.490 1.104–2.405 0.975–1.008 0.539–1.426 0.600–1.350 0.939–2.077 0.754–1.541 0.996–1.003 0.898–3.203 0.993–1.002 0.236–3.916 0.225–1.175 1.063–1.478 1.197–2.936 0.880–1.895
.949 .014 .311 .595 .611 .103 .681 .785 .218 .214 .995 .115 .007 .005 .092
1.625 1.229 1.889 1.142
1.100–2.401 1.040–1.451 1.205–2.961 0.826–1.577
.015 .016 .006 .422
The annual incidence rate of early recurrence (<2 years from surgery) was 24.0% (95% CI, 19.5–28.6). P < .10 at univariate Cox regression were entered in the multivariate backward proportional hazard model. ALT, Alanine aminotransferase; HBsAg, hepatitis B surface antigen; HCV, hepatitis C virus; INR, International Normalized Ratio; MELD, Model for End-Stage Liver Disease; MVI, microvascular invasion.
possibility of cure. Unfortunately, the long-term prognosis remains undermined by the high recurrence rate that is among the most important factors affecting the survival of HCC patients.4-6 The main problem that surgeons have to face while operating on cirrhotic patients is the balance between achieving a radical intervention and, at the same time, preventing the development of postoperative liver failure that could ensue from removal of too much liver parenchyma. This problem is at the basis of the dispute between AR, that should be theoretically more radical from an oncologic point of view, and NAR, which should reduce the risk of postoperative hepatic failure. In the absence of RCTs, results reported in the literature are still conflicting.7 In particular, different proportions of cirrhotic patients, analyzed together with noncirrhotic patients in previous studies, bias the differences observed between AR and NAR outcomes, because AR was mostly performed in noncirrhotic subjects that already have a favorable clinical outcome. The present study shows that even when restrictive criteria were applied to identify a homogeneous study population, namely cirrhotic patients only, belonging to Child–Pugh class A, differences between patients submitted to AR or NAR still exist, because patients submitted to NAR still show more advanced hepatic dysfunction. Thus, the present initial analysis required an in-depth evaluation of outcomes related to the 2 different operative procedures, in which the
potential confounding impact of covariates, typical of previous observational studies, was appropriately dealt with by means of a meticulous statistical approach. When patients were adequately matched through propensity score, the results from the present study show that AR might be considered the preferred operative procedure for early HCC because of the lower recurrence rate and, consequently, greater patient survival. Nevertheless, this type of resection did not determine an absolute advantage compared with NAR because the benefit seems to be limited to the reduction of early recurrences and depends on tumor features. Early recurrences are defined as subclinical metastases originating from the primary tumor and missed during staging and treatment; thus, the present finding that AR could reduce such an inauspicious event well supports its oncologic rationale. However, there is a peculiar finding of the present analysis that deserves particular attention. AR proved superior to NAR, in determining early recurrence-free survival, only in early HCC with unfavorable tumor features such as poorly or undifferentiated grade or in the presence of microscopic vascular invasion. In other words, the present results suggest that the higher the invasiveness of the tumor, the greater the need for the removal of the entire liver functional unit. This is an important aspect that acquires particular clinical interest since tumor grade can be assessed
518 Cucchetti et al
Surgery March 2014
Fig 2. Recurrence-free survival in relationship with type of surgery and histologic findings. Number of patients at risk, and distinction between early and late recurrence, are reported in Table IV. (A) G1–G2 tumor. (B) G3–G4 tumor. (C) Microscopic vascular invasion absent. (D) Microscopic vascular invasion present. (E) Tumor size <2 cm. (F) Tumor size $2 cm.
Cucchetti et al 519
Surgery Volume 155, Number 3
Table IV. Overall, early, and late recurrence-free survival in relationship with anatomic resection (AR), nonanatomic resection (NAR), and tumor histologic features in the matched population of 298 patients Survival (y)
At risk
All patients (n = 298) 1 2 3 5 G1–G2 tumors (n = 102) 1 2 3 5 G3–G4 tumors (n = 196) 1 2 3 5 MVI absent (n = 145) 1 2 3 5 MVI present (n = 153) 1 2 3 5 Size <2 cm (n = 56) 1 2 3 5 Size $2 cm (n = 242) 1 2 3 5
149 113 79 61 25 51 36 29 25 11 98 77 51 35 18 72 56 40 31 11 77 60 40 30 12 32 28 21 15 3 117 85 58 45 19
AR, % (95% CI) 84.5 68.7 57.7 37.2
(77.4–89.5) (59.9–75.8) (48.4–65.9) (27.6–46.7)
81.6 74.5 69.1 45.5
(67.5–90.0) (64.0–82.3) (53.2–80.6) (34.0–56.3)
86.1 65.7 50.3 33.5
(77.3–91.7) (49.9–77.6) (33.5–64.9) (22.6–44.8)
82.7 67.9 58.3 38.5
(71.6–89.8) (55.0–77.9) (44.7–69.6) (23.9–52.9)
84.8 64.7 55.2 32.2
(74.2–91.3) (51.7–74.3) (42.0–66.5) (20.7–43.8)
90.2 72.0 68.0 55.8
(72.5–96.7) (51.5–84.9) (47.2–82.0) (32.7–73.7)
81.9 66.4 53.5 34.2
(73.3–87.9) (56.4–74.7) (42.9–63.0) (24.3–44.4)
At risk 149 105 72 49 20 51 42 33 25 11 98 63 39 26 11 73 56 42 29 10 76 49 33 22 13 24 18 13 6 2 125 87 58 43 17
NAR, % (95% CI)
P value .044
72.7 55.3 48.2 32.2
(64.7–79.3) (46.7–63.0) (39.5–56.4) (23.2–41.5)
83.9 73.4 68.5 46.1
(70.4–91.6) (58.6–83.6) (53.0–79.7) (29.1–61.6)
65.9 45.8 37.6 25.2
(55.5 –74.4) (35.4–55.6) (27.4–47.7) (15.3–36.3)
81.9 65.8 55.4 35.2
(70.8–89.1) (53.4–75.6) (42.5–66.5) (21.8–48.9)
65.4 51.4 41.3 31.9
(53.5–75.0) (38.3–62.9) (29.5–52.6) (20.3–44.1)
78.3 64.9 56.8 47.3
(55.4–90.3) (41.9–80.7) (31.7–75.7) (21.6–69.4)
71.7 53.5 48.6 31.2
(62.9–78.8) (44.2–62.0) (39.2–57.3) (21.8–31.0)
.013 .896 .627 .974 .527 .020 .003 .845 .178 .275 .427 .120 .017 .586 .270 .283 .733 .258 .034 .296
For each subgroup, P values are reported for the entire time interval, for early recurrence (<2 years) and for late recurrence assuming 2 years as the starting point. To obtain late recurrence-free survival (RFS) rates from actuarial data, survival can be calculated as follows: 3-year late RFS = 5-year RFS/2-year RFS. CI, Confidence interval; MVI, microvascular invasion.
preoperatively with reasonable accuracy.16-23 Tumor grade can be evaluated before surgery with tumor biopsy17,18 and there is recent evidence that diffusion-weighted magnetic resonance imaging can identify degrees of tumor differentiation.21-23 In the presence of a preoperative diagnosis of poor differentiation, 1 suggestion could be to study candidates for hepatic resection to plan an AR. On the other hand, in patients without these features, a NAR could be performed without affecting recurrence-free survival. This latter aspect is of particular importance in those patients in whom the future remnant liver volume will be inadequate if an AR will be pursued. These
considerations must be made with caution, because the risks of a liver biopsy can overshadow potential benefits. In particular, the risk of seeding is reported to range around 2–3%, when not associated with percutaneous therapies, but some Authors reported seeding #10%.24 It is clear that liver biopsy cannot be suggested as a routine preoperative tool. In addition, data on the accuracy of modern imaging techniques in predicting tumor differentiation are still scarce. Consequently, preoperative information regarding tumor invasiveness should probably still be based on tumor size. In the present study, and in the largest published experience from Eguchi et al,25 the benefit
520 Cucchetti et al
from AR over NAR, in terms of recurrence-free and disease-free survivals, was not observed in HCC <2 cm. Thus, it can be suggested that when an anatomic approach cannot be pursued owing to inadequate remnant liver volume, NAR for small HCCs will not affect tumor recurrence. That is because tumor size is known to be strictly related to tumor differentiation and presence of MVI.16,19,20 We are aware that the present study is not an RCT, and the present statistical approach is not aimed at replacing one; however, we believe that these results acquire particular strength for 3 main reasons. First, matched patients had similar, or even identical, clinical and tumor features known to influence prognosis, the match was achieved with a statistical approach and not made a priori on the basis of preselected covariates. This approach excluded outlier patients that are more prone to develop postoperative liver failure or, conversely, bring a null risk for postoperative complications. In these patients, the uncertainty of the best operative approach is absent, in particular for patients at high risk for postoperative liver failure, where NAR will be the only approach that can be attempted. Second, the operative procedures were not selected by a predetermined algorithm, but were left to the surgeon’s choice because the superiority of one or the other type of resection was unclear at the time of the study. This aspect could apparently represent a possible bias right at the start, and it should be observed that other features may have affected the operative choice, such as the attempt to obtain greater tumor-free margins, presence of comorbidities, or transplant eligibility. However, the proportion of patients that were suitable for matching was quite large (298/543), suggesting that surgeons preferred AR or NAR approach independent of clinical circumstances and tumor characteristics, providing comparable cases to be analyzed. This second aspect is further highlighted by the fact that the study population derived from 2 surgical centers that provide a good representation of both Eastern and Western HCC epidemiology and surgical perspectives. Third, to the best of our knowledge, this is the largest and most welldefined study conducted on exclusively cirrhotic patients; in fact, most of previous literature results derived from mixed cohorts of cirrhotic and noncirrhotic patients.7 This feature assumes particular importance because cirrhosis represents the principal clinical impasse in the selection of the operative procedure to adopt. Nevertheless, an RCT conducted to clarify the debate is warranted, but
Surgery March 2014
unfortunately, to the best of our knowledge, only 1 RCT has been designed and at present is not yet open for participant recruitment.26 In conclusion, the present analysis suggests that anatomic resection of early HCC can reduce the early recurrence-rate after hepatic resection, and this is true for patients having poorly differentiated HCC or MVI. Nevertheless, AR cannot be applied to all patients with cirrhosis because of the risk of postoperative insufficiency; in this regard, NAR can provide similar results to that of AR in the absence of unfavorable tumor features or for small tumors. A. Cucchetti and G-L. Qiao equally contributed to the present work. The authors thank Ms. Susan West for writing assistance. SUPPLEMENTARY DATA Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10. 1016/j.surg.2013.10.009.
REFERENCES 1. El-Serag HB. Hepatocellular carcinoma: an epidemiologic view. J Clin Gastroenterol 2002;35:S72-8. 2. Parkin D. Global cancer statistics in the year 2000. Lancet Oncol 2001;2:533-43. 3. Fan ST, Lo CM, Liu CL, et al. Hepatectomy for hepatocellular carcinoma: toward zero hospital deaths. Ann Surg 1999; 229:322-30. 4. Cucchetti A, Zanello M, Cescon M, Ercolani G, Del Gaudio M, Ravaioli M, et al. Improved diagnostic imaging and interventional therapies prolong survival after resection for hepatocellular carcinoma in cirrhosis: the university of bologna experience over 10 years. Ann Surg Oncol 2011; 18:1630-7. 5. Park JH, Koh KC, Choi MS, Lee JH, Yoo BC, Paik SW, et al. Analysis of risk factors associated with early multinodular recurrences after hepatic resection for hepatocellular carcinoma. Am J Surg 2006;192:29-33. 6. Kamiyama T, Nakanishi K, Yokoo H, Kamachi H, Tahara M, Kakisaka T, et al. Analysis of the risk factors for early death due to disease recurrence or progression within 1 year after hepatectomy in patients with hepatocellular carcinoma. World J Surg Oncol 2012;10:107. 7. Cucchetti A, Cescon M, Ercolani G, Bigonzi E, Torzilli G, Pinna AD. A comprehensive meta-regression analysis on outcome of anatomic resection versus non-anatomic resection for hepatocellular carcinoma. Ann Surg Oncol 2012; 19:3697-705. 8. Austin PC. A critical appraisal of propensity-score matching in the medical literature between 1996 and 2003. Stat Med 2008;27:2037-49. 9. Huang G, Lai EC, Lau WY, Zhou WP, Shen F, Pan ZY, et al. Posthepatectomy HBV reactivation in hepatitis B-related hepatocellular carcinoma influences postoperative survival in patients with preoperative low HBV-DNA levels. Ann Surg 2013;257:490-505. 10. Edmondson HA, Steiner PE. Primary carcinoma of the liver: a study of 100 cases among 48,900 necropsies. Cancer 1954; 7:462-503.
Surgery Volume 155, Number 3
11. Nzeako UC, Goodman ZD, Ishak KG. Comparison of tumor pathology with duration of survival of North American patients with hepatocellular carcinoma. Cancer 1995;76:579-88. 12. Vauthey JN, Lauwers GY, Esnaola NF, Do KA, Belghiti J, Mirza N, et al. Simplified staging for hepatocellular carcinoma. J Clin Oncol 2002;20:1527-36. 13. Cheng AL, Kang YK, Chen Z, Tsao CJ, Qin S, Kim JS, et al. Efficacy and safety of sorafenib in patients in the AsiaPacific region with advanced hepatocellular carcinoma: a phase III randomised, double-blind, placebo-controlled trial. Lancet Oncol 2009;10:25-34. 14. Llovet JM, Ricci S, Mazzaferro V, Hilgard P, Gane E, Blanc JF, et al. Sorafenib in advanced hepatocellular carcinoma. N Engl J Med 2008;359:378-90. 15. Normand ST, Landrum MB, Guadagnoli E, Ayanian JZ, Ryan TJ, Cleary PD, et al. Validating recommendations for coronary angiography following acute myocardial infarction in the elderly: a matched analysis using propensity scores. J Clin Epidemiol 2001;54:387-98. 16. Cucchetti A, Piscaglia F, Grigioni AD, Ravaioli M, Cescon M, Zanello M, et al. Preoperative prediction of hepatocellular carcinoma tumour grade and micro-vascular invasion by means of artificial neural network: a pilot study. J Hepatol 2010;52:880-8. 17. Colecchia A, Scaioli E, Montrone L, Vestito A, Di Biase AR, Pieri M, et al. Pre-operative liver biopsy in cirrhotic patients with early hepatocellular carcinoma represents a safe and accurate diagnostic tool for tumour grading assessment. J Hepatol 2011;54:300-5. 18. Pawlik TM, Gleisner AL, Anders RA, Assumpcao L, Maley W, Choti MA. Preoperative assessment of hepatocellular carcinoma tumour grade using needle biopsy: implications for transplant eligibility. Ann Surg 2007;245:435-42. 19. Kim BK, Han KH, Park YN, Park MS, Kim KS, Choi JS, et al. Prediction of microvascular invasion before curative resection of hepatocellular carcinoma. J Surg Oncol 2008;97:246-52.
Cucchetti et al 521
20. Pawlik TM, Delman KA, Vauthey JN, Nagorney DM, Ng IO, Ikai I, et al. Tumour size predicts vascular invasion and histologic grade: implications for selection of surgical treatment for hepatocellular carcinoma. Liver Transpl 2005;11: 1086-92. 21. Nasu K, Kuroki Y, Tsukamoto T, Nakajima H, Mori K, Minami M. Diffusion-weighted imaging of surgically resected hepatocellular carcinoma: imaging characteristics and relationship among signal intensity, apparent diffusion coefficient, and histopathologic grade. AJR Am J Roentgenol 2009;193:438-44. 22. Saito K, Moriyasu F, Sugimoto K, Nishio R, Saguchi T, Akata S, et al. Histological grade of differentiation of hepatocellular carcinoma: comparison of the efficacy of diffusion-weighted MRI with T2-weighted imaging and angiography-assisted CT. J Med Imaging Radiat Oncol 2012;56:261-9. 23. Witjes CD, Willemssen FE, Verheij J, van der Veer SJ, Hansen BE, Verhoef C, et al. Histological differentiation grade and microvascular invasion of hepatocellular carcinoma predicted by dynamic contrast-enhanced MRI. J Magn Reson Imaging 2012;36:641-7. 24. Stigliano R, Marelli L, Yu D, Davies N, Patch D, Burroughs AK. Seeding following percutaneous diagnostic and therapeutic approaches for hepatocellular carcinoma. What is the risk and the outcome? Seeding risk for percutaneous approach of HCC. Cancer Treat Rev 2007;33:437-47. 25. Eguchi S, Kanematsu T, Arii S, Okazaki M, Okita K, Omata M, et al. Comparison of the outcomes between an anatomical subsegmentectomy and a non-anatomical minor hepatectomy for single hepatocellular carcinomas based on a Japanese nationwide survey. Surgery 2008;143:469-75. 26. Prognostic impact of anatomical resection vs. nonanatomical resection for HCC (ARversusNAR) [cited 2013 May 31]. Available from: http://clinicaltrials.gov/ct2/ show/NCT01236989.
Background. Whether anatomic resection (AR) for hepatocellular carcinoma (HCC) can really confer a survival advantage over non-AR (NAR), especially for cirrhotic patients, remains unclear. Methods. Prospectively collected data of 543 cirrhotic patients in Child–Pugh class A submitted to AR (n = 228) versus NAR (n = 315) for early HCC in an Eastern (n = 269) and a Western (n = 274) surgical unit, were reviewed. To control for confounding variable distributions, a 1-to-1 propensity score match was applied to compare AR and NAR outcomes (n = 298). Results. The 5-year recurrence-free and overall survivals of the 543 patients were 32.3% and 60.0%, respectively, without differences between the 2 centers (P = .635 and .479, respectively). AR conferred better overall and recurrence-free survival than NAR (P = .009 and .041, respectively), but NAR patients suffered from significantly worse hepatic dysfunction. After 1-to-1 match, AR (n = 149) and NAR (n = 149) patients had similar covariate distributions. In this matched sample, AR still conferred better recurrence-free survival over NAR (P = .044) but the beneficial effect of AR was limited to the reduction of early recurrence (<2 years) of poorly differentiated tumors and of tumors with microvascular invasion (P < .05), resulting in better overall survival (P = .018). Conclusion. In cirrhotic patients, AR for early HCC can lead to a lower early recurrence rate in tumors with unfavorable tumor features, whereas NAR will not worsen the recurrence rate in well/moderately differentiated tumors or in the absence of microvascular invasion. (Surgery 2014;155:512-21.) From the Liver and Multiorgan Transplant Unit,a S. Orsola Hospital, Alma Mater Studiorum – University of Bologna, Bologna, Italy; and the Eastern Hepatobiliary Surgery Hospital,b Shanghai, China
HEPATOCELLULAR CARCINOMA (HCC) is 1 of the 5 most common malignancies worldwide and the third most common cause of cancer related mortality1; although more common in East Asia, the incidence of HCC is increasing in the Western world.2 Liver resection is widely accepted as a safe treatment with a low operative mortality as the result of advances in surgical techniques and perioperative management.3,4 Unfortunately, the high incidence of recurrence remains the major challenge in obtaining long-term results. Most recurrences occur in the liver as the consequence of subclinical metastases, originating from the primary tumor
Supported in part by a grant from Regione Emilia Romagna. Accepted for publication October 8, 2013. Reprint requests: Alessandro Cucchetti, MD, Policlinico Sant’ Orsola-Malpighi, University of Bologna, Via Massarenti 9, 40138 Bologna, Italy. E-mail: [email protected]. 0039-6060/$ - see front matter Ó 2014 Mosby, Inc. All rights reserved. http://dx.doi.org/10.1016/j.surg.2013.10.009
512 SURGERY
growth through microscopic vascular invasion and peripheral spread, and are considered the most important factors associated with poor prognosis.4-6 On this basis, the systematic removal of a hepatic segment, confined by tumor-bearing portal tributaries, namely anatomic resection (AR), was suggested because it should be more effective for eradication of the intrahepatic metastases of HCC. On the contrary, most surgeons prefer to leave a greater portion of parenchyma of this functional unit, such as in non-AR (NAR), focusing on the preservation of a $1-cm tumor-free margin to reduce postoperative liver failure in patients with cirrhosis. It remains unclear whether AR can really confer a survival advantage over NAR. Some authors have described better long-term outcomes after AR compared with NAR, whereas others have not been able to demonstrate these benefits, as outlined by a recent meta-analysis.7 Discrepancies have probably to be considered a consequence of the intrinsic relatively low level of evidence of available literature represented by observational
Surgery Volume 155, Number 3
retrospective studies only. Meta-regression analysis on this topic suggested that both overall survival and disease-free survival after AR seem to be superior to NAR because the worse liver function reserve in the NAR group significantly affects prognosis.7 In particular, comparative data available from pertinent literature showed that patients submitted to NAR were characterized by a significantly higher prevalence of cirrhosis and more advanced hepatic dysfunction compared with patients in the AR group, and that such differences are able to modify postoperative results.7 Although a randomized, controlled trial (RCT) comparing operative approaches would be ideal, a retrospective analysis using a propensity score matching8 patients groups to reduce bias was used herein to better determine the impact of operative approach on recurrence-free and overall survival for cirrhotic patient with HCC undergoing resection. METHODS Patient selection was accomplished through 3 levels of inclusion criteria. First, all patients submitted to portosystemic shunts before or at the same time as hepatic resection, or treated as an emergency, or submitted to preoperative portal vein embolization were excluded from the analysis. Second, we considered eligible for the present study only those patients in whom no evidence of extrahepatic metastasis was present at the time of surgery, and at pathologic examination did not present tumor invasion into a major branch of the portal or hepatic veins, direct invasion of adjacent organs, or spread to the lymph nodes of the hepatic hilum. In addition, we retained only those cases in which a tumor-free margin of $1 cm was confirmed at pathologic examination; consequently, no tumor enucleations were included in the present study and all resections considered in the present analysis were curative resections at histology. Applying these criteria, between February 2001 and August 2010, 508 cirrhotic patients underwent a first curative resection at the Eastern Hepatobiliary Surgery Hospital of Shanghai, and between January 1997 and November 2011, 388 cirrhotic patients underwent the same procedure at the Department of Surgery and Transplantation of the University of Bologna: The policies of the 2 centers regarding indications for hepatic resection have been published elsewhere.4,9 Diagnosis of cirrhosis was confirmed on histologic specimens. To reduce potential confounding nomenclature of AR and NAR, a third level of inclusion criteria was adopted. Patients
Cucchetti et al 513
with large tumors were excluded from the study that was limited to patients with a single nodule of HCC #5 cm or no more than 3 nodules none >3 cm at histologic examination. In addition, the study was also limited to patients belonging to Child–Pugh class A. The final study population thus consisted of 543 cirrhotic patients, in Child– Pugh class A, resected for early HCC: A total of 269 patients came from the Eastern surgical center and 274 from the Western surgical center. All patients underwent intraoperative hepatic ultrasonography and were deemed to have resectable tumors at the time of surgery. AR was defined as the complete removal of $1 Couinaud segment containing the tumor together with the related portal vein branch and the corresponding hepatic territory. The appropriate segment margins were identified with the discoloration of the parenchyma after ligation of the corresponding arterial and portal venous branches and with intraoperative ultrasonography assistance when necessary. NAR was defined as the resection of the tumor with a margin of $1 cm without regard to segmental, sectional, or lobar anatomy. Pathologic and histologic evaluations of the resected specimens were carried out for all cases. The resected tumor, with its surrounding liver, was examined both microscopically and macroscopically for its histopathologic features. The maximal diameter of the tumor was taken as the tumor size. Curative resection was defined as complete macroscopic and microscopic removal of the tumor. Tumor differentiation and microscopic vascular invasion in the resected tumor were also determined.10-12 After discharge, all patients were observed periodically at follow-up to exclude possible recurrence of HCC: Biochemical liver function tests, serum alpha-fetoprotein level measurement and ultrasonography were performed 3 and 6 months after discharge and then according to an annual or semiannual surveillance program. When any recurrence was suspected, a computed tomography or magnetic resonance imaging was performed for confirmation. Recurrent lesions were managed aggressively by a multimodal approach, which included re-resection, transarterial chemoembolization, percutaneous radiofrequency ablation, and percutaneous ethanol injection. The treatment was decided by the pattern of recurrence, liver functional reserve, and the general condition of the patient at the time of recurrence. For selected patients with transplantable recurrence, salvage liver transplantation was also adopted. Since the end of 2008, Sorafenib (Nexavar; Bayer,
514 Cucchetti et al
Leverkusen, Germany) therapy was also adopted, either alone or in combination with percutaneous approaches for treatment of recurrence.13,14 Follow-up data were collected until December 31, 2011. The study was approved by the appropriate institutional review committees and met the guidelines of their responsible governmental agencies. Statistical analysis. Categorical variables were reported as number of cases and prevalence, and differences between subgroups were compared using the Fisher exact test Chi-square analysis. Continuous variables were initially explored for their normal distribution using the KolmogorovSmirnov test. Because normal distribution could not be confirmed for most variables, all were expressed as median and range and differences between subgroups were explored by the MannWhitney test. All analyses were 2-tailed. Overall survival was computed from the day of operation until the most recent follow-up or until patient death; recurrence-free survival was computed from the day of operation until the most recent followup visit or until clinical evidence of tumor recurrence. Patients transplanted for postoperative liver failure or tumor recurrence were censored on the day before transplantation. In particular, at the end of follow-up, in the whole study group of 543 patients, salvage transplantation was performed in 18 cases (3.3%; 11 of them underwent NAR). Survival estimates were obtained with the KaplanMeier method and compared using the log-rank test. To overcome biases owing to the different distribution of covariates among patients submitted to AR and those submitted to NAR, a 1-to-1 match was created using propensity score analysis: The propensity score represents the probability of each individual patient being assigned to a particular condition in a study given a set of known covariates.8 Propensity scores are used to reduce selection bias by equating groups on the basis of these covariates and are used to adjust for selection bias in observational studies through matching. A multivariate logistic regression was built to predict the probability of each individual patient being submitted to AR or NAR (predictive values) on the basis of covariates that showed a different distribution in the 2 groups and taking into account histologic variables that are known to be able to affect postoperative survival or recurrence. The predictive values were then used to obtain a 1-to1 match by using the nearest neighbor matching method: Matching to 5 decimal points was initially performed, followed by 4-, 3-, 2-, and 1-point matching, and cases whose propensity score
Surgery March 2014 deviated >0.10 were considered unmatched. Thus, patients for whom the propensity score was not matchable were excluded from further analysis. The Cox proportional hazard model was finally applied in the matched sample to identify independent predictors of tumor recurrence and overall survival. Effect size was also calculated for each covariate. Effect size is a measure that is independent of the sample size and can give a more robust estimation of a difference in means or proportions: Values
Cucchetti et al 515
Surgery Volume 155, Number 3
Table I. Baseline characteristics of the cirrhotic patients, belonging to Child–Pugh class A, submitted to surgery for early hepatocellular carcinoma in relationship with anatomic resection (AR) and nonanatomic resection (NAR) Variable Eastern Surgical Centre Western Surgical Centre Age (y) Male gender HCV positive serology HBsAg positive serology Serum albumin (g/dL) Platelet count (3103/mm3) Serum bilirubin (mg/dL) ALT (IU) INR MELD score Histologic solitary tumor Histologic size of largest tumor (cm) Histologic tumor size <2 cm Histologic tumor size >3 cm Tumor grading G3–G4 Presence of MVI
All patients (n = 543) 269 274 58 449 204 291 4.0 118 0.85 53 1.07 8 501 3.0 104 230 366 310
(49.5%) (50.5%) (48–66) (82.7%) (37.6%) (53.6%) (3.6–4.3) (86–162) (0.64–1.16) (33–84) (0.98–1.16) (7–9) (92.3%) (2.0–4.0) (19.2%) (42.4%) (67.4%) (57.1%)
AR (n = 228) 128 100 57 192 73 136 4.1 120 0.84 55 1.05 7 216 3.0 48 105 149 124
(56.1%) (43.9%) (48–65) (84.2%) (32.0%) (59.6%) (3.7–4.3) (87–167) (0.64–1.12) (34–85) (0.90–1.13) (7–8) (94.7%) (2.0–4.0) (21.1%) (46.1%) (65.4%) (54.4%)
NAR (n = 315) 141 174 59 257 131 155 3.9 115 0.85 53 1.09 8 285 3.0 56 125 217 186
(44.8%) (55.2%) (49–67) (81.6%) (41.6%) (49.2%) (3.5–4.2) (84–149) (0.64–1.18) (33–84) (1.00–1.17) (7–9) (90.5%) (2.0–4.0) (17.8%) (39.7%) (68.9%) (59.0%)
P value
Effect size
.009 — .115 .491 .025 .019 .035 .179 .521 .983 .001 .001 .074 .409 .377 .159 .405 .293
0.226 — 0.134 0.059 0.193 0.202 0.182 0.163 0.078 0.009 0.298 0.283 0.154 0.070 0.076 0.121 0.072 0.090
Continuous variables are reported as median values and interquartile ranges (25th and 75th percentiles) because normal distribution cannot be confirmed for most variables. Comparisons were performed with Mann-Whitney U test for continuous variables and Fisher exact test for categorical variables. For continuous variables, the effect size was measured after log10 transformation. The following variables, significantly different between the 2 groups and/or known to be able to affect postoperative survival or recurrence, were selected for propensity score calculation: Age, HCV serology, HBsAg serology, albumin, MELD score (INR was not included because it was already present in the MELD formula), solitary tumor, tumor size, tumor grading and MVI. ALT, Alanine aminotransferase; HBsAg, hepatitis B surface antigen; HCV, hepatitis C virus; INR, International Normalized Ratio; MELD, Model for End-Stage Liver Disease; MVI, microvascular invasion.
belonged to Child–Pugh class A, NAR patients showed lower serum albumin levels (P = .035), higher International Normalized Ratio values (P = .001), and worse Model for End-stage Liver Disease scores (P = .001) than AR patients. Histologic characteristics were similar. The 1-, 3-, and 5-year recurrence-free rates were 83.4%, 57.0%, and 37.2%, respectively, after AR, and 76.8%, 49.2%, and 28.4%, respectively, after NAR (P = .041). The 1-, 3-, and 5-year overall survival rates were 93.7%, 82.2%, and 65.1%, respectively, after AR, and 93.1%, 68.4%, and 56.1%, respectively, after NAR (P = .009). Analysis of the matched sample. After the 1-to-1 propensity score match, 298 patients were selected for comparison: 149 were submitted to AR and an equal number to NAR; details regarding propensity score calculation can be found in Supplementary Table II. As can be observed in Table II, clinical and histologic variables had very similar distributions as confirmed by the effect size of <0.1 in all cases. In this matched sample, the 30- and 90-day postoperative mortality rates were 0.3% and 1.3%, respectively. The 1-, 3-, and 5-year recurrence-free survivals were 78.5%,
52.9%, and 34.6%, respectively, and the corresponding overall survivals were 93.1%, 74.9% and 59.2% (Fig 1). As can be noted, recurrence-free survival changed its slope from the second year onward. In particular, the annual incidence rate of recurrence within 2 years of surgery (early recurrence) was 24.0% (95% confidence interval [CI],19.5–28.6) and the annual incidence of recurrence 2 years after surgery (late recurrence) was 19.6% (95% CI, 15.0–24.1). In this matched sample, patients submitted to AR still showed an overall better recurrence-free survival than NAR patients (P = .044). Results from the Cox regression are reported in Table III and showed that when tumor recurrence was stratified into early and late recurrence, the type of surgery significantly affected early recurrence (P = .015) together with tumor diameter (P = .016) and tumor grading (P = .006), but late recurrence was not significantly affected by AR (ExpB = 1.032; 95% CI, 0.644–1.652; P = .896) or the other available variables (full data not reported). These data were confirmed when patients were stratified on the basis of tumor invasiveness to further investigate how type of surgery could influence early
516 Cucchetti et al
Surgery March 2014
Table II. Baseline characteristics of the matched cohort in relationship with anatomic resection (AR) and nonanatomic resection (NAR) Variable Age, y (range) Male gender HCV-positive serology HBsAg-positive serology Serum albumin (g/dL) Platelet count (3103/mm3) Serum bilirubin, mg/dL (range) ALT, IU/L (range) INR MELD score Histologic solitary tumor Histologic size of largest tumor (cm) Tumor grading G3–G4 Presence of MVI
All patients (n = 298) 56 245 107 169 4.0 124 0.84 55 1.07 8 286 3.0 196 153
(48–66) (82.2%) (35.9%) (56.7%) (3.6–4.3) (91–165) (0.63–1.12) (35–85) (0.97–1.14) (7–9) (96.0%) (2.0–4.1) (65.8%) (51.3%)
AR (n = 149) 57 124 54 84 4.0 124 0.84 58 1.06 8 143 3.0 98 77
(48–66) (83.2%) (36.2%) (56.4%) (3.7–4.3) (91–160) (0.64–1.11) (36–89) (0.97–1.14) (7–9) (96.0%) (2.0–4.0) (65.8%) (51.7%)
NAR (n = 149) 56 121 53 85 4.0 124 0.82 54 1.07 8 143 3.0 98 76
(47–66) (81.3%) (35.6%) (57.0%) (3.6–4.4) (93–169) (0.62–1.15) (34–81) (0.97–1.15) (7–8) (96.0%) (2.0–4.1) (65.8%) (51.0%)
P value
Effect size
.507 .560 .860 .868 .784 .892 .909 .399 .996 .702 1.000 .870 1.000 .903
0.034 0.076 0.016 0.015 0.041 0.074 0.022 0.078 0.033 0.011 0.000 0.085 0.000 0.015
Continuous variables are reported as median values and interquartile range (25th and 75th percentiles). Comparisons were performed with Wilcoxon signed-rank test for continuous variables and McNemar for categorical variables. For continuous variables, effect size was measured after log10 transformation. Effect size < 0.1 indicates very small differences. ALT, Alanine aminotransferase; HBsAg, hepatitis B surface antigen; HCV, hepatitis C virus; INR, International Normalized Ratio; MELD, Model for End-Stage Liver Disease; MVI, microvascular invasion.
Fig 1. Overall (upper plot) and recurrence-free (lower plot) survivals of the matched cohort of 298 cirrhotic patients. Recurrence-free survival changes its slope from year 2 after surgery onward, dividing early from late recurrences. Linear interpolation for early recurrence: Constant = 0.977; b1 = 0.016; r2 = 0.984. For late recurrence: constant = 0.801; b1 = 0.008; r2 = 0.988.
and late recurrence. As can be noted from Figure 2 and Table IV, AR lost its prognostic significance in improving the recurrence-free survival in G1–G2 tumors and in the absence of microvascular invasion (MVI). In these 2 categories neither the overall recurrence rate nor the early recurrence rate were affected by the type of operative procedure (P > .05 in all cases). The benefit obtainable with
AR was narrowed to G3–G4 tumors and in tumors with MVI: In these 2 categories, AR provided better early recurrence rates (P = .003 and .017, respectively) even if late recurrence rates remained substantially unaffected (P > .05 in both cases). Similar findings were observed for tumor size (Fig 2 and Table IV). The 1-, 3-, and 5-year patient survivals after AR were 93.1%, 83.3%, and 65.8%, respectively, and after NAR were 93.1%, 66.8%, and 52.9% (P = .018). In particular, the overall survival was similar, between AR and NAR patients, in G1–G2 tumors (P = .388), in tumors without MVI (P = .197) and in tumors <2 cm (P = .113), whereas AR was related to better overall survival in G3–G4 tumors (P = .014); when MVI was present (P = .016) and in tumor sized $2 cm (P = .043). Cox regression results on overall survival are provided in Supplementary Table III. Regarding data from the 245 patients excluded from the match, 79 patients were submitted to AR and 166 to NAR. Patients submitted to AR had a median Model for End-stage Liver Disease score of 7 (range, 6–8) and a tumor diameter of 3.3 (range, 2.0–4.0) whereas those submitted to NAR had a median Model for End-stage Liver Disease of 8 (range, 7–10) and a tumor diameter of 2.8 (range, 2.0–3.7; P = .001 and .041, respectively), confirming that outliers were excluded from the match. DISCUSSION Surgery is the most important therapeutic approach for patients with HCC, offering a
Cucchetti et al 517
Surgery Volume 155, Number 3
Table III. Uni- and multivariate Cox regression model on early recurrence-free survival in the matched cohort of 298 cirrhotic patients Univariate regression model
Multivariate regression model
Variable
Exp(B)
95% CI
P
Exp(B)
95% CI
P
Western (vs Eastern) Surgical Centre Non-anatomic resection (vs Anatomic) Age (per year increase) Male (vs female) HCV positive (vs negative) HBsAg positive (vs negative) Serum albumin (per g/dL increase) Platelet count (3103/mm3 increase) Serum bilirubin (per mg/dL increase) ALT (per IU/L increase) INR (per unit increase) Solitary tumor (vs multiple) Size of largest tumor (per cm increase) Tumor grading G3–G4 (vs G1–G2) Presence of MVI (vs absent)
1.013 1.629 0.992 0.876 0.900 1.396 1.078 1.001 1.951 0.997 0.961 0.515 1.253 1.875 1.291
0.689–1.490 1.104–2.405 0.975–1.008 0.539–1.426 0.600–1.350 0.939–2.077 0.754–1.541 0.996–1.003 0.898–3.203 0.993–1.002 0.236–3.916 0.225–1.175 1.063–1.478 1.197–2.936 0.880–1.895
.949 .014 .311 .595 .611 .103 .681 .785 .218 .214 .995 .115 .007 .005 .092
1.625 1.229 1.889 1.142
1.100–2.401 1.040–1.451 1.205–2.961 0.826–1.577
.015 .016 .006 .422
The annual incidence rate of early recurrence (<2 years from surgery) was 24.0% (95% CI, 19.5–28.6). P < .10 at univariate Cox regression were entered in the multivariate backward proportional hazard model. ALT, Alanine aminotransferase; HBsAg, hepatitis B surface antigen; HCV, hepatitis C virus; INR, International Normalized Ratio; MELD, Model for End-Stage Liver Disease; MVI, microvascular invasion.
possibility of cure. Unfortunately, the long-term prognosis remains undermined by the high recurrence rate that is among the most important factors affecting the survival of HCC patients.4-6 The main problem that surgeons have to face while operating on cirrhotic patients is the balance between achieving a radical intervention and, at the same time, preventing the development of postoperative liver failure that could ensue from removal of too much liver parenchyma. This problem is at the basis of the dispute between AR, that should be theoretically more radical from an oncologic point of view, and NAR, which should reduce the risk of postoperative hepatic failure. In the absence of RCTs, results reported in the literature are still conflicting.7 In particular, different proportions of cirrhotic patients, analyzed together with noncirrhotic patients in previous studies, bias the differences observed between AR and NAR outcomes, because AR was mostly performed in noncirrhotic subjects that already have a favorable clinical outcome. The present study shows that even when restrictive criteria were applied to identify a homogeneous study population, namely cirrhotic patients only, belonging to Child–Pugh class A, differences between patients submitted to AR or NAR still exist, because patients submitted to NAR still show more advanced hepatic dysfunction. Thus, the present initial analysis required an in-depth evaluation of outcomes related to the 2 different operative procedures, in which the
potential confounding impact of covariates, typical of previous observational studies, was appropriately dealt with by means of a meticulous statistical approach. When patients were adequately matched through propensity score, the results from the present study show that AR might be considered the preferred operative procedure for early HCC because of the lower recurrence rate and, consequently, greater patient survival. Nevertheless, this type of resection did not determine an absolute advantage compared with NAR because the benefit seems to be limited to the reduction of early recurrences and depends on tumor features. Early recurrences are defined as subclinical metastases originating from the primary tumor and missed during staging and treatment; thus, the present finding that AR could reduce such an inauspicious event well supports its oncologic rationale. However, there is a peculiar finding of the present analysis that deserves particular attention. AR proved superior to NAR, in determining early recurrence-free survival, only in early HCC with unfavorable tumor features such as poorly or undifferentiated grade or in the presence of microscopic vascular invasion. In other words, the present results suggest that the higher the invasiveness of the tumor, the greater the need for the removal of the entire liver functional unit. This is an important aspect that acquires particular clinical interest since tumor grade can be assessed
518 Cucchetti et al
Surgery March 2014
Fig 2. Recurrence-free survival in relationship with type of surgery and histologic findings. Number of patients at risk, and distinction between early and late recurrence, are reported in Table IV. (A) G1–G2 tumor. (B) G3–G4 tumor. (C) Microscopic vascular invasion absent. (D) Microscopic vascular invasion present. (E) Tumor size <2 cm. (F) Tumor size $2 cm.
Cucchetti et al 519
Surgery Volume 155, Number 3
Table IV. Overall, early, and late recurrence-free survival in relationship with anatomic resection (AR), nonanatomic resection (NAR), and tumor histologic features in the matched population of 298 patients Survival (y)
At risk
All patients (n = 298) 1 2 3 5 G1–G2 tumors (n = 102) 1 2 3 5 G3–G4 tumors (n = 196) 1 2 3 5 MVI absent (n = 145) 1 2 3 5 MVI present (n = 153) 1 2 3 5 Size <2 cm (n = 56) 1 2 3 5 Size $2 cm (n = 242) 1 2 3 5
149 113 79 61 25 51 36 29 25 11 98 77 51 35 18 72 56 40 31 11 77 60 40 30 12 32 28 21 15 3 117 85 58 45 19
AR, % (95% CI) 84.5 68.7 57.7 37.2
(77.4–89.5) (59.9–75.8) (48.4–65.9) (27.6–46.7)
81.6 74.5 69.1 45.5
(67.5–90.0) (64.0–82.3) (53.2–80.6) (34.0–56.3)
86.1 65.7 50.3 33.5
(77.3–91.7) (49.9–77.6) (33.5–64.9) (22.6–44.8)
82.7 67.9 58.3 38.5
(71.6–89.8) (55.0–77.9) (44.7–69.6) (23.9–52.9)
84.8 64.7 55.2 32.2
(74.2–91.3) (51.7–74.3) (42.0–66.5) (20.7–43.8)
90.2 72.0 68.0 55.8
(72.5–96.7) (51.5–84.9) (47.2–82.0) (32.7–73.7)
81.9 66.4 53.5 34.2
(73.3–87.9) (56.4–74.7) (42.9–63.0) (24.3–44.4)
At risk 149 105 72 49 20 51 42 33 25 11 98 63 39 26 11 73 56 42 29 10 76 49 33 22 13 24 18 13 6 2 125 87 58 43 17
NAR, % (95% CI)
P value .044
72.7 55.3 48.2 32.2
(64.7–79.3) (46.7–63.0) (39.5–56.4) (23.2–41.5)
83.9 73.4 68.5 46.1
(70.4–91.6) (58.6–83.6) (53.0–79.7) (29.1–61.6)
65.9 45.8 37.6 25.2
(55.5 –74.4) (35.4–55.6) (27.4–47.7) (15.3–36.3)
81.9 65.8 55.4 35.2
(70.8–89.1) (53.4–75.6) (42.5–66.5) (21.8–48.9)
65.4 51.4 41.3 31.9
(53.5–75.0) (38.3–62.9) (29.5–52.6) (20.3–44.1)
78.3 64.9 56.8 47.3
(55.4–90.3) (41.9–80.7) (31.7–75.7) (21.6–69.4)
71.7 53.5 48.6 31.2
(62.9–78.8) (44.2–62.0) (39.2–57.3) (21.8–31.0)
.013 .896 .627 .974 .527 .020 .003 .845 .178 .275 .427 .120 .017 .586 .270 .283 .733 .258 .034 .296
For each subgroup, P values are reported for the entire time interval, for early recurrence (<2 years) and for late recurrence assuming 2 years as the starting point. To obtain late recurrence-free survival (RFS) rates from actuarial data, survival can be calculated as follows: 3-year late RFS = 5-year RFS/2-year RFS. CI, Confidence interval; MVI, microvascular invasion.
preoperatively with reasonable accuracy.16-23 Tumor grade can be evaluated before surgery with tumor biopsy17,18 and there is recent evidence that diffusion-weighted magnetic resonance imaging can identify degrees of tumor differentiation.21-23 In the presence of a preoperative diagnosis of poor differentiation, 1 suggestion could be to study candidates for hepatic resection to plan an AR. On the other hand, in patients without these features, a NAR could be performed without affecting recurrence-free survival. This latter aspect is of particular importance in those patients in whom the future remnant liver volume will be inadequate if an AR will be pursued. These
considerations must be made with caution, because the risks of a liver biopsy can overshadow potential benefits. In particular, the risk of seeding is reported to range around 2–3%, when not associated with percutaneous therapies, but some Authors reported seeding #10%.24 It is clear that liver biopsy cannot be suggested as a routine preoperative tool. In addition, data on the accuracy of modern imaging techniques in predicting tumor differentiation are still scarce. Consequently, preoperative information regarding tumor invasiveness should probably still be based on tumor size. In the present study, and in the largest published experience from Eguchi et al,25 the benefit
520 Cucchetti et al
from AR over NAR, in terms of recurrence-free and disease-free survivals, was not observed in HCC <2 cm. Thus, it can be suggested that when an anatomic approach cannot be pursued owing to inadequate remnant liver volume, NAR for small HCCs will not affect tumor recurrence. That is because tumor size is known to be strictly related to tumor differentiation and presence of MVI.16,19,20 We are aware that the present study is not an RCT, and the present statistical approach is not aimed at replacing one; however, we believe that these results acquire particular strength for 3 main reasons. First, matched patients had similar, or even identical, clinical and tumor features known to influence prognosis, the match was achieved with a statistical approach and not made a priori on the basis of preselected covariates. This approach excluded outlier patients that are more prone to develop postoperative liver failure or, conversely, bring a null risk for postoperative complications. In these patients, the uncertainty of the best operative approach is absent, in particular for patients at high risk for postoperative liver failure, where NAR will be the only approach that can be attempted. Second, the operative procedures were not selected by a predetermined algorithm, but were left to the surgeon’s choice because the superiority of one or the other type of resection was unclear at the time of the study. This aspect could apparently represent a possible bias right at the start, and it should be observed that other features may have affected the operative choice, such as the attempt to obtain greater tumor-free margins, presence of comorbidities, or transplant eligibility. However, the proportion of patients that were suitable for matching was quite large (298/543), suggesting that surgeons preferred AR or NAR approach independent of clinical circumstances and tumor characteristics, providing comparable cases to be analyzed. This second aspect is further highlighted by the fact that the study population derived from 2 surgical centers that provide a good representation of both Eastern and Western HCC epidemiology and surgical perspectives. Third, to the best of our knowledge, this is the largest and most welldefined study conducted on exclusively cirrhotic patients; in fact, most of previous literature results derived from mixed cohorts of cirrhotic and noncirrhotic patients.7 This feature assumes particular importance because cirrhosis represents the principal clinical impasse in the selection of the operative procedure to adopt. Nevertheless, an RCT conducted to clarify the debate is warranted, but
Surgery March 2014
unfortunately, to the best of our knowledge, only 1 RCT has been designed and at present is not yet open for participant recruitment.26 In conclusion, the present analysis suggests that anatomic resection of early HCC can reduce the early recurrence-rate after hepatic resection, and this is true for patients having poorly differentiated HCC or MVI. Nevertheless, AR cannot be applied to all patients with cirrhosis because of the risk of postoperative insufficiency; in this regard, NAR can provide similar results to that of AR in the absence of unfavorable tumor features or for small tumors. A. Cucchetti and G-L. Qiao equally contributed to the present work. The authors thank Ms. Susan West for writing assistance. SUPPLEMENTARY DATA Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10. 1016/j.surg.2013.10.009.
REFERENCES 1. El-Serag HB. Hepatocellular carcinoma: an epidemiologic view. J Clin Gastroenterol 2002;35:S72-8. 2. Parkin D. Global cancer statistics in the year 2000. Lancet Oncol 2001;2:533-43. 3. Fan ST, Lo CM, Liu CL, et al. Hepatectomy for hepatocellular carcinoma: toward zero hospital deaths. Ann Surg 1999; 229:322-30. 4. Cucchetti A, Zanello M, Cescon M, Ercolani G, Del Gaudio M, Ravaioli M, et al. Improved diagnostic imaging and interventional therapies prolong survival after resection for hepatocellular carcinoma in cirrhosis: the university of bologna experience over 10 years. Ann Surg Oncol 2011; 18:1630-7. 5. Park JH, Koh KC, Choi MS, Lee JH, Yoo BC, Paik SW, et al. Analysis of risk factors associated with early multinodular recurrences after hepatic resection for hepatocellular carcinoma. Am J Surg 2006;192:29-33. 6. Kamiyama T, Nakanishi K, Yokoo H, Kamachi H, Tahara M, Kakisaka T, et al. Analysis of the risk factors for early death due to disease recurrence or progression within 1 year after hepatectomy in patients with hepatocellular carcinoma. World J Surg Oncol 2012;10:107. 7. Cucchetti A, Cescon M, Ercolani G, Bigonzi E, Torzilli G, Pinna AD. A comprehensive meta-regression analysis on outcome of anatomic resection versus non-anatomic resection for hepatocellular carcinoma. Ann Surg Oncol 2012; 19:3697-705. 8. Austin PC. A critical appraisal of propensity-score matching in the medical literature between 1996 and 2003. Stat Med 2008;27:2037-49. 9. Huang G, Lai EC, Lau WY, Zhou WP, Shen F, Pan ZY, et al. Posthepatectomy HBV reactivation in hepatitis B-related hepatocellular carcinoma influences postoperative survival in patients with preoperative low HBV-DNA levels. Ann Surg 2013;257:490-505. 10. Edmondson HA, Steiner PE. Primary carcinoma of the liver: a study of 100 cases among 48,900 necropsies. Cancer 1954; 7:462-503.
Surgery Volume 155, Number 3
11. Nzeako UC, Goodman ZD, Ishak KG. Comparison of tumor pathology with duration of survival of North American patients with hepatocellular carcinoma. Cancer 1995;76:579-88. 12. Vauthey JN, Lauwers GY, Esnaola NF, Do KA, Belghiti J, Mirza N, et al. Simplified staging for hepatocellular carcinoma. J Clin Oncol 2002;20:1527-36. 13. Cheng AL, Kang YK, Chen Z, Tsao CJ, Qin S, Kim JS, et al. Efficacy and safety of sorafenib in patients in the AsiaPacific region with advanced hepatocellular carcinoma: a phase III randomised, double-blind, placebo-controlled trial. Lancet Oncol 2009;10:25-34. 14. Llovet JM, Ricci S, Mazzaferro V, Hilgard P, Gane E, Blanc JF, et al. Sorafenib in advanced hepatocellular carcinoma. N Engl J Med 2008;359:378-90. 15. Normand ST, Landrum MB, Guadagnoli E, Ayanian JZ, Ryan TJ, Cleary PD, et al. Validating recommendations for coronary angiography following acute myocardial infarction in the elderly: a matched analysis using propensity scores. J Clin Epidemiol 2001;54:387-98. 16. Cucchetti A, Piscaglia F, Grigioni AD, Ravaioli M, Cescon M, Zanello M, et al. Preoperative prediction of hepatocellular carcinoma tumour grade and micro-vascular invasion by means of artificial neural network: a pilot study. J Hepatol 2010;52:880-8. 17. Colecchia A, Scaioli E, Montrone L, Vestito A, Di Biase AR, Pieri M, et al. Pre-operative liver biopsy in cirrhotic patients with early hepatocellular carcinoma represents a safe and accurate diagnostic tool for tumour grading assessment. J Hepatol 2011;54:300-5. 18. Pawlik TM, Gleisner AL, Anders RA, Assumpcao L, Maley W, Choti MA. Preoperative assessment of hepatocellular carcinoma tumour grade using needle biopsy: implications for transplant eligibility. Ann Surg 2007;245:435-42. 19. Kim BK, Han KH, Park YN, Park MS, Kim KS, Choi JS, et al. Prediction of microvascular invasion before curative resection of hepatocellular carcinoma. J Surg Oncol 2008;97:246-52.
Cucchetti et al 521
20. Pawlik TM, Delman KA, Vauthey JN, Nagorney DM, Ng IO, Ikai I, et al. Tumour size predicts vascular invasion and histologic grade: implications for selection of surgical treatment for hepatocellular carcinoma. Liver Transpl 2005;11: 1086-92. 21. Nasu K, Kuroki Y, Tsukamoto T, Nakajima H, Mori K, Minami M. Diffusion-weighted imaging of surgically resected hepatocellular carcinoma: imaging characteristics and relationship among signal intensity, apparent diffusion coefficient, and histopathologic grade. AJR Am J Roentgenol 2009;193:438-44. 22. Saito K, Moriyasu F, Sugimoto K, Nishio R, Saguchi T, Akata S, et al. Histological grade of differentiation of hepatocellular carcinoma: comparison of the efficacy of diffusion-weighted MRI with T2-weighted imaging and angiography-assisted CT. J Med Imaging Radiat Oncol 2012;56:261-9. 23. Witjes CD, Willemssen FE, Verheij J, van der Veer SJ, Hansen BE, Verhoef C, et al. Histological differentiation grade and microvascular invasion of hepatocellular carcinoma predicted by dynamic contrast-enhanced MRI. J Magn Reson Imaging 2012;36:641-7. 24. Stigliano R, Marelli L, Yu D, Davies N, Patch D, Burroughs AK. Seeding following percutaneous diagnostic and therapeutic approaches for hepatocellular carcinoma. What is the risk and the outcome? Seeding risk for percutaneous approach of HCC. Cancer Treat Rev 2007;33:437-47. 25. Eguchi S, Kanematsu T, Arii S, Okazaki M, Okita K, Omata M, et al. Comparison of the outcomes between an anatomical subsegmentectomy and a non-anatomical minor hepatectomy for single hepatocellular carcinomas based on a Japanese nationwide survey. Surgery 2008;143:469-75. 26. Prognostic impact of anatomical resection vs. nonanatomical resection for HCC (ARversusNAR) [cited 2013 May 31]. Available from: http://clinicaltrials.gov/ct2/ show/NCT01236989.