Potential conflict of interest: Dr. Schwartz advises Bayer
Asian series have shown a 5-year survival rate of ≈70% after resection of hepatocellular carcinoma (HCC) ≤2 cm. Western outcomes with resection have not been as good. In addition, ablation of HCC ≤2 cm has been shown to achieve competitive results, leaving the role of resection in these patients unclear. Records of patients undergoing resection at two Western centers between January 1990 and December 2009 were reviewed. Patients with a single HCC ≤2 cm on pathologic analysis were included. Thirty clinical variables including demographics, liver function, tumor characteristics, nature of the surgery, and the surrounding liver were examined. An exploratory statistical analysis was conducted to determine variables associated with recurrence and survival. The study included 132 patients with a median follow-up of 37.5 months. There was one (<1%) 90-day mortality. There were 32 deaths with a median survival of 74.5 months and a 5-year survival rate of 70% (63% in patients with cirrhosis). The median time to recurrence was 31.6 months and the 5-year recurrence rate was 68%. Presence of satellites (hazard ratio [HR], 2.46; P = 0.031) and platelet count <150,000/μL (HR, 2.37; P = 0.026) were independently associated with survival. Presence of satellites (HR, 2.79; P = 0.003), cirrhosis (HR, 2.3; P = 0.010), and nonanatomic resection (HR, 1.79; P = 0.031) were independently associated with recurrence. Patients with a single HCC ≤2 cm and platelet count ≥150,000/μL achieved a median survival of 138 months and a 5-year survival rate of 81%, respectively. Conclusion: Resection of HCC ≤2 cm is safe and achieves excellent results in Western centers. Recurrence continues to be a significant problem. Presence of satellites, platelet count, anatomic resection, and cirrhosis are associated with outcomes after resection, even among such early tumors. Resection should continue to be considered a primary treatment modality in patients with small HCC and well-preserved liver function. (HEPATOLOGY 2013)
Hepatocellular carcinoma (HCC) ≤2 cm is regarded as a separate and distinct clinical subgroup by both Eastern and Western experts.1, 2 Detection of tumors at such an early stage has traditionally been rare in the West and as a result, clinicians have had to rely on data almost exclusively from the East. However, as a result of the increased awareness of the need for screening in patients with liver disease and validated criteria for accurate noninvasive diagnosis of such small tumors, the number of HCCs being detected at an early stage will likely increase in North America and Europe.3-5
Patients with such early HCC have a good likelihood of cure with resection, transplantation, or ablation.6-11 Although there have been a significant number of recent publications on the indications and outcomes of both transplantation and ablation in the treatment of early HCC, the literature on which the recommendations regarding the role of surgical resection are based is more dated.
A review of the data collected by the Liver Cancer Study Group of Japan demonstrated a 5-year survival rate of 71% for the 1,318 patients with a single HCC ≤2 cm undergoing surgical resection.12 In contrast, examination of the Surveillance, Epidemiology, and End Results Program database identified only 154 patients with HCC ≤2 cm undergoing resection in the United States over an 8-year period with a 5-year survival rate of only 49%.13 Such differing results leave the role of surgical resection for such early tumors unclear. In addition, such poor results reported by Western series, as well as the lack of well-defined criteria for resection, have led some authors to suggest that radiofrequency ablation may be the treatment of choice for patients with HCC ≤2 cm even when surgical resection is possible.10, 14
The data presented in this study detail the results from two Western centers performing a large volume of HCC resections. It represents the largest Western series to examine the outcomes of patients undergoing resection of a single HCC ≤2 cm. We also provide the results of our exploratory analyses to determine the clinical variables associated with survival and recurrence. These results will hopefully provide new and valuable insight into the role of resection for such early tumors.
Prospectively collected data from two Western centers (Mount Sinai Medical Center, New York, NY, and National Cancer Institute-Istituto Nazionale Tumori, Milan, Italy) were reviewed retrospectively. Patients undergoing resection of a single pathologically proven HCC ≤2 cm in size between 1990 and 2009 were chosen. Approval for conducting the study was obtained from the institutional review board at both centers.
A total of 30 variables including age, sex, and underlying liver disease were recorded. The anatomic location and distance of the recurrent tumor from the cut surface of the liver after resection was recorded by reviewing postoperative imaging studies. We also recorded preoperative platelet count, albumin level, total bilirubin, creatinine, international normalized ratio, Model for End-Stage Liver Disease score, alpha-fetoprotein level, type of resection performed, need for transfusion, tumor size, and number of tumors. Pathological slides were systematically reviewed for tumor differentiation, extent of vascular invasion, and degree of fibrosis in the surrounding liver using the Metavir staging system. Anatomic resection was defined as removal of the entire Couinaud segments involved with the tumor. Patients without satellites and without vascular invasion were classified “very early” HCC: Barcelona Clinic Liver Cancer (BCLC) stage 0 or Japanese T1 based on pathology.1, 2
Criteria for Diagnosis and Resection and Follow-up Protocol.
Criteria for diagnosis and resection have evolved over time but have been similar in the two centers, with only slight differences regarding preoperative diagnosis. Currently in New York, at least two contrast-enhanced imaging studies (computed tomography [CT] and magnetic resonance imaging [MRI]) are required. The tumor has to display arterial enhancement and venous washout. In Milan, the diagnosis of HCC was made on sequential contrast-enhanced imaging studies (CT, MRI, or ultrasound) unless one study conclusively demonstrated HCC with arterial enhancement and venous washout. Prior to the establishment of the criteria for noninvasive diagnosis of HCC at the European Association of the Study of the Liver in 2000, the diagnosis was established by biopsy. Also, in cases without conclusive radiological diagnosis, ultrasound-guided biopsy was used at both centers. Fourteen (11%) patients, 11 prior to 2000 and three after 2000, required biopsy to confirm the diagnosis of HCC.
Only patients with a single tumor and no evidence of extrahepatic spread were included in the study. In New York only patients with Child's A liver disease or no cirrhosis were considered for resection. After 2002, patients with clinical portal hypertension or platelet count <100,000 /μL were excluded from resection. In Milan, platelet count was not the sole criteria defining portal hypertension: patients with Child-Pugh class A liver disease without esophageal varices (≤F1 grade) and with indocyanine green retention <20% at 15 minutes were allowed resection up to two segments even if they had platelet count <100,000/μL.
The patients were followed with either contrast-enhanced CT or MRI scans of the abdomen as well as blood work including alpha-fetoprotein. Contrast-enhanced ultrasound was also used for surveillance in Milan. Non–contrast-enhanced CT of the chest was used to detect lung recurrence irrespective of the modality used to screen for abdominal recurrence. The follow-up schedule consisted of scans every 3 (New York) or 4 (Milan) months for the first year, every 4 months for the second year, and subsequently every 6 months. No adjuvant therapy was used.
Treatment of Recurrence.
“Very early” recurrence was defined as recurrence within the first year after surgery based on previously published data showing this to be a clinically significant cutoff.15 Data on the more conventional cutoff at 2 years for “early” recurrence is also provided. Solitary recurrences were treated with resection. Patients with a solitary liver recurrence (New York) or multiple tumors within Milan criteria (Milan) and Child-Pugh class A liver disease and no evidence of portal hypertension underwent a second hepatic resection. Patients with multiple intrahepatic recurrences or compromised hepatic function were treated with radiofrequency ablation and/or transarterial chemoembolization. Patients with recurrence confined to the liver and without significant comorbidities were also referred for liver transplantation. Patients undergoing liver transplantation were censored at the time of transplantation for the purposes of this study. After 2008, patients not eligible for repeat resection, liver transplantation, or local-regional therapies were treated with sorafenib.
The primary endpoint analyzed was survival. Secondary endpoints included overall, very early (<1 year), and early (<2 year) recurrence. Exploratory analyses were conducted to determine factors associated with survival and time to recurrence. Subgroups analyzed included patients with cirrhosis, pathologically very early tumors (BCLC stage 0/Japanese T1), satellites, and surgery based on the anatomical resection of all involved segments.
The primary endpoints of survival and time to recurrence were calculated using the Kaplan-Meier method. An exploratory analysis was conducted to determine the variables associated with survival and recurrence. Univariate associations between clinical variables and survival as well as time to recurrence were conducted using the log-rank test. All variables found to be significant on univariate analysis (P < 0.05) were entered into a step-down Cox proportional hazard regression analysis.
Categorical data were compared using the chi-square or Fisher's exact test as indicated. Continuous variables were compared using the Student t test or Mann-Whitney test for variables with an abnormal distribution. Receiver operating characteristic (ROC) curve analysis was used to determine the optimal cutoffs of continuous variables by choosing the point along the curve that maximized the sum of sensitivity and specificity. Platelet count was entered as a continuous variable into a Cox model after checking the necessary assumptions.16 We also drew plots of the hazard function to describe the instantaneous rate of death and disease recurrence over the follow-up period. These plots were obtained using the Epanechnikov method.17 SPSS version 16.0 (SPSS, Inc, Chicago, IL) for Windows was used.
During the study period, over 2,000 hepatic resections were performed for HCC at the two centers. Out of this large group, 132 patients with pathologically proven single HCC ≤2 cm (New York, 57; Milan, 75) were identified. There were no instances when a patient was explored for HCC ≤2 cm without cancer being found in the specimen at either center. During the same period, 79 patients (New York, 36; Milan, 43) with HCC ≤2 cm and Child-Pugh class A liver disease underwent radiofrequency ablation (RFA) at the two centers. These patients underwent RFA either as a bridge to liver transplantation because of the presence of significant portal hypertension or as definitive cancer therapy because of the presence of significant comorbidities precluding safe resection. Patient demographics, tumor characteristics, and details of the surgery are summarized in Table 1. All of the patients in the study were Child-Pugh class A without history of decompensation. The median follow-up was 37.5 months. At the time of data collection, there had been 32 deaths, including one (0.7%) perioperative death within 90 days of surgery. The median survival for the entire cohort was 74.5 months, with a 5-year survival rate of 70% (Fig. 1A). ROC curve analysis revealed an optimal cutoff of 148,000/μL for platelet count and 1.1 for international normalized ratio in terms of predicting survival. Variables significantly associated with survival on univariate and multivariate analyses are listed in Tables 2 and 3, respectively. The two variables independently associated with survival for the entire cohort included presence of satellites (hazard ratio [HR], 2.46; P = 0.031) and platelet count <150,000/μL (HR, 2.37; P = 0.026).
Table 1. Patient Demographics, Pathological Characteristics, and Surgical Details
Values are expressed as the mean ± SD or no. (%) unless noted otherwise.
Abbreviations: AFP, alpha-fetoprotein; HBV, hepatitis B virus; HCV, hepatitis C virus; INR, international normalized ratio.
Both the conventional platelet cutoff of 100,000/μL as well as that identified by ROC curve analysis (150,000/μL) were significantly associated with survival on univariate analysis (Table 2 and Fig. 2A). In addition, platelet count used as a continuous variable was also significantly associated with survival at 5 years (regression coefficient, −0.00764 ± 0.00373; P = 0.0404) (Fig. 1D). Other relevant clinical variables that did not reach statistical significance on univariate analysis for survival are listed in Supporting Table 1.
At the time of data collection, there had been 67 (50.7%) recurrences. The median time to recurrence was 31.6 months with a 5-year recurrence rate of 68% (Fig. 1B). Approximately 80% of the deaths were preceded by recurrence. The instantaneous risk of death and cancer recurrence over time are demonstrated in Fig. 1C. Variables significantly associated with recurrence on univariate and multivariate analyses are outlined in Tables 2 and 3. The variables significantly associated with time to recurrence for the entire cohort included presence of satellites (HR, 2.79; P = 0.003), cirrhosis (HR, 2.3; P = 0.010), and nonanatomic resection (HR, 1.79; P = 0.031). Other relevant clinical variables that did not reach statistical significance on univariate analysis for recurrence are listed in Supporting Table 1.
At 1 year, there had been 20 instances of “very early” recurrence with a rate of 17%. At 2 years there had been 38 recurrences resulting in a rate of 29%. Variables significantly associated with recurrence at 1 and 2 years on univariate and multivariate analysis are listed in Tables 3 and 4.
Table 4. Results of Univariate Analysis of Very Early (<1 Year) and Early (<2 Year) Recurrence
Recurrence at 1 year
Very early (BCLC 0/Japanese T1)
Other (BCLC A/Japanese T2)
Recurrence at 2 years
Moderate or well
Cirrhosis (fibrosis 4)
All but one of the 67 patients with recurrent tumor underwent treatment, either with a single modality or a combination of therapies. Treatments included transarterial chemoembolization (n = 38), percutaneous ethanol ablation (n = 3), Yttrium90 radioembolization (n = 1), liver transplantation (n = 6), repeated hepatic resection (n = 21), RFA (n = 17), sorafenib (n = 2), and resection of extrahepatic tumor (lung = 1, omentum = 1).
Several other findings deserve specific mention. Etiology of underlying liver disease did not correlate with survival or recurrence. Laparoscopic resection was performed in 15 (11%) cases and also did not alter these endpoints. There was no relationship between the location of the tumor and the type of resection (anatomic versus nonanatomic). However, we did find a correlation between nonanatomic resection and platelet count <100,000/μL and/or bilirubin >1 mg/dL, probably in an attempt to preserve functioning parenchyma.
Of the two variables found to be significantly associated with survival on multivariate analysis, only platelet count is available preoperatively to help guide patient selection. If resection was limited only to the 66 patients with platelet count ≥150,000/μL, as determined by ROC curve analysis, the median survival increased to 138 months, and the 5-year survival rate increased to 81% (Fig. 2A).
Likewise, the only variable significantly associated with recurrence that can be controlled by the clinician is the type of resection that is performed. Performing anatomic resection was associated with a 20% decrease in the recurrence rate from 80% down to 60% at 5 years (Fig. 2B).
Patients with Cirrhosis.
Patients with Metavir stage 4 fibrosis were chosen in order to analyze results in the cirrhosis subgroup (n = 89, 67% of the overall series). The median survival and 5-year survival rate in this group were 67.1 months and 63%, respectively. There was no significant difference in survival between the cirrhosis and no cirrhosis groups (Fig. 2C). Recurrence rates at 1 and 5 years were 20% and 75%, respectively, for the cirrhosis subgroup. The overall rate of recurrence was significantly higher in the cirrhosis subgroup compared with the no cirrhosis subgroup (Table 2).
The only variable associated with survival on univariate analysis in the cirrhosis population was platelet count. Both a cutoff of 100,000/μL (P = 0.046) and a cutoff of 150,000/μL (P = 0.039) were significantly associated with survival (Table 5). The only variables significantly associated with time to recurrence on univariate analysis among these patients with cirrhosis were performing a nonanatomic resection (P = 0.017) and the presence of satellites (P = 0.035) (Table 5). Multivariate analysis was not conducted in this subgroup.
Patients with no vascular invasion and no satellites (BCLC 0/Japanese T1) on pathology were selected as “true” cases of very early HCC (n = 85). These patients had median and 5-year survivals of 138 months and 76% compared with 65.1 months and 57% (P = 0.137) for those with vascular invasion and/or satellites. Recurrence rates at 1 and 5 years were 12% and 61%, respectively, for this subgroup (Fig. 2D). Recurrence at 1 year was significantly lower for patients with very early tumors and the difference was just at the cutoff for significance for overall recurrence.
The only variable significantly associated with survival on univariate analysis in this subgroup of patients was platelet count <150,000/μL (P = 0.011) and the only variable associated with recurrence was cirrhosis (stage 4 fibrosis) (P = 0.012) (Table 5). Again, multivariate analyses were not performed. Performing an anatomic resection in these patients with no vascular invasion and no satellites did not result in lower early or overall recurrence.
However, for the remaining 47 patients with either vascular invasion or satellites, performing an anatomic resection was associated with a significant reduction in recurrence at 1 year from 50% down to 11% (P = 0.008). Although there was a clear trend toward better overall survival as well as lower overall recurrence with anatomic resection in these 47 patients with either vascular invasion or satellites, the P values did not reach significance.
There were 16 (12%) patients who were found to have satellites on pathology. The presence of satellites was not recognized preoperatively in any of the cases. As demonstrated in Tables 2 and 3, the presence of satellites was an independent predictor of survival, overall recurrence, and early recurrence at 1 year.
By coincidence, half (n = 8) of the patients with satellites underwent anatomic liver resections, whereas the other half did not. Despite the very small sample size, anatomic resection was associated with significantly better survival, lower overall recurrence, and lower early recurrence at 1 year in these patients (Figs. 2E,F). There were no other factors associated with mortality or recurrence in this subgroup of patients even on univariate analysis.
The data from this study confirm that performing hepatic resection for HCC ≤2 cm is safe with the occurrence of only one (0.8%) perioperative mortality. In addition, it demonstrates that the long-term results are excellent, with median and 5-year survivals of 75 months and 70%, respectively. These results from two high-volume Western centers are much more compatible with those reported by the larger Japanese series showing 5-year survivals near 70%.
The presence of satellites and platelet count, with an optimal cutoff of 150,000/μL were the only the only variables independently associated with survival for the overall cohort on our exploratory analyses. Unfortunately, we were not able to detect the presence of satellites on imaging in any of the 16 cases, making it impossible to use this variable preoperatively to select patients for resection.
Portal hypertension has been shown to have a significant impact on survival after hepatic resection for HCC,18 hence it is not surprising that when resection was limited to patients with platelet count ≥150,000/μL, survival improved significantly. The median survival in these patients without significant portal hypertension, as measured by platelet count, was 138 months, with a 5-year survival rate of 81%. Even patients with established cirrhosis and a platelet count ≥150,000/μL achieved a 5-year survival rate of 74%. These outcomes certainly compare very favorably with the 68% survival at 5 years reported for “resectable” patients undergoing RFA of HCC ≤2 cm.10 The inclusion of patients with platelet counts as low as 40,000/μL in the randomized study by Chen et al.19 may be an explanation as to why no difference in survival was detected when comparing RFA with surgical resection for patients with HCC <5 cm.
An interesting finding was that resection of patients with platelet count <150,000/μL or even <100,000/μL was not associated with an increased early perioperative mortality as we had expected. It seems that, in this particular scenario with small tumors, the influence of portal hypertension becomes evident only late after hepatic resection. Finally, we discovered a near linear relationship between platelet count and 5-year survival. Although we identified a platelet count of 150,000/μL as the optimal cutoff in this cohort, there was no point along the curve in Fig. 1D below which the survival at 5 years dropped precipitously. It would appear that incremental decreases in platelet count at the time of surgery will result in incremental decreases in long-term survival.
Eastern reports have shown that even for tumors ≤2 cm, ≈10% of cases will have microvascular invasion of portal branches by tumor, and 3% will have satellite tumors.20-23 Pathological examination from our Western patients with HCC ≤2 cm revealed a more aggressive picture, with 27% of patients having microvascular portal invasion and, very surprisingly, 2% with gross invasion. In addition, we found satellite lesions in 12% of our patients. Overall, 47 (36%) of patients had either vascular invasion or satellite tumors and did not meet the pathological criteria for very early HCC defined as T1 by the Japanese Society of Hepatology or as BCLC stage 0.
The overall recurrence rate of 68% at 5 years and the 1-year recurrence rate of 17% seemed, at first, surprisingly high to us for such small cancers. A recurrence rate of 61% at 5 years for small tumors without vascular invasion or satellites was particularly unexpected. However, a Japanese study of 70 patients with HCC ≤2 cm undergoing resection found an overall recurrence rate of 88% for the entire cohort.24 The same study demonstrated 1- and 5-year recurrence rates of 8% and 53%, respectively, for patients found to have T1 tumors.24 These numbers are very similar to what we have reported (12% at 1 year and 61% at 5 years) for our patients with pathologically proven very early tumors. Again, the recurrence rates for the entire cohort from our study (17% at 1 year and 68% at 5 years) compare favorably with the 1- and 5-year recurrence rates of 34% and 80%, respectively, reported for RFA of similarly sized HCC.10
The vast majority of the recurrences occurred within the first 3 years after surgery after which there were very few events. The pattern of the instantaneous risk of recurrence for these small tumors was also very different from that published for more advanced tumors.25, 26 Instead of the two peaks generally seen for larger tumors—one at approximately 12 months representing early metastatic recurrence and another at approximately 36 months representing late de novo recurrence—we see only a single and delayed peak at 30 months. This pattern may reflect a reduction in early metastatic recurrences given the early stage of the tumors but deserves further investigation.
The presence of satellites, underlying cirrhosis, and nonanatomic resection were associated with time to recurrence. The presence of satellites has been found to be a significant predictor of outcome after resection of HCC in many other studies.27 Likewise, the nature of the nontumoral liver around the HCC has also been shown to be a strong predictor of recurrence of HCC after resection.28 Generally, neither variable is known preoperatively to help guide patient selection or the selection of the most appropriate therapy.
The success of sorafenib in the treatment of advanced HCC has opened the door for the testing of targeted molecules in the adjuvant setting.29 The degree of fibrosis and the presence of satellites can help select or stratify patients who are most at risk for recurrence and who may benefit most from sorafenib after resection if the drug is eventually found to be an effective agent in the adjuvant setting. Alternatively, patients with satellites who are at risk for early metastatic recurrence can be referred for salvage liver transplantation, as has been proposed by the Barcelona group.30 Patients with cirrhosis, who are at risk for later recurrence, can potentially benefit from antiviral therapy, which has been shown to decrease the incidence of late recurrence in one randomized trial.26 However, further studies are needed in this regard.
Anatomic resection was found to be independently associated with both overall and very early recurrence at 1 year. Other authors have also found a benefit from performing anatomic resection for small HCC.31, 32 Whereas anatomic resection was associated with a lower overall recurrence of 60% and very early (<1 year) recurrence of 10% for the entire cohort, it was associated with a very dramatic 40% lower rate of recurrence at 1 year for patients who were found to have either vascular invasion or satellites and thus did not meet the criteria for pathologically very early cancers. In addition, anatomic resection was associated with a very significant improvement in survival as well as lower rates of overall and early recurrence in patients with satellites. These findings support the theory that removing the entire segment supplied by a portal pedicle will result in lower recurrence and, hopefully, better survival, particularly in those patients in whom the tumor has gained access to the microcirculation or developed micrometastases.33 These oncological benefits of an anatomical resection are something that is unique to hepatectomy and cannot be duplicated by percutaneous ablation.
Several shortcomings of this study must be noted so that readers can exercise the appropriate level of caution when interpreting and extrapolating the results. Perhaps the most significant limitation is the relatively small sample size. With only 32 events in terms of survival, the study lacked statistical power leading to an increased possibility of type 2 error. In addition, the low number of events may also result in a type 1 error where some of the variables identified as significant on univariate analyses may have, in fact, not been truly associated with survival. The fact that two centers with such a large volume of hepatic resections for HCC were able to accrue only 132 cases of HCC ≤2 cm over a 20-year period points out the rare nature of these small tumors in the Western experience. A sample size this small also prevented the ideal scenario of constructing and testing predictors of outcomes using separate training and validation cohorts. In addition, the limited number of events made robust multivariate analysis difficult. Consequently, the results of the multivariate analysis must be viewed as an exploratory analysis of this group of patients and clearly need validation in an independent cohort before clinical decisions can be made based on this data.
In conclusion, hepatic resection for HCC ≤2 cm is safe and offers excellent long-term results. Platelet count ≥150,000/μL was associated with survival on multivariate analysis. Recurrence remains a significant problem despite the small size of these tumors. Anatomic resection was associated with lower overall and 1-year recurrence. Even with such early tumors, approximately one-third will present with either vascular invasion, satellite tumors, or both. It is these patients in particular who cannot be identified preoperatively by imaging, in whom anatomic resection is associated with a lower rate of recurrence.