Effect of Background Liver Cirrhosis on Outcomes of Hepatectomy for Hepatocellular Carcinoma

Key Points

Question  What is the pattern and recurrence rate of de novo hepatocellular carcinoma among patients with liver cirrhosis and normal liver after resection of hepatocellular carcinoma?

Findings  The median annual incidence of postoperative recurrence of hepatocellular carcinoma within 5 years after surgery was lower in the group with normal liver (5.9%) compared with the group with liver cirrhosis (12.7%). Multiple recurrences near the resection margin or at extrahepatic sites were more frequent in the group with normal liver (50.0% vs 15.4%), whereas solitary recurrence at a distant site was more common in the group with liver cirrhosis (53.8% vs 5.6%).

Meaning  Comparison of the patterns and annual incidence of recurrence of hepatocellular carcinoma demonstrated that the poorer prognosis in the group with liver cirrhosis was likely owing to a higher hepatocarcinogenic potential among patients with cirrhosis.

Importance  Background hepatocarcinogenesis is considered a major cause of postoperative recurrence of de novo hepatocellular carcinoma (HCC) in patients with liver cirrhosis (LC). The degree of underlying liver injury has reportedly correlated with surgical outcomes of HCC. However, the pattern and annual rate of recurrence of postoperative de novo HCC are still unclear.

Objective  To clarify the pattern and rate of recurrence of de novo HCC in patients with LC.

Design, Setting, and Participants  Data from 799 patients who underwent curative hepatectomy for HCC at Toranomon Hospital and The Johns Hopkins Hospital between January 1, 1995, and December 31, 2014, were retrospectively collected and analyzed. Of the patients who underwent curative hepatectomy for HCC, 424 met inclusion criteria: 73 with normal liver (NL) and 351 with LC. Sixty-four patients who had histologically proven NL parenchyma were matched with an equal number of patients who had established LC, and postoperative outcomes were compared.

Interventions  Hepatectomy in patients with HCC.

Main Outcomes and Measures  Patterns of recurrence of HCC and chronological changes in recurrence rates.

Results  Among 128 matched patients in the study (mean [SD] age, 64.0 [12.7] years; 93 men and 35 women) 1-, 3-, and 5-year cumulative recurrence was 17.2%, 23.0%, and 37.5%, respectively, in the NL group vs 25.0%, 55.5%, and 72.1%, respectively, in the LC group (P = .001). The 3- and 5-year disease-specific survival was 85.7% and 75.4%, respectively, in the NL group vs 74.9% and 59.1%, respectively, in the LC group (P = .04). The median annual incidence of postoperative recurrence of HCC within 5 years after surgery was lower in the NL group (5.9%) compared with the LC group (12.7%) (P = .003). Assessment of recurrence patterns revealed that multiple recurrences near the resection margin or at extrahepatic sites were more frequent in the NL group (9 [50.0%] vs 6 [15.4%]; P = .01), whereas solitary recurrence at a distant site was more common in the LC group (21 [53.8%] vs 1 [5.6%]; P < .001).

Conclusions and Relevance  Comparison of the patterns and annual incidence of recurrence of HCC demonstrated that the poorer prognosis in the LC group was likely owing to a higher hepatocarcinogenic potential among patients with cirrhosis. Annual recurrence rates in the 2 groups indicate that de novo recurrence may continuously occur from the early postoperative period until the late period after resection of HCC.

Hepatocellular carcinoma (HCC) is the fifth most common cancer in men, and seventh among women, with more than half a million new cases diagnosed annually worldwide. In most cases, the etiologic factors of HCC are attributable to underlying liver injury due to chronic hepatitis or liver cirrhosis (LC).^1,2 Although hepatic resection is one of the most effective treatment options for patients with HCC and preserved hepatic functional reserve, the cumulative risk of recurrence during the first 5 years after surgery ranges from 70% to 100%.^3-5 As such, local tumor control remains a major issue in the clinical management of HCC. The high risk of recurrence after curative-intent hepatectomy is attributable to 2 unique patterns: recurrence derived from residual micrometastases and de novo recurrence owing to carcinogenic potential in the underlying liver.^6,7 Oncologic outcomes after resection of HCC have reportedly correlated with the degree of underlying liver injury, especially among patients with LC.^1,2 Given that the annual incidence of HCC arising from established LC has been reported to range between 2.5% and 6.6%,^8-10 it is estimated that the cumulative incidence of recurrence owing to de novo carcinogenesis would result in significant differences in long-term outcomes among patients with LC vs patients who have an undamaged liver.

Previous studies examining the influence of de novo recurrence of HCC attempted to distinguish the type of recurrence by using a time frame (early recurrence, occurring within 2 years after surgery, and late recurrence, occurring more than 2 years after surgery).^6,7,11,12 However, it is difficult to clearly distinguish the causes of recurrence of HCC, and the actual prognostic effect of de novo recurrence remains unclear. To address this issue, our study sought to determine the prognostic difference associated with the carcinogenic potential between patients with LC and those with a normal liver (NL) in a case-matched population adjusted for other oncologic factors.

A total of 799 adult patients without a history of previous HCC treatment who underwent curative-intent surgery for HCC between January 1, 1995, and December 31, 2014, at Toranomon Hospital in Tokyo, Japan, and at The Johns Hopkins Hospital in Baltimore, Maryland, were identified from a multi-institutional database. Only patients with HCC who had either histologically proven LC or NL were included in this study, which was approved by the Toranomon Hospital Human Ethics Review Committee and The Johns Hopkins University Institutional Review Board. No additional informed patient consent specific to this study was required given its retrospective nature.

Hepatitis B virus (HBV) infection was defined by seropositivity for HBV surface antigen and/or HBV DNA; hepatitis C virus (HCV) infection was defined as patients who were seropositive for HCV antibody and/or HCV-RNA. Alcohol abuse was defined as chronic liver injury resulting from alcohol consumption of 60 g/d or more without another clear etiologic factor and was confirmed by pathologic examination of the specimen.¹³ The serum HBV-DNA level was investigated in patients with HCC without definitive etiologic factors to detect occult HBV infection.¹⁴ Conditions such as autoimmune hepatitis, primary biliary cirrhosis, primary sclerosing cholangitis, Wilson disease, and hemochromatosis were included in the category of other etiologies. Types 1 and 2 diabetes were diagnosed based on the 2003 criteria of the American Diabetes Association.¹⁵

Histopathologic variables were defined according to the pathologic classification system of the World Health Organization.¹⁶ Specimens were fixed in 10% formalin and stained with hematoxylin-eosin, Masson trichrome, silver impregnation, and periodic acid–Schiff after diastase digestion. Only the area of noncancerous liver parenchyma that was distant from the tumor was evaluated to avoid influence from the tumor. Fibrosis and status of inflammation were evaluated using the METAVIR score.¹⁷ Patients with no background liver damage, as evidenced by a liver that had had no fibrosis, moderate to severe inflammation, or severe steatosis (>30% of the total liver parenchyma) were assigned to the NL group. Patients with established cirrhosis (stage F4) were assigned to the LC group.

To control for the possibility of recurrence of HCC from residual micrometastases between the 2 groups, 1-to-1 individual paired case matching was performed for clinical factors other than underlying liver injury using R statistical software (R Foundation for Statistical Computing) with the optmatch package, which uses the optimal matching method with no caliper (https://cran.r-project.org/web/packages/optmatch/optmatch.pdf). Distance was calculated using Mahalanobis distance. These factors used for matching included sex, age, type of surgery, surgical margin status, tumor size, tumor differentiation, microvascular invasion, and preoperative serum α-fetoprotein level. After matching, the cumulative rate of recurrence, overall rate of survival, patterns of recurrence, and chronological changes in rates of recurrence were analyzed and compared.

Data were analyzed with SPSS software, version 19 (IBM SPSS), and EZR.¹⁸ Continuous variables were expressed as median and range or interquartile range. Categorical variables were expressed as frequencies and percentages. The Mann-Whitney test was used to compare continuous variables between the 2 groups. The χ² test or Fisher exact test was used to compare categorical variables between the 2 groups as appropriate. Cumulative recurrence and disease-specific survival were determined using the Kaplan-Meier method. Differences between curves were assessed by the log-rank test. P < .05 (2-tailed) was considered statistically significant.

Among 799 adult patients who underwent primary curative-intent hepatectomy for HCC during the study period, 424 patients (285 men and 139 women) met the inclusion criteria. Of these patients, 73 had histologically proven NL parenchyma and 351 had LC. The baseline characteristics of the NL and LC groups are summarized in eTable 1 in the Supplement. Chronic hepatitis infection, including HBV and HCV, was the major cause of LC (HBV, 114 [32.5%] and HCV, 206 [58.7%]). The prevalence of diabetes in the NL group was higher than in the LC group, although the difference was marginal (NL, 17 [23.3%] and LC, 59 [16.8%]; P = .18). Data from baseline liver function tests, including aspartate aminotransferase and alanine aminotransferanse levels, total bilirubin level, platelet count, and Child-Pugh grade, were different between the NL and LC groups. The preoperative median serum α-fetoprotein level was also different between the 2 groups (NL, 5.0 ng/mL; LC, 20.7 ng/mL; P < .001) (to convert to micrograms per liter, multiply by 1.0).

With regard to treatment-associated factors, 36 patients (49.2%) in the NL group underwent anatomical liver resection, whereas 70 patients (19.9%) in the LC group underwent an anatomical resection (P < .001). The proportion of resections with an R1 surgical margin was comparable between the groups (NL, 4 [5.5%] and LC, 32 [9.1%]; P = .49). With respect to tumor-associated characteristics, the proportion of multiple tumors was similar between the 2 groups (NL, 9 [15.5%] and LC, 41 [13.4%]; P = .84). The median maximum tumor diameter in the NL and LC groups was 40.5 mm (range, 20.0-250.0) and 20.0 mm (range, 2.0-150.0 mm), respectively (P < .001). Tumors with poor histologic differentiation were more commonly found in the NL group (20 [27.4%]) than in the LC group (57 [16.2%]) (P = .03). The presence of microvascular invasion tended to be more common in the NL group (20 [27.4%]) than in the LC group (69 [19.7%]) (P = .16). Mild steatosis in the resected liver was noted in 30 patients (41.1%) in the NL group.

The clinicopathologic characteristics of the case-matched cohort are summarized in eTable 2 in the Supplement. Sixty-four patients who had histologically proven NL parenchyma were matched with an equal number of patients who had established LC. Matching variables included age, sex, operative factors, and tumor-associated factors. The proportion of patients with diabetes was similar between the 2 groups (NL, 18 [28.1%] and LC, 16 [25.0%]; P = .84). Despite matching, background liver function, including alanine aminotransferase level and platelet count, remained different between the 2 groups.

The median follow-up time of the matched cohort was 40.2 months (interquartile range, 21.5-62.6 months). The median follow-up period of the NL group was 39.4 months (interquartile range, 19.8-59. months), which was comparable with follow-up time for the LC group (median, 40.8 months; interquartile range, 24.2-67.5 months) (P = .64). Recurrence was observed in 18 patients (28.1%) in the NL group and 39 patients (60.9%) in the LC group (P < .001). Overall, 1-, 3-, and 5-year cumulative recurrence was 17.2%, 23.0%, and 37.5%, respectively, in the NL group vs 25.0%, 55.5%, and 72.1%, respectively, in the LC group (P = .001; log-rank test) (Figure 1A). The 3- and 5-year disease-specific survival rate was 85.7% and 75.4%, respectively, in the NL group vs 74.9% and 59.1%, respectively, in the LC group (P = .04; log-rank test) (Figure 1B). Of 64 patients in the NL group, no patients died of causes associated with liver disease other than recurrence of HCC; of 64 patients in the LC group, 3 patients died of liver failure and 1 patient died of spontaneous bacterial peritonitis.

When comparing patterns of recurrence in the matched cohort, both the number of recurring lesions (NL, 18 [28.1%] and LC, 39 [60.9%]; P < .001) and site of recurrence (NL: near resection site, 7 [38.9%], distant from resection site, 4 [22.2%], and extrahepatic, 7 [38.9%]; LC: near resection site, 13 [33.3%], distant from resection site, 23 [59.0%], and extrahepatic, 3 [7.7%]; P = .005) were significantly different between the 2 groups (Table). In the NL group, most patients (12 of 18 [66.7%]) experienced multiple recurrences, whereas most patients in the LC group (31 of 39 [79.5%]) experienced a solitary recurrence (P = .001). Moreover, recurrence with 4 or more lesions was seen more frequently in the NL group than in the LC group (7 of 18 [38.9%] vs 4 of 39 [10.3%]; P = .03). In total, either multiple recurrences near the resection site and/or extrahepatic recurrence was the most frequent pattern of recurrence in the NL group than in the LC group (9 of 18 [50.0%] vs 6 of 39 [15.4%]; P = .01), whereas intrahepatic solitary recurrence distant from the resection site was the most frequent recurrence pattern in the LC group (NL, 1 of 18 [5.6%] and LC, 21 of 39 [53.8%]; P < .001)

The correlation between time to recurrence and patterns of recurrence is shown in Figure 2 and Figure 3. In the matched cohort, the median time to recurrence among patients with a solitary recurrence was longer than that among patients who experienced multiple recurrences (18.3 vs 6.1 months; P = .001). The difference in median time to recurrence for patients with single vs multiple recurrences persisted in both the NL and LC groups (NL: 29.0 vs 5.7 months; P = .007; LC: 18.3 vs 9.7 months; P = .04). Recurrence at a site distant from the resection tended to develop later after resection compared with other sites of recurrence (median, 19.8 vs 10.0 months; P = .06), a finding that was similar in both NL and LC groups (NL: median, 22.8 vs 7.8 months; P = .92; and LC: median, 19.8 vs 11.2 months; P = .07). Recurrences distant from the resection site were consistently seen from the early period to the late period after resection. Differences in overall patterns of recurrence between the NL and LC groups persisted over time (Figure 3B).

The estimated annual rates of postoperative recurrence for both groups are shown in Figure 4A. The median estimated annual rates of postoperative recurrence during the 5 years following resection were 5.9% in the NL group and 12.7% in the LC group (P = .003). Although the risk of recurrence was bimodal for both the NL and LC groups, patients in the NL group tended to develop recurrences earlier than did patients in the LC group.

Our study aimed to characterize postoperative de novo recurrence of HCC among patients with LC using a case-matched analysis comparing patients with LC and those with NL who had oncologically equivalent tumors. Despite genetic analysis being the criterion standard to establish that recurrent disease is de novo HCC, few studies have used this approach.^19-21 Rather, most previous studies that have investigated postoperative recurrence of de novo HCC were based on the timing of recurrence. Previous investigators assumed that de novo recurrences developed exclusively in the late postoperative period and therefore classified such recurrences as de novo disease.^6,7,11,12 The carcinogenic potential in the liver remnant is, however, theoretically stable; as such, subsequent de novo recurrence can develop at any time, even in the early period after surgery. To mitigate the limitations associated with the assumption that postoperative recurrences of de novo HCC were based on the timing of recurrence, we conducted a case-matched analysis to define differences in postoperative recurrence of HCC among patients with NL and LC. Histologically proven NL parenchyma has a very low carcinogenic potential. Therefore, we hypothesize that the prognostic difference observed between the patients with a NL and those with LC could be attributed to the difference in carcinogenic potential between the 2 groups. We found that there was a marked difference in the patterns of recurrence between patients with NL and those with LC who underwent resection of an index HCC. Specifically, the main difference in the pattern of recurrence between the 2 groups was observed in the incidence of recurrent disease located distant from the resected portion of the liver. The recurrence rate among patients in the LC group remained consistently 6% to 15% higher than that in the NL group 1 to 4 years after resection.

In examining specific patterns of recurrence, the NL group more often had multiple recurrences near the resection site or at an extrahepatic site, mostly during the early period after resection. In contrast, the LC group tended to develop solitary recurrences located at a distance from the resected part of the liver; this pattern of recurrence among patients with LC was consistent from the early period after surgery to the late follow-up period. The differences in patterns of recurrence were also correlated with differences in postoperative recurrence. The reason for the different patterns of recurrence between the NL and LC groups was undoubtedly multifactorial. Theoretically, recurrence derived from residual micrometastases can develop in the relatively early period after surgery at or near the resection site (ie, intrahepatic micrometastases within a tumor-bearing portal territory) or at an extrahepatic site (ie, distant metastases through systemic circulation).²² Moreover, time to recurrence was shorter for patients with multiple recurrences than for those with a single recurrence. Explanations of early multiple recurrences might be dissemination of the resected tumor or growth of multiple tumors at the initial surgery site, with microscopic nodules in the remaining liver being missed by imaging only to become apparent later. Conversely, an intrahepatic recurrence distant from the primary resection site may be considered a typical pattern of de novo recurrence.^23,24 In other words, the observed difference in patterns of recurrence are consistent with the hypothesis that postoperative recurrences in the NL group can be attributed largely to recurrence from residual cancer, whereas those of patients with LC are owing to both recurrence derived from micrometastases and de novo recurrence associated with the higher carcinogenic potential in the underlying liver.

Chronological changes in the annual incidence of recurrence demonstrated that recurrence in the LC group was about 6% to 15% higher per year than the risk of recurrence in the NL group (Figure 4A). Furthermore, after the first postoperative year, the cumulative hazards of recurrence continuously diverged over time (Figure 4B). These findings indicate that de novo recurrence may occur continuously from the early postoperative period until the late period. The annual incidence of primary HCC in patients with LC without a history of HCC is estimated to be 2.5% to 6.6%.^8-10 However, in patients with a history of HCC, a several-fold increase in the risk of developing a second primary HCC has been reported.^11,12 The difference in the annual rates of recurrence observed between the LC and NL groups is compatible with reported results in other patient populations.^{7,10,11,25,26} The high annual incidence of de novo recurrence after resection persisted from the early to the late postoperative period.

The timing and bimodal nature of the rate of recurrence was roughly comparable between the NL and LC groups, although recurrence tended to be slightly earlier in the NL group (Figure 4). The first peak of recurrence consisted of both recurrences owing to dissemination of resected tumor and small lesions that were not detected by imaging at the time of surgery. Considering that solitary recurrence at a distal site is much higher in the LC group, the first peak among the LC group would contain more undetected lesion at the time of surgery (Figure 3). Imamura et al¹¹ reported a similar recurrence pattern in patients with LC and attributed the late peak to de novo recurrence of HCC owing to background LC. In our study, however, there was a late recurrence peak also seen in the NL group. Other investigators have similarly reported late recurrences arising among patients with NL.^25-28 These late recurrences might be attributed to unknown background liver carcinogenesis present in patients with NL and an index HCC. Given the low chance of carcinogenesis in histologically proven NL, late recurrences in a patient with NL may derive from slowly growing recurrences of the resected index tumor. To this end, a previous study that included 15 944 patients with nonviral hepatitis without a history of significant alcohol intake or fatty liver revealed that only 2 patients (0.01%) developed HCC during the 1-year follow-up period.²⁹ Although presence of fatty liver and diabetes are known risk factors of hepatocarcinogenesis, Kawamura et al²⁹ reported that the annual incidence of HCC among patients with nonalcoholic fatty liver disease was only 0.04%, while the reported hazard ratio of patients with diabetes compared with patients who do not have diabetes is about 2.2 to 3.5.^30,31 Even though patients with a history of HCC have a several-fold increase in hepatocarcinogenesis compared with patients without a previous HCC, the rate of carcinogenesis is still negligibly low compared with the risk among patients who have established LC. In the future, genetic investigations comparing primary and recurrent lesions in patients with NL are required to provide more compelling evidence regarding the origin of late recurrences.

Our study had some limitations, including the retrospective study design. Individual paired matching was performed using a large population of patients with LC to generate the matched pairs. Despite the attempt at matching, it is possible that some residual differences in the 2 groups persisted.³² Although the biology of HCC may differ according to background etiologic factors, most traditional factors associated with aggressiveness of the disease were comparable between the 2 groups. However, residual differences of aggressiveness of the disease that arise in healthy vs cirrhotic livers may not have been fully accounted for. Furthermore, the different etiologic factors associated with the underlying LC were not taken into account. However, the incidence of HCC is fairly similar among patients who develop LC owing to different causes.³³ Although some differences may exist in the activity and modes of cancer promotion among patients with HBV infection, HCV infection, and alcohol abuse, the etiologic factors of background HCC were not associated with recurrence. Given the retrospective nature of the study, we also could not assess the presence or absence of a lead time bias or difference in screening protocol between the NL and LC groups.

Our study demonstrated different oncologic outcomes among patients with HCC and an underlying NL compared with patients who had LC. Specifically, patients with LC had a 6% to 15% higher annual risk of de novo recurrence compared with patients who had HCC resected from NL. Given the constant higher risk of de novo recurrence starting in the early postoperative period for patients with LC, preemptive treatment of viral hepatitis and close follow-up are important to decrease the risk of recurrence and detect early recurrent disease.

Accepted for Publication: October 28, 2016.

Corresponding Author: Timothy M. Pawlik, MD, MPH, PhD, Division of Surgical Oncology, Department of Surgery, The Johns Hopkins Hospital, 600 N Wolfe St, Blalock 688, Baltimore, MD 21287 (tpawlik1@jhmi.edu).

Published Online: January 4, 2017. doi:10.1001/jamasurg.2016.5059

Author Contributions: Dr Pawlik had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Sasaki, Shindoh, Margonis, Nishioka, Andreatos, Sekine, Pawlik.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: All authors.

Critical revision of the manuscript for important intellectual content: Sasaki, Shindoh, Margonis, Nishioka, Andreatos, Sekine, Pawlik.

Statistical analysis: Sasaki, Shindoh, Nishioka, Andreatos, Sekine, Pawlik.

Administrative, technical, or material support: Hashimoto, Pawlik.

Study supervision: Shindoh, Margonis, Pawlik.

Conflict of Interest Disclosures: None reported.