Introduction

One life-threatening complication of hepatocellular carcinoma (HCC) is the spontaneous rupture of the tumor, leading to intra-peritoneal hemorrhaging and shock. The incidence of a spontaneous HCC rupture has been reported as 3–26 %1 – 5 and is associated with a high rate of in-hospital death (32–75 %).3 , 4 , 6 – 8 Furthermore, spontaneous HCC rupture is known to occur at the advanced tumor stage in patients with a poor liver function, leading to a poorer prognosis than that of patients with a non-ruptured HCC, even after R0 hepatic resection.9 – 11 Thus, the current TNM staging system for HCC, defined by the American Joint Committee on Cancer/Union for International Cancer Control (AJCC/UICC),12 classifies all ruptured HCCs as T4.

However, the prognostic influence is still controversial. Some articles reported that the rupture itself is not a prognostic factor.13 , 14 Discrepancies in patient backgrounds and the liver function might partially influence the prognosis. Thus, an analysis after matching the patients’ backgrounds and liver function as well as tumor-related factors will be essential to clarify the impact of the ruptured HCC on postoperative survival.

The high incidence of postoperative recurrence is another issue. The details of postoperative recurrence in patients with ruptured HCC are unclear. Patients who have suffered from spontaneous HCC rupture have a higher incidence of peritoneal dissemination after a hepatic resection than those with a non-ruptured HCC.15 However, there are some reports that a ruptured HCC did not increase the risk for dissemination after a hepatic resection.16 , 17 Furthermore, the recurrence patterns, such as intra- and extrahepatic recurrence, also vary.13 , 14. Thus, because of the low incidence of spontaneous HCC rupture, the heterogeneity of patients, and the choice of treatment at presentation (emergency or staged),10 , 14 , 16 , 18 the true impact of tumor rupture on survival or recurrence after a hepatic resection has not been clarified.

In the present study, clinical data were collected from six university hospitals. We investigated the clinical findings and surgical outcomes of all patients who had spontaneous HCC ruptures and compared them with the findings of non-ruptured HCC patients to clarify the characteristics of patients with spontaneous rupture of HCC. In addition, after selection by matching at a ratio of 1:1 according to propensity scores, we compared the surgical outcomes and analyzed the true impact of the rupture on survival after hepatic resection.

Materials and Methods

Patients

From January 2000 to December 2013, a total of 58 consecutive patients with spontaneously ruptured HCC underwent R0 hepatic resection at six university hospitals (Department of Hepato-Biliary-Pancreatic Surgery, Osaka City Graduate School of Medicine, Osaka, Japan; Department of Surgery, Hirakata Hospital, Kansai Medical University, Hirakata, Osaka, Japan; Second Department of Surgery, Wakayama Medical University, Wakayama, Japan; Department of Gastroenterological Surgery, Graduate School of Medicine, Osaka University, Suita, Osaka, Japan; Department of General and Gastroenterological Surgery, Osaka Medical College, Takatsuki, Osaka, Japan; and Department of Surgery, Faculty of Medicine, Kinki University, Osaka-Sayama, Osaka, Japan) (ruptured group). Patients found to have lymph node metastasis, distant organ metastasis, or peritoneal dissemination using preoperative diagnostic imaging or during surgery were excluded from the study. During the same period, 1922 patients underwent curative hepatic resection for a non-ruptured HCC (non-ruptured group). The 58 patients with ruptured HCC were matched at a ratio of 1:1 according to a propensity score19 – 21 based on their age, sex, underlying hepatic diseases, liver function test findings, tumor-related factors, histological variables, and type of hepatic resection to adjust for differences between the two groups.

This study was approved by the institutional review board from each participating institution and conducted in accordance with the mandates of the Declaration of Helsinki. Informed consent was obtained from all patients for their data to be used for research purposes.

Parameters

The clinical data of all patients were retrospectively collected, including the patients’ baseline characteristics (age, sex [number of males/females]) and number and proportion of patients with positivity for hepatitis B surface antigen (HBV) and for anti-hepatitis C virus antibody (HCV), non-B non-C hepatitis (NBC), alcohol abuse, pathologically proven liver cirrhosis (hepatitis activity index score, fibrosis score 422 , 23), liver function test findings (serum concentration of total bilirubin, albumin, prothrombin time, aspartate aminotransferase, and alanine aminotransferase, platelet count, and distribution of patients with Child-Pugh class [A/B/C]), tumor-related factors (serum level of alpha fetoprotein [AFP], tumor size, number of patients with a tumor or tumors [solitary or multiple], number and proportion of patients with major vessel invasion [tumor invasion to main or first branch of portal vein and/or to major hepatic vein or inferior vena cava], and distribution of patients with clinical stage without rupture factor according to AJCC/UICC classification), histological variables (number and proportion of patients according to tumor differentiation [well, moderately, poorly, undifferentiated, and unknown], histological number of patients with a tumor or tumors [solitary or multiple], and number and proportion of patients with portal vein invasion and hepatic vein invasion), and number and proportion of the patients according to types of hepatic resection (partial, segmentectomy, sectionectomy, and hemihepatectomy or more). Additionally, in the ruptured group, the symptoms at rupture (number and proportion of patients who complained of abdominal pain and who developed hypovolemic shock), number and proportion of patients with initial treatment for the rupture (transcatheter arterial embolization [TAE], conservative treatment because of spontaneous hemostasis, and emergency hepatic resection), timing of hepatic resection (number and proportion of staged hepatic resection), and duration from the rupture to surgery were investigated. The overall survival (OS) time was defined as the interval from the hepatic resection to the date of death or the last follow-up examination. The recurrence time was defined as the duration between the day of hepatic resection and the day when recurrence was diagnosed by image assessment. The date of follow-up was censored on July 31, 2015. The details regarding recurrence were as follows: presence or absence of dissemination during the follow-up period, intrahepatic or extrahepatic initial recurrence, and initial recurrence met or beyond the Milan criteria.24

Types of Hepatic Resection

The hepatic anatomy and type of hepatic resection were classified according to The Brisbane 2000 Terminology of Liver Anatomy and Resections.25 R0 hepatic resection was defined as the complete macroscopic removal of the tumor.

Follow-Up Methods

All patients were followed up at least every 3 months after discharge. The follow-up evaluations included a physical examination, liver function tests, chest radiographs to check for pulmonary metastases, and ultrasonography, computed tomography, or magnetic resonance imaging to check for recurrence in the remnant liver or other abdominal organs. Chest computed tomography was performed if the chest radiographs showed any abnormalities. Bone metastasis was diagnosed from magnetic resonance imaging and/or bone scintigraphy. Positron emission tomography was performed if necessary. Patients who developed recurrence were treated by appropriate treatments (repeat hepatic resection, ablation therapy, transcatheter arterial chemoemblizaton, sorafenib, or best supportive care; selection of these treatments depended on the recurrent tumor situation.).

Propensity Score Matching (PSM)

To avoid confounding differences in the treatment outcome due to “rupture” with those stemming from baseline differences between the two groups, we performed our analyses not only with the complete patient cohort but also a propensity score-matched subset. Logistic regression was used to calculate propensity scores for patients in the ruptured and non-ruptured groups. The following variables were entered into the propensity model: age, sex, underlying hepatic diseases (presence or absence of HBV and HCV, NBC, alcohol abuse, and liver cirrhosis), liver function tests (serum concentration of total bilirubin, albumin, prothrombin time, AST, and ALT, platelet counts, and Child-Pugh class), tumor-related factors (serum concentration of AFP, tumor size, tumor number, presence or absence of macrovacular invasion, and AJCC/UICC Stage without rupture factor), histological variables (differentiation, tumor number, and presence or absence of portal vein invasion and hepatic vein invasion), and type of hepatic resection. PSM was performed using a 1:1 ratio without replacement and a caliper width of 0.1. Receiver operating characteristic (ROC) curves were used to assess the accuracy of PSM, as a predictor of “rupture” indicated by a propensity score. The resulting subset of score-matched patients in the ruptured and non-ruptured groups was used in subsequent analyses.

Statistical Analysis

Continuous variables are expressed as the median (range). Baseline characteristics between the patients with spontaneous ruptured HCC and the other patients were analyzed using the Mann–Whitney U test. Differences in categorical variables were analyzed using the chi-squared test or Fisher’s exact test. We used the Kaplan-Meier method to calculate the OS and DFS rates, and differences between the two groups were evaluated using the log-rank test. Cox’s proportional hazard model with stepwise variable selection was used to estimate the risk factors for unfavorable survival and recurrence after hepatic resection in patients with ruptured HCC. In this study, continuous values were converted into categorical data using their median values, and all of the variables of patients’ backgrounds, liver function test findings, tumor-related factors, pathological variables, and types of hepatic resection were investigated univariately. Variables with a p value <0.10 by a univariate analysis were entered into a multivariate analysis. Logistic regression was used univariately and multivariately to estimate the relative risk of in-hospital death in the ruptured group. The variables in this group with a p value <0.10 by a univariate analysis were also entered into a multivariate analysis. A p value of <0.05 was considered statistically significant. All statistical analyses were performed using the SPSS® v.22.0 software program (IBM Corp., Armonk, NY, USA).

Results

Clinicopathological Features and Surgical Outcomes of All Patients with a Ruptured HCC

In the 58 patients in the ruptured group, the median age was 65 years, and 50 (86 %) were male (Table 1). The most common initial symptom was abdominal pain (44 patients, 75 %). Ten patients (17 %) developed hypovolemic shock upon admission or soon after admission. All patients were diagnosed as having suffered a spontaneous rupture by a physical examination, diagnostic imaging, and laboratory tests. Emergency TAE was performed in 41 (71 %) of the 58 patients, and hemostasis was achieved in 40 of them (98 %). This led to staged hepatic resection; however, one patient required emergency hepatic resection for intractable bleeding soon after TAE. Emergency hepatic resection without a TAE was performed in four patients, and the remaining 13 patients achieved spontaneous hemostasis. These patients were managed conservatively, including a transfusion to stabilize their general condition. Thereafter, they underwent staged hepatic resection.

Table 1 Clinical characteristics in patients with ruptured and non-ruptured hepatocellular carcinoma who underwent hepatic resection
Full size table

The duration from diagnosis of the rupture to surgery was 40 days (range 0–129 days). In-hospital death occurred in seven patients (12 %), with details as follows: 3 of the 5 patients (60 %) who underwent emergency hepatic resection, 4 of the 40 patients (10 %) who underwent staged hepatic resection (p = 0.021 vs. emergency hepatic resection), and 0 of the 13 patients (0 %) who were hemodynamically stabilized by conservative treatments followed by hepatic resection (p = 0.012 vs. emergency hepatic resection). The causes of in-hospital death in the ruptured group were liver failure (n = 5), respiratory failure (n = 1), and disseminated intravascular coagulation (n = 1).

A multivariate logistic regression analysis indicated that emergency hepatic resection (p = 0.026, odds ratio 12.4, 95 % confidence interval [CI] 1.3–113.2) was an independent risk factor for in-hospital death (Table 2). The OS rate at 5 and 10 years after hepatic resection was 37 and 34 %, respectively. The multivariate analysis also indicated that Child-Pugh B was an independent risk factor for an unfavorable OS (p = 0.035, hazard ratio 2.80, 95 % CI 1.1–7.3, Table 3). No patients with Child-Pugh class B were alive beyond 37 months after hepatic resection.

Table 2 Risk factors for unfavorable overall survival in patients with ruptured hepatocellular carcinoma after hepatic resection
Full size table
Table 3 Univariate and multivariate logistic analysis for mortality after hepatic resection for ruptured hepatocellular carcinoma
Full size table

Comparisons of the Clinicopathological Features and Surgical Outcome for Patients with a Ruptured HCC and a Non-ruptured HCC (All Populations)

Comparisons between the ruptured group and the non-ruptured group (all populations) are shown in Table 1. Ruptured HCC patients tended to be younger (mean age 65 vs. 68 years old) and have a higher prevalence of HBV infection than the non-ruptured patients. In the liver function tests, the mean serum concentration of albumin and prothrombin time was lower in the ruptured group than in the non-ruptured group. Regarding tumor-related factors, the mean serum level of AFP was higher and tumor size larger in the ruptured group than in the non-ruptured group. According to the AJCC/UICC classification without rupture factor, the proportion of patients with stage III was higher in the ruptured group than in the non-ruptured group. In histological variables, the proportion of patients with poorly differentiated HCC was higher in the ruptured group than in the non-ruptured group. According to the types of hepatic resection, the proportion of patients who underwent a hemihepatectomy or more was higher in the ruptured group than in the non-ruptured group.

The incidence of in-hospital death was higher in the ruptured group than in the non-ruptured group (13 vs. 2 %, p < 0.001). The mean and median follow-up periods were 39 and 22 months (range 0–159 months) in the ruptured group, and 49 and 41 months (range 0–174 months) in the non-ruptured group, respectively. There was no significant difference in the DFS rates between the two groups (p = 0.131, Fig. 1). However, the OS rates at 3, 5, and 10 years after hepatic resection in the ruptured group were 48, 37, and 34 %, respectively, which were significantly worse than those of the non-ruptured group (78, 65, and 42 %, respectively; p < 0.001, Fig. 2). If patients with in-hospital death were excluded, the OS rates at 3, 5, and 10 years after hepatic resection were 55, 42, and 38 % in the ruptured group, and 79, 66, and 43 % in the non-ruptured group, respectively (p = 0.006). During the follow-up period, the proportion of patients with postoperative recurrence was similar, with 34 patients (67 %) in the ruptured group vs. 1157 patients (61 %) in the non-ruptured group (p = 0.471, Table 4). However, peritoneal dissemination was detected in four (8 %) patients in the ruptured group during the follow-up period (p = 0.001), and as the initial recurrence in two of these patients. The incidence of extrahepatic recurrence was higher in the ruptured group than in the non-ruptured group (24 vs. 10 %, p = 0.022). Furthermore, the proportion of patients with postoperative recurrence beyond the Milan criteria was significantly higher in the ruptured group than in the non-ruptured group (79 vs. 55 %, p = 0.005).

Fig. 1
figure 1

The disease-free survival rate after hepatic resection for patients with a ruptured hepatocellular carcinoma (ruptured group) and those with a non-ruptured hepatocellular carcinoma (non-ruptured group) among all populations (before propensity score matching)

Full size image
Fig. 2
figure 2

The overall survival rate after hepatic resection for patients with a ruptured hepatocellular carcinoma (ruptured group) and those with a non-ruptured hepatocellular carcinoma (non-ruptured group) among all populations (before propensity score matching)

Full size image
Table 4 Postoperative recurrence after hepatic resection for ruptured and non-ruptured hepatocellular carcinoma
Full size table

PSM

The logistic regression model of propensity score (within a caliper width of 0.1) in the two groups and the above covariates were appropriately based on the assessment of goodness-of-fit statistics, as proposed by Lemeshow and Hosmer (p = 0.979).26 The ROC area under the curve of the propensity score for rupture was 0.853 (95 % CI = 0.813–0.893). After PSM, 42 patients from each group were selected for further subset analyses.

Comparisons of the Clinicopathological Features and Surgical Outcomes of Patients with a Ruptured HCC and a Non-ruptured HCC (After PSM)

The clinicopathological variables were similar between the two matched groups (Table 1). The incidence of in-hospital death was five patients (12 %) in the ruptured group and one (2 %) in the non-ruptured group (p = 0.202). The mean and median follow-up periods were 44 and 25 months (range 1–157 months) in the ruptured group, and 54 and 49 months (range 1.7–142 months) in the non-ruptured group, respectively. There was no significant difference in the DFS rate (p = 0.624, Fig. 3) or OS rate (p = 0.115, Fig. 4) between the groups. In the rupture group, the OS rate rapidly decreased within 3 years after hepatic resection. In contrast, in the non-ruptured group, the OS rate gradually decreased during the follow-up period (median survival time: 40 months in the ruptured group vs. 74 months in the non-ruptured group). Moreover, if patients with in-hospital death were excluded, the OS rates at 3, 5, and 10 years after hepatic resection were 58, 48, and 43 % in the ruptured group, and 85, 68, and 31 % in the non-ruptured group, respectively (p = 0.321). Furthermore, there was no significant difference in the proportion of patients with postoperative recurrence (65 vs. 73 %, p = 0.427), extrahepatic recurrence (13 vs. 3 %, p = 0.201), and recurrence beyond the Milan criteria (71 vs. 60 %, p = 0.407, Table 4).

Fig. 3
figure 3

The disease-free survival rate after hepatic resection for patients with a ruptured hepatocellular carcinoma (ruptured group) and those with a non-ruptured hepatocellular carcinoma (non-ruptured group) after propensity score matching

Full size image
Fig. 4
figure 4

The overall survival rate after hepatic resection for patients with a ruptured hepatocellular carcinoma (ruptured group) and those with a non-ruptured hepatocellular carcinoma (non-ruptured group) after propensity score matching

Full size image

Discussion

Hemostasis is the prime purpose of the initial treatment of a spontaneously ruptured HCC. Some have recommended that emergency hepatic resection be performed when the patient’s condition permits5 , 27 , 28; however, the mortality rate has been reported to be high and widely varied (range 5.8–71.4 %).5 , 29 , 30 In addition, Yang et al.10 recently reported that the incidence of in-hospital death in patients who underwent emergency hepatic resection (2 of 28 patients [7 %]) was higher than that in patients who underwent a staged hepatic resection (2 of 115 patients [1.5 %]), although not significantly different. However, TAE can effectively induce hemostasis in hemodynamically unstable patients, with the most recent reports citing a hemostasis success rate of more than 90 %.16 , 31 In addition, in the present study, only one patient failed to achieve hemostasis with TAE (success rate of 98 %), and emergency hepatic resection for patients with a ruptured HCC was an independent risk factor for in-hospital death. Even if the patient is hemodynamically stable, non-surgical treatments, such as TAE and transfusion, allow for the improvement of a patient’s general condition and planning for staged hepatic resection.

It is well known that ruptured HCC is associated with larger-sized tumors with higher incidence with vascular invasion than non-ruptured HCC,9 , 14 , 18 leading to the performance of large-volume liver resection, such as hemihepatectomy or more extended resection9; this agreed with the present findings among all populations. Furthermore, in the present study, the presence of major vessel invasion was a possible risk factor for in-hospital death. Such a situation would increase the difficulty and risk of hepatic resection, which may contribute in part to the high incidence of in-hospital death.

In some cases of a ruptured HCC, an elaborate approach should be considered, as ruptured HCCs are generally large bulging tumors walled off by a hematoma and/or adhesion; non-meticulous mobilization of the tumor could lead to re-rupture or hemorrhaging. Given this evidence, TAE and/or conservative treatment (even if spontaneous hemostasis is achieved) might be an effective and minimally invasive modality as an initial treatment procedure for a spontaneously ruptured HCC, and clinicians should be careful when selecting emergency hepatic resection. Some cases of ruptured HCC with a poor liver function or liver cirrhosis have been associated with an increased 30-day mortality as well as unfavorable OS.11 , 32 The findings from our study also suggest that surgical treatment for ruptured HCC patients with a poor liver function (Child-Pugh class B) should not be performed out of consideration of the poor prognosis.

A Japanese nationwide survey of patients with HCC found that spontaneous HCC rupture occurred in a heterogeneous population of patients where the risk factors for rupture differed according to the underlying hepatic diseases, such as positivity for the hepatitis B and hepatitis C virus antigens, and non-B non-C hepatitis.18 In addition, a comparison of the OS according to each TNM stage of the AJCC/UICC and LCSGJ classifications (without the rupture factor) indicated that the TNM stage of spontaneously ruptured HCCs corresponded to an additional 0.5–2.0 TNM stage without the rupture factor. However, these results included patients who underwent all treatments, including hepatic resection, ablation therapy, TAE, and best supportive care. On closer inspection, we found that the prognosis of patients who underwent curative treatments (hepatic resection or ablation therapy) was much better than that of patients who underwent other treatments. Chan et al.9 also reported that the OS of patients with ruptured HCC (n = 84) and non-ruptured HCC (n = 1254) and stage I and II ruptured HCC according to the AJCC/UICC classification (without a rupture factor) corresponded with those values in stage II and III non-ruptured HCC, respectively.

In contrast, Uchiyama et al.13 reported that the OS of patients with a ruptured HCC (n = 27) divided by TNM stage according to LCSGJ and AJCC/UICC classifications (without a rupture factor) was better than those with a non-ruptured HCC (n = 1004). This evidence seems to suggest that not all ruptured HCC should be defined as T4. However, in these previous studies, the number of patients, background, and tumor-related factors were very different between the ruptured HCC and non-ruptured HCC patients. Thus, our case-matched study decreased the bias between the ruptured and non-ruptured groups and allowed us to investigate the true impact of rupture on survival.

In the present study, the OS rate at 5 years was 37 % (42 % after PSM), which corresponded with the values in recent reports (approximately 30 %, range 10–55 %,).9 – 11 , 13 , 14 , 16 , 18 , 31 Some reports have indicated that a ruptured HCC is a risk factor for an unfavorable OS,9 , 10 which agreed with our results among all populations even if patients with in-hospital death were excluded. However, after PSM, there was no significantly difference in OS rate between the two groups (p = 0.115). Moreover, if patients with in-hospital death were excluded, OS rates at 5 and 10 years was 48 and 43 % in the ruptured group, and 68 and 31 % in the non-ruptured group, respectively (p = 0.321). The ruptures were more frequently observed in patients with a high AFP concentration, large tumor, and poorly differentiated HCC before PSM, which have been established as risk factors for an unfavorable OS after hepatic resection for patients with HCC,33 – 39 indicating a ruptured HCC tends to be associated with an advanced grade of malignancy. Indeed, a large tumor is known to increase the risk for extrahepatic recurrence.37 , 40 , 41 In addition, in our study, the proportion of patients with extrahepatic recurrence and with recurrence beyond the Milan criteria was higher in the ruptured group than in the non-ruptured group among all populations, which reflects the postoperative recurrence pattern of a large tumor.37 , 40 However, after adjusting for the incidence of these known prognostic factors by PSM, there was no significant difference in the pattern of recurrence between the groups.

Given this evidence, the rupture itself does not appear to be a prognostic factor for unfavorable OS. Rather, advanced tumor-related factors such as large tumor, high AFP concentration, and poorly differentiated HCC seem to be more closely associated with a poor prognosis. Two previous studies using PSM also found no marked difference in the DFS and OS between patients with a ruptured HCC and non-ruptured HCC14 , 16; however, the number of matched patients was very small (n = 18,16 and 1414). Our results (n = 42) emphasize these prior findings. A high incidence of peritoneal dissemination may also be a characteristic of a ruptured HCC,15 , 42 which agreed with our results before PSM. However, there was no significant difference in the incidence of peritoneal dissemination between the groups after PSM. Yang et al.10 suggested that early resection may reduce the rate of occurrence of peritoneal dissemination; however, the influence of postoperative peritoneal dissemination on survival could not be investigated in the current study because the number of patients was small (n = 4).

Several limitations associated with the present study warrant mention. This study was retrospective in nature, and the number of patients with ruptured HCC was not very large, despite this being a multicenter study. In addition, the initial treatment for the rupture differed among ruptured patients. Further studies are needed to evaluate the detailed surgical outcome of patients with ruptured HCC.

Conclusions

This multicenter analysis with a case-controlled study by PSM showed that a rupture itself was not a risk factor for postoperative recurrence and unfavorable survival. However, clinicians should be careful when selecting emergency hepatic resection and must keep in mind that the prognosis of patients classified as Child-Pugh class B is poor.