Potential conflict of interest: Nothing to report.
Outcome of variceal bleeding (VB) in patients with hepatocellular carcinoma (HCC) is unknown. We compared outcomes after VB in patients with and without HCC. All patients with HCC and esophageal VB admitted between 2007 and 2010 were included. Follow-up was prolonged until death, transplantation, or June 2011. For each patient with HCC, a patient without HCC matched by age and Child-Pugh class was selected. A total of 292 patients were included, 146 with HCC (Barcelona Classification of Liver Cancer class 0-3 patients, A [in 25], B [in 29], C [in 45], and D [in 41]) and 146 without HCC. No differences were observed regarding previous use of prophylaxis, clinical presentation, endoscopic findings, and initial endoscopic treatment. Five-day failure was similar (25% in HCC versus 18% in non-HCC; P = 0.257). HCC patients had greater 6-week rebleeding rate (16 versus 7%, respectively; P = 0.025) and 6-week mortality (30% versus 15%; P = 0.003). Fewer patients with HCC received secondary prophylaxis after bleeding (77% versus 89%; P = 0.009), and standard combination therapy was used less frequently (58% versus 70%; P = 0.079). Secondary prophylaxis failure was more frequent (50% versus 31%; P = 0.001) and survival significantly shorter in patients with HCC (median survival: 5 months versus greater than 38 months in patients without HCC; P < 0.001). Lack of prophylaxis increased rebleeding and mortality. On multivariate analysis Child-Pugh score, presence of HCC, portal vein thrombosis, and lack of secondary prophylaxis were predictors of death. Conclusions: Patients with HCC and VB have worse prognosis than patients with VB without HCC. Secondary prophylaxis offers survival benefit in HCC patients. (Hepatology 2013; 58:2079–2088)
International Classification of Diseases, Ninth Revision
Model for End-Stage Liver Disease
portal vein thrombosis
In the last few years, there has been an increasing incidence of hepatocellular carcinoma (HCC). The majority of these tumors develop in patients who have liver cirrhosis. The development of HCC has an effect in the natural history of liver disease. In a systematic review of studies that evaluated the natural history of cirrhosis, the presence of HCC was identified as a predictor of death in 66% of the studies in patients with decompensated cirrhosis that included this variable in their analysis. However, an increasing variety of therapeutic options are available for patients with HCC. Many of these options have survival benefit, so it is conceivable that these patients with HCC with longer survival will have greater chances of developing complications of end-stage liver disease (ESLD).
Variceal bleeding (VB) is one of the complications that characterize decompensated cirrhosis. In the last 30 years, there has been a substantial improvement in the survival of patients with VB as a result of the use of vasoactive drugs, the introduction of endoscopic band ligation, and the use of antibiotic prophylaxis.[4, 5] Presently, further efforts are targeted at developing individualized therapeutic strategies to adjust the approach to the risk the patient has.[6, 7]
Several prognostic studies have identified the presence of HCC as a negative prognostic factor in VB.[5, 8, 9] However, many studies in the context of VB were performed at times when the incidence of HCC was much lower.[10, 11] Furthermore, most observational and experimental studies in the setting of secondary prophylaxis excluded patients with HCC,[12-25] whereas other studies have excluded only patients with advanced HCC[26-28] or HCC outside of the Milan criteria.[6, 29] Therefore, it is unclear whether or not secondary prophylaxis is useful in these patients. A recent study in patients admitted because of VB demonstrated greater in-hospital mortality in those patients with HCC, compared to patients without HCC. However, this study was performed on a large database, based on International Classification of Diseases, Ninth Revision (ICD-9), diagnosis, with the limitations these studies have. Given the lack of information, the management of the acute VB (AVB) episode and then the use of secondary prophylaxis in these patients is most likely very heterogeneous across different centers. This gap in knowledge is becoming increasingly relevant, given the rising incidence of HCC, mainly associated with viral cirrhosis, which is expected to peak within the next 10 years. Therefore, the aim of this study was to evaluate the management and long-term outcomes, as defined by rebleeding and death, of patients with HCC and esophageal VB (EVB) in comparison to patients without HCC.
Patients and Methods
This retrospective observational study was performed in 10 centers in Spain (Hospital Vall d'Hebron [Barcelona], Hospital Clinic [Barcelona], Hospital Santa Creu i Sant Pau [Barcelona], Hospital del Mar [Barcelona], Hospital Germans Trías i Pujol [Badalona], Hospital Arnau de Vilanova [Lleida], Hospital Puerta de Hierro [Madrid], Hospital Ramón y Cajal [Madrid], Hospital Gregorio Marañón [Madrid], and Hospital Universitario de Canarias [Tenerife]). Patients meeting the following inclusion criteria were included: (1) AVB episode resulting from esophageal varices between January 2007 and December 2010; (2) liver cirrhosis as diagnosed according to clinical signs, laboratory, and imaging tests or by liver biopsy; and (3) HCC as diagnosed by current criteria, previously known at the time of the VB or diagnosed at the time of the bleeding episode. Patients with gastrointestinal (GI) bleeding not confirmed by diagnostic upper GI (UGI) endoscopy were not included.
For every patient with EVB and HCC, a patient with EVB without HCC was included. Patients were paired according to age (±5 years) and Child-Pugh class (A/B/C).
Follow-up of all patients was prolonged until June 2011. Patients who received liver transplantation (LT) during the follow-up were censored at this time point. Data regarding demographics, liver disease, bleeding episode, and follow-up were registered. In patients with HCC, information regarding tumoral disease was collected.
Bleeding was considered from esophageal variceal origin when the emergency endoscopy, performed within 12 hours after admission, showed any of the accepted criteria defining VB. Baveno V definition of events associated with the bleeding episode was used: failure to control bleeding; 6-week rebleeding; 6-week death and failure of secondary prophylaxis, which includes any significant bleeding resulting from portal hypertension after day 5 during the complete follow-up, that leads to hospitalization; and drop in 3 g of hemoglobin, blood transfusion, or death within 6 weeks of the rebleeding episode. Previous decompensation was defined by the presence of ascites, hepatic encephalopathy (HE), or VB.
Parametric and nonparametric variables are described with means (standard deviation) and medians (interquartile range; IQR), respectively. Categorical variables are described with proportions. Chi-square, Student t, and Mann-Whitney's tests were used according to variable characteristics. Patients who received LT were censored at the time of transplant. Kaplan-Meier's curves were constructed and compared with the log-rank test or Breslow's test, as appropriate. Cox's multivariate stepwise regression analysis was performed to analyze the independent effect of each variable on survival. The presence of statistical and biological interaction and confusion were analyzed by stratified analysis and inclusion of the product term of the interaction. The study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki. Ethics committee approval was obtained.
During the study period, a total of 146 patients were admitted because of EVB and HCC in the 10 centers (see Appendix 1). Patients had a median age of 68 years (IQR, 59-74) and were predominantly Child-Pugh class B (A in 30, B in 76, and C in 40) with a median Model for End-Stage Live Disease (MELD) score of 14 (range, 11-17; Table 1). HCC was diagnosed a median of 4 (range, 0-18) months before the VB episode. Thirty-seven (25%) patients were diagnosed at the time of the bleeding episode, whereas 109 (75%) were diagnosed previously. At the time of the VB, among HCC patients, BCLC staging was as follows: 0 in 3 patients; A in 27; B in 28; C in 45; and D in 41. Previous treatments performed in these patients were surgery (3 patients), radiofrequency ablation (14 patients), percutaneous alcoholization (10 patients), transcatheter arterial chemoembolization (43 patients), radioembolization (1 patient), and sorafenib (17 patients). As planned, 146 patients who were admitted because of VB during the same period without HCC were included with a median age of 67 (range, 56-74) and Child-Pugh class distribution A in 30, B in 79, and C in 37 with a median MELD of 14 (range, 10-17; P = 0.691, in comparison with HCC). Expectedly, viral etiology was proportionally more frequent among patients with HCC than in control patients. Furthermore, they more frequently had previous decompensation than the control group (73% versus 60%; P = 0.025). This finding was observed despite the fact that patients were matched by Child-Pugh class and had comparable MELD scores. Finally, HCC patients had more frequently portal vein thrombosis (PVT) than control patients. Most patients had not had previous VB and were eligible for primary prophylaxis (96 in HCC patients and 111 in non-HCC patients). From these patients, 44 (43%) with HCC had primary prophylaxis, compared to 40 (36%) without HCC (P = 0.186). Similarly, from patients who were eligible for secondary prophylaxis, no significant differences were observed between those with HCC (37 of 44; 84%) versus those without HCC (30 of 34; 88%; P = 0.755).
Table 1. Baseline Characteristics
Non-HCC Patients n = 146
HCC Patients n = 146
Variables are presented in proportion or medians (IQR). Denominator of the proportions is the number of patients in the group, unless otherwise stated. Chi-square and Mann-Whitney's U tests were used, as appropriate. Proportions of primary and secondary prophylaxis are calculated according to the number of eligible patients.
Abbreviations: HBV, hepatitis B virus, HIV, human immunodeficiency virus; INR, international normalized ratio.
Alcohol + HCV
HIV infection (%)
Child-Pugh class (%)
Previous decompensation (%)
Previous primary prophylaxis*
Previous secondary prophylaxis*
Time from HCC diagnosis, months
Portal vein trombosis (%)
Management and Outcomes of VB
No differences were observed regarding clinical presentation, endoscopic findings, and initial pharmacological and endoscopic treatment (Table 2). Five-day failure was similar (25% and 18% in patients with and without HCC; P = 0.257), although more patients with HCC died in this period (11% versus 4%; P = 0.025). Within the first 6 weeks, HCC patients had greater rebleeding rate (17% versus 7%, respectively; P = 0.022) and mortality (30% versus 15%; P = 0.003). Significantly fewer HCC patients received secondary prophylaxis after bleeding (83% versus 93%; P = 0.015) and, among those who received prophylaxis, standard therapy (combination of drugs and endoscopic band ligation [EBL]) was used less frequently (59% versus 70%; P = 0.098). As expected, patients with greater Barcelona Classification for Liver Cancer (BCLC) stages (C and D) had less frequently secondary prophylaxis (47 of 71; 66%), whereas almost all patients with lower BCLC stages (0, A, and B) had secondary prophylaxis (55 of 57; 96%; P < 0.001). Overall, lack of secondary prophylaxis was significantly associated with 6-week rebleeding (25% of those without prophylaxis, compared to 9% of those with prophylaxis; P = 0.016) and mortality (59% of those without prophylaxis, compared to 8% of those with prophylaxis; P < 0.001). PVT (none, benign, or malignant, respectively) was not associated with 5-day failure (20%, 24%, and 30%; P = 0.385), although it was associated with 5-day mortality (5%, 0%, and 23%; P < 0.001) and 6-week rebleeding (8%, 7%, and 29%; P = 0.001).
Table 2. Characteristics of VB and Outcomes (According to Baveno)
Non-HCC Patients n = 146
HCC Patients n = 146
Variables are presented in proportions. The denominator of the proportions is the number of patients in the group in whom the information was available or applicable,* unless otherwise stated. Small varices are grade I, large varices are >grade I varices. Chi-square tests were applied.
No significant differences in rebleeding after 6 weeks were observed between patients with and without HCC (19% versus 17%; P = 0.714; Table 3). However, overall failure of secondary prophylaxis was more frequent in patients with HCC than controls (32% versus 21%; P = 0.05). Expectedly, lack of secondary prophylactic measures was associated with secondary prophylaxis failure (data not shown; P < 0.001). Similarly, PVT was associated with secondary prophylaxis failure (none, 25%; benign, 21%; malignant, 35%; P < 0.001). During follow-up, 3 patients from each group received LT. Most patients without HCC died of decompensated liver disease (40 of 49), whereas those with HCC died of decompensated liver disease (34 of 109), tumoral disease (7 of 109), or a combination of both (61 of 109). Seven patients from each group had nonhepatic deaths.
Table 3. Long-Term Outcomes
Non-HCC Patients n = 140
HCC Patients n = 130
Patients who died within the first 5 days are excluded. Variables are presented in proportion or medians (IQR). The denominator of the proportions is the number of patients in the group, unless otherwise stated. Chi-square, Mann-Whitney's U, and log-rank tests were used, as appropriate.
Proportions are calculated according to the number of eligible patients, that is, the patients who were still alive 6 weeks after the original bleeding episode.
Transplant-free survival was significantly shorter in patients with HCC (median survival of 5 months versus over 38 months in patients without HCC; log rank: P < 0.001; Fig. 1A). This difference was maintained in each Child-Pugh class (log rank: P < 0.001; Fig. 1B-D). Previous decompensation was significantly associated with survival in the overall group; however, in patients with HCC, no significant differences were observed according to this variable (Fig. 2). Survival curves of patients with HCC according to BCLC classification is shown in Supporting Fig. 1. To simplify the statistical analysis and according to these survival curves, patients were divided in two groups of BCLC classification (0, A, and B and C and D). Expectedly, patients with BCLC 0, A, and B had better survival rates (median survival: 17.3 months; IQR, 9.6-36.1) than patients with BCLC C and D (1.5 months; IQR, 0.3-3.7), and both groups presented a worse outcome than patients without HCC (median survival: >60 months; Fig. 3).
Given the uneven distribution of well-known prognostic markers of rebleeding and death, multivariate analysis was performed to evaluate the adjusted effect of HCC on survival (Table 4A). Even when considering the other variables, HCC and lack of secondary prophylaxis remained independent predictors of death.
Table 4. Cox's Uni- and Multivariate Analysis of Predictors of Death
Univariate HR (95% CI)
Adjusted HR (95% CI)
Cox's multivariate stepwise regression analysis was used to evaluate the independent predictors of death (n = 142). Etiology was recoded in two categories (nonviral, which was the reference category, and viral disease). Variables initially included in the multivariate analysis were age, HCC, MELD, previous decompensation, secondary prophylaxis, etiology, PVT, and Child Pugh score. Variables included in MELD score were not included in the multivariate analysis to avoid colinearity. Child-Pugh class was not included because of the fact that this was a matching variable.
Abbreviation: INR, international normalized ratio.
Lack of secondary prophylaxis
Stratified analysis was performed to evaluate specifically the effect of use of secondary prophylaxis in patients according to BCLC. In patients with BCLC 0, A, and B, most had secondary prophylaxis. However, lack of secondary prophylaxis was associated with death (log rank: P < 0.001) with a median survival of 0.9 months in patients without prophylaxis (2 of 57; 4%), compared to 22 months in patients with prophylaxis (55 of 57; 96%). Similarly, in patients with BCLC C and D, and despite their dismal prognosis, lack of secondary prophylaxis was also associated with death (log rank: P < 0.001) with a median survival of 0.7 months (24 of 71; 34%), compared to 3 months in patients who had secondary prophylaxis (47 of 71; 66%; Fig. 4).
A second model was performed to analyze the predictors of death in the subpopulation of patients with HCC and, specifically, the effect of the use of secondary prophylaxis, taking into account BCLC classification stage. In this multivariate analysis, Child-Pugh score, PVT, BCLC classification, and use of secondary prophylaxis remained independent predictors of death (Table 5B). When the independent predictors of failure of secondary prophylaxis were evaluated, only BCLC classification (hazard ratio [HR]: 1.78; 95% confidence interval [CI]: 1.23-2.59), presence of PVT (benign HR: 1.70; 95% CI: 0.61-4.74; malignant HR: 4.62; 95% CI: 1.96-10.90), and use of secondary prophylaxis (HR, 0.33; 95% CI: 0.14-0.75) were independently associated with outcome.
Table 5. Multivariate Predictors of Death Among HCC Patients
Cox's multivariate stepwise regression analysis was used to evaluate the independent predictors of death (n = 99). Variables initially included in the multivariate analysis were BCLC (classified in two categories: 0, A, and B and C and D), previous decompensation, secondary prophylaxis, etiology, Child-Pugh score, and PVT.
Lack of secondary prophylaxis
Taking into account that the differences in the use of secondary prophylaxis were mainly in patients with BCLC C and D, further analysis was performed to compare these patients with and without prophylaxis (see Supporting Table 1). Patients who received no prophylaxis had more-severe liver disease, as shown by greater Child-Pugh score and MELD score, although there were no differences in severity of the HCC, as shown by the proportion of patients with BCLC C or D, PVT, or metastasis.
In this study, a significantly lower survival rate was observed in patients who had HCC at the time of bleeding than patients who did not have HCC, despite the fact that patients were matched for Child-Pugh class and age. This issue is of utmost interest because many studies that evaluated the treatment of acute bleeding episode and prophylaxis of rebleeding had excluded patients with HCC.[12-25] Furthermore, given the increasing incidence of HCC, as a result of rising hepatitis C virus (HCV)-associated advanced liver disease, which is expected to peak in 2020, HCC and VB are an increasingly common clinical problem that clinicians have to deal with. On the other hand, with further improvement in the management of patients with HCC with survival benefit,[33-37] these patients have more probabilities to present with complications of ESLD. A previous study based on ICD-9 diagnostic codes suggested similar results, although as a result of the design of the study, no in-depth analysis could be performed.
Interestingly, patients with HCC were less likely to have secondary prophylaxis than patients without HCC, and there was a trend for a less-frequent use of standard secondary prophylaxis with combination of beta-blockers and endoscopic band ligation in those patients with HCC. The reason why HCC patients were not offered standard therapy is unclear from this study. It is likely that this was because of the assumption, by the attending physician, that this would not result in a clinical benefit. This is also suggested by the fact that patients with HCC without secondary prophylaxis seemed to have more-severe liver disease. However, because lack of secondary prophylaxis was associated with a greater probability of failure and death in models adjusted for severity of liver disease, our results support offering patients with HCC the same treatment after variceal bleeding because it is done for patients without HCC. Although there were no differences in rebleeding rate after 6 weeks, when comparing HCC to non-HCC patients, more patients with HCC died in this period. Indeed, most patients with HCC who died, died of progressive tumoral disease and decompensated liver disease. In addition, when the specific predictors of failure of secondary prophylaxis and death were evaluated in patients with HCC, including BCLC classification, use of secondary prophylaxis had an independent protective effect on the development of rebleeding and death, further suggesting that use of this treatment should be prolonged as long as the clinical condition of the patient allows it.
Despite the fact that the groups were matched by Child-Pugh class and had similar MELD score, patients with HCC had more frequently previous decompensation than patients without HCC. Belonging to the compensated or decompensated phase of the liver disease is of utmost relevance, given the well-known survival differences between these groups. Indeed, after introduction of MELD score, it had been remarked that different survival rates could be noted in patients with the same MELD score according to the presence or absence of clinical decompensation. In the present study, it should be underlined that from the moment they experience VB, all patients are in the decompensated phase. For this reason, this variable was not chosen initially as a matching variable. Also, as expected, patients with HCC had more commonly a viral etiology of their liver disease. Viral etiology has been identified as a negative prognostic factor for 5-day failure in AVB. Given the possible confusion that these variables could introduce, they were included in the multivariate analysis. On multivariate analysis, both the etiology of liver disease and the presence of previous decompensation were not identified as independent predictors of survival.
PVT was also distributed unevenly between patients with HCC and control patients. This variable was significantly associated with outcomes of VB and survival. Previous studies have associated the presence of PVT with negative outcomes in VB. Interestingly, the prognostic information derived from the presence of PVT was independent from the BCLC classification.
Among patients with HCC, survival was mainly influenced by disease stage, best described by the BCLC classification. So, patients in class C and D had a much greater likelihood of dying within 6 months (79%), compared to class 0, A, and B (14%). Nevertheless, lack of secondary prophylaxis was an independent predictor of death, taking into account BCLC classification. Therefore, use of secondary prophylaxis in these patients, even in those with the most advanced tumoral disease (BCLC C and D), had survival advantages. Logically, patients with less-advanced tumors are the ones who have the most to benefit from the use of secondary prophylaxis. Physicians taking care of patients with advanced HCC after a VB episode should individualize therapies according to clinical practice, common sense, and patient needs. Some may judge that the survival benefit in these BCLC C and D patients who received secondary prophylaxis is not clinically relevant (average, 3 months) and that more-interventional therapies (banding ligation) should be avoided, taking into account the possible adverse effects. Nevertheless, this survival benefit is similar to the survival benefit offered with sorafenib treatment in BCLC C patients, which also has side effects, which may affect quality of life. The present study, showing a global survival effect of prophylaxis patients with advanced HCC, provides further evidence to indicate prophylaxis in this subgroup of patients as long as their clinical condition allows them to do so.
There are several setbacks to the study. Some patients with very advanced HCC and UGI bleeding were not included in the study because no endoscopy was performed. This could lead to some bias in the results, because it is probable that these patients who were not included would be the ones who would be most likely to die. However, the decision to exclude these patients from the study was based on several reasons. First, although suspected, the cause of the bleeding was not proven because endoscopy was not performed. It is well established that approximately one third of UGI bleeding episodes in patients with cirrhosis are the result of other causes, rather than esophageal varices.[40, 41] Second, most likely, the patients who would not receive endoscopy would probably be the sickest ones and therefore with the most dismal outcome. Therefore, inclusion of these patients in the analysis might further enhance the differences in the outcomes of VB in patients with and without HCC. Furthermore, and although it seems that patients with HCC without secondary prophylaxis were more sick than the ones who received secondary prophylaxis, which may have influenced the physician's opinion, it could be that there are other factors that influenced this decision that are not included in the analysis. Unfortunately, the study design does not allow analysis of the effect of sorafenib treatment on variceal bleeding. It has been established, both in animal and human studies, that sorafenib has a portal hypotensive effect, perhaps through an inhibition of angiogenesis.[42, 43] Therefore, there could be an effect of the administration of this drug on the outcomes. In the present study, sorafenib was administered exclusively to patients with advanced HCC; therefore, it is logical to speculate that lack of sorafenib could further worsen the outcome of these patients, who already have a dismal prognosis. Another limitation of the study is the uneven distribution of the etiologies among patients with and without HCC. Although, on multivariate analysis, viral disease was not identified as an independent predictor of death (and therefore nonviral disease, which was mainly alcohol, was not identified as a predictor of survival), it could be that non-HCC patients with alcoholic liver disease ceased alcohol consumption after the VB and therefore had a better outcome. Finally, the design of this study does not allow evaluation of the effect of VB in the natural history of HCC.
In conclusion, patients with HCC with VB have worse outcomes than patients without HCC. These differences are only partially explained by differences in secondary prophylaxis measures, as in patients with variceal hemorrhage and HCC. Use of secondary prophylaxis has survival benefit in patients with HCC, irrespective of BCLC stage.
Appendix 1: Patient Distribution According to Centers