Prognostic Impact of Varices and Hemorrhagic Events in Hepatocellular Carcinoma Patients Treated with Atezolizumab Plus Bevacizumab.

INTRODUCTION
The IMbrave150 trial has demonstrated that patients treated with atezolizumab plus bevacizumab have significantly better overall survival (OS) and progression-free survival outcomes than those treated with sorafenib, leading to an improvement in life expectancy for patients with unresectable hepatocellular carcinoma (HCC) by expanding first-line systemic therapy options from tyrosine kinase inhibitors to immunotherapy.1 Recently, tremelimumab plus durvalumab has been approved for use in unresectable HCC and found to significantly improve OS compared to sorafenib.2,3 As treatment options for unresectable HCC expand from tyrosine kinase inhibitors to immunotherapies,4 it is necessary to establish specific prognostic factors for optimal therapeutic efficacy in immunotherapy.
Currently, atezolizumab plus bevacizumab is considered an effective first-line treatment for advanced HCC. However, its use is often exercised with caution in patients at high risk of bleeding, particularly those with varices related to portal hypertension.1,5 In addition, advanced liver cirrhosis may also lead to the bleeding risk elsewhere owing to lower platelet count and prolonged prothrombin time.6,7 While the IMbrave150 trial included only patients with fully treated or absent esophageal or gastric varices, delaying treatment with atezolizumab and bevacizumab until varices are fully addressed can be challenging in practice.8,9 Recent efforts have thus focused on administering atezolizumab and bevacizumab concurrently with variceal treatment, following the Baveno VII recommendations for portal hypertension.10,11
Given this context, accurately assessing bleeding risk during atezolizumab plus bevacizumab treatment is an important step for managing patients with unresectable HCC. Additionally, the prognostic effect of varices on patients treated with atezolizumab plus bevacizumab has not been fully addressed yet. Therefore, this study aimed to evaluate prognostic effects of varices assessed by either computed tomography (CT) or esophagogastroduodenoscopy (EGD) on bleeding risk, treatment responses, and OS in patients with unresectable HCC treated with atezolizumab plus bevacizumab.

MATERIALS AND METHODS

1. Study population
Consecutive patients who received atezolizumab plus bevacizumab as the first-line systemic therapy for treating advanced HCC between October 2020 and July 2022 at Seoul National University Hospital (Seoul, South Korea) were retrospectively reviewed. HCC was diagnosed radiologically and/or histologically according to the latest updated guidelines.12,13 This study conformed to ethical guidelines of the World Medical Association Declaration of Helsinki. It was approved by the Institutional Review Board (IRB) of Seoul National University Hospital (IRB number: H-2311-073-1483). The requirement to obtain informed consent from patients was waived by the IRB due to the retrospective nature of this study.
Patients were included in the study based on the following criteria: (1) age >18 years; (2) unequivocal diagnosis of advanced HCC based on radiological and/or histological findings; (3) treatment with atezolizumab plus bevacizumab as first-line systemic therapy; (4) endoscopic evaluation of varices prior to administration of atezolizumab plus bevacizumab; (5) treatment of high-risk varices according to the Baveno VII consensus prior to administration of atezolizumab plus bevacizumab,11 and (6) Eastern Cooperative Oncology Group performance status of 0 or 1. Patients who met any of the following criteria were excluded: (1) presence of another concurrent malignancy; (2) failure to conduct initial response assessment after administration of atezolizumab plus bevacizumab; and (3) prior recipient of liver transplantation.

2. Treatment regimens
The atezolizumab plus bevacizumab regimen was administered intravenously. Atezolizumab was given at a dose of 1,200 mg. Bevacizumab was then given at a dose of 15 mg/kg. Both were administered on the same day every 3 weeks.14 When an adverse event occurred, treatment was resumed according to the National Comprehensive Cancer Network guidelines.15 Treatment with atezolizumab plus bevacizumab was discontinued in patients who failed to recover from fatal adverse events, showed disease progression, or developed liver failure.12,13,16
For high-risk varices, non-selective beta-blockers (NSBBs) and endoscopic variceal ligation (EVL) are considered equally effective in preventing first and acute variceal bleeding.11,17-19 NSBBs are preferred over EVL in patients with small high-risk varices, while a combination of NSBBs and EVL is recommended for patients with large, red-spotted high-risk varices.11 NSBB included propranolol or carvedilol, with the dosage individually adjusted to maintain a heart rate of 55 to 60 beats per minute and to avoid SBP <90 mm Hg. In cases with a persistently low blood pressure (systolic blood pressure <90 mm Hg or mean arterial pressure <65 mm Hg) or development of hepatorenal syndrome associated with acute kidney injury, the dosage was reduced or discontinued.20 EVL was preferred as the standalone treatment in cases with persistent intolerance or contraindications20 to NSBBs. In patients with high-risk varices, EVL was performed repeatedly until the varices were no longer classified as high risk. It was performed urgently when there were signs of impending variceal rupture or active variceal bleeding. After the occurrence of variceal bleeding, combination therapy with NSBBs and EVL was implemented for secondary prophylaxis until no more varices with high-risk features were detected.11

3. Assessments and outcomes
The presence of varices was defined as the detection of esophageal or gastric varices on EGD (GIF-H290, Olympus Co., Japan) or the identification of esophageal, paraesophageal, gastric, splenic, or umbilical varices on CT imaging. Varices detected on CT were assessed by experienced radiologists following standardized imaging criteria to ensure diagnostic accuracy and consistency. High-risk varices requiring treatment, as observed in EGD, were defined as follows: (1) medium to large varices (>5 mm), or (2) small varices with red wale marks.11,17 Patients treated with EVL for high-risk varices underwent follow-up endoscopic examinations at intervals of 1 to 6 months depending on the severity of their varices. EVL was repeated until no high-risk varices were detected.
The index date was defined as the date on which the patient started treatment with atezolizumab plus bevacizumab. Patients were censored on the earliest of the following dates: the date of the last follow-up, death, or the data cutoff date (December 31, 2023). The primary endpoint was the incidence of the first bleeding event during treatment. Secondary endpoints were OS, time-to-progression (TTP), and disease control rate (DCR). Radiological response to treatment was assessed every 9 to 12 weeks based on dynamic CT or magnetic resonance imaging in accordance with the Response Evaluation Criteria in Solid Tumors, version 1.1 (RECIST 1.1). Tumor markers including serum levels of alpha-fetoprotein and protein induced by vitamin K absence or antagonist-II (PIVKA-II) were measured every 9 to 12 weeks using ARCHITECT i2000SR (Abbott Laboratories, IL, USA).

4. Statistical analysis
Baseline clinical and demographic characteristics were compared between the two groups according to the presence of varices. Categorical variables are presented as frequencies (%) and continuous variables are presented as medians with interquartile ranges. Categorical variables were compared using the Pearson chi-square test or the Fisher exact test while continuous variables were compared using the Student t-test or the Mann-Whitney U test. Inverse probability of treatment weighting (IPTW) was applied to minimize potential confounding in the propensity score calculation by incorporating variables such as age, sex, the fibrosis-4 index, albumin-bilirubin score, international normalized ratio of prothrombin time, and presence or absence of liver cirrhosis, portal vein tumor thrombosis (PVTT), or extrahepatic metastasis.21,22 The Kaplan-Meier estimator and log-rank test were used to analyze time to first bleeding event, OS, and TTP. The Firth method was employed to address issues of separation and small sample sizes in our logistic regression analysis, adjusting the likelihood function to reduce bias in the maximum likelihood estimate.23 For risk factor analysis, multivariate analysis was performed using Cox regression analysis for covariables found to be significant in the univariable analysis or had a low degree of collinearity.
All statistical analyses were performed using R version 4.2.3 (R Foundation for Statistical Computing, Vienna, Austria). p-values less than 0.05 indicated statistically significant differences.

RESULTS

1. Study population
A total of 123 patients were included in this study. EGD showed that 36 patients had varices, including 25 with esophageal varices only, four with gastric varices only, and seven with both esophageal and gastric varices. Of these patients, five received only NSBB treatment, seven underwent only EVL, and seven received both NSBB and EVL as combination therapy. There were no untreated high-risk varices in this study. All patients underwent a contrast-enhanced liver CT prior to treatment, which revealed varices in 78 patients. Among these, 33 patients also had varices detected on EGD, leaving 45 patients with varices identified only on CT. Overall, 81 patients had varices on either EGD or CT, while 42 patients had no varices on either modality (Fig. 1).
The median follow-up duration was 11.1 months (interquartile range, 6.1 to 14.9 months). Baseline characteristics of the study population are shown in Table 1. Before and after IPTW adjustment, there were no significant differences in age (p=0.27) or sex (p=0.56) between the two groups. HCC etiologies, Barcelona Clinic Liver Cancer stages, alpha-fetoprotein, and PIVKA-II levels also showed no significant differences; however, PVTT (p=0.04) and liver cirrhosis (p=0.002) were more prevalent in the varices group. After IPTW adjustment, there were no longer any significant differences between the two groups in any variables, including PVTT and liver cirrhosis (Table 1).

2. Bleeding events
Of the 42 patients without baseline varices, only one experienced bleeding event at 23 months after study enrollment due to esophageal varices developed after starting atezolizumab plus bevacizumab, leading to treatment cessation. The cumulative incidence of bleeding in this group was 0.0% at 6, 12, and 18 months (Fig. 2). Among the 36 patients with varices found on EGD, 10 experienced bleeding events: variceal (n=4); ulcer (n=2); tumor (n=1); hemobilia from a biliary stent (n=1); intracranial hemorrhage (n=1); and unknown cause (n=1). Among these 10 patients, seven had received EVL and/or NSBB prior to treatment initiation due to high-risk varices, of whom six (85.7%) achieved complete recovery and resumed therapy. In contrast, the remaining four patients discontinued treatment due to performance decline. Their cumulative bleeding incidence was 17.1%, 28.6%, and 64.3% at 6, 12, and 18 months, significantly higher than in those without varices (p<0.001) (Fig. 2A and B). In 45 patients with varices detected on CT but not on EGD, four patients experienced bleeding events: variceal (n=2); ulcer (n=1); and tumor (n=1). One continued their treatment after complete recovery, switching to sorafenib due to concomitant disease progression. Meanwhile, three discontinued treatment because of deteriorated performance status following a bleeding event. The cumulative incidence of bleeding in this group was 0.0%, 15.1% and 15.1% at 6, 12, and 18 months, and was also significantly higher than in the non-varices group after IPTW adjustment (p=0.001) (Fig. 2A and B). Detailed information on all bleeding events and their management is provided in Supplementary Table 1. Overall, the cumulative bleeding incidence in 81 patients with varices on either EGD or CT was 7.7%, 21.3%, and 32.6% at 6, 12, and 18 months, respectively, significantly higher than in those without varices (p=0.001) (Fig. 2C).
The varices group had a significantly higher incidence of bleeding events both before IPTW (hazard ratio [HR], 18.06; 95% confidence interval [CI], 2.41 to 2,311; p=0.001) (Fig. 2C) and after IPTW (HR, 7.22; 95% CI, 1.37 to 325.92; p=0.01) (Fig. 2D). Univariable analyses before IPTW showed that the presence of varices, a Child-Pugh score of A6, a high albumin-bilirubin score, and PVTT were significantly associated with bleeding events (Table 2, Supplementary Figs 1-3). However, in multivariate analyses, only the presence of varices remained significant (adjusted HR [aHR], 13.56; 95% CI, 1.79 to 2,311.08; p=0.006).

3. Treatment response and TTP
Among 81 patients with varices on either EGD or CT, 1.2%, 17.3%, 61.7%, and 19.8% achieved complete response, partial response, stable disease, and progressive disease, respectively. Among 42 patients without varices, 7.1%, 21.4%, 26.2%, and 45.2% achieved complete response, partial response, stable disease, and progressive disease, respectively. This difference was statistically significant (p<0.001) (Table 3). Although objective response rates were not significantly different (18.5% in those with varices vs 28.6% in those without varices, p=0.30), the DCR was significantly higher in patients with varices (80.2% vs 54.8%, p=0.006). In addition, DCR was significantly higher in patients who experienced a bleeding event during the atezolizumab plus bevacizumab treatment than in patients who did not (100.0% vs 67.6%, p=0.006) (Table 3). When divided into three groups, both the varices on CT and varices on EGD groups showed a significant increase in DCR compared to the non-varices group (75.6% vs 86.1% vs 54.8%, p=0.007) (Supplementary Table 2).
Disease progression occurred in 77 of 123 patients: 56.8% in the varices group and 73.8% in the non-varices group. The median TTP was 9.0 months (95% CI, 7.2 to 13.1 months) for the varices group and 5.0 months (95% CI, 3.4 to 13.7 months) for the non-varices group (p=0.14) (Supplementary Fig. 4). Before IPTW, TTP differences were not significant (HR, 0.71; 95% CI, 0.45 to 1.12; p=0.14). After IPTW, the varices group had a significantly longer TTP (HR, 0.61; 95% CI, 0.37 to 0.99; p=0.047) (Supplementary Fig. 4B). TTP was also prolonged in patients who experienced bleeding events (HR, 0.40; 95% CI, 0.37 to 0.99; p=0.02 after IPTW) (Supplementary Fig. 5). In multivariate analyses, bleeding events were independently associated with prolonged TTP, even after IPTW (aHR, 0.39; 95% CI, 0.17 to 0.90; p=0.03) (Supplementary Table 3). The result of TTP analysis across three groups was given in Supplementary Fig. 6.

4. Overall survival
During the follow-up period, 77 of the 123 patients died, including 50 (61.7%) in the varices group and 27 (64.3%) in the non-varices group. The majority of deaths were attributed to progression of HCC, accounting for 45 patients (90.0%) in the varices group and all 27 patients (100.0%) in the non-varices group (Supplementary Table 4). The median OS was 11.6 months (95% CI, 9.5 to 14.6 months) in the varices group and 11.8 months (95% CI, 9.3 months to not estimated) in the non-varices group (p=0.75 by log-rank test) (Fig. 3A). The OS between the two groups showed no statistically significant difference before IPTW (HR, 1.08; 95% CI, 0.67 to 1.74; p=0.75). Even after IPTW, OS was not significantly different between the two groups (HR, 0.84; 95% CI, 0.49 to 1.46; p=0.54) (Fig. 3B). Multivariate analyses showed no significant differences in OS between the two groups either before IPTW (aHR, 0.92; 95% CI, 0.56 to 1.51; p=0.74) or after IPTW (aHR, 1.04; 95% CI, 0.62 to 1.73; p=0.89) (Supplementary Table 5). An increase of neutrophil-to-lymphocyte ratio and an elevation of PIVKA-II level were only factors independently associated with decreased OS (Supplementary Table 5). According to bleeding events, the median OS was 12.7 months (95% CI, 11.2 months to not estimated) in the bleeding group and 11.5 months (95% CI, 9.3 to 14.6 months) in the non-bleeding group (p=0.28 by log-rank test) (Supplementary Fig. 7A). Bleeding events and OS after IPTW showed no significant associations (HR, 0.68; 95% CI, 0.35 to 1.33; p=0.26) (Supplementary Fig. 7B). There was also no significant difference in OS among the groups with varices on CT, varices on EGD, and the non-varices group (Supplementary Fig. 8).

5. Safety
Treatment-related adverse events other than variceal bleeding were observed in 14 patients (17.3%) in the varices group and seven patients (16.7%) in the non-varices group, with no statistically significant difference between the groups (p=1.00) (Supplementary Table 6). The most frequently reported adverse event in both groups was hypothyroidism (4 [4.9%] vs 3 [7.1%], p=0.69), followed by dermatitis (3 [3.7%] vs 1 [2.4%], p=1.00) and proteinuria (3 [3.7%] vs 2 [4.8%], p=1.00). Other adverse events, including gastrointestinal perforation, hepatitis, hypertension, and pneumonitis, also did not differ significantly between the groups.

DISCUSSION
Based on results of the IMbrave150 trial, atezolizumab plus bevacizumab has been approved as a first-line treatment for unresectable HCC, following sorafenib.1 This marks the beginning of the immunotherapy era for HCC treatment. Unlike in other cancers, in HCC patients with cirrhosis, varices are highly susceptible to bevacizumab-induced bleeding, which can be life-threatening, and therefore limits the use of atezolizumab plus bevacizumab in clinical practice.8,16 In addition, advanced cirrhosis might lead to the bleeding tendency elsewhere owing to the lower platelet count and prolonged prothrombin time, and it could increase the risk of bevacizumab-induced bleeding during the atezolizumab plus bevacizumab treatment.6,7 However, studies about their impact on OS and treatment response are scarce. Our study assessed varices using both CT and EGD and managed them according to guidelines before treatment. The varices group had a significantly higher incidence of bleeding events, and 85.7% of patients who received proactive management resumed subsequent treatments after recovery. The varices group demonstrated improved TTP and DCR, and multivariate analyses revealed that the presence of varices was the sole predictive factor for bleeding events. Bleeding events also improved DCR and were independently associated with prolonged TTP, but there was no difference in OS compared to cases without bleeding. Therefore, with appropriate management and monitoring, we cautiously suggest that atezolizumab plus bevacizumab can be used in patients with varices, and that bleeding events may indicate improvements in clinical outcomes.
In this study, both the presence of varices and the occurrence of bleeding events prolonged TTP and improved DCR. Bleeding events in this context are more plausibly attributed to bevacizumab-induced on-target vascular effects rather than immune-related mechanisms. Previous studies have identified bevacizumab-induced hypertension as a potential surrogate marker of therapeutic efficacy, demonstrating improved survival outcomes among patients who developed hypertension during treatment.24,25 Although the underlying mechanisms of bleeding and hypertension differ, both may reflect pharmacodynamic effects of anti-vascular endothelial growth factor (VEGF) pathway inhibition. Bevacizumab exerts its antitumor activity by targeting VEGF, thereby disrupting the integrity of tumor-associated microvasculature and potentially leading to vascular manifestations such as bleeding or hypertension.26 Conversely, resistance to bevacizumab has been associated with the compensatory upregulation of alternative pro-angiogenic pathways, resulting in sustained angiogenesis, reduced treatment efficacy, and attenuation of VEGF-related adverse events.27-29 Based on this rationale, the observed bleeding events—though conventionally regarded as adverse—may, in certain cases, indicate active VEGF pathway blockade. However, given the absence of direct mechanistic evidence, this interpretation remains speculative, and further investigation is warranted to determine whether bevacizumab-associated adverse events such as bleeding could serve as surrogate indicators of antitumor response.
The presence of varices and the occurrence of bleeding events were not significantly associated with OS even after IPTW adjustment (p=0.54 and p=0.26, respectively). Previous studies have shown that bevacizumab can increase the risk of bleeding in patients with varices, a life-threatening complication in those with cirrhosis and HCC.1,8,16 However, few studies have addressed its correlation with OS. Findings of this study suggest that when high-risk varices are appropriately treated and monitored, the presence or absence of varices and the occurrence of bleeding do not significantly affect OS. In this study, PIVKA-II levels and neutrophil-to-lymphocyte ratio were the only two factors associated with OS. PIVKA-II, an abnormal prothrombin found in malignant cells, can stimulate tumor growth and angiogenesis, worsening behavior and prognosis in HCC patients.30-32 A recent study has shown that rising PIVKA-II levels during atezolizumab and bevacizumab treatment can independently predict poorer OS.33
Both extraluminal varices (identified only on CT) and intraluminal varices (visible on EGD) share a common pathophysiological basis attributed to portal hypertension.34 Since bevacizumab may increase bleeding risk in both intraluminal and extraluminal varices through the same mechanism, we subdivided the varices group into two categories: those with varices identified on EGD and those detected on CT, to evaluate the differential impact on clinical outcomes. Compared to the non-varices group, both subgroups showed a significantly increased cumulative incidence of bleeding events. The results for TTP, DCR, and OS also showed consistent trends with prior comparisons between varices and non-varices groups. Additionally, previous studies indicate that, beyond traditional markers of clinically significant portal hypertension, extraluminal varices found on CT can be an important predictive indicator for prognosis.35,36 By adjusting for factors that may affect clinically significant portal hypertension, such as PVTT and liver cirrhosis, using IPTW, our study would isolate the impact of varices on clinical outcomes. These findings suggest that extraluminal varices detected on CT, similar to intraluminal varices seen on EGD, may also act as surrogate markers for clinically significant portal hypertension and serve as prognostic factors in HCC patients treated with atezolizumab plus bevacizumab. Given the routine use of CT scans for tumor burden evaluation in clinical practice, particular attention may be warranted for findings detectable exclusively on CT, such as paraesophageal varices, in addition to endoscopic evaluation for varices. These findings could provide additional prognostic information and aid in clinical decision-making.
This study has several limitations. First, the retrospective design, single-center setting, and relatively small sample size inherently introduce the possibility of selection bias and may limit the generalizability of the findings. To address these issues, future prospective, multicenter studies are warranted to enhance the external validity and reduce the risk of selection bias associated with retrospective single-center analyses. Second, the short follow-up period, resulting from the relatively recent introduction of atezolizumab plus bevacizumab in South Korea, may have limited the assessment of long-term outcomes. Extended follow-up will be necessary to more comprehensively evaluate long-term outcomes. To enhance the robustness of our conclusions, it would be beneficial to undertake additional studies featuring extended follow-up durations and larger participant cohorts.
In conclusion, the presence of varices in unresectable HCC patients treated with atezolizumab plus bevacizumab requires caution due to an increased bleeding risk. However, with appropriate treatment and monitoring for varices, the presence of varices and the occurrence of bleeding events were not associated with OS but may indicate a better anticancer effect. Thus, the presence of varices and bleeding events, when managed properly, could serve as positive prognostic factors.