Varices and Hemorrhagic Events during Atezolizumab-Bevacizumab Treatment: Reassurance with Caveats for Prognosis

The widespread use of atezolizumab and bevacizumab (A/B), the first immune checkpoint inhibitor-based regimen approved for hepatocellular carcinoma (HCC), has confirmed its efficacy but has also raised safety concerns in real-world practice, particularly regarding bevacizumab-related bleeding.^1-4 In a recent issue of Gut and Liver, Park et al.⁵ provide important evidence addressing whether variceal status and bevacizumab-related bleeding events influence disease progression or overall survival (OS) in patients receiving A/B. Their findings suggest that there is no detrimental effect on survival or disease control, even in the context of increased bleeding risk. The prevalence of bleeding during A/B therapy is non-negligible. In the pivotal IMbrave150 trial, which enrolled a well-defined population with preserved liver function and adequately managed varices, bleeding occurred in 7% of patients and was the most common cause of treatment discontinuation.⁶ In a comprehensive meta-analysis, the pooled incidence of bleeding was reported at 8.42%, with grade ≥3 bleeding occurring in 4.42% of patients. However, importantly, the incidence of grade 3-4 bleeding events did not differ significantly between A/B and tyrosine kinase inhibitors.⁷,8 Furthermore, multicenter real-world data indicated that the overall rate of bleeding with A/B was not significantly higher than that with lenvatinib.⁹ Therefore, while a higher prevalence of bleeding is evident during A/B treatment, the findings presented by Park et al.⁵ suggest that such events do not compromise survival, reinforcing the position of A/B as a reasonable therapeutic choice in appropriately selected patients.

The authors demonstrate that the presence of varices is associated with a significantly increased risk of bleeding, with a cumulative incidence of 32.6% at 18 months in patients with varices compared with 0% in those without varices. In multivariable analysis, varices were an independent predictor of bleeding events (adjusted hazard ratio, 13.56). In other studies, impaired liver function, main portal vein invasion, and a history of bleeding were identified as significant predictors of bleeding complications.¹⁰ Identifying risk factors for bleeding is crucial, as adequate pretreatment and vigilant monitoring can mitigate the negative prognostic impact of hemorrhagic complications, allowing patients to maintain the survival benefit of A/B therapy.

Despite these insightful findings, several limitations warrant consideration. The relatively small sample size (n=123) limits the statistical analyses, particularly for infrequent but clinically consequential bleeding events. In addition, although inverse probability of treatment weighting was employed, residual imbalance remained in key clinical variables, with standardized mean differences exceeding 0.30 for both Barcelona Clinic Liver Cancer stage and Child-Pugh score, suggesting incomplete covariate balance. Patients in the variceal group appeared to have a relatively lower tumor burden, which may partially explain an OS benefit despite higher rates of bleeding and treatment discontinuation. This imbalance is further compounded by differences in baseline liver function, as patients with varices tended to have more impaired hepatic reserve. Given that these factors are well-established prognostic indicators associated with survival, the findings of this study need validation in larger, multicenter cohorts to confirm the reproducibility and external validity of the reported associations. Supplementary Table 1 of Park et al.⁵ reveals that a substantial proportion of patients permanently discontinued A/B following bleeding events. Among the 15 patients who experienced bleeding, nine discontinued A/B, corresponding to approximately 11% of the entire variceal group. This observation warrants particular caution in interpreting the association between bleeding events and favorable oncologic outcomes. Notably, despite this relatively high treatment discontinuation rate, time to progression remained longer in the variceal group compared with the non-variceal group, a finding that raises the possibility of confounding by baseline disease characteristics rather than a direct therapeutic effect. To reinforce the results of this study, additional information regarding post-bleeding treatment trajectories, such as whether patients who resumed systemic therapy did so immediately or after a prolonged interruption, and whether alternative treatments were initiated during this interval, would be helpful.

Furthermore, Supplementary Fig. 1 of Park et al.⁵ demonstrates a trend toward increased bleeding risk with worsening liver function, raising the possibility that a proportion of bleeding events may be attributable not only to bevacizumab-related toxicity but also to progressive hepatic decompensation during continued A/B therapy. Consequently, it remains difficult to disentangle bleeding driven by bevacizumab treatment from bleeding related to the natural history of cirrhosis and declining liver function. The short median follow-up duration (11.1 months) also constrains the evaluation of long-term survival outcomes. In the context of an evolving therapeutic landscape for HCC with increasing availability of subsequent lines of systemic therapy, longer follow-up will be critical to determine whether the observed neutral effect of bleeding events on OS persists over time. Taken together, these considerations underscore the need for cautious interpretation of bleeding events as a surrogate marker of treatment efficacy. Unresolved questions regarding the optimization of A/B therapy, such as the optimal timing for withholding and resuming bevacizumab and determining whether and when to switch to alternative systemic regimens after a bleeding event, remain to be elucidated.