Editorial: Variceal Bleeding in Patients Receiving Atezolizumab–Bevacizumab for Hepatocellular Carcinoma—Don't Ignore the Risk Factors! Authors' Reply

Abstract

We appreciate the interest of Salimi and Liu in our study on variceal bleeding (VB) risk in patients receiving atezolizumab–bevacizumab (Atezo–Bev) for hepatocellular carcinoma (HCC) and their insightful comments highlighting the importance of risk stratification [1, 2]. We fully concur that VB risk factors should not be overlooked and that a patient-specific risk approach is crucial in clinical decision-making.

One key finding from our study that we would like to emphasise is that 14 of 45 patients (31%) who experienced VB had no varices observed on pretreatment oesophagogastroduodenoscopy (OGD) [2]. Of note, 12 of these patients (85.7%) had portal vein invasion (PVI), with seven cases (50%) involving extensive main PVI. Our multivariable analysis also identified PVI as a strong predictor of VB risk. This finding suggests that even patients classified as low risk based on standard OGD criteria may still face a substantial risk of VB due to tumour-related factors, particularly PVI. Therefore, the traditional stratification approach based solely on OGD findings may be inadequate in this population, necessitating a more comprehensive assessment incorporating tumour burden and vascular involvement.

Another important point is the observed higher incidence of VB (22% vs. 3%) in patients who received prophylaxis compared to those who did not. As noted by Salimi and Liu, this should be interpreted with caution. Rather than suggesting a lack of efficacy of prophylaxis, this likely reflects a selection bias, where patients receiving prophylaxis were inherently at higher risk of VB at baseline. As we discussed in our paper, patients who received prophylaxis had more significant VB risk factors, including a history of gastrointestinal bleeding, extensive PVI and advanced liver cirrhosis. Thus, the difference in VB incidence does not necessarily indicate prophylaxis failure but rather highlights the need for improved risk stratification and timely intervention in the highest risk patients.

In addition, we initially sought to develop a patient-specific VB risk prediction model based on our findings. However, due to the relatively limited number of VB events and the lack of external validation, constructing a statistically robust model proved challenging. We believe that future efforts incorporating a larger sample size and multicentre cohorts will be essential in establishing this model. A validated risk stratification model could significantly enhance clinical decision-making, allowing for more precise VB risk prediction and personalised prophylaxis strategies in patients undergoing Atezo–Bev treatment.

Given the lack of strong consensus guidelines for VB prevention during the Atezo–Bev treatment [3], we hope our study contributes to the development of evidence-based risk prediction and prophylaxis strategies. Based on our findings, patients with risk factors such as low platelet count, PVI, a history of gastrointestinal bleeding and varices needing treatment should undergo vigilant monitoring for VB.