Multicenter study of bleeding and thromboembolic events with durvalumab tremelimumab vs. atezolizumab and bevacizumab in advanced HCC

Background & Aims: Atezolizumab/bevacizumab (A/B) and durvalumab ± tremelimumab (DT/D) are preferred first-line regimens for the treatment of hepatocellular carcinoma (HCC). Although both therapies share an anti-PD-L1-backbone, only the A/B-combination is targeting vascular-endothelial-growth-factor (VEGF). This difference may have safety-implications, as the anti-VEGF component may elevate the risk of bleeding and thromboembolic events. This multicenter study analyzes bleeding or thromboembolic events in patients receiving therapy with DT/D, compared with A/B. Methods: Patients with HCC treated with A/B exclusively before the approval of DT/D, or treated with DT/D, were recruited from 11 tertiary centers and analyzed retrospectively. Information on baseline characteristics was collected. Kaplan–Meier analyses and multivariate Cox regression modeling was used to compare event rates. Results: This study included 490 patients (n = 165 DT/D; n = 325 A/B). The DT/D group showed a higher prevalence of gastroesophageal varices and increased spleen size. Median follow-up time was similar in both groups. A total of 74 patients (22.8%) treated with A/B showed a bleeding event of any grade vs. 22 patients (13.3%) treated with DT/D (p = 0.016). High-grade ≥3 bleeding episodes were more frequent in the A/B group with 47 patients (14.5%) vs. 12 patients (7.3%) in the DT/D group (p = 0.027). Kaplan–Meier analysis showed significantly shorter time to first bleeding (p = 0.037) with A/B as compared with DT/D. Multivariate regression confirmed that treatment with A/B was independently associated with an increased bleeding risk (p = 0.010). Variceal bleeding and thromboembolic events did not differ significantly between both groups. Conclusions: Overall and high-grade bleeding events were more frequent and occurred earlier in patients treated with A/B vs. DT/D, whereas variceal bleeding, grade 4–5 bleeding, and thromboembolic events did not differ significantly between groups. These findings should be confirmed in randomized trials or, if not feasible, through meta-analyses to provide more robust evidence for treatment decisions in advanced HCC. Impact and implications: This multicenter retrospective study was scientifically justified by the need to clarify whether the addition of VEGF inhibition to PD-L1–based immunotherapy confers differential bleeding or thromboembolic risks in patients with advanced HCC. The results are particularly relevant for clinicians and trialists, as they demonstrate a higher incidence and earlier onset of overall and high-grade bleeding events with atezolizumab/bevacizumab compared with durvalumab ± tremelimumab, despite similar rates of variceal bleeding and thromboembolic events. These findings may inform treatment selection, risk stratification, and surveillance strategies in routine practice, especially in patients with portal hypertension or other bleeding risk factors. Given the retrospective design and baseline-imbalances between treatment groups, the results should be interpreted cautiously and warrant confirmation in prospective randomized studies or, if not feasible, through meta-analyses.