3396 - Clinical and Dosimetric Predictors for Radiation-Induced Acute Upper Gastrointestinal Bleeding in Patients with Hepatocellular Carcinoma Undergoing Proton or Photon Radiation Therapy

Purpose/Objective(s): This study aimed to identify clinical and dosimetric predictors of post-irradiation acute grade = 3 upper gastrointestinal bleeding (aUGIB³⁺) in patients with hepatocellular carcinoma (HCC) undergoing proton or photon radiation therapy (RT).

Materials/Methods: We retrospectively analyzed 379 patients with HCC who were treated with proton (n = 306) or photon (n = 73) RT at Linkou Chang Gung Memorial Hospital between 2014 and 2021. The primary endpoint was aUGIB³⁺, defined as upper gastrointestinal (UGI) hemorrhage requiring blood transfusion, invasive intervention, or hospitalization (grade 3) or being complicated by life-threatening consequences or requiring urgent operative intervention (grade 4) within four months post-RT. The UGI structure was delineated from the lower third of the esophagus to the fourth portion of the duodenum, and UGI bleeding sites were confirmed via panendoscopy. Radiation doses were converted to equivalent doses in 2 Gy (or cobalt Gy equivalent, CGE) per fraction (EQD2; a/ß = 10 Gy or CGE). Dosimetric indices, including mean dose, maximum dose, aVX (absolute volume receiving = X Gy or CGE), VX (percentage of volume receiving = X Gy or CGE) of the UGI tract, along with clinical parameters, were analyzed for their correlation with aUGIB³⁺ incidence using logistic regression.

Results: 27 patients (7%) experienced aUGIB³⁺, including grade 3 in 20 cases and grade 4 in 7 cases. Among them, esophageal, gastric, and duodenal hemorrhage was documented in 9, 17, and 10 cases, respectively. On multivariate analysis, independent predictors of aUGIB³⁺ included V15 (P < 0.001), advanced portal vein thrombosis (Vp3/4 PVT, P = 0.025), and splenomegaly (splenic volume > 300mL, P = 0.044). Mean and maximum UGI doses were significantly associated with a higher incidence of aUGIB³⁺ in univariate analysis but not in the multivariate model. Notably, V15 of the UGI tract demonstrated the highest predictive value among all dosimetric indices in receiver operating characteristic (ROC) analysis, outperforming aV15, as well as mean and maximum UGI doses. In patients without splenomegaly and Vp3/4 PVT, aUGIB³⁺ incidence was 0.7%, 4.8%, and 12.5% for V15 <10%, 10-25%, and >25%, respectively (P = 0.036). Among those with splenomegaly or Vp3/4 PVT, the corresponding incidences were 3.1%, 19.4%, and 38.2% (P < 0.001). Importantly, patients treated with proton RT had significantly lower V15 of the UGI tract (6% vs. 22%, P < 0.001) compared to those undergoing photon RT, along with a significantly lower risk of aUGIB³⁺ (4% vs. 19%, P < 0.001).

Conclusion: In the context of HCC RT, V15 of the UGI tract independently predicts aUGIB³⁺. In patients without or with splenomegaly or Vp3/4 PVT, V15 of the UGI tract <25% and <10%, respectively, was associated with minimal risk of aUGIB³⁺, whereas >25% and >10% were associated with excessive risk. Proton RT mitigates low to intermediate dose-bath associated with photons and reduces V15 of the UGI tract, thereby significantly lowering the risk of aUGIB³⁺.