Early Prediction of Response Focused on Tumor Markers in Atezolizumab plus Bevacizumab Therapy for Hepatocellular Carcinoma

SIMPLE SUMMARY: The combination of atezolizumab and bevacizumab was introduced as a first-line therapy for patients with unresectable hepatocellular carcinoma in 2020. Although some patients have shown a treatment response, there have also been those with disease progression. Such cases should be appropriately transitioned to second-line or later treatment. Thus, this study investigated early predictors of response and disease progression categorized into two groups based on a baseline alpha-fetoprotein (AFP) of 20 ng/mL. As a result, we found that changes in AFP and baseline des-gamma-carboxy prothrombin levels were useful predictors of treatment response. Tumor markers are useful in predicting treatment response and prognosis. ABSTRACT: Despite the promising efficacy of atezolizumab plus bevacizumab (atezo/bev), some patients with unresectable hepatocellular carcinoma (HCC) experience disease progression. This retrospective study, which included 154 patients, aimed to evaluate predictors of treatment efficacy of atezo/bev for unresectable HCC. Factors associated with treatment response were examined, focusing on tumor markers. In the high-alpha-fetoprotein (AFP) group (baseline AFP ≥ 20 ng/mL), a decrease in AFP level > 30% was an independent predictor of objective response (odds ratio, 5.517; p = 0.0032). In the low-AFP group (baseline AFP < 20 ng/mL), baseline des-gamma-carboxy prothrombin (DCP) level < 40 mAU/mL was an independent predictor of objective response (odds ratio, 3.978; p = 0.0206). The independent predictors of early progressive disease were an increase in AFP level ≥ 30% at 3 weeks (odds ratio, 4.077; p = 0.0264) and the presence of extrahepatic spread (odds ratio, 3.682; p = 0.0337) in the high-AFP group and up-to-seven criteria, OUT (odds ratio, 15.756; p = 0.0257) in the low-AFP group. In atezo/bev therapy, focusing on early AFP changes, baseline DCP, and tumor burden of up-to-seven criteria are useful in predicting response to treatment.