Response Prediction Model of Atezolizumab plus Bevacizumab in Patients with Unresectable Hepatocellular Carcinoma: Multicenter Retrospective Study

Introduction: Atezolizumab plus bevacizumab (ATZ+BV) treatment has become the first-line regimen for unresectable hepatocellular carcinoma (u-HCC). Prediction of response to it might be clinically beneficial. Using peripheral blood parameters, we aimed to construct a prediction model for ATZ+BV treatment.

Methods: Clinical records of 119 patients with u-HCC treated by ATZ+BV were retrospectively analyzed. The primary outcome measurement was defined as any-size reduction at the initial image evaluation. Using baseline values of peripheral blood parameters, a prediction model was constructed by univariate and multivariate logistic regression analysis. Validation was performed internally by bootstrap method.

Results: The primary outcome was achieved in 46 patients. Univariate analysis showed that C-reactive protein (CRP), alpha-fetoprotein (AFP), des-gamma-carboxy prothrombin (DCP), neutrophil-to-lymphocyte ratio (NLR), and platelet-to-lymphocyte ratio (PLR) were possible predictors. CRP and DCP, and NLR and PLR had correlation (correlation coefficient >0.3), so we used CRP and NLR as representative factors, respectively. Multivariate analysis constructed the following prediction model: Logit = 1.62-0.61×[CRP] -0.38×[Log10AFP] -0.37×[NLR]. Bootstrapped median (95% confidence interval) of coefficients of CRP, Log10AFP, NLR were -0.64 (-1.46 ∼ -0.11), -0.40 (-0.82 ∼ -0.03), and -0.38 (-0.74 ∼ -0.05), respectively. The area under the receiver operating characteristic curve (95% confidence interval) was 0.73 (0.60-0.80). Median overall survival of the favorably and unfavorably predicted groups were 17.0 and 11.0 months (p = 0.03), respectively.

Discussion: In patients with u-HCC treated by ATZ+BEV, a prediction model constructed using baseline values of CRP, AFP, and NLR had impact on any-size reduction at the initial image evaluation and on prognosis.