Usefulness of alpha-fetoprotein response in patients treated with sorafenib for advanced hepatocellular carcinoma

Background & Aims: Tumor shrinkage has been considered a fundamental surrogate efficacy measure for new cancer treatments. However, in patients treated with sorafenib for advanced hepatocellular carcinoma (HCC), tumor shrinkage rarely accompanies increased survival, thereby questioning the prognostic value of imaging-based Response Evaluation Criteria in Solid Tumors (RECIST). We investigated the prognostic usefulness of a decrease in serum alpha-fetoprotein (AFP) and compared it to RECIST. Methods: In HCC patients treated with sorafenib with baseline AFP >20 ng/ml, AFP response was defined as a >20% decrease in AFP during 8 weeks of treatment. Patients were also assessed by RECIST and were categorized as having radiologically proven progressive disease or disease control (consisting of complete or partial responses and stable disease). Comparisons of survival by RECIST and AFP response were corrected for guarantee-time bias by the landmark method. Results: We evaluated 85 patients for AFP response, among them, 82 were also evaluated by RECIST. In the analysis of AFP response, 32 out of 85 patients (37.6%) were responders, whereas 58 out of 82 patients (70.7%) achieved disease control. In landmark analysis, the hazard ratios (HR) for survival according to AFP response and disease control were 0.59 (p = 0.040) and 1.03 (p = 0.913), respectively. In multivariate analysis, only AFP response (HR = 0.52; p = 0.009) and Cancer of the Liver Italian Program dichotomized stage (HR = 0.42; p = 0.002) were prognostic factors of survival. Conclusions: Assessment of AFP response may be considered as an alternative to RECIST to capture sorafenib activity in HCC.