For a study, researchers sought to examine the on-treatment alpha-fetoprotein (AFP) response as a potential replacement biomarker of prognosis for the combo therapy in an exploratory investigation. Patients were divided into responders and nonresponders, as determined by the independent review facility-assessed RECIST (IRF-RECIST) version 1.1, and patients with disease control and primary progressors. These distinctions were made using data from Group A of the phase Ib GO30140 trial. Investigators used these cutoffs to distinguish patients based on overall survival (OS) and progression-free survival (PFS) per IRF-RECIST 1.1 and best-confirmed response per IRF-RECIST 1.1 using independent data from the atezolizumab+bevacizumab arm of the phase III IMbrave150 study. To identify responders and those with disease control at 6 weeks, they set AFP cutoffs of more or around 75% decrease and less than or around 10% increase from baseline, respectively. In GO30140, these cutoffs demonstrated good sensitivity and specificity. For the AFP cutoff of more or around 75% decline, sensitivity was 0.59, specificity was 0.86, and for the AFP cutoff of less than 10% rise, specificity was 0.44. Particularly in patients with hepatitis B virus etiology, both AFP cutoffs were linked to the prolonged OS and PFS (HR<0.5; P<0.01). A promising surrogate biomarker of prognosis for HCC patients receiving atezolizumab+bevacizumab is AFP response at 6 weeks into treatment.