Disclosed are a PIVKA-II measurement method that achieves better serum-plasma correlation than conventional methods, and a reagent and a kit therefor. The PIVKA-II measurement method according to the present invention comprises measuring PIVKA-II in a sample by an immunoassay using a mixture of an anti-F1 antibody that specifically binds to prothrombin fragment F1 or an antigen-binding fragment thereof and an anti-F2 antibody that specifically binds to prothrombin fragment F2 or an antigen-binding fragment thereof; and an anti-PIVKA-II antibody that specifically binds to PIVKA-II or an antigen-binding fragment thereof.