To investigate the impact of different setup methods, vertebral body matching (VM), diaphragm matching (DM), and marker matching (MM), on the dose distribution in proton therapy (PT) for hepatocellular carcinoma (HCC). Materials and Methods: Thirty-eight HCC lesions were studied retrospectively to assess changes in the dose distribution on two computed tomography (CT) scans. One was for treatment planning (1st-CT), and the other was for dose confirmation acquired during the course of PT (2nd-CT). The dose coverage of the clinical target volume (CTV-D 98) and normal liver volume that received 30 Gy relative biological effectiveness (RBE) (liver-V 30) were evaluated under each condition. Initial treatment planning on the 1st-CT was defined as reference, and three dose distributions recalculated using VM, DM, and MM on the 2nd-CT, were compared to it, respectively. In addition, the relationship between the CTV-D 98 of each method and the distance between the center of mass (COM) of the CTV and the right diaphragm top was evaluated. Results: For CTV-D 98 , significant differences were observed between the reference and VM and DM, respectively (P = 0.013, P = 0.015). There were also significant differences between MM and VM and DM, respectively (P = 0.018, P = 0.036). Regarding liver-V 30 , there was no significant difference in any of the methods, and there were no discernable difference due to the different setup methods. In DM, only two out of 34 cases with a distance from right diaphragm top to COM of CTV of 90 mm or less that CTV-D 98 difference was 5% or more and CTV-D 98 was worse than VM were confirmed. Conclusion: Although MM is obviously the most effective method, it is suggested that DM may be particularly effective in cases where the distance from right diaphragm top to COM of CTV of 90 mm or less.