The purpose of the study was to investigate the dosimetric effect defining the body structure with various Hounsfield unit (HU) threshold values on the dose distributions of volumetric-modulated arc therapy (VMAT) plans. Twenty patients with prostate cancer and twenty patients with head and neck (H&N) cancer were retrospectively selected. For each patient, the body structure was redefined with HU threshold values of -180 (Body180), -350 (Body350), -700 (Body700), and -980 (Body980). For each patient, dose-volumetric parameters with those body structures were calculated using identical VMAT plans. The differences in dose-volumetric parameters due to the varied HU threshold values were calculated. For the prostate boost target volume, the maximum dose, mean dose, D95%, and D5% with Body180 were higher than those with Body980 by approximately 0.7% (p < 0.001). For H&N target volumes, the changes in D95% of the targets receiving 67.5 Gy, 54 Gy, and 48 Gy between Body180 and Body980 were -1.2%, -0.9%, and -1.2%, respectively (p < 0.001). The differences were larger for H&N VMAT plans than for prostate VMAT plans due to the inclusion of an immobilization device in the irradiated region in H&N cases. To apply all attenuating materials to dose calculation, the body structure would be defined with -980 HU. Otherwise, systematic error of about 1%, resulting in underdosage of the target volume, can occur.