A quality assurance (QA) procedure was developed to evaluate the congruence between the cone-beam computed tomography (CBCT) image center and the radiation isocenter on a Varian Trilogy linac. In contrast to the published QA procedures, this method did not require a ball bearing (BB) phantom to be placed exactly at the radiation isocenter through precalibrated room lasers or light field crosshairs. The only requirement was that the BB phantom be in a stationary position near the radiation isocenter during the image acquisition process. The radiation isocenter was determined with respect to the center of the BB using a Winston-Lutz test. The CBCT image center was found to have excellent short-term positional repro-ducibility (i.e., less than 0.1 mm of wobble in each of the x (lateral), y (vertical), and z (longitudinal) directions) in 10 consecutive acquisitions. Measured over a seven-month period, the CBCT image center deviated from the radiation isocenter by 0.40 ± 0.12 mm (x), 0.43 ± 0.04 mm (y), and 0.34 ± 0.14 mm (z). The z displacement of the 3D CBCT image center was highly correlated (ρ = 0.997) with that of the 2D kV portal image center. The correlation coefficients in the x and y directions were poor (ρ = 0.66 and -0.35, respectively). Systematic discrepancies were found between the CBCT image center and the 2D MV, kV portal image centers. For the linear accelerator studied, we detected a 0.8 mm discrepancy between the CBCT image center and the MV EPID image center in the anterior-posterior direction.This discrepancy was demonstrated in a clinical case study where the patient was positioned with CBCT followed by MV portal verification. The results from the new QA procedure are useful for guiding high-precision patient positioning in stereotactic body radiation therapy.