The Delta4DVH Anatomy 3D quality assurance (QA) system (ScandiDos), which converts the measured detector dose into the dose distribution in the patient geometry was evaluated. It allows a direct comparison of the calculated 3D dose with the measured back-projected dose. In total, 16 static and 16 volumetric-modulated arc therapy (VMAT) fields were planned using four different energies. Isocenter dose was measured with a pinpoint chamber in homogeneous phantoms to investigate the dose prediction by the Delta4DVH Anatomy algorithm for static fields. Dose distributions of VMAT fields were measured using GAFCHROMIC film. Gravitational gantry errors up to 10° were introduced into all VMAT plans to study the potential of detecting errors. Additionally, 20 clinical treatment plans were verified. For static fields, the Delta4DVH Anatomy predicted the isocenter dose accurately, with a deviation to the measured phantom dose of 1.1% ± 0.6%. For VMAT fields the predicted Delta4DVH Anatomy dose in the isocenter plane corresponded to the measured dose in the phantom, with an average gamma agreement index (GAI) (3 mm/3%) of 96.9± 0.4%. The Delta4DVH Anatomy detected the induced systematic gantry error of 10° with a relative GAI (3 mm/3%) change of 5.8% ± 1.6%. The conventional Delta4PT QA system detected a GAI change of 4.2%± 2.0%. The conventional Delta4PT GAI (3 mm/3%) was 99.8% ± 0.4% for the clinical treatment plans. The mean body and PTV-GAI (3 mm/5%) for the Delta4DVH Anatomy were 96.4% ± 2.0% and 97.7%± 1.8%; however, this dropped to 90.8%± 3.4% and 87.1% ± 4.1% for passing criteria of 3 mm/3%. The anatomy-based patient specific quality assurance system predicts the dose distribution correctly for a homogeneous case. The limiting factor for the error detection is the large variability in the error-free plans. The dose calculation algorithm is inferior to that used in the TPS (Eclipse).