BACKGROUND AND PURPOSE: To demonstrate the feasibility of back-projection portal dosimetry for accurate 3D dosimetric verification of volumetric-modulated arc therapy (VMAT), pre-treatment as well as in vivo. MATERIALS AND METHODS: Several modifications to our existing approach were implemented to make the method applicable to VMAT: (i) gantry angle-resolved data acquisition, (ii) calculation of the patient transmission, (iii) compensation for detector 'flex' and (iv) 3D dose reconstruction and evaluation. RESULTS: Planned and EPID-(Electronic Portal Image Detector)-reconstructed dose distributions show good agreement for pre-treatment verification of two prostate, a stereotactic lung and a head-and-neck VMAT plan and for in vivo verification of VMAT treatments of prostate and lung cancer. Averaged over pre-treatment verifications, planned and measured isocentre dose ratios were -1.2% (range [-4.7%,1.8%]). 3D gamma analysis (3% maximum dose, 3mm) revealed mean gamma gamma(mean)=0.37 [0.34,0.39], maximum 1% gamma gamma(1%)=0.72 [0.66,0.81] and percentage of points with gamma1 P(gamma)(1)=99% [97%,100%]. For in vivo verification, the average isocentre dose ratio was -1.2% [-0.8%,-1.7%], gamma(mean)=0.52 [0.40,0.64], gamma(1%)=0.92 [0.76,1.08] and P(gamma)(1)=96% [93%,100%]. CONCLUSIONS: Our portal dosimetry method was successfully adapted for verification of VMAT treatments, pre-treatment as well as in vivo.