A method was evaluated to reconstruct the 3D dose distribution in patients using their planning CT-scan in combination with a Monte Carlo calculation, and the energy fluence of the actual treatment beams measured pre-treatment with an EPID without the patient or a phantom in the beam. Nine plans of lung cancer patients treated with a 3D conformal technique, calculated using a simple convolution algorithm (CA), as well as five IMRT treatments of head-and-neck cancer patients, calculated with a more advanced superposition algorithm (SA), were verified. Differences between planned and reconstructed dose distributions were quantified in terms of DVH parameters. For the lung cancer group, differences between the reconstructed mean PTV dose and the values calculated with the TPS were 5.0+/-4.2% (1SD) and -1.4+/-1.5% for the CA and SA algorithm, respectively. No large differences in the lung and spinal cord DVH parameters were found. For the IMRT treatments, the average dose differences in the PTV were generally below 3%. The reconstructed mean parotid gland dose was 3.2+/-1.2% lower, while the maximum spinal cord dose was on average 3.1+/-1.9% higher. EPID dosimetry combined with 3D dose reconstruction is a useful procedure for patient-specific QA of complex treatments. DVH parameters can be used to interpret the dose distribution delivered to the patient in the same way as during standard treatment plan evaluation.