Monte Carlo calculations for breast dosimetry in Digital Breast Tomosynthesis (DBT) may require experimental validations, e.g. via the assessment of the calculated and measured dose volume distribution in the breast. This might also take into account any uneven distribution of the beam intensity in the entrance plane, not usually considered in Monte Carlo simulations, e.g. as determined by the heel effect. We measured the 3D dose distribution in a breast phantom, using XR-QA2 radiochromic films calibrated in free-in-air air kerma. Film sheets were positioned at the entrance surface, at the bottom surface as well as at four depths between adjacent slabs in the 5-slabs, 5-cm-thick phantom simulating a compressed breast with 50% glandular fraction. By varying the irradiation angle, the basic requirements of a DBT scan were simulated. The irradiations were made at 40 kV (HVL 1.1 mm Al) for three angular positions of the beam central axis (θ = ±25 deg and θ = 0 deg normal incidence, simulating a craniocaudal view). Results show that it is possible to determine the transverse and longitudinal distribution of the average dose in the phantom (in terms of kerma in simulated breast tissue 50/50 normalized to the entrance kerma), showing the angular dependence of the depth-resolved dose. In the direction of the beam axis, the dose decreases down to about 26% of the entrance value without the phantom. The backscatter fraction was 8%. In transverse planes the maximum dose variations are between 6% and 18% at θ = 0 deg, whereas the dose varies up to 22% in angular views.