In cone-beam computed tomography dedicated to the breast (BCT), the mean glandular dose (MGD) is the dose metric of reference, evaluated from the measured air kerma by means of normalized glandular dose coefficients (DgN ). This work aimed at computing, for a simple breast model, a set of DgN values for monoenergetic and polyenergetic X-ray beams, and at validating the results vs. those for patient specific digital phantoms from BCT scans. We developed a Monte Carlo code for calculation of monoenergetic DgN coefficients (energy range 4.25-82.25 keV). The pendant breast was modelled as a cylinder of a homogeneous mixture of adipose and glandular tissue with glandular fractions by mass of 0.1%, 14.3%, 25%, 50% or 100%, enveloped by a 1.45 mm-thick skin layer. The breast diameter ranged between 8 cm and 18 cm. Then, polyenergetic DgN coefficients were analytically derived for 49-kVp W-anode spectra (half value layer 1.25-1.50 mm Al), as in a commercial BCT scanner. We compared the homogeneous models to 20 digital phantoms produced from classified 3D breast images. Polyenergetic DgN resulted 13% lower than most recent published data. The comparison vs. patient specific breast phantoms showed that the homogeneous cylindrical model leads to a DgN percentage difference between -15% and +27%, with an average overestimation of 8%. A dataset of monoenergetic and polyenergetic DgN coefficients for BCT was provided. Patient specific breast models showed a different volume distribution of glandular dose and determined a DgN 8% lower, on average, than homogeneous breast model.