We investigate the need for anisotropic materials when simulating X-ray mammograms from real 3D MR breast images employing biomechanical models. We previously observed on 3D MRI that the breast in the prone position elongates very little in the anterior-posterior direction even when applying large lateral-to-medial compressions. Improved accuracy was achieved for these 3D deformations when employing transverse-isotropic materials, where the tissue in anterior-posterior direction was stiffer than in the other two directions. We investigate here whether this also holds when simulating cranio-caudal mammograms where the patient is standing. The realism of the simulated breast compressions was judged by comparing the anterior breast shapes of simulated and real mammograms. The anterior breast shape was quantified by the log ratio of the medial-lateral to anterior-posterior diameter of an ellipse fitted to the anterior breast edge. The breast shapes on real and simulated mammograms were on average statistically significantly different (0.48 versus 0.27, P<0.01) when employing isotropic materials. No such difference was observed for transverse-isotropic materials (0.47). The estimated breast thickness, to achieve the breast shape observed on the corresponding real mammogram, was on average very unrealistic for isotropic materials (8.7 mm) while reasonable for anisotropic materials (49.5 mm).