The accuracy of BMD is how close the measurement by densitometry is to actual calcium content of the bone (the ash weight of bone). Because a BMD measurement in trabecular bone incorporates both bone and marrow, the accuracy of BMD measurement by QCT is sometimes questioned with reference to the relative proportions of red and fatty marrow content of the bone. Whereas the conversion from red to fatty marrow tends to finish by the mid-20s in the hip and peripheral skeleton, the vertebrae show a gradual age-related increase in the proportion of fat in the bone marrow, which starts in youth and continues through old age. An increase in fatty marrow is indistinguishable from a decrease in bone. The inclusion of fatty marrow in the vertebral QCT BMD measurement results in accuracy errors ranging from 5% to 15% depending on the age group. In comparison, the BMD measurement accuracy of DXA lies in the region of 10-15%; the error in accuracy being related to the same issue regarding marrow fat content of the bone and to the assumptions DXA makes in taking soft tissue as a reference. DXA makes assumptions about body composition and soft tissues that means that inaccuracies may occur with inhomogeneities in soft tissue composition and with large changes in weight between scans. Exactly how this may corrected for in adults in uncertain. In the cases of both DXA and QCT BMD measurements, comparing this measurement to a normal reference population makes systematic accuracy errors (as opposed to precision errors) of questionable importance in most situations.