Effect of increasing vertebral marrow fat content on BMD measurement, T-Score status and fracture risk prediction by DXA.

Quantitative examination of iliac crest bone biopsies shows that as subjects become older bone and functional marrow are replaced by adipose tissue. Studies of vertebral marrow fat using nuclear magnetic resonance spectroscopy ((1)H-MRS) show that subjects with lower spine T-scores have significantly higher marrow fat content. These findings suggest that the ability of DXA scans to determine fracture risk may be partly explained by the effect of increased marrow fat on BMD. However, a proper evaluation of the relationship between WHO spine T-score status and marrow fat content requires that the BMD data are first corrected for the bias caused by a selection effect in which subjects with higher marrow fat are more likely to be identified as having osteoporosis. In this study we have therefore reanalysed previously published data for 185 elderly Hong Kong Chinese subjects (103 women, mean age 73 y; 82 men, mean age 73 y) who had spine DXA scans and (1)H-MRS measurements of L3 marrow fat. The effect of varying marrow fat on BMD was modelled using vertebral body thicknesses measured in 50 men and women. Spine T-scores in each individual were adjusted for the measured marrow fat. Subjects were assigned to WHO categories based on their corrected T-scores, and the relationship between marrow fat and T-score status evaluated using regression analysis and analysis of variance. The average change in percent marrow fat per T-score unit was used to infer the fraction of the spine BMD fracture discrimination explained by marrow composition. The mean (SD) of the L1-L4 vertebral body thickness was 30.2 (2.1) mm for Hong Kong women and 33.4 (2.5) mm for men. A change in marrow fat content from 0 to 100% was estimated to produce a BMD decrease of 0.14 g/cm(2) (1.3 T-score units) in women and 0.16 g/cm(2) (1.3 T-score units) in men. Although adjusting spine BMD for marrow fat reduced the significance of the correlation, there was still a trend for marrow fat to increase with decreasing T-score with a slope of -1.2+/-0.7% per T-score unit (p=0.078) for women and -1.4+/-0.6% per T-score unit (p=0.023) for men. When the effect of marrow composition on fracture discrimination was evaluated the results showed that the higher vertebral marrow fat content found in osteoporotic subjects made a negligible contribution to the ability of spine BMD measurements to predict fracture risk.