We compare mammography breast density (BD(MD)) to the measure of breast composition using a clinical dual energy absorptiometry (DXA) system (BD(DXA)) calibrated to measure breast density. A DXA scanning protocol was developed to scan breasts isolated in the DXA scan field in either a prone pendulous or decubitus mediolateral position. A total of 17 participants were recruited among women undergoing clinical mammography examinations. Each participant had duplicate DXA scans and duplicate craniocaudal-view mammograms of their right breast with repositioning between each scan and one DXA and one craniocaudal-view mammogram of their left breast. The in vivo repeatability (RMS SD) of BD(DXA) and BD(MD) on duplicate scans was found to be 1.2% for BD(DXA) and 1.4% for BD(MD) when repeat BD(MD) measures were made on the same day. When repeat BD(MD) measures of the same breast were made more than 50 days apart, the repeatability decreased to 5.5%. Left and right breast measurements were highly correlated with both techniques at r2 = 0.98 for BD(DXA) and r2 = 0.86 for BD(MD). Moderate correlation (r2 = 0.52) was found between BD(DXA) and BD(MD) measurements. However, after recalibrating the DXA system to mammography reference materials, negative percent fibroglandular values were measured for the most fatty breasts. Thus, our results are reproducible and accurate to common mammography tissue standards, but did not accurately reflect true percent fibroglandular levels and further development of phantom standards are necessary. We conclude that breast composition can be precisely evaluated and assessed with clinical DXA densitometers at a lower dose than with mammographic breast density methods.