PIXImus DXA with different software needs individual calibration to accurately predict fat mass

To validate GE PIXImus2 DXA fat mass (FM) estimates by chemical analysis, to compare previously published correction equations with an equation from our machine, and to determine intermachine variation.

C57BL/6J (n = 16) and Aston (n = 14) mice (including ob/ob), Siberian hamsters (Phodopus sungorus) (n = 15), and bank voles (Clethrionomys glareolus) (n = 37) were DXA scanned postmortem, dried, then fat extracted using a Soxhlet apparatus. We compared extracted FM with DXA-predicted FM corrected using an equation designed using wild-type animals from split-sample validation and multiple regression and two previously published equations. Sixteen animals were scanned on both a GE PIXImus2 DXA in France and a second machine in the United Kingdom.

DXA underestimated FM of obese C57BL/6J by 1.4 +/- 0.19 grams but overestimated FM for wild-type C57BL/6J (2.0 +/- 0.11 grams), bank voles (1.1 +/- 0.09 grams), and hamsters (1.1 +/- 0.13 grams). DXA-predicted FM corrected using our equation accurately predicted extracted FM (accuracy 0.02 grams), but the other equations did not (accuracy, -1.3 and -1.8 grams; paired Student's t test, p < 0.001). Two similar DXA instruments gave the same FM for obese mutant but not lean wild-type animals.

DXA using the same software could use the same correction equation to accurately predict FM for obese mutant but not lean wild-type animals. PIXImus machines purchased with new software need validating to accurately predict FM.