Impact of marrow fat on accuracy of quantitative CT.

An assessment of the role of marrow fat as a source of error in single energy quantitative CT (SEQCT) measurements of the spine is presented. Fat-induced shifts and spreads in SEQCT findings are discriminated and quantified. Studies using chemical analysis have reported an average vertebral bone mineral (MIN) loss of approximately 1.3 mg/ml (0.9%) per year from a peak MIN of 150 mg/ml at age 30 years with a standard error of the estimate (SEE) of approximately 23 mg/ml [coefficient of variation (CV) = 20%]; intraosseous vertebral fat content increases by approximately 2.4 mg/ml (1.4%) per year from a fat density of 175 mg/ml at age 30 years with an SEE of approximately 87.5 mg/ml (CV = 36%). If one does not adjust for the average age-related fat change, SEQCT underestimates MIN and overestimates MIN loss by 10-30% (at a scanner setting of approximately 80 kVp). Application of correction procedures reduces the average underestimation to 1-3%. The impact of residual vertebral fat variability on the accuracy of MIN measurements accounts for only 12-24% of the total accuracy error of typically 10-15 mg/ml. The impact of fat changes between zero and twice the normal age-related rate on MIN loss measurements on individual patients and patient groups (treatment versus control) accounts for only 1-2% of the total precision error of typically 1.5-3 mg/ml. From 94 to 99% of the spread of SEQCT versus age data is caused by sources of variability other than fat. All fat-related errors are higher than the above estimates by a factor of approximately 1.5-2.5 when scanning at higher voltages of up to 130 kVp. Given the limited impact of fat variability, we conclude that for most clinical applications, MIN and bone loss measurements should use SEQCT. Adjustment for the average under/overestimation (by either explicitly correcting SEQCT findings or by comparing them with normative data obtained on a similar scanner), measuring at low voltages of approximately 80 kVp, and strict compliance to quality assurance procedures are highly recommended to achieve optimum accuracy.