Background: Assessment of femoral anteversion in children with cerebral palsy with two or three-dimensional computed tomography scans may be limited by both positional and anatomic variables. Three-dimensional computed tomography techniques are considered to be more accurate than two-dimensional imaging when the femur is not optimally positioned in the gantry or when the neck-shaft angle is increased.

Methods: Computed tomography scanning was performed on a series of nine model femora with anteversion ranging from 20° to 60° and neck-shaft angles ranging from 120° to 160°. Each femoral model was scanned in two holding devices, the first of which held the femur in optimal alignment (normal model) and the second of which held the femur in flexion, adduction, and internal rotation (cerebral palsy model) relative to the gantry. Femoral anteversion was calculated for each model from two and three-dimensional computed tomography scans by four examiners on two separate occasions. The intraobserver and interobserver reliability, the accuracy, and the effect of increasing the neck-shaft angle on the accuracy of the measurements made on the two and three-dimensional scans of the normal and cerebral palsy models were then examined.

Results: The mean differences in the measurements of femoral anteversion made by the same examiner (intraobserver reliability) were <2° for the two-dimensional scans of the normal and cerebral palsy models and the three-dimensional scans of the normal models, and the mean difference was <4° for the three-dimensional scans of the cerebral palsy models. The mean differences among examiners (interobserver reliability) were <3° for the two-dimensional scans of the normal and cerebral palsy models and the three-dimensional scans of the normal models, and the mean difference was <6° for the three-dimensional scans of the cerebral palsy models. The accuracy of the assessments of femoral anteversion of the normally aligned models was comparable between the two and three-dimensional scans. However, the three-dimensional assessment was significantly more accurate than the two-dimensional assessment for measurement of anteversion of the cerebral palsy models (p = 0.003). Accuracy within 5° was comparable between the two and three-dimensional scans for measurement of the normally aligned models, with 86% of the two-dimensional measurements and 78% of the three-dimensional measurements falling within 5° of the actual measurements. However, the accuracy within 5° was significantly compromised when the models were placed in cerebral palsy alignment. Only 3% of the two-dimensional measurements and 14% of the three-dimensional measurements fell within 5° of the actual measurements, with three-dimensional assessment being significantly better than two-dimensional assessment (p = 0.006). Increasing the neck-shaft angle did not significantly compromise the accuracy of measurement of femoral anteversion with either the two-dimensional or the three-dimensional technique (p > 0.05 for all comparisons).

Conclusions: When adequate alignment of the femur in the computed tomography scanner was possible, a simple two-dimensional technique exhibited excellent intraobserver and interobserver reliability and clinically acceptable accuracy within the relevant ranges of anatomic variability tested (neck-shaft angles of 120° to 160° and femoral anteversion of 20° to 60°). When optimal alignment of the femur in the scanner was not possible, neither two-dimensional nor three-dimensional techniques exhibited clinically acceptable accuracy for the measurement of femoral anteversion.