Femoral offset is underestimated on anteroposterior radiographs of the pelvis but accurately assessed on anteroposterior radiographs of the hip

Patients and Methods

We retrospectively reviewed a consecutive series of 152 patients who had undergone a primary THR with a custom-made cementless femoral component¹⁵ for primary OA of the hip between July 2009 and December 2009. For each patient, a pre-operative AP radiograph of the pelvis and an AP radiograph and CT scan of the affected hip had been obtained. All images were retrieved in DICOM format from a picture archiving and communication system (PACS).

We excluded patients with secondary OA of the hip, those on medication which affected bone metabolism and those whose radiographs did not have a calibration marker. Patients who underwent bilateral THR during this period only had the first procedure included in the study. In all, 52 patients were excluded, leaving 100 patients in the study (43 men and 57 women; mean age 61 years (45 to 74); mean body mass index (BMI) 28 kg/m² (20 to 45)). The study was approved by the institutional review board.

In each case, a standardised protocol was used to achieve reproducible projection. Low-centred AP radiographs of the pelvis and hip were taken in the supine position (Fig. 1). In order to correct for the effects of magnification, a metal calibration sphere of 25 mm was placed on the inner thigh at the level of the femoral head in AP projection. During the study period, two different x-ray tubes were in use: a Canon CXDI series (Canon Inc., Tokyo, Japan) and a Philips Bucky Diagnost VE VT (Royal Philips Electronics Inc., Amsterdam, Netherlands). The tube-to-film distance was 1150 mm, with the tube orientated perpendicular to the table.

Fig. 1

Standardised anteroposterior radiographs of the pelvis (left) and the hip (right) of a 71-year-old man with primary osteoarthritis of the left hip, showing the measurements of femoral offset. The difference between the two measurements is 5.2 mm.

For AP radiographs of the pelvis, both legs were internally rotated by 15° using a leg retainer and the crosshair of the beam was centred on the symphysis pubis.

For AP radiographs of the affected hip, the crosshair of the beam was sited over the midpoint of a line between the anterior superior iliac spine and the symphysis pubis to position it over the centre of the femoral head. The affected leg was internally rotated and retained so that the prominence of the greater trochanter could be palpated in its most lateral position thereby aligning the femoral neck to the coronal plane. When the leg could not be internally rotated sufficiently because of an external rotation contracture, the whole affected hip was additionally elevated with a wedge placed under the buttock.

All CT scans were performed pre-operatively using a Toshiba Aquilion 16 CT scanner (Toshiba Corp., Tokyo, Japan). Each patient was positioned supine with their legs in neutral rotation as confirmed by scout views. The scans were obtained in three sets: the first from the cranial aspect of the acetabulum to below the lesser trochanter, the second from below the lesser trochanter to a point 50 mm distal to the femoral isthmus and the third, which consisted of four to six slices of the knee. The slices were spaced at 4 mm, 8 mm and 2 mm, respectively. Each scan was recorded with gantry tilt 0°, 120 kV and a field of view (FOV) of 250 mm.

A validated MATLAB program (version 7.10; The MathWorks Inc., Natick, Massachusetts) was used to determine the centre (HC) and diameter of the femoral head (HD) and the axis of the femoral shaft on AP projections of the pelvis and hip. Two points were defined on the medial and lateral cortex of the femoral diaphysis 20 mm below the lesser trochanter, and a further two points on the medial and lateral cortex at the level of the femoral isthmus. The midpoints of these pairs determined the axis of the femoral shaft. A circle tool was used to define the diameter of the head and the co-ordinates of its centre. The femoral offset (FO) was calculated as the perpendicular distance from the centre of the femoral head to the axis of the femoral shaft. These measurements were labelled FO_p and HD_p on AP radiographs of the pelvis, and FO_h and HD_h on AP radiographs of the hip (Fig. 1). We also calculated the distance from the midpoint of both tear drops to the centre of the femoral head (x) on the AP radiographs of the pelvis.

For three-dimensional (3D) measurements, a further validated MATLAB programme was used to measure FO and FA. This programme enabled the user to select points from pre-selected axial CT slices, and performed calculations in the 3D co-ordinate system of the CT scanner (Fig. 2).

Fig. 2

Diagram showing selection of CT slices of the proximal femur for three-dimensional calculation of femoral offset and femoral anteversion (s1, centre of the femoral head; s2, femoral neck axis; s3, centroid of the proximal femoral metaphysis; s4, centre of femoral isthmus).

For the 3D calculation of offset (FO_c) and head diameter (HD_c), three axial slices were selected (s1, s3, s4; Fig. 3). HD_c and the centre of the femoral head were determined on the slice with the femoral head at its largest diameter (s1) using a circle tool. The axis of the femoral shaft was defined by the centroid of the proximal femoral metaphysis (s3)^16,17 and the centre of the isthmus (s4); FO_c was then calculated as the perpendicular distance from the axis of the femoral shaft to the centre of the femoral head.

Fig. 3

Series of CT slices for three-dimensional calculation of femoral offset and femoral anteversion: from left to right: head diameter (s1), femoral neck axis (s2), centroid of proximal femoral metaphysis (s3), centre of isthmus (s4), and posterior condylar axis (s5).

For the calculation of FA, the axis of the femoral neck (s2, Fig. 3) was defined using the single slice method as described by Sugano,¹⁸ and the posterior axis of the condyles (s5) by the most posterior aspects of the lateral and medial condyles. The angle between the axis of the femoral neck and the axis of the posterior condyles represented the FA.

Intra- and interobserver reliabilities for 20 randomly-selected AP radiographs of the pelvis, AP radiographs of the hip and CT scans were evaluated by two independent blinded observers using single-measure intra-class-correlation coefficients (ICC) with a two-way random effects model for absolute agreement.

Statistical analysis

The distributions of variables were examined in exploratory data analysis, and tested for normality using Kolmogorov-Smirnov tests. All variables (FO_p, FO_h, FO_c, HD_p, HD_h, HD_c and FA) were normally distributed (p = 0.06 to 0.20). For descriptive analysis, absolute mean values and differences of FO were expressed in millimetres with 95% confidence intervals (CIs). FA was expressed in degrees with 95% CI. Distributions of FO values were compared using paired-samples t-tests for paired observations and independent-samples t-tests for unpaired observations. Pearson’s correlation coefficient (r) was used to evaluate associations between continuous variables. Correlation was characterised as poor (0.00 to 0.20), fair (0.21 to 0.40), moderate (0.41 to 0.60) good (0.61 to 0.80), or excellent (0.81 to 1.00). Results with p-values < 0.05 were considered to be significant: p-values of < 0.001 were considered to be highly significant. Differences in FO measurements on all three modes of radiological imaging were analysed using Bland-Altman plots.¹⁹ Linear regression was performed to relate corresponding measurements of FO. Statistical analysis was carried out using PASW Statistics 18 (SPSS Inc., Chicago, Illinois) and Sigmaplot 12.0 (Systat Inc., San Jose, California).

Results

Excellent intra-observer ICC was seen for all measurements: FO_p (0.993), FO_h (0.990), FO_c (0.990), HD_p (0.886), HD_h (0.906), HD_c (0.933) and FA (0.984).

Interobserver ICC also showed similar correlations for all measurements: FO_p (0.986), FO_h (0.977), FO_c (0.991), HD_p (0.812), HD_h (0.848), HD_c (0.866) and FA (0.950).

The mean FO_p of the whole cohort was 39.0 mm (95% CI 37.4 to 40.6) and the mean FO_h was 44.0 mm (95% CI 42.4 to 45.6). The mean FO_p was significantly lower than the mean FO_h (p < 0.001, paired samples t-test) with a mean difference of 5.0 mm (95% CI 4.3 to 5.7) or 11% (Fig. 4). The difference between FO_p and FO_h was slightly higher in men (5.8 mm, 14%) than in women (4.4 mm, 12%; p = 0.45, independent samples t-test).

Fig. 4

Boxplots showing the differences (in mm) of femoral offset (FO) between anteroposterior radiographs of the hip and pelvis (FO_h - FO_p), anteroposterior radiographs of the pelvis and CT (FO_c - FO_p), and anteroposterior radiographs of the hip and CT (FO_c - FO_h). The box represents the median and interquartile range (IQR), whiskers represent range within 1.5×IQR, circles represent outliers.

We found no correlation between the difference in mean FO on corresponding radiographs and FA (r = 0.18, p = 0.07). The observed mean difference between HD_p and HD_h (1.0 mm (95% CI -0.6 to 1.4)) was not statistically significant (p = 0.068, paired samples t-test). There was also excellent correlation between HD_p and HD_h (r = 0.89, p < 0.001).

The mean FO_c was 44.7 mm (95% CI 43.5 to 45.9). FO_p was underestimated by 5.7 mm (95% CI 4.5 to 6.9) or 13% with reference to FO_c (p < 0.001, paired samples t-test). FO_c and FO_h were similar (p = 0.191, paired samples t-test) with a mean difference of 0.7 mm (95% CI -0.3 to 1.7) (Fig. 4). The differences between FO_h and FO_c were similar in both men and women (p = 0.79, independent samples t-test).

We noted a good correlation between FO_p and FO_c (r = 0.687, p < 0.001) and between FO_h and FO_c (r = 0.767, p < 0.001) (Fig. 5). When compared with the CT measurements, disagreement in assessment of FO was within ± 5 mm in 44% (44 patients) of cases for AP radiographs of the pelvis and 74% (74 patients) of cases for AP radiographs of the hip (Fig. 6).

Fig. 5

Overlay scatter plot illustrating the relationship between femoral offset as measured on anteroposterior radiographs of the pelvis (FO_p), hip (FO_h) and CT (FO_c) in mm. The least square fit lines suggest a good correlation of anteroposterior hip and CT values (R² = 0.588).

Figs. 6a - 6b

Bland-Altman plots illustrating the agreement between measurements of femoral offset (FO) as measured on anteroposterior radiographs of a) the pelvis (FO_p) and CT (FO_c) and b) the hip (FO_h) and CT (FO_c), in mm.

The mean diameter of the femoral head as measured by CT (HD_c) was less than the mean HD_p (mean difference -1.2 mm (95% CI -2.0 to -0.4), p = 0.003, paired samples t-test) and HD_h (mean difference -2.2 mm (95% CI -3.0 to -1.4), p < 0.001, paired samples t-test). Correlations between HD_c and HD_p (r = 0.62, p < 0.001) and HD_c and HD_p (p = 0.63, p < 0.001) were similar.

Discussion

Accurate and reliable assessment of femoral offset is a key element in the pre-operative planning of a THR as it determines the selection and positioning of the prosthetic components. Restoration of offset is essential to achieve a stable articulation,²⁰ a good range of movement,¹ good muscle function^1,2 and equal limb length.³ The importance of the restoration of offset in preventing long-term adverse effects related to impingement²¹ and wear^8,22 is well accepted.

Despite the lesser accuracy of plain radiography when compared with 3D techniques based on CT,¹⁰ AP radiographs of the pelvis are widely used for pre-operative templating in primary hip OA as they provide essential information about the anatomy of the pelvis and contralateral hip and allow evaluation of leg-length discrepancy. It is well established that femoral rotation influences the radiological appearance of the proximal femur,²³ and that failing to correct for femoral anteversion and external rotation contracture will result in underestimation of femoral offset.^10,24 Consequently, clear recommendations have been made about how to position the femoral neck in the coronal plane by internally rotating the lower limb.²⁵ Little attention has been given, however, to potential differences in the accuracy of assessment of offset on comparable AP radiographs of the pelvis and hip. In contrast to AP views of the pelvis, AP radiographs of the hip theoretically allow full correction of individual femoral version, even in the presence of an external rotation contracture, as the hip can be elevated and symmetry is not an issue. Moreover, the centre of the beam is directed to the centre of rotation of the diseased hip, which minimises the effects of radiographic beam divergence.

The present study evaluated differences in measurements of femoral offset performed on standardised AP radiographs of the pelvis and hip in patients with primary end-stage hip OA and compared these measurements to 3D offset values obtained from corresponding CT scans. We observed a significant underestimation of offset on AP radiographs of the pelvis with a mean difference of 5.7 mm (13%) when compared with CT based measurements. The measurements of offset performed on AP radiographs of the hip were comparable to CT-based measurements with a mean difference of 0.7 mm (2%), suggesting that measurements of offset performed on standardised AP views of the hip are more accurate and allow a better representation of 3D offset.

We acknowledge the following limitations of this study: first, we cannot, in retrospect, separately quantify the effects of positioning and centring of the beam. Although a standardised positioning protocol was used, the method for obtaining the radiographs remains a potential source of bias as both femoral rotation and positioning of the calibration marker depend on the judgment of the technician. Similarly, internal rotation of a retroverted femur during positioning of the patient may have exaggerated measurement errors of offset. There were, however, only seven retroverted femora in our cohort with a mean retroversion of -2.7° (95% CI -4.5 to -0.9)).

Secondly, the diameter of the femoral head as determined by CT was less than on the corresponding radiographs. It has been shown that the femoral head is not a perfect sphere²⁶; consequently, the demonstrated differences may be explained by the fact that HD was measured in the transverse plane on CT, but in the frontal plane on radiographs. Moreover, the slice spacing of the CT protocol (4 mm) aimed to reduce radiation exposure and to minimise artefacts caused by contralateral implants. This may have impeded the selection of the true level of the largest head diameter in some cases, potentially contributing to the lower HD values obtained from CT.

Thirdly, in order to evaluate 3D FA, the true axis of the femoral neck can only be determined by means of a 3D reconstruction of the neck. This was a limitation of our CT protocol as a 1 mm slice interval is needed for this purpose. However, the selected single slice method chosen in this study has been shown to be sufficiently accurate to measure FA when the slice for determining the femoral neck axis is chosen from a point just below the femoral head.¹⁸ Moreover, we excluded patients with head or neck deformity associated with secondary OA.

Lastly, we did not use the long axis of the femur (centroid of metaphysis to centre of knee) as previously reported for measurements of femoral offset on CT.²⁷ In the present study, measurements of offset were based on the longitudinal axis of the proximal femur (centroid of metaphysis to centre of isthmus) in order to replicate the measurements of offset made on the plain radiographs.

Because the accuracy of conventional radiographs of the pelvis in measuring femoral offset is limited, it has been suggested that CT scans are used to measure it pre-operatively.^10,11 As a result of this study we would question this premise for patients with primary OA, as we did not see a clinically relevant difference in offset values between AP radiographs of the hip and those made on CT. The mean values which we obtained for femoral offset (44.0 mm) on AP views of the hip and offset (44.7 mm) and FA (14.9°) on CT scans are comparable to those reported in the literature.^10,28,29 Moreover, the present study suggests that the 3D measurement of offset may potentially be predicted from measurements performed on AP views of the pelvis as a good correlation between corresponding measurements of offset was seen. However, a larger series of patients would be needed to confirm this finding and to obtain more powerful prediction equations for men and women separately.

A decrease in femoral offset of approximately 12% following THR has been reported to cause abductor weakness² which suggests that the observed underestimation of offset between AP views of the pelvis and hip is of clinical relevance. The difference between the offset values on AP views of the hip and CT is likely to be of minor clinical importance. The observed differences in measurements were consistent when considering men and women separately, which suggests that our findings are independent of gender.^30,31

In conclusion, this study shows that, when compared with CT, femoral offset is significantly underestimated on AP radiographs of the pelvis, but can be reliably measured with a higher level of accuracy on AP radiographs of the hip in patients with primary OA. AP radiographs of the hip allow better control of femoral rotation, even in the presence of an external rotation contracture, and minimise the adverse effects of radiological projection. Instead of the suggested routine performance of CT for the pre-operative assessment of offset, AP radiographs of the hip reduce the patient's exposure to radiation, are more cost-effective and are easily available. Although the ultimate choice of implant design, size and position must still be made intra-operatively, we recommend that an additional AP radiograph of the hip is obtained routinely for the pre-operative assessment of femoral offset in patients with primary OA.