For this retrospective study patients were randomly recruited from the radiology clinic’s existing database over a period of four years from 2004 to 2008. Demographic data are shown in Table 1. The computed tomography scans (CT) ranged from the iliac crest to the proximal femur and included at least to the lesser trochanter.

The exclusion criteria were hip osteoarthritis Kellgren-Lawrence score ≥ 2, presence of hip endoprosthesis and fractures and tumors in the lower extremities.

CT scans were conducted in supine position using a Toshiba Aquilion 64^© and a Philips Mx8000 IDT 16^© Scanner. The slice thickness created was between 3 mm and 5mm at a Gantry inclination of 0°. Pixel size ranged from 0.583 mm × 0.583 mm to 0.885 mm × 0.885 mm.

CT data were analyzed using ImageJ and GnuPlot (both Public Domain). The method has been described to a certain extent already [12]. In brief: to avoid rounding during the 3D reconstruction, bony landmarks were marked in the 2D slices using image-processing software. Both the right and the left sides were analyzed.

Center of femoral head (CFH): coordinates of the femoral head center were marked manually in the slices of the greatest diameter in the horizontal plane (Fig. 1).

Fig. (1). The 3D calculation of the femoral offset is schematically shown in a 3D-reconstruction of the pelvic bone and the proximal femur produced by the Amira 4.1.1 (Amira®, San Diego, California, USA) software package. Ellipses have been fitted automatically into the medullary canal (shown here for the right side). The femoral axis was calculated by the best line of fit through the ellipse-center points. A virtual plane was created containing the femoral axis and the center of the femoral head. The femoral offset was calculated as the distance between the femoral axis and the femoral head center perpendicular to the femoral axis.

Axis of the proximal femur (FPA): the proximal femoral axis was calculated as best line of fit for the centers of the ellipses that were automatically fitted to femoral medullary canal on each slide (Fig. 1).

Femoral offset (FO) was determined as the length of the vector perpendicular to the FA from the FA to the CFH (Fig. 1).

Data were analyzed using the Kolmogorov-Smirnov-Z-test and Levene-Test for normal distribution and homogeneity of the variances. Correlation analyses as well as scatterplots were performed for demographic data and measurements of FO and BWLA were obtained. Pearson correlation analysis was used for parametric and non-parametric data as appropriate. A p-value less than .05 was considered to be significant.

The FO measured in our patient population without OA yielded slightly lower values than Atkinson et al., and Maruyama et al., found in their investigations [6, 10]. A possible explanation for this might be that in their studies, the femoral axis was determined by the diaphyseal part of the femur, thus by the long femoral axis (FLA). Maruyama [10] has already described the influence of the deviation of those axes and the anterolateral bowing of the femur on the femoral offset.

Our study showed almost identical results compared to the data retrieved from the cadaveric samples in the study of Unnanuntana et al. [9]. Interestingly, the average height of the populations investigated almost completely match even though the average height of the general population has increased since those cadaveric samples were collected [13]. However, as the age limit in Unnanuntana's study was set at 40 years at the time of death the decrease of height with age was possibly excluded and might have assimilated the populations in terms of height. Furthermore, that population lived at a time when systemic bone diseases were more frequent and skeletal deformities altering the FO could have occurred more often. Although the measuring methods used in the study of Unnanuntana et al., were different, they also focused on the proximal part of the femur when determining the FO. Just as in our study neither Atkinson nor Unnanuntana found differences in FO between the patient’s left and right side [6, 9].

Studies based on 3D CT analysis of the FO in patients suffering from idiopathic osteoarthritis showed rather similar results [7, 8]. The virtual absence of differences between the values measured within these two populations is quite remarkable considering the influences of FO on the joint reaction forces [11]. Moreover, it has been shown that the FO is associated with the development of degenerative hip joint diseases [14, 15]. From investigations concerning joint wear and tribology we know that a decrease [16] in FO increases the wear between the articulating partners by increasing the joint reaction forces. Assuming that the femoral offset directly influences the development of osteoarthritis, we would have expected the FO to be higher in our patient population without hip osteoarthritis. Other parameters influencing the resultant joint reaction forces like pelvic geometry and bodyweight, however, were not captured. Therefore no conclusion could be drawn from the comparison of our values with those from Husmann et al., [8] and Sariali et al., [7] regarding the development of osteoarthritis.

The sex-specificity of the FO measured in our study agrees well with earlier investigations [6, 9]. The absolute differences between the FO of males and females, however, vary between the study of Atkinson et al., and Maruyama et al., (both 7 mm, both 14%), Unnanuntana et al., (2.9 mm, 7%) and our study (4.1 mm, 9%). This deviation in the absolute expression of the sex-associated differences in FO can be explained by the different methods for determining FO. While in our study and Unnanunata´s the FPA was used to determine the FO, the FLA was used by Atkinson et al. and Maruyama et al. If the influence of the anterolateral bowing of the femur is taken into account, which is more pronounced in males than in females [10], these varying sex-specific relative differences can be accounted for.

We do not think that the significant correlation between height and FO found in males, but not females, should be attributed to the sex differences. Rather, we believe that the population investigated was too small and that in a larger population such a correlation would have been observed in females, too. The results of this study show that the contralateral side is the most accurate template for the reconstruction of the femoral offset for both men and women. Where the contralateral side cannot serve as a template, the FO values measured and the correlation found in our study might help to identify possible reasons for symptoms that occur postoperatively, as FO has been shown to be closely linked to the incidence of a variety of alterations [1, 3, 4, 17].