The Modified Femoral Neck-Shaft Angle: Age- and Sex-Dependent Reference Values and Reliability Analysis

Background. The femoral neck-shaft angle (NSA) is of high importance for the diagnostics and treatment of various conditions of the hip. However, rotational effects limit its precision and applicability using plain radiographs. This study introduces a novel method to measure the femoral NSA: the modified NSA (mNSA), possibly being less susceptible against rotational effects compared to the conventional NSA. Patients and Methods. The method of measurement is described and its applicability was tested in 400 pelvis computed tomography scans (800 hips). Age- and gender-dependent reference values are given and intra- and interrater reliability are analyzed. Results. The mean age of all 400 patients (800 hips) was 54.32 years (18–100, SD 22.05 years). The mean mNSA was 147.0° and the 95% confidence interval was 146.7°–147.4°. Differences of the mNSA between sexes, age groups, and sides were nonsignificant. The absolute difference between NSA and mNSA was 16.3° (range 3–31°; SD 4.4°); the correlation was high (0.738; p < 0.001). Overall, the intra- and interrater reliability were excellent for the mNSA. Interpretation. We introduced a novel concept for the analysis of the neck-shaft angle. The high reliability of the measurement has been proven and its robustness to hip rotation was demonstrated.


Background
The femoral neck-shaft angle (NSA) is one of the most frequently applied measures of hip anatomy [1]. It is essential in the diagnosis of various pathological conditions of the hip and femur including developmental dysplasia of the hip (DDH) [2,3], cerebral palsy (CP) in children and adults with consecutive pathologies of the hip [4], femoroacetabular impingement (FAI), Perthes disease and femoral head necrosis (FHN) [5], epiphysiolysis capitis femoris (ECF), osteogenesis imperfecta (OI), and proximal femoral fractures [6]. In particular the planning of operative interventions including osteotomies, hip replacement surgery, and internal fixation of fractures requires adequate preoperative and postoperative measurements of hip parameters.
To correctly assess the NSA, highly standardized anteroposterior radiographs are mandatory to avoid projectional errors of the projected NSA caused by hip rotation (Figure 1) [7][8][9]. Accounting for the suspected antetorsion of the femoral neck, 15 to 20 ∘ of internal rotation of the feet is recommended for standardized anteroposterior radiographs of the pelvis and hip. However, in patients with contractures (i.e., osteoarthritis, cerebral palsy) and those following musculoskeletal trauma, the overall mobility of the affected limb is usually severely restricted and the required internal rotation cannot be achieved [10].
The effect of internal and external hip rotation is an overestimation of the NSA (i.e., valgization), while flexion does lead to an underestimation (i.e., varization) [7,8]. Figure 1 depicts the effect of hip rotation. Whereas complex biplanar radiographs were developed to account for these effects, their use is usually restricted to individual cases in paediatric orthopedics and associated with relatively high radiation exposure [11]. Additionally, the reliability of the 2 BioMed Research International Figure 1: Effect of hip rotation on the projected neck-shaft angle (NSA). The axial view of the femur in a CT reconstruction of a right hip is shown (a). Three coronal reconstructions are projected. Internal hip rotation of 0 ∘ , 30 ∘ , and 60 ∘ is demonstrated as shown in the axial view. The NSA is included and the overestimation of the projected NSA due to the rotation is visualized. measurement of the NSA has been challenged in recent publications and there exists no standard method of measurement in the current literature [8,[12][13][14]. Notably, a current systematic analysis of its application in healthy adults revealed a considerable heterogeneity in the distribution of the NSA [1].
This study aimed to present a novel method to measure the femoral NSA: the modified NSA (mNSA). We assumed that the modified NSA is less susceptible to rotational effects than the conventional NSA and is highly reproducible due to the exact definition of the anatomic landmarks. Thus, the mNSA might prove to be beneficial in the anatomic description of the hip for diagnostics and planning of surgery.
The method of measurement is described and the application was tested in 400 pelvis computed tomography scans (800 hips). Age-and gender-dependent reference values are given and intra-and interrater reliabilities were analyzed.

Patients and Methods
A previously described cohort of 400 patients undergoing whole-body trauma computed tomography scans was retrospectively included [15]. All CT scans were screened for inclusion and exclusion criteria. Here, the presence of fractures, deformities, implants, and immature skeleton served as exclusion criteria. Patients were stratified for further analysis into four groups consisting of 100 scans (200 hips) each: (A) male under 60 years of age, (B) male over 59 years of age, (C) female under 60 years of age, and (D) female over 59 years of age. A detailed description of the cohort has been published previously [15,16].
Computed tomography scans were acquired using a Brilliance iCT 256 scanner (Philips Healthcare, Cleveland, OH, USA) and a standardized trauma-scan protocol was used. Radiographic data was deposited in a picture archiving and communication system (PACS) and analyzed using a PACS client (IMPAX EE; AGFA HealthCare GmbH, Bonn, Germany). Two independent observers performed radiologic measurements after a training-series of 10 cases. In 10% of all cases repeated measurements were performed by observer one and two 3 months after the first analysis, blinded to the previous results; intra-and interrater reliabilities were calculated.

Planes and Reconstructions for Measurements.
To simulate nonstandardized radiographs of the pelvis, measurements were performed in the CT scout-view (Scout). To simulate semistandardized radiographs of the pelvis without optimal hip rotation, original axial CT images were reformatted in the anterior pelvic plane and a simulated radiograph was constructed (APP). Finally, a coronal reconstruction of each proximal femur in the plane of the femoral neck (FNP) was created, which eliminated rotational influences. All measurements followed a standardized protocol [15,16]. The detailed methods of the reconstruction of the APP and FNP were described previously [15,16].

The Modified Neck-Shaft Angle (mNSA).
The long axis of the femur (FLA) was defined by a line crossing the centre of two circles placed around the outer margins of the subtrochanteric femur at two distinct positions: the centre of the upper circle was positioned at the lower boundary of the minor trochanter and the lower circle was placed two centimeters below the first due to the end of the scan at this height. In radiographs and lower-reaching CT scans, the FLA is defined as depicted in Figure 2.
The modified femoral neck axis (mFNA) was defined as the line connecting the centre of rotation and the FLA at the height of the apex of the minor trochanter. Thus, a circle defined by three points around the margin of the femoral head was drawn, determining the centre of rotation. The cutting point with the FLA was found by drawing a perpendicular line from the FLA to the apex.
The modified NSA (mNSA) is the angle between the FLA and the modified FNA (mFNA). Figures 2(a) and 2(b) depict the NSA and the mNSA, respectively.  (Table 1), divided by sex (Table 2), by side (Table 3), by age (Table 4), and by the combination of age and sex (Table 5).
Overall, the mean mNSA was 147.0 ∘ and the 95% confidence interval was 146.7 ∘ -147.4 ∘ . The variance was 25.3 ∘ and the standard deviation was 5.0 ∘ . Figure 4 shows a histogram of the frequencies of the mNSA in all 800 hips.
Differences of the mNSA between sexes were nonsignificant in all planes (Scout: = 0.649; APP: = 0.065; FNP: = 0.468). The differences between age groups were significant ( < 0.001 in all planes). The spearman rho correlation  The Kruskal-Wallis Test showed significant differences between the four age-and sex-stratified groups in all planes ( < 0.001) while pair-wise post hoc analysis showed nonsignificant results between old males and females as well as young males and females.
The comparison of the mNSA in the FNP and the scout or APP showed significant differences (both < 0.001) (Box-Plot; Figure 3).
ICCs are given in Tables 6 and 7 for repeated measurements of one and two independent observers. The ICC values for mNSA and NSA are given for each plane. Overall, the intra-and interrater reliability were excellent in all planes.

Discussion
The femoral neck-shaft angle is of highest clinical relevance for orthopedic and trauma surgeons and remains to be a subject of interest in current research [17,18]. However, this specific parameter of proximal femoral anatomy has multiple limitations. First, the reproducibility has been questioned and various methods of measurement were described [1,18].
While the method of Müller as presented by Boese et al. seems to allow for a sufficient reliability, a current systematic analysis of its application in the published literature emphasized the necessity to improve the method of its measurement [1]. Second, the NSA shows significant differences between ethnic groups, sexes, and age [19]. Of note, historical specimens showed significant differences compared to current populations [1,19]. Third, the NSA is highly dependent from    hip rotation and the projected NSA can differ significantly from the true (maximum) NSA ( Figure 1) [7,8,15]. Even the application of highly standardized radiographs does not completely eliminate this effect [9]. Consequently, a number of recent studies aimed to provide novel reference values [1,20].
The development of the modified NSA addresses all three limitations of the NSA.
First, the method of measurement is clearly defined. In particular, the problem to assess the femoral neck axis is avoided and a reproducible modified axis is used. The excellent intraclass correlation coefficients confirmed the high reproducibility of the method.
Second, the analyzed large cohort of 400 Caucasian patients without known pathologies to the hip covers both sexes and sides and the age distribution includes a wide range of adult patients. Thus, a representative sample for Caucasian patients was generated. Reference values for other ethnic groups can be provided easily by future studies. However, no