Introduction

A precise understanding of the biomechanics of the human hip joint is essential for the correct interpretation of diagnostic radiographs and treatment decisions. To reliably differentiate between normal hip anatomy and pathological conditions, the comparison of individual measurements to reference values is of high importance to radiologists and orthopedic surgeons.

Proximal femoral geometry is known to play a key role in a multitude of pathologies of the hip and the femoral neck-shaft angle (NSA) is one of the most commonly measured parameters. It is routinely assessed in pediatric orthopedics during the management of dysplasia of the hip, Perthes disease, slipped capital femoral epiphysis and cerebral palsy, as well as in the treatment of a variety of musculoskeletal conditions in adults (planning of femoral osteotomies, joint replacement surgery, fracture fixation, etc.).

Currently available reference values for the NSA in adults are based on measurements of plain radiographs or relatively small series of anatomic specimens [1]. However, due to rotational influences and imprecise positioning of the femoral shaft and neck axis, the precision and reliability of the NSA measured on plain anteroposterior radiographs has been challenged [1–5].

The primary research goal of the present study was to provide reliable reference values for measuring the NSA in adult patients. For this purpose, rotation corrected coronal computed tomography reconstructions of 800 adult hips were measured. Today, CT represents the best radiological means of analyzing osseous hip anatomy and the NSA measured in the exactly reconstructed coronal plane was therefore considered to equal the true anatomic NSA. The present study included NSA measurements from simulated anteroposterior roentgenograms of the pelvis (in the anterior pelvic plain) to compare the CT-based reference values with the currently employed clinical standard (secondary research question).

Materials and methods

An anatomical analysis of the NSA in pelvic computed tomography scans of adult patients was performed at a university hospital in central Europe. The study design was approved by the local ethical review committee (No. 15-035). All procedures performed were in accordance with the ethical standards of the institutional research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Due to the retrospective design of the study, no informed consent was necessary. All data was pseudonymized.

A total of 400 patients who were treated between 16.08.2010 and 30.11.2014 at a single emergency department and received whole-body trauma CT scans were retrospectively enrolled. Exclusion criteria were: age below 18 years, fractures to the pelvis and femur, posttraumatic or other deformities of the pelvis and proximal femur, existence of metallic implants in the pelvis or proximal femur, immature skeletal system, incomplete availability of medical or radiologic data and insufficient depiction of the proximal femur. The study was performed as part of a larger investigation. A flow-chart of the inclusion process is given in Fig. 1.

Fig. 1
figure 1

Inclusion/exclusion process and grouping of 400 patients with pelvic computed tomography

Full size image

Four groups were generated: (1) male < 65 years, (2) female < 65 years, (3) male ≥ 65 years, (4) female ≥ 65 years. For each group, available cases were screened and tested for inclusion/exclusion criteria until 100 CT scans were included per group. Each hip (n = 800) was analyzed separately. For reliability analysis, repeated measured were performed by observer 1, 2 and 3 months after the first measurement blinded to the previous results in 10 % of the cohort (10 random patients of each group).

Computed tomography scans were performed on a Brilliance iCT 256 scanner (Philips Healthcare, Cleveland, OH, USA). The detector collimation was 2 × 128 × 0.625 mm, creating 256 overlapping slices via a dynamic z-flying focal spot. The pitch was 0.492 and the rotation speed was 0.33 s. As a reconstruction kerne, the “D” (bone) filter was chosen. A fixed tube current-time product of 250 mAs was used, the tube voltage was 120 kVp. The examinations were reconstructed with the filtered back projection algorithm (FBP) by Phillips.

All patients underwent contrast enhanced CT scans in the portal venous phase using bolus tracking method. Iodine contrast medium (Accupaque, GE Healthcare 350 mg iodine/ml, Braunschweig, Germany) was administered at a rate of 3.5 ml/s with a total amount of 100 ml using a power injector. All scans reached below the lower boundary of the lesser trochanter. The radiographic data was stored in a picture archiving and communication system (PACS), and analyzed using the corresponding PACS client (IMPAX EE; AGFA HealthCare GmbH, Bonn, Germany).

Planes and reconstructions for measurements

CT scans were reformatted to three-dimensional pelvic models simulating standardized radiographic views of an anteroposterior roentgenogram of the pelvis in the anterior pelvic plane (suffix AAP) [6]. Furthermore, coronal reconstructions of each proximal femur in the plane of the femoral neck (suffix FNP), neutralizing femoral neck torsion, hip rotation and flexion were acquired. All measurements were performed according to a standardized protocol.

Reconstruction of the anterior pelvic plane (AAP)

The reconstruction was performed using the extended multi planar reconstruction (EMPR) plug-in of IMPAX EE. First, the view was centered on the pelvis in all three planes. The coronal plane was defined as the plane between the anterior margin of the pubic symphysis and the anterior superior iliac spines (ASIS). Therefore, the sagittal and axial views were centered on the symphysis and the coronal and axial planes were rotated, until both ASIS were just visible in the coronal reconstruction (Fig. 2a). Next, a parallel plane through the femoral head was chosen. The slice thickness of the EMPR was changed from 1 to 200 mm, thus creating a pseudo anteroposterior roentgenogram of the pelvis (Fig. 2b).

Fig. 2
figure 2

a Extended multiplanar reconstruction of the anterior pelvic plane. b Simulated anterioposterior pelvic radiograph in the anterior pelvic plane

Full size image

Coronal reconstruction of the proximal femur (FNP)

For measurements of the proximal femur with neutralization of the femoral antetorsion, a reconstruction in the coronal plane of each hip was performed. The coronal plane was defined as the plane between the femoral neck axis (FNA) in the axial reconstruction and the long axis of the femur (FLA) in the sagittal reconstruction (Fig. 3).

Fig. 3
figure 3

Extended multiplanar reconstruction of the femoral neck plane

Full size image

The FNA was drawn according to Müller [7]. First, to define the center of rotation, a circle determined by three points around the circumference of the femoral head was placed. Next, a circle with the same center was drawn, cutting the most waisted part of the cranial margin of the femoral neck. A line connecting the cranial and pedal cutting-points of this circle and the femoral neck was inserted and a perpendicular line drawn through the center of rotation, thus defining the FNA (Fig. 4a). The FLA was defined by a line crossing the centre of two circles placed in the femur at two positions. The center of circle one was positioned at the lower boundary of the lesser trochanter, the second circle 2 cm below the first. The circles lay within the outer margins of the femur (Fig. 4b). The NSA is the angle between the FLA and FNA (Fig. 4c). Femora were categorized according to Tönnis as coxa vara (NSA < 120°), physiologic (NSA ≥ 120° to < 135°), and coxa valga (NSA ≥ 135°) in the FNP and AAP [8].

Fig. 4
figure 4

a Construction of the femoral neck axis (FNA) in the femoral neck plane of the proximal femur. b Construction of the long axis of the femur (FLA) in the femoral neck plane of the proximal femur. c Construction of neck-shaft angle between the FLA and FNA

Full size image

Statistics

For descriptive analysis, absolute mean values, ranges and standard deviations (SD) of the measured variables were reported. Variables were tested for normality using the Kolmogorov-Smirnoff test. Correlations of non-Gaussian distributed variables were described with the Spearman correlation coefficient (ρ). Exploratory analysis was performed using the two-tailed Wilcoxon matched pair test for non-normally distributed variables. For analysis of factors influencing the NSA (age, sex, side, groups), the non-parametric Mann–Whitney-U test or the Kruskal-Wallis test were performed. Intra- and inter-rater reliabilities were evaluated using intraclass correlation coefficients. To detect differences between NSAAAP and NSAFNP, the two-tailed nonparametric Wilcoxon test for matched pairs was employed and the level of significance was set at p < 0.05. For statistical analysis, IBM SPSS Statistics 22 (IBM Corporation, Armonk, NY, USA) and Microsoft Excel 2008 for Mac version 12.3.6 (Microsoft Corporation, Redmond, WA, USA) were used.

Results

The mean age of all 400 patients was 54.3 years (18–100, SD 22.1 years); females had a mean age of 55.4 years (18–100; SD 22.4 years) and males of 53.2 years (18–89; SD 22.6 years). Age distribution of female and male patients showed no significant difference (p = 0.424). Subgroup-specific age distribution is given in Table 1.

Table 1 Age distribution in years of male and female patients by age group
Full size table

The NSA values of the study population measured in two views are given in Table 2. Age-, sex-, and side-dependent NSA measurements are shown in Tables 3, 4, and 5. The group-specific NSA is provided in Table 6. The relation between age and NSA is depicted in Fig. 5 and Spearman’s rank correlation coefficients of hip and age are given in Table 7.

Table 2 Descriptive results of neck-shaft angle (NSA) measurements in two views of 800 hips
Full size table
Table 3 Descriptive results of neck-shaft angle (NSA) measurements in two views of 800 hips by sex
Full size table
Table 4 Descriptive results of neck-shaft angle (NSA) measurements in two views of 800 hips by side
Full size table
Table 5 Descriptive results of neck-shaft angle measurements in two views of 800 hips by age
Full size table
Table 6 Descriptive results of neck-shaft angle measurements in two views of 800 hips by group
Full size table
Fig. 5
figure 5

Scatterplot depicting the relation between the neck-shaft angle and age. Four groups of age and gender are coded by color and linear regression lines are included for each group

Full size image
Table 7 Spearman’s rank correlation coefficients between NSA in two different projections and age of patients
Full size table

In the coronal reconstruction (NSAFNP), 41 (5.1 %) femora were categorized as coxa vara (NSA < 120°), 558 (69.8 %) as physiologic (NSA ≥ 120° to < 135°), and 201 (25.1 %) as coxa valga (NSA ≥ 135°). The group-specific categorization is shown in Fig. 6. The categorization of the neck shaft angles measured on the pseudo anteroposterior roentgenogram of the pelvis (NSAAAP) revealed the following distribution: coxa vara 17 (2.1 %), physiologic 456 (57 %), and coxa valga 327 (40.9 %) hips.

Fig. 6
figure 6

Diagram showing group-specific distributions of varus, physiologic and valgus hips according to Tönnis

Full size image

Intra- and inter-observer reliability were tested and intraclass-correlation coefficients (ICC) are given for repeated measurements of one and two independent observers (Tables 8 and 9). The ICC values for NSA are given for both planes and all measurements combined. The intra-rater reliability for both reconstructions was excellent.

Table 8 Intraclass correlation coefficient (ICC) for repeated measurements of one observer (intra-rater reliability) of 80 hips
Full size table
Table 9 Intraclass correlation coefficient for repeated measurements of two independent observers (inter-rater reliability) of 80 hips
Full size table

The mean difference between NSAAAP and NSAFNP was 2.87° (−21 to 32°, SD 5.84°). The two-tailed nonparametric Wilcoxon test for matched pairs showed significant differences between both measurements (p < 0.001).

Discussion

Both, the precise radiologic measurement and the correct interpretation of the NSA are integral parts of the diagnosis and management of various conditions of the hip joint [2, 9–11]. Today, NSA reference values are mainly based on historic anatomic series and the analysis of non-rotation corrected plain radiographs [1, 8]. We aimed to provide reliable reference values for the NSA by analyzing a large adult cohort with computed tomography scans of the hip.

Primary research question

The present study reports CT-based NSA values of 129.6° for adult men and 131.9° for adult women. These values and the corresponding standard deviations are in accordance with previous reports on the NSA in adults based on plain radiographs [1, 2]. Our data confirmed homogeneity between the right and left hip, and supports the significant influences of age and gender on the NSA. The occurrence of varus hips increases with age in both sexes (Fig. 6) while valgus hips are decreased, which might be explained by physiologic changes and decreased bone mineral density. While no age effect was found in previous studies on NSA, these publications were based on DXA images and therefore are limited in their precision and comparability. In concordance with the data of Reikaraas et al., we found higher mean NSA-values in males compared to females [12, 13]. Again, most studies on NSA and sex were based on DXA images and did not show significant differences [14]. The present study has the advantage of high precision due to the exact reconstruction in the FNP with elimination of positional effects. Applying the classic threshold values for varus, physiologic and valgus NSA measurements, resulted in 69.8 % patients graded as physiologic.

Secondary research question

While reconstruction (FNP) does eliminate rotational influences by definition, the NSA measurements in (AAP) are susceptible to projectional errors due to hip rotation and flexion. Consequently, we found significantly higher NSA values in the simulated anteroposterior roentgenograms of the pelvis (in the anterior pelvic plain) when compared to the exact coronal reconstructions. Furthermore, the categorization of coxa vara/valga and physiologic hips revealed an overestimation of cases with coxa valga in the simulated anteroposterior roentgenograms.

While plain radiographs remain to be the current clinical standard in pre- and postoperative evaluation of the hip, our data underlines the importance of the elimination of projectional errors during the measurement of the NSA on anteroposterior radiographs. Despite optimal imaging techniques involving standardized patient positioning and internal leg rotation, relevant values of hip rotation have been described recently in highly standardized pelvic anteroposterior radiographs [3]. While there exists a number of published methods to estimate rotation of the proximal femur on plain radiographs [3, 4, 15, 16], their applicability and precision have not yet been tested for the correction of the NSA in hips without internal radio-opaque markers. Here, the irregularity of the proximal femur, masking of anatomic landmarks due to hip rotation and influences of hip flexion interfere with the precise measurement of the NSA and its interpretation.

Finally, we report excellent intra- and inter-observer reliability for the presented cohort for both views. While previous researchers achieved comparable intra-rater reliabilities [2, 17], the inter-rater ICC values ranged from 0.58 to 0.89 in plain radiographs. Our result for the inter-rater ICC underlines the exact definition of the anatomic landmarks, the consistent placement of the digital measuring tools and the excellent image quality of our study. Based on this data, an advantage of measurements in standardized reconstructions of CT images over plain radiographs can be assumed. In combination with a clearly defined measurement protocol, CT allows for an optimal measuring tool placement, free choice of the reconstructed plain and independence from external/internal calibration markers resulting in a high precision and reliability of the performed measurements.

The authors are aware of the limitations of the present report

The study was performed at a single institution in Central Europe, thus limiting the applicability of our data to cohorts with Caucasian ethnic backgrounds. Furthermore, no pediatric cases were analyzed, limiting the provided reference values to adult patients. Next, we assessed the NSA on CT-based anteroposterior roentgenograms of the pelvis and coronal rotation corrected reconstructions of the proximal femur. While the latter allowed the highly standardized analysis of the NSA, projection errors due to hip rotation and flexion must be expected for the anteroposterior roentgenograms. Finally, our report does not provide functional data or information regarding preexistent patient perceived abnormalities.

In conclusion, we report a mean NSA for male adults of 129.6° (range 113.2° to 148.2°; SD 5.9°) and 131.9° (range 107.1° to 151.9°; SD 6.8°) for females. Correlation analysis revealed a significant relation between NSA and age. While anteroposterior pelvis radiographs are susceptible to rotational errors, derotated coronal reconstructions of the proximal femur in the neck plane allow the correct measurement of the NSA. We provide highly reliable age and sex dependent NSA reference values for adult patients.