Introduction
Upright human spinal biomechanics and the sagittal shape of the spine play an important role in the development and progression of spinal deformities such as adolescent idiopathic scoliosis (AIS) [
1‐
7]. In 1985, During et al. [
8] described the pelvic sacral angle, and later on, Duval-Beaupère et al. [
9] used the term pelvic incidence (PI) for a parameter that defines the relationship between the pelvic anatomy and spinal alignment. Several other authors described the spino-pelvic balance in normal children and adolescents during growth and in patients with different pathologies [
1,
10‐
21]. Dickson et al. [
5] described that differences in the sagittal plane (reversal of the normal thoracic kyphosis) during growth could initiate a progressive idiopathic scoliosis. Moreover, previous authors suggested a link between the spino-pelvic morphology and spinal deformities in the sagittal plane [
1,
10‐
21]. Most of the previous studies obtained two-dimensional (2D) referential values using standing lateral radiographs that could lead to a deviation from the true sagittal plane and to a certain degree of distortion of the image. Three-dimensional (3D) images are not affected by the characteristics of radiographic imaging, like the projection plane, and can thus better observe the relationship between the sacrum, pelvis and hips [
22]. Pasha et al. [
14] introduced novel pelvic parameters based on 3D reconstructed radiographs, and Vrtovec et al. [
22] measured the 3D PI in a non-scoliotic population using computed tomography (CT) scans, demonstrating the improved accuracy of this method as compared to the traditional measurements on plain sagittal X-rays. In previous studies, the exact difference in pelvic morphology between AIS subjects and non-scoliotic controls has remained unclear. Therefore, the purpose of this study was to analyze, using a previously validated accurate 3D technique, the differences in sagittal pelvic morphology between AIS patients and asymptomatic adolescents.
Discussion
AIS is a disease of the human spine and trunk in which fully upright sagittal spinal biomechanics is known to play an important role [
1‐
7,
10,
27]. Since its introduction in 1985, the PI has been used to describe pelvic morphology and its relation to the spino-pelvic alignment, as well as the relationship between the spino-pelvic alignment and spinal deformities [
1,
8‐
10]. However, previous studies used predominantly 2D imaging that could initiate a certain degree of distortion of the image and influences the spino-pelvic measurements [
25]. Therefore, the true relationship between the 3D pelvic morphology and spinal deformity has remained unclear up to now [
22].
Using the accurate 3D measurement on CT scans, the PI was on average higher in patients with a Lenke type 5 curve as compared to controls, whereas the PI in patients with a Lenke type 1 curve did not differ from controls. In previous studies, the mean PI varied between 41° and 49° for non-scoliotic adolescents and between 42° and 57° for AIS patients, using radiographs [
1,
10,
13‐
15,
17,
19‐
22,
28]. In these studies, the patients with scoliosis had a moderate to severe thoracic curve (mean Cobb angles varied between 35° and 60°); however, Schlösser et al. [
15] included only mild curves (Cobb angle 20° or lower) and found similar results (PI for thoracic AIS patients: 47°, PI for (thoraco)lumbar AIS patients: 42° and PI for controls: 43°; without significant differences). It has been described that the PI in AIS is higher as compared to the non-scoliotic controls; however, others described no significant differences between AIS patients and non-scoliotic adolescents [
1,
10,
13,
19,
21,
27,
28]. Slight PI differences among studies could be explained by the differences in age, since the PI increases with age, but also to the inaccuracy of the 2D method [
16,
29].
The 3D PI of the non-scoliotic population was already described by Vrtovec et al. [
16] and Schlösser et al. [
22]. Pasha et al. [
14] introduced novel pelvic parameters, based on 3D reconstructed radiographs of AIS patients and controls. However, the PI of AIS patients was only based on 2D radiographs, whereas the PI of our study was measured using low-dose CT data and a computerized method based on accurate image processing techniques [
22]. Mac-Thiong et al. [
10] and Farshad et al. [
13] found that the scoliotic curve type was not associated with a specific pattern of sagittal pelvic morphology and balance. However, the results of the present study showed a higher PI in Lenke curve type 5 as compared to Lenke curve type 1. By using low-dose CT data and accurate image processing techniques, we were able to quantify the PI on 3D images with high accuracy and reproducibility in AIS patients
versus controls. Using this 3D CT measurement method, bias due to image acquisition or subject positioning was avoided.
Radiographical measurements may be inaccurate due to the projective nature of radiograph acquisition, because it is usually impossible to obtain the superposition of the two femoral heads [
22,
30,
31]. Moreover, the inclination of the sacral end plate in the sagittal plane may be altered by its architecture and inclination in the coronal plane. 3D reconstructed images are not affected by the characteristics of radiographic imaging and generate true sagittal views by aligning the centers of the femoral heads in 3D, resulting in a lower variability as compared to the radiographical measurements [
22]. However, if the PI is measured from 3D CT images, it is important to consider that measurements in 2D radiographic images may be overestimated by approximately 5° [
22]. CT measurements provide superior imaging precision; however, this study does not imply that measurements of PI from plain radiographic images for clinical use should be replaced by CT.
The PI was correlated with the LL, but not with the Cobb angle or the TK. Most studies described a relation between the PI and the (thoraco)lumbar sagittal alignment, but not between the PI and the thoracic alignment [
10,
13,
17,
19,
21,
27,
30]. This suggests that the PI influences the (thoraco)lumbar region of the spine, but the thoracic part is independent of the PI. However, the LL is correlated with the TK, indicating that the TK is not completely independent of the PI. Additionally, we found a relation between the location of the curve, thoracic or lumbar, and the PI. The PI was higher in primary lumbar curves as compared to controls, whereas the primary thoracic curves did not differ significantly from the controls. This suggests a correlation between the PI and the pathogenesis of the scoliosis, even more because the PI is not influenced by the severity of the curve, suggesting that the PI is not influenced by spinal deformities. However, the exact role of the PI could not be determined in this cross-sectional study. In previous studies it was shown that an excess of posterior shear leads to a decrease in the rotational stiffness of the involved segments, that only posteriorly inclined vertebrae take part in the development of different scoliotic curve patterns, and that more posteriorly inclined vertebrae are rotationally less stable and could lead to initiation and progression of AIS [
7,
15,
17,
32]. The fact that the PI was higher in AIS patients with primary (thoraco)lumbar curves as compared to controls, in contrary of patients with AIS with primary thoracic curves, is consistent with this theory and suggests that the PI is part of the etiopathogenesis of AIS as well as a determining factor for the curve type.
The examinations in this study were made in supine positions. Philippot et al. [
33] has shown that positioning (standing, supine or sitting position) does not influence the PI. Because all scans were acquired preoperatively, only moderate to severe AIS curves, no mild AIS curves, were included in this study. Visualizing the femoral heads is not part of the standard low-dose (20 times lower than that of a standard CT for trauma) imaging protocol [
23,
24]. Therefore, only a small number of scans could be included for the purpose of this study. Minor, not significant, differences were observed between the Lenke type 1 group and the controls and between the Lenke type 1 group and the Lenke type 5 group. Using an alpha error probability of 0.05, a power (1-beta error probability) of 0.80 and the means and standard deviations as described in this study, the power analysis showed a minimum sample size of 101 patients in each group to determine the differences between these groups. However, this study highlighted that the PI was higher in the Lenke type 5 curves, as compared to controls, whereas the Lenke type 1 curves did not differentiate from the normal anatomy.