Statistical modelling of how the sagittal alignment of the cervical spine is affected by adolescent idiopathic scoliosis and how scoliosis surgery changes that

Abstract

The relationship between the sagittal shape of the cervical spine and that of the thoracolumbar spine is established in the normal spine. Adolescent idiopathic scoliosis (AIS) is recognised as a change in the shape of the spine in both the coronal and sagittal planes. The effects of AIS on the alignment of the cervical spine, including the effects of surgery, has been less well studied. The objective of this study was to identify, using regression analysis, the significant relationships between the alignment of the thoracolumbar spine, in both the coronal and sagittal planes, and the sagittal alignment of the cervical spine in AIS. This study used coronal and sagittal radiographic measures from a group with AIS, both pre and post‐operatively, which were analysed using multiple linear regression methods to identify significant parameters that explain the sagittal shape of the cervical spine. There were 51 pairs of pre and post‐operative radiographs analysed, 40 of which were Lenke 1 curves and 11 Lenke 3 curves. Posterior spinal fusion was performed for all. The significant parameters pre‐operatively were T1 slope, thoracic kyphosis, lumbar lordosis and SVA with an R² value of 78%. Post‐operatively, the significant parameters were T1 slope, thoracic kyphosis, lumbar lordosis and thoracolumbar scoliosis with an R² of 63%. The sagittal alignment of the cervical spine in AIS is related to the shape of key parameters in the rest of the spine. Changes in the cervical sagittal shape occur to compensate for changes in shape to the rest of the spine that occur as a consequence of surgery. This has implications for the understanding of how the compensatory mechanisms of the spine are used to maintain a horizontal gaze, along with prediction of the effects of surgery on the shape of the spine.

The sagittal measures of alignment of the spine used to define the relationships between the cervical and thoracolumbar spines.

1. INTRODUCTION

The sagittal alignment of the whole spine (the view of the spine from the side) is anatomically described as a series of balanced curves that comprise kyphosis (forward bend) of the thoracic spine (the spine found within the chest cavity between T1 and T12 vertebral bodies) with lordosis (backwards bend) of both the cervical (the spine in the neck between the occiput and the C7 vertebral body) and lumbar spines (the spine found within the abdomen between the L1 vertebral body and the sacrum) when in a static and neutral position (upright stance when stationary) (Sinnatamby, 2011). The coronal view of the spine (the view of the spine from the front) is anatomically described as straight without any lateral deviation. It is recognised that the description of the cervical spine as being lordotic does not always reflect the alignment of the spine. In those classed as ‘without spinal pathology’, there is a range of normative values for cervical sagittal alignment (Khalil et al., 2018; Virk et al., 2020). In the asymptomatic population, a ‘normal’ cervical spine can be classified as lordotic, neutral or kyphotic, although lordotic is the most common finding (Khalil et al., 2018; Kim et al., 2018). In the adult literature that describes spinal shape in those without spinal deformity, it is recognised that the sagittal shape of the cervical spine is coupled to that of the thoracolumbar spine and pelvis (Roussouly et al., 2005).

Adolescent idiopathic scoliosis (AIS) is a complex three‐dimensional (3D) deformity of the spine (Stokes, 1994) in individuals aged 10–18 years, that presents as an alteration in spinal shape in both the coronal and sagittal planes, along with rotation in an axial plane (top down view) in between the vertebral bodies. In AIS, the spine is routinely imaged with standing radiographs in two views at 90° to each other (orthogonal views), matching the coronal and sagittal aspects of the spine. The changes in the shape of the thoracolumbar spine in AIS are seen as a bend in the coronal view, a reduction in the thoracic kyphosis and change in the size of the lumbar lordosis when compared to normative data (Newton et al., 2019). This can also be associated with changes in cervical spine alignment (Garg et al., 2021; Tang et al., 2019). The surgical correction of AIS is a procedure where, through the implantation of pedicle screws into the vertebral bodies, with connection to rods which act as internal scaffolding, the shape of the deformed spine is altered to recreate normal anatomy as closely as possible. A bony fusion is also formed between the instrumented vertebral bodies which, with the rods, abolishes all movement over that segment of spine (Choudhry et al., 2016). Correction of an AIS spinal deformity is performed through minimising both the amount of scoliosis and the axial rotation of the vertebral body, alongside normalising the values of thoracic kyphosis and lumbar lordosis (Choudhry et al., 2016). Following scoliosis surgery, it is recognised that the change in the three‐dimensional shape of the thoracolumbar spine can result in compensatory changes occurring elsewhere in the spine to maintain a horizontal gaze (Pepke et al., 2019). Whilst the relationships between the different parameters in the spine, along with the changes in those parameters that occur secondary to scoliosis surgery, have been explored using correlation analysis (Garg et al., 2021; Shimizu et al., 2019), a greater understanding of what parameters have the most effect on an individual's cervical spine alignment with AIS is lacking. The sagittal alignment of the cervical spine has import as a contributing factor to the subsequent development of myelopathy and spinal cord compression (Ames et al., 2013; Scheer et al., 2013) amongst other conditions that include difficulties with swallowing and respiratory function (Theologis et al., 2019). An understanding of the different alignment between the cervical and thoracic spines are also critical to the prevention of proximal junctional kyphosis and failure, where, following surgery for AIS there is segmental kyphosis over the cervicothoracic junction which in the worst case can lead to spinal cord injury (Hart et al., 2013). More advanced mathematical modelling techniques have been used previously to describe the cervical spine alignment in an adult population without spinal deformity (Diebo et al., 2016) and with regards to the 3D shape of the torso in AIS (Gardner, Berryman, & Pynsent, 2021). These techniques have not, however, been applied to cervical spine alignment in the setting of AIS. An investigation of these issues will add to the understanding of the global sagittal shape of the spine in AIS, and how the addition of a stiff instrumented fusion, that fixes the shape of a section of the spine, may change that relationship, illustrating the ability of the cervical spine to compensate for changes in the coupled relationship between the cervical and thoracolumbar spines.

2. METHODS

The aim of this study is to examine the relationship between the coronal and sagittal alignment of the thoracolumbar spine and the sagittal alignment of the cervical spine, in both pre and post‐operative AIS. This was a retrospective review of prospectively collected data from radiographs of adolescents with AIS from a single centre. This study was performed in accordance with all the relevant guidelines and regulations. Ethical approval for this work was given by the National Research Ethics Service UK (ref: Northampton 15/EM/0283). As this is a retrospective review of previously collected routine care data, there was no requirement to seek individual informed consent to access the images used in this study.

Whole spine postero‐anterior and lateral radiographs, taken in a standardised erect position with a horizontal gaze using a Siemens system (Siemens Healthcare GmbH), were reviewed on a digital PACS system (GE Healthcare Centricity). Measures were taken of the sagittal alignment of the spine (Figure 1) in pre‐operative and post‐operative radiographs of the same individuals.

• The cervical sagittal alignment using the Cobb angle (Cobb, 1948) was measured from the occiput (measured as McRae's line [McRae & Barnum, 1953]) to the inferior endplate of C2.

• The cervical sagittal alignment using the Cobb angle was measured from the inferior endplate of C2 to the inferior endplate of C7.

• The cervical sagittal vertical alignment (SVA) was measured as the horizontal distance from the centre of the body of C2 to the postero‐superior corner of T1.

• The angle of the slope of the T1 superior endplate to the horizontal (T1 slope).

• The thoracic kyphosis was measured as a Cobb angle from the superior endplate of T1 to the inferior endplate of T12.

• The lumbar lordosis was measured as a Cobb angle from the superior endplate of L1 to the superior endplate of S1.

• The pelvic incidence was measured as an angle between a line joining the centre of the femoral heads to the midpoint of the superior end plate of S1 and a line perpendicular to the midpoint of the superior endplate of S1.

• Global SVA was measured as the horizontal distance from the centre of the body of C7 to the postero‐superior corner of S1.

FIGURE 1.

An illustration of the measures of sagittal alignment used

The coronal shape of the cervical spine was not used as part of this analysis. This was because the radiographs reviewed did not allow an appreciation of the coronal shape of the superior cervical spine due to the overlying mandible. Analysis of the coronal shape of the full cervical spine requires an MRI or CT scan. What was seen of the coronal view of the cervical spine was variable and could not be standardised.

All angular measures were deemed positive when in kyphosis and negative when in lordosis. The measures of cervical and global SVA were deemed positive when the superior vertebral body was located anterior to the inferior vertebral body. In the coronal plane, the scoliotic curves were measured with the Cobb method (Cobb, 1948) at the points of inflection for the proximal thoracic, mid‐thoracic and thoracolumbar curves making no distinction between the side of convexity. In the post‐operative radiographs, the levels of superior and inferior instrumentation were also recorded. Radiographs were rejected for review if the necessary anatomy was not adequately visualised, which in the coronal view was from superior to T1 and inferior to S1, and in the lateral view, superior to the occiput and inferior to the femoral heads. Rejection of the radiograph also occurred if it appeared that the subject was not looking straight ahead, with either cervical flexion or extension when the radiograph was taken, as previously reported in the literature (Khalil et al., 2018; Kim et al., 2018; Lee et al., 2012; Li et al., 2021; Theologis et al., 2019). Finally, rejection occurred if the lateral radiograph was not taken with the forearms at approximately 60° relative to the vertical axis with the fists resting lightly on the clavicles as described by Marks et al. (2009).

Additive multiple linear regression models were created using R statistical software (R Core Team R, 2020) in the following format:

$$\begin{array}{c}{\mathrm{CSA}}_{\mathrm{pre}}=\text{constant}+\mathrm{a}\left({\text{CSVA}}_{\mathrm{pre}}\right)+\mathrm{b}\left(\mathrm{T}{1}_{\mathrm{pre}}\right)+\mathrm{c}\left({\mathrm{TK}}_{\mathrm{pre}}\right)\hfill \\ +\mathrm{d}\left({\mathrm{LL}}_{\mathrm{pre}}\right)+\mathrm{e}\left({\mathrm{PT}}_{\mathrm{pre}}\right)+\mathrm{f}\left({\mathrm{MT}}_{\mathrm{pre}}\right)+\mathrm{g}\left({\mathrm{TL}}_{\mathrm{pre}}\right)+\mathrm{h}\left({\text{GSVA}}_{\mathrm{pre}}\right)\hfill \\ +\mathrm{i}\left({\mathrm{PI}}_{\mathrm{pre}}\right)+E\hfill \end{array}$$

$$\begin{array}{c}{\mathrm{CSA}}_{\text{post}}=\text{constant}+\mathrm{a}\left({\text{CSVA}}_{\text{post}}\right)+\mathrm{b}\left(\mathrm{T}{1}_{\text{post}}\right)+\mathrm{c}\left({\mathrm{TK}}_{\text{post}}\right)\hfill \\ +\mathrm{d}\left({\mathrm{LL}}_{\text{post}}\right)+\mathrm{e}\left({\mathrm{PT}}_{\text{post}}\right)+\mathrm{f}\left({\mathrm{MT}}_{\text{post}}\right)+\mathrm{g}\left({\mathrm{TL}}_{\text{post}}\right)\hfill \\ +\mathrm{h}\left({\text{GSVA}}_{\text{post}}\right)+\mathrm{i}\left({\mathrm{PI}}_{\text{post}}\right)+\mathrm{j}\left(\mathrm{SUI}\right)+\mathrm{k}\left(\mathrm{NIL}\right)+E\hfill \end{array}$$

CSA, cervical sagittal alignment measured from occiput to C7; CSVA, cervical SVA; T1, T1 slope; TK, thoracic kyphosis; LL, lumbar lordosis; PT, proximal thoracic scoliosis; MT, mid‐thoracic scoliosis; TL, thoracolumbar scoliosis; GSVA, global sagittal vertical alignment; PI, pelvic incidence; SUI, superior uninstrumented thoracic vertebral levels (number of vertebral levels between C7 and superior instrumented level); NIL, number of instrumented levels; 𝛦, error. The CSA is the sum of the sagittal alignment between the occiput and C2 and C2 and C7.

Multiple linear regression models were also created for cervical sagittal alignment measured from C2 to C7 having removed the contribution of occiput to C2 to the cervical sagittal alignment.

Collinearity was examined using the R vif package (Fox & Weisberg, 2011). Models were simplified using the R stepAIC package (Venables & Ripley, 2002), removing all non‐significant parameters (p > = 0.05) in a stepwise fashion. The model that best explained the data was assessed using the R² parameter and the Akaike Information Criterion (AIC) (Akaike, 1974).

The model that best explained the data was plotted as the measured value from the radiograph against the predicted value from the model, with both inner and outer 95% confidence intervals included. The coefficients for the final models were also plotted using the sjPlot package in R (Lüdecke, 2021).

3. RESULTS

There were 56 pairs of whole spine postero‐anterior and lateral radiographs, both pre and post‐operatively, for review. There were five pairs of radiographs that were rejected as the radiographs were inadequate to visualise the required anatomy for measurement, leaving 51 pairs of radiographs for analysis. The demographics of the group are shown in Table 1. There were 40 main thoracic curves (a main thoracic curve with smaller thoracolumbar curve, Lenke 1) and 11 double major curves (two curves of equal magnitude, Lenke 3) (Lenke et al., 2001). All individuals in this study underwent posterior instrumentation and correction of scoliosis using a modern pedicle screw system. Correction was obtained through a combination of translation, in both the coronal and sagittal planes, along with derotation of the apex of the spinal deformity. The exact sequence of corrective manoeuvres, the upper (most superior) instrumented level (UIV) and the lower (most inferior) instrumented level (LIV), the rod material and diameter used, types and pattern of pedicle screws used and the use of other techniques including posterior column osteotomies, was a decision made on a case by case basis to maximise the correction obtained. The measures taken from the radiographs are found in the Data S1.

TABLE 1.

The demographics of the analysed group

	n
Males	10
Females	41
Mean age at pre‐operative radiograph (SD, range)	14.4 years (1.4, 9.5 to 17.0)
Mean age at post‐operative radiograph (SD, range)	17.5 years (1.7, 13.8 to 21.7)
Mean difference between pre and post‐operative imaging (SD, range)	2.5 years (1.6, 0.9 to 7.8)

Figures 2 and 3 show the equations that best explain the data, following simplification of the model, eliminating parameters not found to be significant. Figures 4 and 5 show the coefficients of the final models as the mean and 95% confidence interval for each coefficient. The value and significance coefficients for every term in the equation for that model are found in the figures and in Table 2. In both the pre and post‐operative equations, there were no parameters deemed to be co‐linear. Non‐significant coefficients are marked as NS. Not applicable parameters are marked NA.

FIGURE 2.

The plot of the pre‐operative parameters comparing the measured and predicted value for the model. The diagonal blue line across the plot represents the line of best fit if the model was 100% accurate for reference. Each point is an individual patient. The line of best fit for the model is the solid black line, surrounded by the limits of prediction (green lines) and the 95% confidence limits of the line of best fit (red lines). The equation of the model is shown. CSA, cervical sagittal alignment; GSVA, global sagittal vertical axis; LL, lumbar lordosis; T1, T1 slope; TK, thoracic kyphosis

FIGURE 3.

The plot of the post‐operative parameters comparing the measured and predicted value for the model in a similar way to that described in Figure 2. CSA, cervical sagittal alignment; LL, lumbar lordosis; T1, T1 slope; TK, thoracic kyphosis; TL, thoracolumbar scoliosis

FIGURE 4.

The plot of the pre‐operative coefficients. Each solid dot represents the mean effect, with the line the 95% confidence intervals, of the coefficient, with red a negative and blue a positive effect. GSVA, global sagittal vertical axis; LL, lumbar lordosis; T1, T1 slope; TK, thoracic kyphosis

FIGURE 5.

The plot of the post‐operative coefficients. Each solid dot represents the mean effect, with the line the 95% confidence intervals, of the coefficient, with red a negative and blue a positive effect. LL, lumbar lordosis; T1, T1 slope; TK, thoracic kyphosis; TL, thoracolumbar scoliosis

TABLE 2.

The coefficients of the models

	Pre‐operative	Post‐operative
Intercept	−21.02 (0.002)	10.09 (0.079)
CSVA	NS	NS
T1	0.58 (0.012)	−1.02 (<0.001)
TK	−1.43 (<0.001)	−0.70 (0.001)
LL	−0.44 (0.001)	0.13 (0.034)
PI	NS	NS
PT	NS	NS
MT	NS	NS
TL	NS	0.54 (0.031)
GSVA	−0.29 (<0.001)	NS
SUI	NA	NS
NIL	NA	NS

Note: NS is not significant (p ≥ 0.05). NA is not applicable. The level of significance is seen in the brackets.

Abbreviations: CSVA, cervical sagittal vertical axis; GSVA, global sagittal vertical axis; LL, lumbar lordosis; MT, main thoracic scoliosis; NIL, number of instrumented levels; PI, pelvic incidence; PT, proximal thoracic scoliosis; SUI, number of levels superior to the upper limit of instrumentation; T1, T1 slope; TK, thoracic kyphosis; TL, thoracolumbar scoliosis.

It was found that the models that best explained the data were those that measured cervical sagittal alignment as the sum of the alignment from the occiput to C7, rather than just from C2 to C7. Consequently, the C2 to C7 models are not further described.

The pre‐operative model of

$$\begin{array}{c}{\mathrm{CSA}}_{\mathrm{pre}}=-21.02+0.58\left(\mathrm{T}{1}_{\mathrm{pre}}\right)-1.43\left({\mathrm{TK}}_{\mathrm{pre}}\right)-0.44\left({\mathrm{LL}}_{\mathrm{pre}}\right)\hfill \\ -0.29\left({\text{GSVA}}_{\mathrm{pre}}\right)+E\hfill \end{array}$$

showed that cervical sagittal alignment was related to the T1 slope, thoracic kyphosis, lumbar lordosis and the sagittal vertical axis with an R² value of 78%. The pre‐operative model was not related to a measure of scoliosis at any anatomical spinal level. As the model demonstrates, an increasing T1 slope, increasing kyphosis, a decreasing lordosis (noting that lordosis is by definition a negative value) and increasing SVA lead to a more negative cervical spinal alignment (cervical lordosis). As can be seen, the largest component is thoracic kyphosis.

In the post‐operative model,

$$\begin{array}{c}{\mathrm{CSA}}_{\text{post}}=10.09–1.02\left(\mathrm{T}{1}_{\text{post}}\right)-0.70\left({\mathrm{TK}}_{\text{post}}\right)+0.13\left({\mathrm{LL}}_{\text{post}}\right)\hfill \\ +0.54\left({\mathrm{TL}}_{\text{post}}\right)+E\hfill \end{array}$$

the cervical sagittal alignment was related to the T1 slope, thoracic kyphosis, lumbar lordosis and the thoracolumbar scoliosis with an R² value of 63%. Again, the model demonstrates that increasing thoracic kyphosis negatively increases the cervical sagittal alignment (cervical lordosis). However, increasing T1 slope and increasing lumbar lordosis now has the opposite effect to that of the pre‐operative model. In the post‐operative model, there is also a positive effect from the amount of thoracolumbar scoliosis, but no effect from scoliosis either in the proximal or mid‐thoracic spine. Of note, the levels of instrumentation, as either the number of un‐instrumented or instrumented vertebrae, were not found to be significant. Global SVA was not found to be a significant parameter in the post‐operative model. The size of the cervical SVA was not significant in either the pre or post‐operative models.

4. DISCUSSION

The anatomical description of the spine, as a series of curves, is noted to be lordotic in both the cervical and lumbar spines and kyphotic in the thoracic spine (Sinnatamby, 2011). Reassessment of the cervical sagittal alignment in those without any spinal deformity has noted that there is a range of alignment that includes kyphotic, neutral and lordotic cervical spines, all of which would be described as representing normality (Khalil et al., 2018; Kim et al., 2018; Theologis et al., 2019; Virk et al., 2020). It is also relevant to note the coupled relationship between the shapes of the cervical and thoracolumbar spine such that a change in the shape of the thoracolumbar spine is seen to have a direct effect on the shape of the cervical spine (Gardner, Archer, et al., 2021; Roussouly et al., 2005).

Adolescent idiopathic scoliosis (AIS), as a three‐dimensional deformity of the spine (Stokes, 1994), is associated with a change in the sagittal alignment of the thoracic and lumbar spines (Schlösser et al., 2016). The whole spine acts to maintain the head over the pelvis (Gardner, Archer, et al., 2021). The cervical spine is reported to be less lordotic in AIS than when compared to those without spinal deformity and can be classed as kyphotic (Tang et al., 2019). A number of papers in the literature have reported correlation analysis between the cervical sagittal alignment and other parameters that describe the sagittal alignment of the spine in AIS (Fan et al., 2020; Pepke et al., 2019; Yagi et al., 2014; Yan et al., 2018), demonstrating a variable amount of positive correlation between the amount of cervical sagittal alignment, thoracic kyphosis and lumbar lordosis. Unfortunately, these analyses do not explain the totality of the relationship with all parameters viewed together as ‘one spine’, rather they are a series of correlations of one parameter against another. It is also of note that the presentation of mean and median values across a group of individuals is of little use when performing an analysis of multiple parameters (as used in this paper), as mean and median values do not predict an individuals result.

The use of a multiple regression technique allows the identification of the statistically significant parameters, and the magnitude of their effects, on the cervical sagittal alignment in both the pre‐operative and post‐operative adolescent with AIS. The use of AIS adds another factor, in addition to the known relationship between the cervical and thoracolumbar spines in those without scoliosis, to further investigate any compensatory response of the cervical spine to changes of shape in the anatomically inferior thoracolumbar spine. Further, the adding a fixed segment, as occurs during surgery, adds to this analysis.

In the model of pre‐operative individuals, cervical sagittal alignment, measured between the occiput and the C7 vertebral body, is associated with other measures of sagittal alignment, namely the T1 slope, thoracic kyphosis, lumbar lordosis and global SVA. Of note, cervical SVA, pelvic incidence and all measures of scoliosis were not found to be significant in this model. The model for the cervical sagittal alignment in pre‐operative individuals with AIS has an adjusted R² value of 78%. In post‐operative individuals, the model explains the cervical sagittal alignment less well, with an adjusted R² value of 63%. The significant parameters were the post‐operative T1 slope, thoracic kyphosis, lumbar lordosis and thoracolumbar scoliosis. Non‐significant parameters were both cervical and global SVA, proximal and mid‐thoracic scoliosis and pelvic incidence. The number of mobile or fused vertebral segments was also not found to be significant.

It is understandable that the sagittal alignment of the thoracic and lumbar spines have a direct effect on the sagittal alignment of the cervical spine. This concept is encapsulated in the ‘cone of economy’ described by Dubousset (Dubousset, 1994) and subsequently demonstrated using pressure analysis of stance (Haddas & Lieberman, 2018). The greatest effect in both models is from the thoracic kyphosis, which is unsurprising given the anatomical proximity to the cervical spine. Whilst not a co‐linear variable, the amount of T1 slope would also be related to the thoracic kyphosis, although noting that the T1 slope is a more focal measure of the upper thoracic spine. A positive global SVA would be represented with an increasing lordosis of the cervical spine, as when an individual leans forwards, there is a requirement to extend the cervical spine to maintain horizontal gaze. Overall, this demonstrates that despite the abnormal 3D spinal shape that occurs with AIS, the sagittal shape of the cervical spine is related to the size of the sagittal curves inferior (thoracic kyphosis and lumbar lordosis) along with the amount of forward lean of the trunk. The T1 slope is also key, noting that a more vertical angle leads to a greater curve in the cervical spine.

In the post‐operative model, the largest covariate is that of T1 slope. This is likely to reflect the fact that inferior to T1 is an instrumented fusion of a large amount of the thoracic and upper lumbar spine. As this spine is instrumented, the amount of thoracic kyphosis, and to some degree, lumbar lordosis is fixed and cannot change. Thus, any compensation must occur elsewhere across the spine and also accommodate the fused segment. This may well also explain why the effect of lumbar lordosis in the post‐operative model is opposite to that of the pre‐operative model. This is also the likely reason for the difference of fit of the models, as reflected by the R² value. It is likely that the measures of the number of instrumented or un‐instrumented vertebrae were not found to be significant in the post‐operative model as the number of levels is not discriminative enough when compared to the range of values for kyphosis and lordosis. Again, this demonstrates the ability of the whole spine to adjust shape allowing for a segment that has become immobile. It is not clear why the amount of scoliosis in the thoracolumbar spine is a coefficient in this model. This may reflect the practice of placing the end of the fusion in the mid‐lumbar spine (L3) to maximise movement, accepting a residual amount of deformity. This is in the knowledge that the more of the lumbar spine is fused, the less chance there is of the individual returning to their pre‐operative levels of sport and activity (Fabricant et al., 2012).

The ability to conduct a similar analysis of the parameters of shape of the thoracolumbar spine and the relationship to the shape of the cervical spine in those without a spinal deformity is more difficult to define. This is partly due to the ethical issues of the use of ionising radiation in those without pathology (Kleinerman, 2006). However, the literature does assess the alignment of the cervical spine in asymptomatic individuals. A systematic review concentrating on the shape of the cervical spine based on the sagittal shape from C2 to C7 and the size of the T1 slope identified three subtypes labelled as the kyphotic curve cohort, the medium lordosis cohort and the large lordosis cohort (Virk et al., 2020). Unfortunately, this review did not examine the rest of the thoracolumbar spine. Theologis et al. (2019) analysed the cervical sagittal alignment based on the previously published sagittal shapes of the normal thoracolumbar spine defined by Roussouly et al. (2005). This analysis was based on an adult group (mean age 49 ± 16 years) and demonstrated that there was a difference between the cervical spine alignment and the subtype of thoracolumbar alignment of the spine. This relationship was not explored further. Kim et al. (2018) also examined an adult population (mean ages between 38 and 47 years) noting that there were a number of sagittal shapes to the cervical spine which were correlated with lumbar lordosis. Kim states in this paper that the cohort sampled are ‘currently asymptomatic’. No conclusions are drawn as to whether any of the defined shapes are more likely to become symptomatic over time and what those symptoms might be.

Analysis of the sagittal shape of the cervical spine is less common in children and adolescents. Lee et al. (2012) analysed 181 asymptomatic children correlating cervical shape to the shape of the thoracolumbar spine. In this cohort, cervical kyphosis was found in approximately 40% and that cervical shape was correlated with the amount of thoracic kyphosis.

Were the data available in a form that could be analysed in the way performed in this paper, then there would be relevant comment that could be made around a more statistically rigorous analysis of spinal shape in asymptomatic normals, rather than just correlation. However, what is apparent is that the sagittal shape of the cervical spine can be correlated to that of the rest of the thoracolumbar spine in those without AIS.

The clinical relevance of this work is related to the understanding of the shape of the spine in AIS and how surgery changes that. Upright stance relies on a harmonic relationship between all components of the sagittal profile. With an alteration in sagittal profile, secondary to the reduction in thoracic kyphosis caused by AIS (Newton et al., 2019) and then changed by any subsequent surgery (Schlösser et al., 2021), the post‐operative uninstrumented and still mobile spine will be recruited to maintain upright stance with an appropriate head position and horizontal gaze. Whilst statistical in nature, these models will allow a greater understanding of how the spine has the ability to compensate, even whilst performing a scoliosis correction, through the aspects of the spine that are directly modifiable by the surgeon. Proximal junctional kyphosis and failure will occur when the sagittal shape of the spine is not balanced and harmonious (Hart et al., 2013).

The models reflect cervical sagittal alignment as a measure between the base of the skull and the inferior end plate of the vertebral body of C7 rather than the inferior end plate of C2 to the inferior endplate of C7. Using the R² value, the models explained the cervical sagittal alignment more completely when all of the cervical spine was included rather than only the sub‐axial spine. Previous papers have used C2 as the upper limit to measurement and drawn conclusions on the relationships of the sagittal parameters based on that (Hayashi et al., 2017; Norheim et al., 2015; Yan et al., 2018). The work presented here would suggest that this approach may be flawed and measures of sagittal alignment parameters need to include the cervical spine to the occiput; this agrees with the approach taken by Berger et al. (2018).

The findings of the work presented here are in line, although presented in a slightly different fashion, to the previous literature. Clement et al. (2021) only measured sagittal parameters in those with AIS, without any measure of the size of the scoliosis, and divided the shape of the cervical, thoracic and lumbar spines into two different arcs based on horizontal lines. The surgery described only fused the thoracic spine. However, most of the change in inferior cervical spine shape (60%) was related to the thoracic kyphosis and this was defined as a regression equation documenting how the change of thoracic kyphosis altered the inferior cervical spine. Li et al. (2021) identified that in 124 AIS patients with major scoliotic curves in the thoracic spine, the post‐operative sagittal alignment of the spine is related to the thoracic kyphosis and T1 slope. Failure to restore thoracic kyphosis and T1 slope along with a pre‐operative cervical kyphosis are factors for cervical decompensation. The work by Roussouly et al. (2013) confirms the pre‐eminence of the shape of the thoracic spine in determining the shape of the cervical spine and any decompensation of a balanced spinal alignment.

The strengths of this paper are the use of regression techniques, rather than correlation analysis, to identify and quantify statistically significant parameters that explain the relationship with the whole spine as a ‘linked system’, which can also be used for prediction, as opposed to correlation analysis of individual parameters with no predictive ability. It is acknowledged, however, that a different data set may change the parameters that are significant and the size of the coefficients in the model and that other, as yet unidentified, factors must play a role in explaining the remaining variance for both models, particularly in the post‐operative setting. It must also be acknowledged that the size of the coefficients may well change with the addition of different curve morphologies (Lenke types) as the models presented here are based on the most common types of scoliosis, the main thoracic curve (Lenke 1) and the double major curve (Lenke 3) (Lenke et al., 2001). As the other curves types are far less commonly seen, a multi‐centre data collection would be required to have a large enough group from which to draw any sensible conclusions.

5. CONCLUSION

In conclusion, a more thorough understanding of the relationships that inform the sagittal shape of the cervical spine in AIS is key to the appreciation of the effects of scoliosis, and the subsequent effects of surgery. This will add help to clarify surgical interventions of the future and potentially help to prevent longer‐term issues in the future ageing population from sagittal misalignment of the cervical spine (Scheer et al., 2013).

CONFLICT OF INTEREST

There are no competing interests to report.

AUTHOR CONTRIBUTIONS

AG designed the study, acquired the data, conducted the analysis and wrote the first draft of the paper. FB designed the study and critically revised the manuscript. PP provided statistical analysis guidance and oversight and critically revised the manuscript. All authors approved the final submission.

CONSENT FOR PUBLICATION

No consent for publication is required.

ETHICS APPROVAL AND CONSENT TO PARTICIPATE

This study was performed in accordance with all the relevant guidelines and regulations. Ethical approval for this work was given by the National Research Ethics Service UK (ref: Northampton 15/EM/0283). As this is a retrospective review of previously collected routine care data, there was no requirement from the Research Ethics Service to seek individual informed consent to access the images used in this study.

Supporting information

Data S1

Gardner, A. , Berryman, F. & Pynsent, P. (2022) Statistical modelling of how the sagittal alignment of the cervical spine is affected by adolescent idiopathic scoliosis and how scoliosis surgery changes that. Journal of Anatomy, 241, 437–446. Available from: 10.1111/joa.13662

DATA AVAILABILITY STATEMENT

The datasets used and/or analysed during the current study are available as the supplementary file.