Introduction

The cervical spine is under a series of loads and allows for a large range of motion, in the location between the inflexible thoracic spine and the cranium with absorbing torsional, axial and shear loads. The cervical alignment is important for the maintenance of proper horizontal gaze and for the compensation of more caudal spinal alignment changes [1, 2]. It is vital to recognize the normal variations in cervical spine sagittal profile which in turn will help determine the optimal cervical spine alignment to maximize function while minimizing complications [3].

Numerous studies have described the “normal” shape of the cervical spine and defined the normal alignment as lordosis [4, 5]. However, some reversed curve on radiographs was emphasized in asymptomatic patients by historical studies. A meta-analysis of 15,364 asymptomatic subjects even found that approximately 36% of them had non-lordotic cervical spines [6]. Hence, unlike the lumbar spine which should almost invariably be lordotic, cervical spine alignment varies considerably [7]. In addition, the physical cervical alignment in normal subjects could be various by gender and with degeneration in different age groups, while little of the literature had focused on demographic factors linked with cervical lordosis [8].

Furthermore, restoration of the cervical spinal sagittal profile has been recognized as an important objective, but cervical spine alignment targets are less well established partly due to the large variation in cervical spine alignment as well as the conflicting evidence [2, 9, 10]. In contrast, early studies have focused on achieving lumbar spinal balance as determined by the relationship between lumbar lordosis and pelvic incidence [11]. Here, thoracic inlet angle (TIA) [12, 13], reported as an intrinsic parameter analogous to pelvic incidence, was expected to be introduced, together with C2-7 CL, to characterize the cervical spinal balance. Therefore, this study was to propose and characterize a novel classification for cervical spine morphologies with Chinese asymptomatic subjects in cervical spine, and to address the knowledge gap for cervical balance based on the classification.

Material and methods

It was a single-center retrospective study from January 2020 to December 2022. The cohort of healthy individuals (from general population) by physical examination having received cervical spine X-ray was selected from our center. The included criteria were participants (1) from the asymptomatic group without any discomfort on the spine (e.g., neck pain, radiculopathy or back pain); (2) with the age of over 20 years; (3) standing with the head in a neutral position (horizontal gaze), and the knees and hips fully extended; (4) with the lateral cervical spine radiographs in a neutral weight-bearing position. The excluded criteria were: (1) cervical fusion; (2) cases with neck pain or cervical-related symptom; (3) the T1 superior endplate and manubrium sterni not clearly visible on radiographs; (4) a history of spinal surgery especially on cervical segments; (5) with spinal trauma, deformity or malignancy; (6) other systemic spinal diseases such as ankylosing spondylitis and rheumatoid spondylitis and (7) pregnancy. In addition, neck disability index (NDI) (0 ~ 30 score) was also introduced in this study to reflect the asymptomatic nature of this population. No strict cutoffs were used for baseline NDI because the exclusion criteria were sufficient in identifying relevant cervical spine pathology.

All patients have signed informed consent and this study was approved by the Institutional Ethics Committee of our hospital.

Sagittal radiological parameters in this study characterized whole cervical spine alignment, segmental morphology and cervical balance status. Cervical alignment was expressed by C2-7 cervical lordosis (C2-7 CL). Segmental morphology was measured by C1-2 CL, C2-4 CL and C5-7 CL as well as C2 slope (C2S). Cervical balance status was featured by T1 slope (T1S), T1S minus C2-7 CL (T1S-CL), cervical sagittal vertical axis (CSVA), TIA and neck tilt (NT), as well as CBVA (Fig. 1). C2-7 CL was regarded as the primary outcomes and others were the secondary outcomes. For all angle parameters, the lordosis angle was set as the positive values.

Fig. 1
figure 1

Measurement of radiological parameters of cervical spine in neutral positions. C2-7 CL: the angle between the lower endplate of C2 and the lower endplate of C7; C1-2 CL: the angle between C1 and the lower endplate of C2; C2-4 CL: the angle between the lower endplate of C2 and C4; C5-7 CL the angle between the upper endplate of C5 and the lower endplate of C7; C2S: the angle between a horizontal line and the inferior endplate of C2; CSVA: axis from upper posterior corner of C7 body to plumb of C2 center; TIA: the angle between a vertical line from the center of the T1 upper endplate, and a line connecting the center of the T1 upper endplate and the upper end of the sternum; NT: the angle formed by a vertical line from the sternum tip and a line connecting the center of the T1 upper endplate and the upper end of the sternum

Full size image

All parameters, acquired on the picture archiving and communication system (PACS), were measured three times by two independent authors and the average score was used in analysis. Intra-observer reproducibility of these measurements was explored with the intraclass correlation coefficient (ICC). On inter-observer reliability, the ICC with 95% CI was also identified, comparing the mean of all three measurements from three observers. ICC <|0.40| indicated poor results; |0.40| to |0.75| was fair to good, and |0.75| to |1.00| was excellent reliability.

Eventually, a total of 632 participants (F:M = 364:268) were eligible for this study. The group was with a mean age of 54.82 ± 16.56 (y) (20–92 y), where the age group of 60–70 y was the most (28.4%) while the 40–50 y group was the least (10.6%). Intra-observer reproducibility and inter-observer reliability using ICC for all radiological parameters showed good to excellent agreement (Table 1).

Table 1 Intra-observer reproducibility and inter-observer reliability using ICC for all parameters
Full size table

All measurement data were expressed by mean ± standard deviation. In addition, all subjects were divided into six age groups for analysis to outline the trend in cervical spinal parameters with age, which contains 20–29 year age group, 30–39, 40–49, 50–59, 60–69 and over 70 year age groups.

To identify groups of patients with similar cervical alignment parameters, a 2-step cluster analysis was performed using a combination of hierarchal cluster and k-mean cluster analysis for all participants. Of note, the 2-step cluster analysis is a natural way to select the number of groups and criteria for groups within a dataset, where measurement data were automatically standardized. The distance was calculated with the log-likelihood method and the number of clusters was determined automatically with use of the Bayesian information criterion. The only limitation that we imposed was a maximum number of clusters of 15 for computation timing purposes. The clustering average contour value was addressed by Silhouette value, which larger than 0.5 means the clustering was reasonable.

The one-way analysis of variance (ANOVA) method was used to compare all parameters among various age-group participants and among different clusters. Pearson’s correlation analysis was utilized for the C2-7 CL and secondary outcomes, and then, the multiple linear regression analysis (C2-7 CL as the dependent variable) was performed to determine the influencing factor for cervical spine alignment. Notably, as TIA was regard as the intrinsic anatomical parameters [12], like pelvic incidence in lumbo-pelvic region, so the post hoc analysis on the relationship between C2-7 CL and TIA, analogous to pelvic incidence and lumbar lordosis, was established for various clusters. And then, the multinomial logistic regression analysis was performed to identify the disparities of these parameters for these clusters. The statistical analysis was performed using SPSS 22.0 (International Business Machines Corporation, Armonk, NY, USA) and statistical significance was defined as P value < 0.05.

Results

The mean C2-7 CL was 12.9 ± 12.2 and C1-2 was 32.1 ± 6.2, and the mean T1S, CSVA and TIA was, respectively, 25.3 ± 8.4, 20.2 ± 9.7 and 73.3 ± 10.9. Totally, the parameters of C2-7 CL, T1S, CSVA, C2-4 CL, TIA and NT increased by age while T1S-CL and C2S decreased by age, but not for C1-2 CL (P = 0.178) and C5-7 CL (P = 0.826). The mean value of C2-7 CL, T1S, CSVA, TIA and NT was larger in male than female while C1-2 CL (P = 0.009) was larger in female group (Table 2 and Fig. 2).

Table 2 The cervical spinal radiological parameters and the asymptomatic nature of participants in four cluster groups
Full size table
Fig. 2
figure 2

The measurements of cervical spinal parameters in various age groups, by gender. Data were expressed by mean + standard error

Full size image

We included sagittal radiological parameters as well as gender and age as the variables in our 2-step cluster analysis. The most important predictor variables were gender and C2-7, followed by C2S and T1S-CL, while age and NT were less important. Eventually, we found four unique clusters of subjects mainly based on gender and C2-C7 CL, which were female lordotic cluster (FLC), female kyphotic cluster (FKC), male lordotic cluster (MLC) and male kyphotic cluster (MKC) (Fig. 3a). Four similar clusters were also found when the 2-step cluster analysis was verified with use of the Akaike information criterion. The silhouette measure of cohesion was > 0.5, which was indicative of a good fit for our clusters. A scatterplot detailing the spread of each cluster in terms of gender and C2-C7 CL is shown in Fig. 3b. Among the four groups, C2-C7 CL was larger in male, so was the trend of T1S and TIA. The mean age was larger in lordosis groups than kyphosis groups. The mean NDI for all subjects was 2.91 ± 2.85 and there were no differences among the four clusters (P = 0.787). Other information is shown in Table 2 and Fig. 3b (Table 2 and Fig. 3c).

Fig. 3
figure 3figure 3

The classifications of cervical spinal alignment by cluster analysis. a The ranking of importance of all predictor variables based on radiological parameters as well as gender and age; b The distribution of all individuals by cluster analysis based on gender and C2-7 CL; c The differences of cervical spinal parameters among four clusters of FLC, FKC, MLC and MKC. FLC: female lordotic cluster; FKC: female kyphotic cluster; MLC: male lordotic cluster; MKC: male kyphotic cluster. “—” means P < 0.001; “–-” means P < 0.01; “…..” means P < 0.05

Full size image

For all cluster groups, C2-7 CL was positively related to T1S, C2-4 CL and C5-7 CL, as well as to TIA, while it negatively related to T1S-CL and C2S. The C2-7 CL was correlated to the age in FLC, FKC and MLC but not to MKC. NDI was not correlated to C2-7 CL for all population (Table 3). Furthermore, it showed that T1S was the independent influencing factor (P < 0.001) for C2-7 CL in all individuals after collinearity discrimination with multiple linear regression analysis (Table 4), which held the similar trend in separate cluster (data not shown). In addition, T1S was only determined by TIA, the intrinsic anatomical parameters, for all clusters. Therefore, there was the formula for all individuals that C2-7 CL =  − 14.25 + 1.07 × T1S and T1S =  − 13.46 + 0.53 × TIA, so there was C2-7 CL =  − 28.65 + 0.57 × TIA. For FLC group, C2-7 CL =  − 0.42 + 0.7 × T1S and T1S =  − 1.7 + 0.38 × TIA, and C2-7 CL =  − 1.61 + 0.27 × TIA. The similar computational process was performed in the other three clusters, and the formula were C2-7 CL =  − 16.65 + 0.28 × TIA in FKC, C2-7 CL =  − 10.89 + 0.248 × TIA in MLC and C2-7 CL =  − 22.76 + 0.375 × TIA in MKC, respectively (Fig. 4).

Table 3 Correlation between C2-7 CL and secondary parameters by Pearson analysis
Full size table
Table 4 Influencing factors on C2-7 CL by multiple linear regression analysis for all individuals
Full size table
Fig. 4
figure 4

The scatter diagram on the relationship among C2-7 CL, T1S and TIA. a The relationship between C2-7 CL and T1S of 4 clusters; b The relationship between T1S and TIA of 4 clusters. FLC: female lordotic cluster; FKC: female kyphotic cluster; MLC: male lordotic cluster; MKC: male kyphotic cluster

Full size image

When compared to MKC by multinomial logistic regression analysis, it showed that T1S was the key positive factor in FLC and MLC group, which was, respectively, 1.620 and 1.736 times in forming lordotic cervical shape in the two clusters than MKC. C1-2 CL was the negative factor in lordotic groups compared to MKC (RR = 0.793 and RR = 0.730, respectively), while C2-4 was the positive factor for lordotic groups. In addition, C2S was relatively less in FLC (P = 0.019) and MLC (P = 0.006) compared to kyphotic groups. In FLC group, the possibility for forming lordotic cervical shape in 70 + age group was as high as 250 times than that in 20 + age group (P < 0.001), almost 17 times and 13 times than 30 + and 40 + age groups (P = 0.012 and P = 0.016, respectively). In MLC group, the possibility in 70 + age group was 66.7 times higher than 20 + age group (P = 0.001) but not significant than other age groups. In FKC groups, it was, respectively, 3.9 and 3.4 times higher than 20 + and 30 + age groups (P = 0.012 and P = 0.023, respectively) (Table 5).

Table 5 The risk factors for various clusters compared to MKC cluster group by multinomial logistic regression analysis
Full size table

Discussion

The physiological significance of cervical spinal alignment has long been a topic of interest in the medical community. It was previously believed that there was a strict correlation between the absence of symptoms and the C-shaped curvature of the cervical spine [14, 15]. However, recent literature suggests that asymptomatic individuals exhibit a variety of curvature types, and attempts have been made to characterize these different types [16]. In that condition, the cervical spine took one of five distinct sagittal profiles with lordosis, neutral, kyphosis, S-shaped and inverted S-shaped by some authors [17, 18]. Another recent study performed cluster analysis and identified three morphotypes of kyphotic curve cohort, medium lordosis cohort and large lordosis cohort based on C2-7 CL and T1S [16]. In our study, the asymptomatic individuals of different age groups in China were conducted to utilize cluster analysis to divide the population into four groups based in terms of gender and C2-7 CL.

As reported by most literature, the cervical lordosis, especially C2-7 CL, was the most commonly used descriptive parameter. Overall, we found that C2-7 CL gradually increased with age while C5-7 CL remained relatively stable in each age group, as well as the upper cervical spine. Firstly, the degenerative of cervical disk, facet and narrowed intervertebral foramen in the elderly resulted to a relative shortened posterior column [19]. In addition, with the population aging, the thoracic kyphosis increased, where there was the interaction between enlarged cervical curvature and the degenerative thoracic kyphosis for the compensatory of the sagittal balance and horizontal gaze [20]. Then, the straightening of cervical spinal curvature in younger cohort was nation widely proved. In this study, the individuals were with 7.8 ± 11.9 C2-7 CL in 20–30 age group while 17.1 ± 13.8 in 70 plus group, and the former group was significantly lower than ever reported although they were regarded as “normal” range. Probably, the postural habits of modern society such as the need for computer gazing and reading may be responsible for the change [21]. However, further research is needed to determine whether the straightened curvature in young people would be associated with clinical symptoms with prolonged time.

In our data, four groups were naturally classified by cluster analysis and C2-7 cervical alignment and gender were the primary factors, where most parameters also showed significant difference among clusters. In other words, neither the lordotic nor non-lordotic sagittal profiles are pathological to an extent. Hey et al. [18] concluded that only 27% of their asymptomatic cohort had cervical spine lordosis. Okada et al. [22] classified their cohort of asymptomatic volunteers into lordotic and non-lordotic subtypes, they found only 45% of subjects under 40 years and 56% of total subjects had lordotic cervical spines. It might be linked with the concept of ligamentous muscular counterbalancing, where the natural tendency of every individual was proposed to adopt energy-conserving postures and there was a resultant impetus toward kyphosis to rely on the posterior tension band to the spine. Therefore, cervical lordosis may not truly be physiological in every individual [23], and better outcomes may be achieved with patient-specific realignment targets.

Few reports on gender-related differences among cervical spine alignment have been published, although more differences among sagittal spinal and spino-pelvic parameters were determined [8, 24]. Here, we featured a gender disparity on cervical alignment and C1-2 contributes more to cervical curvature in female, while C2-7 contributes more in male. Totally, the cervical sagittal parameters in female were slightly smaller than the corresponding parameters in males. It was assumed that this phenomenon might due to disparity in physiological structure, body size and degree of spinal degeneration. In a previous study, Baoge et al. [25] conducted a small sample cohort study, demonstrating that cervical alignment parameters in female subjects were lower than that in male subjects. Greaves et al. [26] showed that cervical spine kinematics in females have a more anterior helical axis of motion in flexion–extension compared to males. Another analysis on histological osteoarthritic changes in cervical spine facet joints demonstrated that the osteocartilaginous junction and the subchondral bone affected males more severely, which would result to the larger thoracic kyphosis and cervical lordosis [27].

As described before, the wide range of the cervical spine mobility allows it to compensate for alignment changes caudally. T1S is consistently regarded to correlate with cervical alignment, where a larger T1S corresponds to more lordotic curvature in asymptomatic individuals [18, 28]. T1S is not a simple extension of adjacent cervical spine, but also reflects the degree of thoracic kyphosis, thus many literature characterized cervical spinal profile by both C2-7 CL and T1S. Similar to others, we also found that T1S was closely correlated with C2-7 CL and proved as the independent influencing factor for cervical lordosis. The cervical spinal balance was expressed by known parameters such as T1S, CSVA and T1S-CL. T1S-CL is an angle that is used to compare cohorts but with no papers providing normative values, so it is not a useful parameter in clinical practice [29]. T1S, with the average value 20, must not be higher than 40 while CSVA must be less than 40 mm. In this study, the mean CSVA was 20.2 ± 9.7 mm and most individuals (97.5%) was in the normal range, but there was unidentified relationship between CSVA and C2-7 CL, which was probably attributed to the variety in different age groups and to the cross-sectional survey instead of the cohort detection. There has also been interest in finding a morphological parameter for the cervical spine, as we extrapolate the idea of the pelvic incidence or cervical incidence to the cervicothoracic region. So TIA equaling NT plus T1S described this angle was introduced [30]. Here, we established a formula between cervical lordosis and TIA for various clusters, which may provide reference for certain individuals in cervical reconstruction surgery.

The multinomial logistic regression generalized the characteristics for different clusters for further step. In this process, we found that T1S played more important role in lordotic group (FLC and MLC), which implied a more familiar relationship between T1S and cervical lordosis in commonly believed “normal” cohorts. C1-2 lordosis angle seemed more prominent and functional in lordotic groups to maintain horizontal gaze, the similar trend for C2-4 CL with larger proportion and flexibility in these individuals [31]. In addition, another general regular was revealed that the degree of lordosis increased more rapidly with unit age group in female and lordotic cluster, while the further mechanism needed to be determined. Totally, it is speculated that the alignment of the cervical spine is more coordinated among different parts in lordotic groups since more positive parameters have been taken part in this process.

This study characterized different cervical curvature by cluster analysis and further emphasized the non-lordotic sagittal profiles as the “normal” and “non minority” pattern. Furthermore, the use of cluster analysis to classify individuals, especially based on gender, provided a meaningful framework for future research. It revealed the perspective features and influencing factors for each cluster, and provided guidelines for correcting the alignment into reasonable range by personalized cervical surgery. Then, a greater understanding of cervical alignment and its relationship will promote the interaction and biomechanics among different parts and the overall balance.

There were limitations to be noted: The individuals enrolled were without symptoms in cervical spine but not the real healthy condition, which may put impact on the mechanical chain of the whole sagittal alignment as well as the cervical region. Then, the dynamic parameters instead of static balance of the cervical spine were emphasized recently while the measurements in this study only revealed the static regular. In addition, only the cervical parameters data were collected while the parameters of the whole spinal alignment and the pelvis were missing, which may discount on the elaboration of the whole spine and spino-pelvic bio-mechanical features. Finally, although the relationship between cervical curvature and symptoms is not as straightforward, recent research has implied the importance of cervical curvature when assessing an individual's risk for developing neck pain especially in kyphotic groups, which need a certain cohort to be observed.

Conclusion

The cervical sagittal profile, especially C2-7 curvature, gradually varied with age, and the mean age of C2-7 CL and T1S was larger in male than female group. Four clusters of female lordotic cohort, female kyphotic cohort, male lordotic cohort and male kyphotic cohort were naturally classified based on C2-7 CL and gender. The cervical balance status was addressed by linking C2-7 CL and TIA with C2-7 CL =  − 28.65 + 0.57 × TIA, which varied among four different clusters. Although there were interactions among cervical alignment parameters, it seemed that the alignment was more coordinated among different parts in lordotic groups.