The morphology of cervical deformities: a two-step cluster analysis to identify cervical deformity patterns

J Neurosurg Spine. 2019 Nov 15;32(3):353-359. doi: 10.3171/2019.9.SPINE19730.


Objective: Cervical deformity (CD) is difficult to define due to the high variability in normal cervical alignment based on postural- and thoracolumbar-driven changes to cervical alignment. The purpose of this study was to identify whether patterns of sagittal deformity could be established based on neutral and dynamic alignment, as shown on radiographs.

Methods: This study is a retrospective review of a prospective, multicenter database of CD patients who underwent surgery from 2013 to 2015. Their radiographs were reviewed by 12 individuals using a consensus-based method to identify severe sagittal CD. Radiographic parameters correlating with health-related quality of life were introduced in a two-step cluster analysis (a combination of hierarchical cluster and k-means cluster) to identify patterns of sagittal deformity. A comparison of lateral and lateral extension radiographs between clusters was performed using an ANOVA in a post hoc analysis.

Results: Overall, 75 patients were identified as having severe CD due to sagittal malalignment, and they formed the basis of this study. Their mean age was 64 years, their body mass index was 29 kg/m2, and 66% were female. There were significant correlations between focal alignment/flexibility of maximum kyphosis, cervical lordosis, and thoracic slope minus cervical lordosis (TS-CL) flexibility (r = 0.27, 0.31, and -0.36, respectively). Cluster analysis revealed 3 distinct groups based on alignment and flexibility. Group 1 (a pattern involving a flat neck with lack of compensation) had a large TS-CL mismatch despite flexibility in cervical lordosis; group 2 (a pattern involving focal deformity) had focal kyphosis between 2 adjacent levels but no large regional cervical kyphosis under the setting of a low T1 slope (T1S); and group 3 (a pattern involving a cervicothoracic deformity) had a very large T1S with a compensatory hyperlordosis of the cervical spine.

Conclusions: Three distinct patterns of CD were identified in this cohort: flat neck, focal deformity, and cervicothoracic deformity. One key element to understanding the difference between these groups was the alignment seen on extension radiographs. This information is a first step in developing a classification system that can guide the surgical treatment for CD and the choice of fusion level.

Keywords: cervical deformity; cervicothoracic deformity; cluster analysis; extension alignment; flat neck; focal deformity.