Radiographic Characteristics of Cervical Deformity (CD) Using a Discriminant Analysis: The Value of Extension...

This was a retrospective review of a prospectively collected database.

The aim of this study was to delineate radiographic parameters that distinguish severe cervical spine deformity (CSD).

Our objective was to define parameters that distinguish severe CSD using a consensus approach combined with discriminant analysis as no system currently exists in the literature.

Twelve CSD surgeons reviewed preoperative x-rays from a CSD database. A consensus was reached for categorizing patients into a severe cervical deformity (sCD), non-severe cervical deformity (non-sCD), or an indeterminate cohort. Radiographic parameters were found including classic cervical and spinopelvic parameters in neutral/flexion/extension alignment. To perform our discriminant analysis, we selected for parameters that had a significant difference between the sCD and non-sCD groups using the Student t test. A discriminant function analysis was used to determine which variables discriminate between the sCD versus non-sCD. A stepwise analysis was performed to build a model of parameters to delineate sCD.

A total of 146 patients with cervical deformity were reviewed (60.5±10.5 y; body mass index: 29.8 kg/m2; 61.3% female). There were 83 (56.8%) classified as sCD and 51 (34.9%) as non-sCD. The comparison analysis led to 16 radiographic parameters that were different between cohorts, and 5 parameters discriminated sCD and non-sCD. These parameters were cervical sagittal vertical axis, T1 slope, maximum focal kyphosis in extension, C2 slope in extension, and number of kyphotic levels in extension. The canonical coefficient of correlation was 0.689, demonstrating a strong association between our model and cervical deformity classification. The accuracy of classification was 87.0%, and cross-validation was 85.2% successful.

More than one third of a series of CSD patients were not considered to have a sCD. Analysis of an initial 17 parameters showed that a subset of 5 parameters can discriminate between sCD versus non-sCD with 85% accuracy. Our study demonstrates that flexion/extension images are critical for defining severe CD.