The Correlation Between Gender, Age, Curvature, and Symptom-related Changes in C6 and C7 Slope in 10,000 Subjects.

Abstract

Study design: A cross-sectional analysis of 10,000 cervical spine X-rays.

Objective: This study investigates the variations in C6S and C7S across demographic factors (gender, age, cervical curvature, and symptoms) and explores their correlation. In addition, machine learning models are applied to improve the accuracy of C7S prediction.

Summary of background data: The C7S is crucial for assessing cervical balance but is often limited by visibility issues. This study uses a large sample to validate the feasibility of the C6S as a substitute for C7S across diverse populations with varying ages, genders, symptoms, and cervical curvatures.

Materials and methods: A retrospective study was conducted on 10,000 subjects who underwent cervical sagittal X-ray imaging. Four orthopedic specialists labeled key points, which were cross-validated, and an algorithm was then used to measure C6S and C7S. Pearson correlation coefficients were calculated to assess the relationship between C6S and C7S, and linear regression derived a predictive equation for C7S. Various machine learning models were compared with improve C7S prediction accuracy.

Results: The average angles for C6S and C7S were 15.4° (16.8° in males, 14.7° in females) and 19.1° (21.1° in males, 18.2° in females), respectively, with C7S generally larger than C6S, except in Sigmoid 1 curvature. Males exhibited higher values for both C6S and C7S, and both slopes increased after age 20. Both angles increased significantly with age from 20 to 90 years. A strong positive correlation was found between C6S and C7S ( r >0.75, P <0.001), confirmed by linear regression ( R2 ​​​​​​=0.688). Among the machine learning models, both Ridge regression and linear regression performed better than the others, with R2 =0.855 in predicting C7S.

Conclusion: The strong correlation between C6S and C7S suggests that C6S can substitute for C7S when visibility is limited. Machine learning models further enhance prediction accuracy, demonstrating promising clinical potential.