A Nomogram for Predicting Late-Onset Neurological Deficits in the Natural Course of Kyphosis or Kyphoscoliosis.

STUDY DESIGN

Retrospective single-center comparative analysis.

OBJECTIVE

To develop a nomogram model for predicting late-onset neurological deficits (LONDs) in patients with kyphosis or kyphoscoliosis.

BACKGROUND

Patients with kyphosis or kyphoscoliosis might suffer from LONDs, and surgical correction may improve neurological function. Nevertheless, there exists a significant gap in the identification of predictive factors for LONDs in these patients.

PATIENTS AND METHODS

A consecutive series of 244 patients with kyphosis or kyphoscoliosis who underwent corrective surgery between April 2010 and June 2024 were included in our study. Relevant measurements, including the Cobb angle, deformity angular ratio, and level of the apex were assessed and calculated using x-ray imaging. Spinal cord morphology at the apex of the major curve was evaluated using preoperative axial T2-weighted magnetic resonance imaging to categorize patients into 3 types based on the spinal cord shape classification system (SCSCS). To identify independent risk factors associated with LONDs, we employed univariate analysis, followed by backward stepwise multivariate logistic regression analysis. A nomogram was established based on the identified independent risk factors to predict the likelihood of LONDs in patients with kyphosis or kyphoscoliosis.

RESULTS

The mean age of the 244 patients was 46.4 ± 17.8 years, with an observed incidence of LONDs at 57.8%. The backward stepwise multivariate logistic regression analysis indicated that age, etiological diagnosis, and SCSCS were independent predictors of LONDs. Utilizing these independent risk factors, we constructed a nomogram model to estimate the probability of LONDs. The concordance index of the model was 0.912 (95% CI: 0.876-0.947), indicating a satisfactory level of accuracy in predicting the likelihood of LONDs.

CONCLUSION

The predictive factors for LONDs include age, etiological diagnosis, and SCSCS. We developed a nomogram model to predict LONDs, which could be useful for patient counseling and facilitating treatment-related decision-making.