A Predictive Model to Identify Treatment-related Risk Factors for Odontoid Fracture Nonunion Using Machine Learning

Iris Leister, Thomas Haider, Matthias Vogel, Jan Vastmans, Patrick Langthaler, Georg Mattiassich, Alexandra Christ, Martin Etschmaier, Alexander Eijkenboom, Julia Burghuber, Harald Kindermann, Orpheus Mach, Doris Maier, Florian Högel

Research output: Contribution to journalArticlepeer-review

5 Citations (Scopus)

Abstract

Study Design. Multicenter retrospective analysis of routinely collected data. Objective. The underlying aim of this study was to identify potential treatment-related risk factors for odontoid fracture nonunion while accounting for known patient- and injury-related risk factors. Summary of Background Data. Type II and III odontoid fractures represent the most common cervical spine fracture in elderly patients and are associated with a relatively high nonunion rate. The management of odontoid fractures is controversial and treatment strategies range from conservative treatment to extensive surgical stabilization and fusion. Methods. A total of 415 individuals who sustained odontoid fracture and were treated in either of four tertiary referral centers in Austria and Germany were included in the study. We included the following potential contributing factors for fracture nonunion in cross-validated extreme gradient boosted (XGBoost) and binary logistic regression models: age, gender, fracture displacement, mechanism of injury (high vs. low energy), fracture classification (Anderson II vs. III), presence of comorbidities (Charlson comorbidity index), and treatment (conservative, anterior screw fixation with one or two screws, posterior C1/C2 spondylodesis, cervico-occipital C0-C4 fusion). Results. In our cohort, 187 (45%) had radiologically confirmed odontoid nonunion six months postinjury. The odds for nonunion increase significantly with age, and are lower in type III compared to type II fractures. Also, odds for nonunion are significantly lower in posterior C1/C2 spondylodesis, and C0-C4 fusion compared to conservative treatment. Importantly, odds are not statistically significantly lower in the group treated with anterior screw fixation compared to conservative treatment. The factors gender, fracture displacement, mechanism of injury, and the presence of comorbidities did not produce significant odds. Conclusion. Higher age, type II fractures, and conservative treatment are the main risk factors for odontoid nonunion. Anterior screw fixation did not differ significantly from conservative treatment in terms of fracture union. Level of Evidence. 3.

Original languageEnglish
Pages (from-to)164-171
Number of pages8
JournalSpine
Volume48
Issue number3
DOIs
Publication statusPublished - 1 Feb 2023

Keywords

  • Aged
  • Fracture Fixation, Internal
  • Fractures, Bone
  • Humans
  • Machine Learning
  • Odontoid Process/diagnostic imaging
  • Retrospective Studies
  • Risk Factors
  • Spinal Fractures/diagnostic imaging
  • Spinal Fusion
  • Treatment Outcome
  • risk factors
  • cervical spine
  • fracture nonunion
  • odontoid fracture

Fingerprint

Dive into the research topics of 'A Predictive Model to Identify Treatment-related Risk Factors for Odontoid Fracture Nonunion Using Machine Learning'. Together they form a unique fingerprint.

Cite this