Related Experiment Video
Updated: Oct 16, 2026
Establishing a Competing Risk Regression Nomogram Model for Survival Data
Published on: October 23, 2020
Development of a Multivariable Prognostic Model for Risk Stratification in C1 Fractures
Patrick K Cronin1, Kaitlyn E Holly1, Rishi Virani2
1Department of Orthopaedic Surgery, Mass General Brigham, Harvard Medical School, 75 Francis Street, Boston, MA 02115 USA.
Study Design:
Retrospective analysis of adult patients with radiographically confirmed C1 fractures treated across a single healthcare system including multiple level 1 trauma centers and multiple community hospitals (2015-22). The primary outcome was instability. Peri-operative outcomes and complications were also considered.
Objective:
To evaluate radiographic and clinical predictors of instability or surgical management in patients with C1 fractures and to develop a multivariable prognostic model.
Summary Of Background Data:
Determining instability in C1 fractures remains challenging. Traditional criteria rely heavily on radiographic displacement thresholds, which may not fully capture the multifactorial nature of atlas fracture instability.
Methods:
Demographic, clinical, and radiographic variables were analyzed including lateral mass displacement, dens fracture, neurologic deficit, age, sex, and body mass index. Multivariable Firth penalized logistic regression was used to adjust for confounders and identify factors significantly associated with instability/operative treatment. Receiver operating characteristic analysis assessed model discrimination. A point-based instability risk score was derived from model coefficients.
Results:
In adjusted analysis of Two hundred and seventy-nine patients, Lateral mass displacement (OR 1.20 per mm, 95% CI 1.09-1.33, P<0.001), dens fracture (OR 2.67, 95% CI 1.35-5.38, P=0.0047), and neurologic deficit (OR 5.22, 95% CI 1.69-16.1, P=0.0047) were significantly associated with instability or surgical management. The traditional 7-mm displacement threshold performed poorly (AUC 0.64, P=0.012). The multivariable model demonstrated good discrimination (AUC 0.75. Based on the final multivariable regression model, a point-based instability risk score was derived to estimate patient-specific risk with good predictive accuracy.
Conclusions:
Instability in C1 fractures was significantly associated with several factors including increasing lateral mass displacement, dens fracture, and neurologic deficit. These parameters outperformed legacy assessment tools based on radiographic measurement. Our prognostic utility demonstrated improved predictive performance and may assist clinicians in risk stratification and treatment decision-making.