Related Experiment Video
Updated: Mar 21, 2026
Establishing a Competing Risk Regression Nomogram Model for Survival Data
Published on: October 23, 2020
Development and Validation of a Nomogram for Predicting Poor Outcome in Spontaneous Cervical Artery Dissection
Shimeng Chen1, Zhicheng Yang2, Lijuan Yang2
1Department of Ultrasound, the Affiliated Central Clinical Hospital of Baotou Medical College, Inner Mongolia University of Science and Technology, Baotou, 014040, People's Republic of China.
Objective:
This exploratory study aimed to develop a preliminary nomogram for risk assessment of poor prognosis in spontaneous cervical artery dissection (sCAD) and evaluate its statistical performance using internal validation.
Methods:
We retrospectively analyzed 75 patients with sCAD (mean age 51.8 ± 14.0 years; 41 males [55%] and 34 females [45%]) diagnosed between November 2013 and April 2024. Poor prognosis was defined as imaging-confirmed acute cerebral infarction or hemorrhage (n=38); the remaining 37 patients comprised the good prognosis group. Due to the small sample size, variables with p<0.2 in univariate analysis (sex, extracranial CAD type, hypertension, hyperhomocysteinemia) were considered for multivariate modeling, though none were statistically significant predictors (all p>0.05). A multivariate logistic regression-based nomogram was constructed and internally validated using 1000 bootstrap resamples.
Results:
The final model included the four variables above. Only non-intramural hematoma (other) CAD type showed statistical significance in the multivariate model (OR=13.41, 95% CI: 2.89-62.17, P<0.01), while sex, hypertension, and hyperhomocysteinemia did not, likely reflecting statistical instability from inadequate power. In bootstrap internal validation, the model demonstrated moderate discrimination (AUC=0.788, 95% CI: 0.686-0.891) with a Brier score of 0.185. Hosmer-Lemeshow test indicated acceptable calibration (χ2=8.11, P=0.23). Mean AUC across bootstrap samples was 0.763 (95% CI: 0.662-0.863), suggesting minimal overfitting within this dataset, though this does not imply generalizability.
Conclusion:
This pilot study generates the hypothesis that ultrasonographic CAD type, combined with clinical variables, may aid in predicting sCAD outcomes. However, due to the small sample size (9.5 events/variable), lack of external validation, and inclusion of non-significant predictors, this model is not ready for clinical application. Multi-center prospective validation in a cohort of at least 400 patients is required before any clinical utility can be claimed.
