Abstract: Background: Cerebral cavernous malformations are central nervous system vascular abnormalities, susceptible to intra-cerebral hemorrhages, with considerable clinical consequences. Objective: This study aimed to predict the risk of hemorrhage in patients with cerebral cavernous malformations. Methods: We conducted a retrospective analysis of 1,219 patients with cerebral cavernous malformations at a high-volume medical center between January 1, 2003, and December 31, 2018. We employed the Least Absolute Shrinkage and Selection Operation regression model to identify possibly relevant features while developing a novel model through multivariate Cox proportional hazards analysis. Results: The model included surgical procedures, mean cerebral cavernous malformations volume, intracerebral hemorrhage occurrence, and brainstem localization. We achieved accurate calibration and discrimination in the training dataset using these factors. The training set produced areas under the curve values of 0.881, 0.867, and 0.820 for predicting the probability of hemorrhage at 1, 3, and 5 years, respectively. The model demonstrated strong calibration and discrimination in the validation set, with area under the curve values of 0.877, 0.881, and 0.859 for predicting the 1-, 3-, and 5-year hemorrhage risks, respectively. Decision curve analysis revealed that the model had significant clinical utility. In addition, we developed a web-based calculator (https://rehablitation.shinyapps.io/CCMs/) to display the prediction findings visually. Conclusions: The nomogram integrating the four demographic and clinical parameters yielded an accurate hemorrhage prediction in individuals with cerebral cavernous malformations. This predictive model can guide physicians in clinical decision-making.
Bi et al. (2026) studied this question.