Abstract WP126: A Predictive Model for Intracranial Aneurysm Growth Trajectory

Abstract

Introduction: The rate of growth of a particular intracranial aneurysm (IA) is relevant for assessing the risk posed by the IA, and determining an appropriate management strategy. However, the rate of growth cannot currently be estimated when an IA is detected. By determining the association of various factors with IA growth rate based on longitudinally followed aneurysm data, we have developed a predictive model for the patient-specific IA growth trajectory. Hypothesis: If a particular IA will grow, its course of growth is largely determined by the characteristics of the patient and the IA itself. Methods: This study utilized a retrospective dataset of 382 patients with 520 IA diagnosed between 2005-2015. Medical images and electronic medical records were used to build the predictive model. IA were monitored for 32.7 ± 25.4 months, with a mean of three CTA image studies monitoring IA growth or lack thereof, comprising 1,636 data points. Multivariate adaptive splines for analysis of longitudinal data (MASAL) was utilized to guide the development of multivariate linear mixed models for two-level data. Results: MASAL showed that initial size was significantly associated with stroke, hypertension, coronary artery disease, atrial fibrillation, and diabetes mellitus. Therefore, these patient-specific factors were eliminated to avoid overfitting. Smoking history (previous or current) and diagnosis of thyroid disease were found to be independent variables for growth rate prediction and were used to build the model. Our preliminary model found different growth patterns for IA with different initial sizes. We also found that the growth rate of large IA (>7 mm) was more strongly influenced by smoking and thyroid disease. Growth of IA in patients with thyroid disease was found to slow down significantly after 24 months of follow-up, suggesting IA do not grow linearly. Conclusions: We present the first attempt to model IA growth based on medical data and 3D image data for aneurysm cases collected longitudinally over the past 10 years. Our results show it may be possible to predict rate of growth on an individual and aneurysm-specific basis.