Prognostic aspects of the outcomes of intracranial meningiomas

Meningiomas are stratified depending on the extent of the tumor and the degree of resection, often in isolation from other clinical variables. The aim of the work is to integrate demographic, clinical, radiological and pathological data for the development of prognostic models of meningioma outcomes. Material and methods. Authors have developed a comprehensive database containing information on 235 patients who underwent surgery for 257 meningiomas in one facility from 2013 to 2023. The median follow-up was 4.3 years, and the resection samples were re-evaluated in accordance with modern diagnostic criteria, resulting in 128 grade II meningiomas and 25 grade III meningiomas according to WHO. A series of machine learning algorithms have been trained and configured using nested resampling to create models based on preoperative functions, conventional postoperative functions, or both. Results. Authors compared the accuracy of different algorithms, as well as the unique information they provided in the data. Machine learning models limited to preoperative information such as patient demographics and radiological characteristics had the same accuracy in predicting local insufficiency or overall survival as models based on the degree of meningioma and the degree of resection. Integrated models, including all available demographic, clinical, radiological and pathological data, allowed us to obtain the most accurate estimates. Based on these models, authors have developed decision trees and nomograms to assess the risks of local insufficiency or overall survival in meningioma patients. Conclusion. Clinical information has historically been underused in predicting meningioma outcomes. Prognostic models trained on the basis of preoperative clinical data work comparably with conventional models in terms of the degree of meningioma and the degree of resection. Combining all the available information can help to more accurately stratify meningioma patients