Predicting survival in malignant meningiomas: A machine learning approach.

Abstract

2087 Background: Intracerebral meningiomas account for over 90% of all meningioma cases, with only 1–3% classified as malignant. Malignant meningiomas remain understudied compared with other brain tumors. This study is the first to apply machine learning (ML) to identify prognostic factors and improve outcomes of malignant intracerebral meningiomas. Methods: Data were obtained from the SEER database (2004-2021). Patients who met any of the following criteria were excluded: diagnosis not confirmed by histology; previous history of cancer or other concurrent malignancies; or unknown data. To identify prognostic variables, we conducted Cox regression analysis and constructed prognostic models using ML algorithms to predict the 5-year survival. Patient records were randomly divided into training (70%) and validation (30%) sets. A validation method incorporating the area under the curve (AUC) of the receiver operating characteristic curve was used to validate the accuracy and reliability of the ML models. We also investigated the role of multiple therapeutic options using Kaplan-Meier survival analysis. Results: A total of 1,363 patients were included. Most patients were White (71.8%) or female (56.7%). The median patient age was 62 years, and the median tumor size was 4.8 cm. Most of the tumors were localized (67.8%). Adjuvant radiation therapy was administered to 50.2% of the patients. Patients aged < 62 years exhibited better 5-year survival rates, with an overall survival (OS) of 79.4% and cancer-specific survival (CSS) of 82%, compared to those aged ≥ 62 years, who had an OS of 40.9% and CSS of 50.6%. Tumors smaller than 4.5 cm were associated with higher survival rates (OS: 67.7%, CSS: 72.4%) than larger tumors (OS: 53.6%, CSS: 61.9%). The impact of adjuvant radiation therapy showed an OS of 59.9% and 64.5%, respectively, compared with those who did not receive radiation, with an OS of 59.5% and CSS of 68.7%. Multivariate Cox regression analysis identified older age (HR: 3.6, 95% CI: 3.03–4.4) and large tumor size (HR: 1.4, 95% CI: 1.22–1.7) as poor prognostic factors. The Random Forest and MLP classifiers were the most accurate models. The ML models identified age as the most significant prognostic factor. The performance metrics for all the ML algorithms are summarized in Table. Conclusions: This study underscores the transformative potential of ML in enhancing personalized medical approaches for malignant intracerebral meningiomas. Furthermore, whether the benefits of adjuvant radiotherapy outweigh the risks remains unclear, indicating the need for further targeted research to investigate its therapeutic impact on these rare tumors. ML Model Accuracy Precision Recall F1 score AUC LR 63% 50.8% 61.2% 55.6% 0.696 KNN 63.3% 51.2% 45.1% 48% 0.660 RFC 69.1% 58.9% 60.2% 59.5% 0.743 GBC 66.2% 55.6% 52.6% 54.1% 0.723 MLP Classifier 67.48% 57.47% 53.76% 55.56% 0.716