Motivated by recent research findings on glioblastoma multiforme (GBM) highlighting the insufficiency of relying solely on imaging data for predicting the O6-methylguanine-DNA methyltransferase (MGMT) methylation status, this study takes an innovative approach by integrating both preoperative clinical and non-invasive acquisition data. A stacked dataset, combining predictions from models developed on these two distinct data sources, is employed for the analysis. Leveraging the Hard Voting Classifier ensemble (Random Forest (RF), XGBoost (XGB), Support Vector Machine (SVM)) on a resampled dataset with replacement, the models demonstrated an average accuracy of 85.9%, emphasizing the robustness of the predictive performance. The non-invasive nature of the approach, relying solely on structured data from diverse datasets, provides a comprehensive view of GBM dynamics. Results, consolidating clinical and acquisition data insights, are presented, underscored by their reliance solely on non-invasive structured data, providing valuable implications for non-invasive diagnostic and therapeutic advancements in GBM research.