2025 Proffered Presentations
S194: RADIOGRAPHIC SIGNATURE OF TRAF7 MUTANT MENINGIOMAS
Aymen S Kabir, BA1; Abraham Dada, BA1; Rithvik Ramesh, BA1; Daniel Quintana, BA1; Christian Jimenez, BS1; Wesley Shoap, MD2; Robert C Osario, MD3; Ezequiel Goldschmidt, MD, PhD3; 1School of Medicine, University of California, San Francisco; 2Department of Neurosurgery, Louisiana State University; 3Department of Neurological Surgery, University of California, San Francisco
Introduction: Meningiomas with different somatic DNA mutations have been found to be associated with varying recurrence rates and driving biology. A multimodal analysis of the radiological features of tumors followed by next-generation DNA sequencing can highlight preoperative findings that may suggest specific genetic mutations. Around 25% of all meningiomas carry a TRAF7 missense single nucleotide variant, whose biological significance is uncertain. Understanding whether there is a correlation between TRAF7 mutations and imaging patterns may reveal aspects of the biology driven by this gene.
Objective: To evaluate the association between various radiomic features and TRAF7-mutated meningiomas.
Methods: The authors retrospectively reviewed the records and imaging of patients with meningiomas with available genetic panels resected between 2021-2024 at a single institution. Patient characteristics examined in the analysis included patient demographics, tumor location and size, somatic DNA mutations through capture-based next-generation sequencing, and imaging characteristics including the presence and extent of peritumoral edema and enhancing pattern. CT images were analyzed for features including bone involvement, presence of osteolysis or hyperostosis, and type of hyperostosis. Type 1 hyperostosis was defined as hyperostosis with destruction of cortical architecture while Type 2 hyperostosis was defined by the preservation of cortical structure. These variables’ associations with TRAF7 mutations were analyzed via univariate t-tests, chi-square tests, Fisher’s exact tests, and multivariate logistic regression. Furthermore, Random Forest, Gradient Boosting Machine (GBM), and Support Vector Machine (SVM) learning models were used to assess non-parametric relationships.
Results: 322 meningiomas were included in the final analysis, with TRAF7 mutations occurring in 71 (22%) of the tumors. Younger age (55.4 vs 60.9, p=0.002) and female sex (77.5% vs 64.5%, p=0.038) were associated with TRAF7 mutations. Univariate analysis of MRI features showed that TRAF7 mutations were less likely to have parenchymal T2 signal change (38.0% vs 55.0%, p=0.022) and more likely to be homogenously enhancing (93.0% vs 64.9%, p<0.001), located in the skull base (78.9% vs 57.0%, p<0.001), and have a smaller maximal tumor dimension (3.0 vs. 3.9, p<0.001). Furthermore, TRAF7 mutations were more likely to involve the bone (31.0% vs 16.7%, p=0.008), be hyperostotic (26.8% vs 9.2%, p<0.001), and exhibit type 1 hyperostosis (73.7% vs 8.7%, p<0.001). Multivariate logistic regression analysis showed that smaller maximal tumor dimension (OR: 0.68, p=0.004), homogenous enhancement (OR: 4.30, p=0.010), and type 1 hyperostosis (OR: 21.69, p=0.003) were significantly associated with TRAF7 mutations.
The Random Forest model achieved the highest performance on test data (AUC: 0.975, Accuracy: 96.9%). The three most important features included tumor dimension, hyperostosis pattern, and T2 signal change. A subgroup analysis of lesions with Type 1 hyperostosis and no T2 signal change showed a TRAF7 mutation rate of 91.7%.
Conclusions: Radiological features that best predict TRAF7 mutation status in meningiomas include the absence of brain signal change on T2-weighted images, hyperostosis with disrupted cortical architecture, and maximal tumor dimension. The phenotype of type 1 hyperostotic meningiomas without peritumoral edema seems to be highly suggestive of a TRAF7 mutation.