MD/PhD Student The Warren Alpert Medical School of Brown University, Providence RI
Introduction: Glioblastoma is the most prevalent primary malignant brain tumor with no routine screening tests. Identification of blood biomarkers of disease could enable cost-effective screening and earlier treatment. Circulating miRNAs are promising candidates, though disease specificity is poorly understood.
Methods: N=188 patients undergoing resection of newly diagnosed brain tumors were recruited at a single center (2019–2024) for research blood sampling. Twenty-four otherwise healthy patients undergoing surgery for common degenerative spinal diseases (DSD, e.g. discectomy) were recruited as controls. Patients with pathology-confirmed glioblastoma, meningioma, or metastases with pre-operative sample availability were included for analysis. Following RNA isolation (miRNeasy Serum/Plasma kit, Qiagen), 798 miRNAs per sample were quantified (Human V.3 miRNA Expression Assay, Nanostring). After normalization, JADBio®’s Auto-ML platform was employed to test binary classification models by diagnosis and extract predictive miRNA biomarkers. Optimal model performances are reported as area under receiver operating characteristic (AUROC) and mean average precision (MAP).
Results: N=21 glioblastoma patients, n=27 meningioma, n=9 metastases and n=24 DSD controls were included for analysis. Age was similar between tumor and controls (63.2 vs 66.3, p=0.35), though more controls were male (67% vs 33%). Classification of brain tumors (n=57) versus controls yielded excellent performance (AUROC=0.93 [0.88–0.98], MAP=0.95 [0.91–0.98]). Classification of glioblastoma and malignancy (glioblastoma + metastasis, n=30) subgroups versus controls performed similarly (AUROC=0.94 [0.86–1.00], MAP=0.95 [0.91–1.00]; AUROC=0.92 [0.85–0.97], MAP=0.95 [0.91–0.98]). Across models, miR-1283 and miR-150-5p under-expression in disease typically drove performance. Classification of glioblastoma versus other brain tumor (n=36) performed worse (AUROC=0.69 [0.57–0.81], MAP=0.75 [0.67–0.83]).
Conclusion : miRNA quantification from preoperative serum samples may have the capacity to accurately classify age-matched patients according to brain tumor diagnosis. Performance was strongest when classifying tumors from controls. Performance-driving miRNAs tended to be lower in tumor cases. Further analyses including more individuals and controlling for additional clinical variables are needed.
.jpg)
