BIOM-54. ANALYTICAL VALIDATION OF A TARGETED GENE EXPRESSION BIOMARKER PREDICTING MENINGIOMA OUTCOMES AND RADIOTHERAPY RESPONSES

Abstract

Abstract BACKGROUND Improvements in meningioma risk stratification are needed to guide postoperative management. We previously developed and externally validated a targeted gene expression biomarker predicting meningioma outcomes and radiotherapy response using fresh frozen meningiomas. Here, we present the analytical validity, test-retest and cross-platform reproducibility, intra-tumor heterogeneity, and formalin fixed paraffin embedded (FFPE) concordance for this biomarker. METHODS Matched FFPE and fresh frozen samples from 50 meningiomas underwent RNA extraction for concordance testing using a custom Nanostring gene expression panel comprised of 34 meningioma genes and 7 housekeeping genes. Matched Nanostring gene expression profiling and RNA-sequencing was available from 173 fresh frozen meningiomas. Nanostring batch, machine, and technician variability was tested on 10 FFPE meningiomas using 2 independent batches of biomarker reagents that were processed at 2 independent laboratories. To study intra-tumor heterogeneity, 68 spatially distinct meningioma biopsies obtained under stereotactic guidance from 13 tumors were analyzed using the Nanostring gene expression biomarker. RESULTS Matched FFPE and fresh frozen meningioma samples (N=50) demonstrated high gene expression concordance (Pearson R=0.89, F-test P< 2.2e-16) and risk score concordance (R=0.78, F-test P< 1.8e-11, residual standard error 0.12). The biomarker achieved high performance on FFPE samples (local freedom from recurrence [LFFR] c-index 0.75±SEM 0.05, 5y LFFR AUC 0.84, 95% interval 0.67-0.96). Test-retest concordance on FFPE samples (N=10) across independent reagent batches, machines, and technicians was high (gene expression concordance: R=0.97; risk score concordance R=0.94). Transcript counts were concordant across Nanostring and RNA sequencing approaches (N=173 frozen samples, R=0.90). Stereotactic intra-operative sampling (N=68 samples from 13 patients) revealed low levels of intra-tumor heterogeneity (median inter-quartile risk score within tumor: 0.10±0.01). CONCLUSIONS We demonstrate analytical reproducibility and robustness of a gene expression biomarker predicting meningioma outcomes and radiotherapy responses in FFPE and frozen samples across multiple conditions and platforms.