NIMG-55. PREDICTING SITES OF TUMOR PROGRESSION IN THE INVASIVE MARGIN OF GLIOBLASTOMAS (PRAM-GBM STUDY): AN INTERIM ANALYSIS OF A MULTI-CENTER IMAGING BIOMARKER STUDY

Abstract

Abstract INTRODUCTION Although glioblastoma cells extend throughout the brain, virtually all patients die from local progressive disease. This progressive tumor occurs at the resection margins and usually within the high-dose radiotherapy field. If we could identify where the tumor would progress, we could consider personalizing radiotherapy and surgical targets. Previous work has shown diffusion tensor MR as a potential biomarker of occult invasion. Biopsy studies reveal that the q-anisotropic component correlates with high-density tumor, and the extent of resection correlates with improved survival. We hypothesize that this is a biomarker that can predict where glioblastomas may progress. STUDY DESIGN: A multicentre, pragmatic, longitudinal observational cohort study from 6 centers. This is a planned interim analysis of the first cohort. METHODS Patients with glioblastomas suitable for maximal resection were imaged pre-operatively with a diffusion tensor MRI protocol. This was processed using a tensor decomposition method to generate the q-anisotropic component. Masks of the abnormal q-component were overlaid on the masks of new areas of contrast enhancement at progression (ground truth). Intersecting voxels were classified as true positive, true negative, false positive, and false negative, and the sensitivity and specificity were calculated. RESULTS 139 patients were recruited, but 50 were withdrawn. The interim analysis was planned for the first 2/3rds of the cohort (55 patients). Combining q and contrast enhancement provided the best predictor of site of tumor progression with a mean sensitivity of 79% [95%CI: 71-88%] compared to contrast enhancement of only 50% [95%CI: 39-61%]. Specificity remained high in both groups (CE+q mean 94% [95%CI: 93-95%] vs. CE alone 98% [95%CI: 97-98%]). CONCLUSION The anisotropic diffusion component, q, provides the best biomarker for identifying ‘residual tumor.’ This biomarker will form the basis of an early phase surgical trial IDEAL Stage 1 and 2a study to change surgical resection target.