Local fractional anisotropy is reduced in areas with tumor recurrence in glioblastoma

Purpose: To analyze if fractional anisotropy (FA) in nonenhancing peritumoral regions (NEPTRs) at baseline is associated with later tumor recurrence in glioblastoma. Materials and Methods: Ethical approval was obtained for this retrospective, HIPAA-compliant study. FA was measured in 70 patients with glioblastoma in five regions of interest (ROIs) per patient in the NEPTR at preoperative magnetic resonance (MR) imaging with (166 regions) or without (184 regions) local contrast-enhancing tumor recurrence at follow- up MR imaging (median, 7.3 months; range, 0.9-46.6 months). ROIs were classified according to their location (white matter, cortex, fiber tracts, basal ganglia). Ratio of FA in the ROI of the NEPTR to that in the contralateral side (FA_contra) and to that in the internal capsule (FA_int) was calculated. A generalized linear mixed model was performed. Ten-fold cross-validation was used for the receiver operating characteristics (ROC) analysis. Results: FA_contra and FA_int were significantly lower in regions with later tumor recurrence than in regions without (median FA_contra: 0.29 [interquartile range {IR}, 0.22-0.36] vs 0.46 [IR, 0.38-0.57]; median FA_int: 0.20 [IR, 0.16-0.24] vs 0.29 [IR, 0.22-0.36], respectively). ROC analysis revealed an area under the ROC curve of 0.893 for FA_contra and of 0.815 for FA_int, resulting in respective sensitivity and specificity of 85.5% and 84.2% for FA_contra and 86.7% and 66.8% for FA_int. Conclusion: Local tumor recurrence in the NEPTR may be predicted by FA metrics at baseline in patients with glioblastoma. This might be important for surgery or radiation planning.