FLAIR extending beyond a quadrant (FEQ) of the cerebrum as a qualitative imaging biomarker: a conceptual proposition and proof-of-principle.

Significant variability exists in the use of corticosteroids for treating adverse radiation effects (ARE) after stereotactic radiosurgery (SRS) of brain metastasis (BM). Here, we determine the diagnostic utility of a quadrant-based, visual assessment of magnetic resonance (MR) FLAIR as an imaging biomarker for steroid-dependent ARE. FLAIR was assessed at four axial levels along the rostral-caudal axis of the cerebrum, defined by standard landmarks of superior temporal line, third ventricle, temporal horn, and fourth ventricle. Each axial level was divided into four quadrants, defined by 12, 3, 6, and 9 on a clock face. New, post-SRS FLAIR hyperintensity extending beyond any quadrant was defined as FEQ+. FEQ+ was then correlated with corticosteroid treatment instituted within a month of the MRI. To establish intra- and inter-rater reliability of FEQ, MR images from 20 patients (10 FEQ+ and 10 FEQ-) were assessed by three clinicians (a radiation oncologist and two neurosurgeons) for FEQ positivity. These results showed an > 85% intra- and inter-rater reliability (Cohen's Kappa and Fleiss' Kappa of 0.970 and 0.785, respectively, both p < 0.001). We tested the hypothesis that FEQ+ is associated with corticosteroid use post-SRS in an initial cohort of 40 patients. The sensitivity, specificity, positive predictive value, and negative predictive value of FEQ for corticosteroid treatment were 75.0%, 96.4%, 75.0%, and 90.0%, respectively. To validate these findings, we examined the association of FEQ and corticosteroid use in an independent cohort of 214 SRS-treated BM patients. The sensitivity, specificity, positive predictive value, and negative predictive value of FEQ for corticosteroid treatment in this validation cohort were 94.6%, 74.0%, 43.2%, and 98.5%, respectively. We conclude that FEQ is an imaging marker with high intra- and inter-rater reliability, with a high negative predictive value (90.0-98.5%) for steroid treatment in SRS-treated BM patients. These results lay the foundation for future studies of FEQ for research and clinical applications.