152 - A Longitudinal Study to Correlate Neurocognitive Changes of IDH-Mutant and IDH-Wildtype Glioma Patients after Chemoradiotherapy with Changes on Resting-State MRI

Z. Liu¹, T. Mitchell¹, C. Luo², J. S. Shimony^3,4, K. Y. Park⁵, R. Fucetola⁶, S. M. Perkins^1,4, E. C. Leuthardt^4,5, A. Snyder³, T. Zhu¹, and J. Huang^1,4; ¹Department of Radiation Oncology, Washington University School of Medicine, St. Louis, MO, ²Department of Surgery, Division of Public Health Sciences, Washington University School of Medicine, St. Louis, MO, ³Department of Radiology, Division of Medical Oncology, Washington University School of Medicine, St. Louis, MO, ⁴Brain Tumor Center, Siteman Cancer Center, Washington University School of Medicine, St. Louis, MO, ⁵Department of Neurological Surgery, Washington University School of Medicine, St. Louis, MO, ⁶Department of Neurology, Washington University School of Medicine, St. Louis, MO

Purpose/Objective(s): This prospective observational study investigates mechanisms behind neurocognitive function (NCF) decline after partial-brain irradiation in malignant glioma patients. We aim to use resting-state functional MRI (RS-fMRI) to identify the dominant network-level disturbances post-radiation therapy (RT) that correlate with NCF changes. Materials/

Methods: Adult patients with IDH-wildtype or IDH-mutant gliomas underwent NCF test using the NIH Toolbox Cognitive Function Battery at baseline and 6 months post RT. The battery includes fluid cognition tests: dimension change card sort test (executive function), flanker test (attention), picture sequence test (episodic memory), list sorting test (working memory), and pattern comparison test (processing speed). The five test scores were then combined into an age-normalized composite score, from which the percent change of composite (PCC) was calculated relative to the baseline. To determine a potential correlation between NCF and changes in brain functional connectivity (FC), we used seed-based FC analysis from a 12-minute RS-fMRI scan. A split-sample approach was used for analysis, with a 26-patient training set and a 6-patient validation set, iterated 200 times. Within each run, connectivity-regression analysis within the training set was first used to identify which intra- or inter-network FC change was most significantly associated with PCC, and a linear regression was used to predict FC change of the selected networks using the validation set. Permutation test was used to evaluate the significance of network selection, and R² value was used to evaluate the predictive performance.
Results: From September 2020 to December 2023, 43 patients had baseline data, while 32 patients completed 6 month follow-ups and were evaluable. The mean NCF composite changed from 88.8 (±16.2) at baseline to 91.1 (±19.4) at 6 months. The mean PCC was 2.9 (±13.7), including 12 patients with negative PCC (Decline cohort) and 20 patients with positive PCC (Non-decline cohort). The clinical, treatment, and dosimetric characteristics between two cohorts were not significantly different among 24 variables examined. The mean R2 was 0.36 (±0.27). The most significant correlations with PCC were consistently observed in the inter-network FC changes between Default Mode Network to Medial Temporal Lobe (DMN-MTL, P = 0.005) and Parietal Memory Network to MTL (PMN-MTL, P = 0.02). Sensitivity analyses using only FC maps from the contra-lateral side of the tumor confirmed the same finding.
Conclusion: RS-fMRI changes post RT correlated with NCF decline, suggesting its potential as an imaging biomarker. Specifically, disruption of inter-network FC between DMN-MTL and PMN-MTL may be key mechanism underlying RT-induced NCF decline, warranting further investigation.