Dosimetric Predictors of Cognitive Decline in Attention and Processing Speed after Fractionated Brain Radiation Therapy

Abstract

Neurocognitive decline, including changes in attention/processing speed, can be an unfortunate consequence of brain radiation therapy (RT). The corpus callosum (CC) and right-sided subcortical white matter (WM) are known to subserve this cognitive function. We aimed to determine dosimetric predictors of attention/processing decline at 6 months after fractionated RT in patients with primary brain tumors. Twenty-three patients with primary brain tumors were enrolled on a prospective cohort study. Patients had neurocognitive assessments and high-resolution diffusion and volumetric MRI imaging pre-RT and 6 months post RT. Subjects were tested on Delis-Kaplan Executive Function System trail making number and letter sequencing. Reliable change indices accounting for practice effects (RCI-PE) scores were calculated as a normalized measure of cognition change from pre-RT to post-RT. Practice effects accounts for expected improvements in tasks with repetition. Binary variables decline and substantial decline were defined as RCI-PE <0 and <-1, respectively. Brain WM regions were segmented using a validated probabilistic diffusion tensor WM atlas. Regions of interest (ROIs) were CC and total right-sided subcortical WM with and without CC. Tumor and edema were censored from ROIs. Data including Vx (structure volume receiving ≥ x Gy) in 5 Gy intervals, minimum, maximum, and mean doses to each ROI were extracted from RT plans. A general linear model on dosimetric/volumetric data was created to identify predictors of decline (significance defined as p<0.05). Most patients (n=15, 62%) had benign or low-grade tumors. Median age at RT was 51 years (range 20-75) and 48% of patients were male. Fifteen patients (65%) underwent resection pre-RT and 9 (39%) patients received concurrent chemotherapy. Between 60-69% of patients experienced a decline; substantial number and letter sequencing decline was seen in 5 (22%) and 2 (9%) patients, respectively. Minimum dose to CC predicted for substantial letter sequencing decline (p<0.001), but there were no significant predictors of number sequencing. In right sided subcortical WM excluding CC, V30 predicted for substantial number sequencing decline (p<0.001) while minimum dose (p<0.001) and V30-V40 (all p<0.001) predicted for substantial letter sequencing decline. Within right-sided subcortical WM including CC, V30 predicted for substantial number sequencing decline (p<0.001) while V30-V35 (p<0.001) predicted for substantial letter sequencing decline. In this prospective cohort study, increasing volume receiving 30-40Gy within the right-sided total subcortical WM predicted for decreased attention/processing speed at 6 months after RT. Minimum dose to CC also predicted for decline. This suggests that dose-dependent subcortical WM effects contribute to cognitive decline. Further complex modeling is needed to generate normal tissue parameters for WM structures to guide cognitive sparing brain RT.