Prospective Trial of Quantitative Neuroimaging Correlates of Verbal and Nonverbal Memory Decline

Abstract

To analyze clinical and quantitative neuroimaging parameters associated with verbal and nonverbal memory changes in a longitudinal, prospective clinical trial of brain tumor patients undergoing radiotherapy (RT). We hypothesized that volume and white matter (WM) diffusion changes in memory-associated regions would correlate with post-RT cognitive decline. We prospectively evaluated 25 patients with primary brain tumors undergoing fractionated partial-brain RT. Patients underwent high-resolution 3D volumetric brain MRI and diffusion-weighted imaging, along with neurocognitive testing prior to and 6 months after RT. Semi-automated software was used to segment discrete neuroanatomic regions, quantifying volume and diffusion biomarkers of WM damage (MD = mean diffusivity; FA = fractional anisotropy). Tumor, surgical effects, and edema were censored. Memory-associated regions of interest included the hippocampus and WM subadjacent to entorhinal, parahippocampal, and temporal pole neocortex. Reliable change indices (RCI) were calculated to measure changes in verbal (Hopkins Verbal Learning Test [HVLT]) and nonverbal (Brief Visuospatial Memory Test [BVMT]) learning and memory from pre- to post-RT. Pearson correlation coefficients were calculated to assess associations between clinical and imaging parameters (change from baseline in hippocampal volume, MD, FA) and memory outcomes. Statistical significance was set at α = 0.05 for two-tailed tests. Patients showed a significant decline in nonverbal learning (BVMT total recall, P = 0.009, one sample t-test) but not in verbal learning or memory/HVLT. Glioma patients had a greater decline in nonverbal learning and memory (BVMT total and delayed recall) at 6 months post-RT (r = -.44, P = 0.047 for both tests), as did patients with seizures (r = -.58, P = 0.035; r = -.44, P = 0.035) and those receiving concurrent (r = -.57, P = 0.005; r = -.56, P = 0.005) or adjuvant chemotherapy (r = -.44, P = 0.035 for both tests). Right hippocampal volume significantly decreased after RT (mean change -.036; one sample t-test; P = 0.01). While hippocampal volume loss was not significantly associated with memory decline, changes in WM microstructure were associated with verbal memory decline. Increased left parahippocampal WM FA was associated with decline in verbal learning and memory (HVLT total and delayed recall, r = -.68, P = 0.008; r = -.70, P = 0.004, respectively) and increased MD was associated with decline in verbal learning (HVLT total recall, r = -.65, P = 0.01). Increased mean dose to left entorhinal and temporal pole WM was associated with verbal memory decline (HVLT delayed recall, r = -.54, P = 0.02; r = -.54, P = 0.02, respectively). WM microstructure in the medial temporal lobe, but not hippocampal atrophy, is associated with memory decline in brain tumors patients receiving RT. These data suggest that damage to medial temporal lobe WM may serve as an important biomarker for RT-associated memory decline.