Fig. 2
Mean neuropsychological test performance over time by white matter hyperintensity quartiles. The figure shows selected associations between white matter hyperintensity quartiles and clinical measures. Comparisons were made between quartiles of WMH (using raw values) amount at baseline. Higher quartiles reflect greater burden of white matter hyperintensities. Scores shown on the y-axis are raw neuropsychological test scores. For Trail Making Test B, higher scores reflect worse performance (i.e., WMH fourth quartile performed the worse). For all other measures, higher scores reflect better performance. For Trail Making Test B, individuals in the highest two quartiles (i.e., greatest WMH burden) experienced a mean increase of 2.9 s per year, whereas individuals in the lowest quartile only slowed by 0.9 s per year. On the WAIS Digit Symbol Coding, the highest quartile had a mean decline of 0.58, whereas the lowest quartile experienced a mean score increase of 0.11 per year. The mean slope for decline in Semantic Fluency was twice as steep among participants in the highest WMH quartile compared to the lowest WMH quartile (mean raw score decline of 0.61 words compared to 0.30 per year, respectively). For CDR Sum of Boxes, the participants in the lowest WMH quartile had a net slope of zero over time, whereas participants in the highest two quartiles had a mean raw score increase of 0.11 per year. GEE models showed the displayed relationships to be statistically significant after controlling for age, sex, race, years of education, APOE ε4 carrier status, total brain volume, and hippocampal volume. Although not displayed, GEE models also showed statistically significant findings for Trail Making Test A, Logical Memory Immediate and Delayed Recall, and the 15-item Geriatric Depression Scale