White matter hyperintensities in strategic tracts for cognitive impairment: A multicenter lesion‐symptom mapping study in 3525 memory clinic patients

Abstract Background White matter hyperintensities (WMH) due to small vessel disease (SVD) are associated with dementia risk. Previous studies suggest that the cognitive impact of WMH depends on location. However, these studies were potentially limited by moderate brain lesion coverage. We aimed to identify strategic WMH locations for cognitive impairment in memory clinic patients by performing a large‐scaled multicenter study providing unprecedented lesion coverage. Method Individual patient neuropsychological and MRI data from eleven memory clinic cohorts were harmonized. Voxel based lesion symptom mapping‐ (VLSM) and region of interest (ROI) based analyses were performed to identify strategic WMH locations for four cognitive domains (attention and executive functioning, information processing speed, verbal memory, and language). A strategic WMH score was calculated based on quartiles of WMH volumes in the strategic white matter tracts and was related to cognition using Lasso regression, accounting for multicollinearity, and multivariable linear mixed models. Result The combined sample consisted of 3525 patients (age: 71.6 ± 9.0 years; 49.9% female) (Table 1). The WMH prevalence maps are shown in Figure 1. In the merged study sample, 63.0% of the supratentorial white matter was affected by WMH in ≥5 patients and could be included in the VLSM analysis. In the VLSM analysis, statistically significant voxels were found for each of the four cognitive domains, after correction for total WMH volume and multiple testing. These voxels were mostly located in the left and right anterior thalamic radiation and the forceps major. Independent ROI based analysis confirmed these strategic tracts and additionally identified the left inferior fronto‐occipital fasciculus as significant tract (Table 2). The strategic WMH score was significantly associated with lower performance on attention & executive functioning (B‐0.146; p<0.001), information processing speed (B‐0.090, p = 0.023), language (B‐0.124, p = 0.003), and verbal memory (B‐0.128, p<0.001), independent of total WMH volume, age, sex, education, brain parenchymal fraction, and study site. Conclusion This large multicenter study provides strong evidence that the cognitive impact of WMH in memory clinic patients depends on location and identified four significant white matter tracts. With further refinement, WMH location might become a useful diagnostic biomarker for SVD‐related cognitive impairment.