The effect of white matter hyperintensities on verbal memory: Mediation by temporal lobe atrophy

Abstract

Objective: To determine the relationship between white matter hyperintensities (WMH) presumed to indicate disease of the cerebral small vessels, temporal lobe atrophy, and verbal memory deficits in Alzheimer disease (AD) and other dementias.

Methods: We recruited groups of participants with and without AD, including strata with extensive WMH and minimal WMH, into a cross-sectional proof-of-principle study (n = 118). A consecutive case series from a memory clinic was used as an independent validation sample (n = 702; Sunnybrook Dementia Study; NCT01800214). We assessed WMH volume and left temporal lobe atrophy (measured as the brain parenchymal fraction) using structural MRI and verbal memory using the California Verbal Learning Test. Using path modeling with an inferential bootstrapping procedure, we tested an indirect effect of WMH on verbal recall that depends sequentially on temporal lobe atrophy and verbal learning.

Results: In both samples, WMH predicted poorer verbal recall, specifically due to temporal lobe atrophy and poorer verbal learning (proof-of-principle -1.53, 95% bootstrap confidence interval [CI] -2.45 to -0.88; and confirmation -0.66, 95% CI [-0.95 to -0.41] words). This pathway was significant in subgroups with (-0.20, 95% CI [-0.38 to -0.07] words, n = 363) and without (-0.71, 95% CI [-1.12 to -0.37] words, n = 339) AD. Via the identical pathway, WMH contributed to deficits in recognition memory (-1.82%, 95% CI [-2.64% to -1.11%]), a sensitive and specific sign of AD.

Conclusions: Across dementia syndromes, WMH contribute indirectly to verbal memory deficits considered pathognomonic of Alzheimer disease, specifically by contributing to temporal lobe atrophy.

© 2018 American Academy of Neurology.

Figures

Figure 1. Processing of neuroimaging data to obtain atrophy and white matter hyperintensity volumetrics

MRI-derived 3D segmentation and regional parcellation generated by lesion explorer. Top row: axial slices of skull-stripped coregistered MRI from (A) T2-weighted, (B) T1-weighted, and (C) proton density. Middle row: (D) demarcation of the Semi-Automatic Brain Region Extraction (SABRE) regional parcellation of the left temporal lobe region overlaid on axial T1, (E) tri-feature tissue segmentation for gray (light gray) and white matter (dark gray), ventricular (yellow) and sulcal CSF (blue), and white matter hyperintensities (purple), (F) white matter hyperintensities (purple) overlaid on proton density. Bottom row: 3D surface volume rendering of (G) SABRE regional parcellation (left temporal lobe in red). (H) Transparent surface shows white matter hyperintensities (purple).

Figure 2. Brain volumetrics and memory measures from the proof-of-principle sample

(A) White matter hyperintensity volumes (WMH), (B) left temporal lobe brain parenchymal fraction, (C) California Verbal Learning Test, 2nd ed. (CVLT-II) verbal learning trials, (D) CVLT-II long delayed free recall trials in the proof-of-principle sample. WMH volume (F1,117 = 22.591, p < 0.001), left temporal brain parenchymal fraction (F1,117 = 17.534, p < 0.001), verbal learning (F1,117 = 20.57, p < 0.001), and verbal recall (F1,117 = 40.057, p < 0.001) differed significantly between groups. AD = Alzheimer disease; CAD = coronary artery disease.

Figure 3. Path models show direct and indirect relationships between white matter hyperintensity (WMH) volume and verbal memory performance

Path model tested in (A) proof-of-principle sample (n = 118) and (B) confirmation sample (n = 702). Thick lines are part of a significant pathway. Thin lines are part of a nonsignificant pathway. For each connection, the coefficient (a, b, c, a2 and b2), its standard error, and significance (p) are shown. The mediation effect of the hypothesized serial pathway (a × b × c) and its 95% CI are shown in red, the mediation effect of learning alone (a2 × c) in green, and the mediation effect of atrophy alone (a × b2) in blue. Direct effects are shown in orange; d is the direct effect adjusted only for covariates and d′ is the direct effect adjusted for covariates and indirect effects. Covariate effects on mediators or outcomes with p < .05 are shown in grey. SE = standard error; TIC = total intracranial capacity.