Contributions of enlarged perivascular spaces and white matter hyperintensities to cognitive performance in the Alzheimer’s disease continuum – shared and unique effects

Abstract

AbstractBackgroundIndependent contributions of enlarged perivascular spaces (ePVS) to cognition remain elusive due to their intertwining with white matter hyperintensities (WMH). We studied the unique and shared effects of ePVS and WMH on cognitive performance in 397 subjects along the Alzheimer’s disease (AD) continuum.MethodsWe quantified baseline volumes of WMH, basal ganglia (BG) and centrum semiovale (CSO) ePVS in participants of the multicentre DELCODE study (Controls = 81/Subjective Cognitive Decline = 151/Mild Cognitive Impairment = 79/AD = 45/AD relatives = 41) and regressed age, sex, pTau181, and image quality out of them.We established cognitive domain scores (memory, working memory, language, executive, visuospatial) based on a comprehensive cognitive test battery and confirmatory factor analysis, and assessed preclinical Alzheimer’s cognitive composite (PACC5) scores.Using principle component (PC) analysis, we estimated unique and shared variances. We began by extracting the first PC (PC1)—a weighted average of ePVS and WMH volumes—which indicated overall cerebrovascular dysfunction (Figure 1). We then regressed PC1 out of ePVS and WMH volumes to estimate their unique variances, which should reflect glymphatic function and white matter (WM) damage, respectively.For each cognitive domain, we compared null linear regression models (years of education, age, sex, pTau181, and image quality) to those enriched with the shared (PC1) and unique variances of BG‐/CSO‐ePVS and WMH.ResultsOverall cerebrovascular dysfunction—PC1—substantially affected all cognitive domains negatively.BG‐ePVS’ unique variance—surrogate for glymphatic function—contributed to better memory (B = 0.19, 95%‐CI[0.02;0.36]), language (B = 0.25, 95%‐CI[0.09;0.42]) and PACC5 performance (B = 0.25, 95%‐CI[0.07;0.43]) beyond overall cerebrovascular dysfunction. However, WM damage—WMH unique variance—contributed more substantially and negatively to the same cognitive domains (memory: R²PC1 = 0.365<R²PC1+BG‐ePVS = 0.372<R²PC1+WMH = 0.382; language: R²PC1 = 0.353<R²PC1+BG‐ePVS = 0.366<R²PC1+WMH = 0.370; PACC5: R²PC1 = 0.300<R²PC1+BG‐ePVS = 0.313<R²PC1+WMH = 0.324; Table 1), and moreover to executive (B = ‐0.10, 95%‐CI[‐0.18;‐0.02]) and visuospatial performance (B = ‐0.07, 95%‐CI[‐0.14;‐0.00]).CSO‐ePVS’ unique variance did not correlate with any cognitive domain.ConclusionsOverall cerebrovascular dysfunctions and WM damage—reflected by the shared variance between ePVS and WMH as well as the unique WMH variance—contributed distinctively to poorer cognitive performance. The contrasting positive impact of unique BG‐ePVS variance could represent a “glymphatic or vascular reserve” in a region important for different cognitive processes—this hypothesis, however, requires further investigation.