AbstractBackgroundWhite matter hyperintensities (WMHs) are prevalent vascular pathologies that often co‐occur with amyloid plaque deposition in Alzheimer’s disease (AD). Previous research has shown that antemortem WMHs correlate well with post‐mortem findings of cerebral amyloid angiopathy (CAA). PET‐amyloid imaging is widely used to assess in‐vivo amyloid load, however the extent to which CAA‐binding of amyloid tracers contributes to global measures of amyloid load is unclear. The aim of this meta‐analysis was to determine the strength of the correlation between amyloid load and WMHs in AD subjects.MethodThirteen studies that reported either a correlation or regression coefficient between amyloid load and WMHs were included. Age, education level, sample size, correlation, study setting (community, clinic, Alzheimer’s Disease Neuroimaging Initiative [ADNI]), amyloid tracer type (florbetapir vs. Pittsburgh compound B [PiB]), and method of WMH quantification (visual vs. volumetric) were extracted. Random effects results were reported for the main and subgroup analyses. Between‐study heterogeneity was quantified using I2 with 95% confidence intervals.ResultThe overall correlation for amyloid load and WMH load was r=.25, 95% CI[.15, .35], p<.001; I2=80%, 95% CI[66%, 88%]. The combined effect size of community and clinic‐based studies (r =0.30) was larger than that of ADNI studies (r=0.21) and had substantially less heterogeneity (I2=67% vs. I2=86%). Florbetapir had a stronger correlation with WMHs (r=0.28) than PiB (r=0.20), with less heterogeneity (I2=71%, I2=87%). Visual methods of WMH quantification yielded a stronger correlation with amyloid load (r=0.38) than volumetric methods (r=0.20), however between‐study heterogeneity was similar for both methods (I2=80%, I2=82%).ConclusionWMHs correlate significantly with amyloid load, but this relationship is strongly influenced by amyloid tracer type and WMH quantification method. A high level of heterogeneity was noted for ADNI studies which further indicates the impact that methodologic and analytic techniques may have on a study’s findings. Although CAA‐related binding cannot be distinguished from parenchymal binding in amyloid‐PET studies, these results provide a general estimate of CAA’s contribution to PET‐based estimates of global amyloid load.
Read full abstract