脳画像表現型から血管性認知症サブタイプへ：包括的なメンデルランダム化研究

Vascular dementia (VaD) is a heterogeneous disorder with distinct subtypes, each exhibiting unique neuropathological profiles. Although neuroimaging studies have identified some subtype-specific structural brain alterations, a systematic investigation establishing causal relationships between imaging-derived phenotypes (IDPs) and the development of specific VaD subtypes remains absent from current literature.

Two-sample Mendelian Randomization (MR) analyses were conducted to assess causal relationships between 3,935 brain IDPs from UK Biobank neuroimaging and four VaD subtypes (multiple infarctions, subcortical, sudden onset, and mixed) from FinnGen. Significant findings from the primary inverse variance weighted analysis were validated using Bayesian Weighted MR (BWMR) and MR Robust Adjusted Profile Score (MR-RAPS) methods to mitigate potential pleiotropy. Sensitivity analyses and reverse MR assessed robustness and directionality.

Initial analyses identified highly significant causal associations for 33, 27, 25, and 30 brain IDPs with multiple infarctions, subcortical, sudden onset, and mixed VaD subtypes, respectively. Validation with BWMR and MR-RAPS confirmed 22, 17, 17, and 21 robustly causal IDPs for each subtype. Key findings included causal roles for gray matter volume/surface area changes in cognition-related regions specific to each subtype, as well as axonal and myelin damage with distinct anatomical localization in each subtype. Reverse MR found no evidence that any VaD subtype causally influenced the identified brain IDPs.

Our study provides robust genetic evidence for distinct causal relationships between specific patterns of brain structural alterations and different VaD subtypes. These subtype-specific neuroimaging signatures highlight divergent neuroanatomical substrates underlying VaD heterogeneity and offer potential targets for developing diagnostic biomarkers.