Vascular cognitive impairment neuropathology guidelines (VCING): the contribution of cerebrovascular pathology to cognitive impairment

Abstract

There are no generally accepted protocols for post-mortem assessment in cases of suspected vascular cognitive impairment. Neuropathologists from seven UK centres have collaborated in the development of a set of vascular cognitive impairment neuropathology guidelines (VCING), representing a validated consensus approach to the post-mortem assessment and scoring of cerebrovascular disease in relation to vascular cognitive impairment. The development had three stages: (i) agreement on a sampling protocol and scoring criteria, through a series of Delphi method surveys; (ii) determination of inter-rater reliability for each type of pathology in each region sampled (Gwet’s AC2 coefficient); and (iii) empirical testing and validation of the criteria, by blinded post-mortem assessment of brain tissue from 113 individuals (55 to 100 years) without significant neurodegenerative disease who had had formal cognitive assessments within 12 months of death. Fourteen different vessel and parenchymal pathologies were assessed in 13 brain regions. Almost perfect agreement (AC2 > 0.8) was found when the agreed criteria were used for assessment of leptomeningeal, cortical and capillary cerebral amyloid angiopathy, large infarcts, lacunar infarcts, microhaemorrhage, larger haemorrhage, fibrinoid necrosis, microaneurysms, perivascular space dilation, perivascular haemosiderin leakage, and myelin loss. There was more variability (but still reasonably good agreement) in assessment of the severity of arteriolosclerosis (0.45–0.91) and microinfarcts (0.52–0.84). Regression analyses were undertaken to identify the best predictors of cognitive impairment. Seven pathologies—leptomeningeal cerebral amyloid angiopathy, large infarcts, lacunar infarcts, microinfarcts, arteriolosclerosis, perivascular space dilation and myelin loss—predicted cognitive impairment. Multivariable logistic regression determined the best predictive models of cognitive impairment. The preferred model included moderate/severe occipital leptomeningeal cerebral amyloid angiopathy, moderate/severe arteriolosclerosis in occipital white matter, and at least one large infarct (area under the receiver operating characteristic curve 77%). The presence of 0, 1, 2 or 3 of these features resulted in predicted probabilities of vascular cognitive impairment of 16%, 43%, 73% or 95%, respectively. We have developed VCING criteria that are reproducible and clinically predictive. Assuming our model can be validated in an independent dataset, we believe that this will be helpful for neuropathologists in reporting a low, intermediate or high likelihood that cerebrovascular disease contributed to cognitive impairment.