P3-140: Age differences in the association of white matter lesions with the occurrence of dementia: The memento cohort

Brain MRI studies of biomarkers predicting dementia development in patients with mild cognitive impairment have focused on neurodegenerative biomarkers such as whole brain or hippocampal atrophy. Neuropathological studies have shown that, in older individuals, vascular and neurodegenerative abnormalities coexist and could have synergistic effects on dementia progression. In a large naturalistic cohort, we evaluated vascular and neurodegenerative MRI markers as possible predictors of dementia onset across age categories. Memento is a multicenter longitudinal study on determinants of cognitive evolution (including dementia) in participants consecutively enrolled in French memory clinics, presenting either isolated cognitive complaints or mild cognitive impairment. Throughout France, 2319 participants have been enrolled and will be followed at least five years with yearly clinical examinations and brain imaging (MRI) every two years. The “CATI-project” developed automated software to analyze baseline MR images: SACHA for hippocampal volume (Hippv), WHASA for white matter lesions volumes (WMLv). All dementia cases diagnosed during follow-up were validated by an events review committee. Cox regression models investigated the likelihood of progression to dementia according to baseline MRI markers adjusting for age, gender, education, baseline CDR score and vascular factors. Mean baseline age was 70.4 years (Standard Deviation (SD) =8.7), 62% were women and 40% had a clinical dementia rating scale scored 0. During follow-up, 85 participants became demented out of 3007 person-years (PY) (2.8 per 100PY). In those aged less than 75 (n=1591), higher Hippv was significantly associated with a lower dementia risk (Hazard Ratio (HR) per SD in Hippv=0.51, p<0.001) whereas no association was observed with WMLv (HR per SD=0.96, p=0.84). In those aged 75+, higher Hippv was associated with lower dementia risk (HR per SD=0.54, p<0.001) whereas higher WMLv was significantly associated with increased dementia risk (HR per SD=1.55, p=0.001). The age*Wmlv interaction was significant (p=0.17). In this large naturalistic cohort, higher baseline WMLv predicted dementia onset, independently of Hippv, in participants aged over 75 years but not in participants aged less than 75. These results are consistent with post-mortem brain studies underlining the role of vascular pathologies in the oldest old and their rising impact with age.