BACKGROUND AND OBJECTIVESCerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL) is the most frequent small artery brain disease caused by pathogenic variants of the NOTCH3 gene. During the disease, we still do not know how the various deficits progress and develop with each other at different stages of the disease. We aim to model disease progression and identify possible progressive subgroups and the effects of different covariates on clinical worsening.METHODSData were obtained from patients followed in the French CADASIL referral center, who were aged 25-80 years and had completed at least 2 visits and one of 14 clinical scores. Progression and variability were assessed using a disease course model (Leaspy). A Gaussian mixture model was used to identify different progression subgroups. Logistic regressions were used to compare the characteristics between groups.RESULTSIn 395 patients along 2,007 visits, the follow-up ranged from 6 months to 19 years, with a mean of 7.5 years. They were 45% men with a mean age of 52.2 years. The evolution curves of the different scores showed that clinical manifestations develop heterogeneously and can vary considerably depending on the disease stage. We identified an early-onset, rapidly progressing subgroup of patients with earlier motor symptoms and focal neurologic deficits (median time shift 59 [Q1-Q3 48.9-66.3], median acceleration rate 0.84 [0.07-1.31]) and a late-onset slowly progressing group with earlier cognitive symptoms (median time shift 69.2 [63.4-75.1], median acceleration rate -0.18 [-0.48 to 0.14]). Male sex, lower education level, hypertension, and NOTCH3 pathogenic variant location within epidermal growth factor-like repeat (EGFr) 1-6 were found to be associated with this group difference.DISCUSSIONOur results suggest a gradual and heterogeneous decline in different clinical and cognitive performances over the lifetime of patients with CADASIL. Two progression profiles-one rapid and early and the other, more delayed and slower-are possible after the onset of symptoms. A major limitation of our study is that the clusters were assessed post hoc, which may induce some bias. Overall, male sex, low level of education, pathogenic variant location in EGFr 1 to 6 domains, smoking, and/or arterial hypertension may affect the clinical progression of the disease.