Arjen Lindenholz

139 Intracranial Atherosclerotic Burden and Cerebral Parenchymal Changes at 7T MRI 6 in size and scored according to a recent consensus paper. 20 The periventricular and deep white matter hyperintensities on FLAIR imaging were also defined according to the STRIVE criteria and were scored with the Fazekas four-point scale for periventricular (0: absent WMH lesions, 1: ‘caps’ or pencil-thin lining, 2: smooth ‘halo’ and 3: irregular hyperintensities extending into the deep white matter) and for deep WMH (0: absence or a single punctate WMH lesion, 1: multiple punctate lesions, 2: beginning confluency of lesions and 3: large confluent lesions). 27,28 The presence of any cerebral infarct in the flow territory of the anterior circulation and the WMH score (dichotomized into high; Fazekas 2 or 3 and low; Fazekas 0 or 1) were used as primary outcomes in the analyses. The number of cerebral infarcts were used as outcomes for the secondary analysis. Statistical Analysis Descriptive baseline statistics are presented as proportions or means. To estimate associations between intracranial vessel wall lesion burden and presence or number of cerebral parenchymal changes, appropriate regression analyses for modeling count data were performed. Age and sex were included as covariates. In the primary analyses, the association between the number of intracranial vessel wall lesions (as continuous and independent variable) and the presence of any scored cerebral parenchymal change (dichotomized outcome as dependent variable) were individually and all together investigated with a negative log- binomial regression model. A composite variable was composed with the infarcts that are often considered to be manifestations of cerebral small vessel disease, and included small subcortical and deep grey matter infarcts and lacunes of presumed vascular origin. 27 In the secondary analyses, the associations between the total number of intracranial vessel wall lesions (independent variable) and the total number of infarcts (count data; classified by infarct type, dependent variable) were investigated with a negative log-binomial regression model for the number of infarcts. To assess the association between the number of enhancing intracranial vessel wall lesions (independent variable) and presence and number of the scored cerebral parenchymal changes as dependent variables we used the same methods as in the primary and secondary analysis. Negative log-binomial regression generally provides rate ratios, which are interpretable as relative risks. Therefore, associations are presented in more commonly used relative risks. For all analyses 95% confidence intervals (CI) are given. A two-sided p-value < 0.05 was considered statistically significant. Statistical analyses were performed using SPSS version 21.0 (IBM SPSS Statistics, IBM Corp., Armonk, NY, USA).