Arjen Lindenholz

148 CHAPTER 6 Several etiologies have been described that attempt to explain different categories of WMH, yet it remains unclear if there indeed exist distinct mechanisms for periventricular and deep WMH development. 19,43,44 The presence of both periventricular and deep WMH may be a result of underlying vascular changes (and subsequent loss of vascular integrity) of both large and small vessel disease. 19 Nevertheless, different associations between periventricular and deep WMH were found in our study and reasons for these discrepancies have yet to be elucidated. In contrast to previous studies investigating ICAS as a risk factor for (large artery) ischemic stroke, in our study no association was found between intracranial vessel wall lesion burden and presence and number of (large) cortical infarcts. This may be due to the setup of the IVI study, for which specifically ischemic stroke and TIA patients were selected, not limited to large-artery atherosclerosis according to the TOAST criteria, andwithout inclusion of healthy volunteers. Large numbers of patients are needed to detect small differences in this type of study population. Another explanation (especially for the number of cortical infarcts), may be the difference in definition of ICAS. Previous studies often used intracranial stenosis as a proxy for ICAS, but this represents an advanced state of ICAS. 45,46 In early to moderate stages of ICAS, vessel wall changes may be its only (early) sign; due to vascular remodeling, luminal narrowing often occurs in advanced stages of the disease. 47 The strength of this study is that data has been derived from a relatively large number of patients who received 7T MR vessel wall imaging; high resolution 7T MRI has been shown superior to 3T MRI in visualizing the intracranial vessel wall because of its high contrast-to-noise ratio, and has a better detection rate of especially small symptomatic and asymptomatic cerebral parenchymal changes. 12,48 Still, for analyzing all associations in this study, the absolute number of included patients is limited and the inherently induced selection bias due to the design of the study may have affected the analyzed associations underestimating the true association due to reversed causality. Also, the smaller size of the subgroup analysis may have accounted for the lack of an association, which also applies for the analyses with the enhancing vessel wall lesion burden as included variable. When periventricular and deep WMH share the same etiology, this may also be an explanation for the difference in associations between the presence of periventricular versus deep WMH and vessel wall lesion burden. Overall, there was an overrepresentation of patients with a Fazekas score of 1 in both periventricular and deep WMH, implicating relatively small subgroups of patients with a different Fazekas score. Therefore, (ordinal) regression analyses could not be reliably performed among the individual Fazekas scores. Also, we based our results on multivariable models and tested 7 variables of cerebral parenchymal changes with the presence of (enhancing) vessel wall lesions. The Bonferroni correction for multiple testing would mean that our p-value of 0.05 should be divided by the 9 outcome variables = 0.009, which indicates statistical