A Novel Approach to White Matter Hyperintensity Analysis Jasmin Annica Kuhn-Keller IN SHAPE
IN SHAPE: A Novel Approach to White Matter Hyperintensity Analysis Jasmin Annica Kuhn
Cover artwork: Valeria Kuhn-Oğuz Layout and printing: Ridderprint | www.ridderprint.nl ISBN: 978-94-6506-384-3 This PhD project was supported by Alzheimer Nederland grants WE.03-2019-08 and WE.25-2020-05. Moreover, thesis printing was subsidized by an Alzheimer Nederland grant (WE.04-2024-97). © Jasmin Annica Kuhn-Keller, 2024 All rights reserved. No part of this thesis may be reproduced, stored or transmitted in any form or by any means without prior permission of the author, or the copyrightowning journals for previously published chapters.
IN SHAPE: A Novel Approach to White Matter Hyperintensity Analysis Proefschrift ter verkrijging van de graad van doctor aan de Universiteit Leiden, op gezag van rector magnificus prof.dr.ir. H. Bijl, volgens besluit van het college voor promoties te verdedigen op donderdag 21 november 2024 klokke 14:30 uur door Jasmin Annica Kuhn geboren te Böblingen, Duitsland in 1995
Promotor Prof. Dr. Ir. M.J.P. van Osch Copromotor Dr. J.H.J.M. de Bresser Promotiecommissie Prof. Dr. A. G. Webb Dr. S. van Veluw, Massachusetts General Hospital, Harvard Medical School Dr. E.E. Bron, Erasmus Medical Center Prof. Dr. E. Richard, Radboud Medical Center
Studiere nur und raste nie, Du kommst nicht weit mit deinen Schlüssen; Das ist das Ende der Philosophie, Zu wissen, daß wir glauben müssen. Emanuel Geibel
CONTENTS CHAPTER 1 GENERAL INTRODUCTION 9 CHAPTER 2 DIFFERENT CARDIOVASCULAR RISK FACTORS ARE RELATED TO DISTINCT WHITE MATTER HYPERINTENSITY MRI PHENOTYPES IN OLDER ADULTS 19 CHAPTER 3 WHITE MATTER HYPERINTENSITY SHAPE IS RELATED TO LONG-TERM PROGRESSION OF CEREBROVASCULAR DISEASE IN COMMUNITY-DWELLING OLDER ADULTS 37 CHAPTER 4 A MORE IRREGULAR SHAPE OF WHITE MATTER HYPERINTENSITIES IS ASSOCIATED WITH COGNITIVE DECLINE OVER 5 YEARS IN COMMUNITY-DWELLING OLDER ADULTS 63 CHAPTER 5 WHITE MATTER HYPERINTENSITY SHAPE IS ASSOCIATED WITH LONG-TERM DEMENTIA RISK 83 CHAPTER 6 IDENTIFICATION OF DISTINCT BRAIN MRI PHENOTYPES AND THEIR ASSOCIATION WITH LONG-TERM DEMENTIA RISK IN COMMUNITY-DWELLING OLDER ADULTS 111 CHAPTER 7 STUDY PROTOCOL OF THE WHIMAS: IDENTIFICATION OF NOVEL 7T MRI WHITE MATTER HYPERINTENSITY SHAPE AND BRAIN CLEARANCE MARKERS FOR CEREBRAL SMALL VESSEL DISEASE 141 CHAPTER 8 GENERAL DISCUSSION 161 CHAPTER 9 SUMMARY 169 APPENDICES LIST OF PUBLICATIONS 182 CURRICULUM VITAE 185 ACKNOWLEDGEMENTS 186
CHAPTER1
GENERAL INTRODUCTION
10 Chapter 1 1.1 AGEING AND DEMENTIA With an ageing population age-related diseases such as dementia will continue to increase in the coming years.1 This will create an increasing burden for society and health care systems.1 There are several types of dementia, the most common ones being Alzheimer’s disease, vascular dementia, frontotemporal dementia and Lewy body dementia.2 Multiple co-existing diseases contribute to the dementia phenotype and cardiovascular risk factors are an important contributor. Cerebrovascular disease is an umbrella term for a range of conditions that result in pathological changes in or surrounding the cerebral blood vessels.3 Large vessel disease, a type of cerebrovascular disease, is caused by atherosclerosis in the upstream arteries leading to the brain and is a major cause of ischemic stroke.4,5 Cerebral small vessel disease (SVD) refers to a group of pathological changes affecting the cerebral small arteries, arterioles, capillaries and venules of the brain.3 SVD is a major contributor to ischemic stroke, cognitive decline, and dementia.6 SVD cannot be referred to as a single disease, but should be considered a combination of radiological features that can be caused by different genetic and non-genetic diseases.3 Examples of genetic SVD forms are Dutch-type cerebral amyloid angiopathy (D-CAA) and cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL). Genetic forms of SVD typically have a more early time of onset compared to the sporadic forms. The main types of cerebral SVD that occur in older adults are ischemic SVD (including e.g. arteriolosclerosis)4 and sporadic cerebral amyloid angiopathy (CAA), where amyloid is deposited in the walls of the cerebral blood vessels increasing the risk for hemorrhage.3 SVD often starts with asymptomatic changes in the vasculature and parenchyma, which can be captured best in population-based studies or in targeted studies in genetic cases. To date, treatment options are limited for vascular dementia. There are some preventive life-style changes and several new pharmaceutical options that make it important to select patients at an early stage. For example, a new pharmaceutical trial on CAA will start in 2024, also including patients with Dutch-type cerebral amyloid angiopathy7 For patient selection and especially when aiming for the earlier disease stages, specific markers are currently lacking. This thesis aims to identify and characterize novel specific SVD markers. 1.2 NEUROIMAGING IN CEREBRAL SVD Magnetic resonance imaging (MRI) has developed a lot since its introduction in the early 1980s. Today, MRI is a versatile technique that allows the investigation of brain changes in humans in a non-invasive manner both in clinical as well as in research settings. Changes in blood vessel properties caused by SVD, such as arteriolosclerosis, lipohyalinosis, and fibrinoid necrosis are difficult to image directly
11 General Introduction 1 with MRI, but they will eventually also lead to pathology of the brain parenchyma that can be made visible by MRI.3,6 The most common SVD related brain changes are white matter hyperintensities (WMH), lacunes, microbleeds, enlarged perivascular spaces and also atrophy.6 WMH appear as hyperintense lesions on fluid-attenuated inversion recovery (FLAIR) MRI.6,8 WMH can be categorized into three types based on their location and extent: periventricular, confluent and deep WMH. Periventricular and confluent WMH surround the margins of the lateral ventricles, while deep WMH are punctual lesions located in the deep white matter. In clinical practice assessments of brain MRI scans are usually based on visual inspection and scoring. However, automated segmentation techniques can provide more objective results and are especially useful for research and even more when working with larger datasets. For example, WMH volumes or brain atrophy can be calculated based on automated segmentations.9–12 Different MRI markers of parenchymal changes and their distribution over the brain can be used to discriminate different SVD types.6,13 SVD is a heterogeneous disease including many possible underlying pathologies. Some brain changes in SVD might be the result of impaired clearance of waste products, which has been associated with aging and dementia.14 It is postulated that the brain clearance process is partly driven by the glymphatic system, where cerebrospinal fluid and interstitial fluid ‘flush’ brain tissue and transport metabolic waste out of the parenchyma via perivascular spaces. In cerebral amyloid angiopathy15 and Alzheimer’s dementia, glymphatic function might be impaired.16 Currently, brain clearance related processes are mainly studied invasively in humans, for example by contrast-enhanced MRI following intrathecal injection.17 In this thesis a study including non-invasive MR imaging techniques of the glymphatic system is proposed. 1.3 WMH SHAPE Traditionally, WMH were investigated in research settings by visual rating scales or volume measurements.18 While WMH volume is an objective measure that can be obtained automatically, it is also a rather crude measure. When inspecting MRI scans visually, WMHs can appear very different from each other in shape and location, even if their calculated volumes may be roughly the same. However, measures to objectively quantify such differences that may easily be caught by the eye of a neuroradiologist were lacking. For example, the borders of WMH can in some cases look smooth while in other cases they are irregular and complex. To automatically quantify shape differences of WMHs, several WMH shape markers were introduced previously.19,20 For periventricular/confluent WMH solidity, convexity, concavity index, and fractal dimension specific measures were introduced using the formulas shown
12 Chapter 1 in Figure 1.1.20 High solidity and convexity, as well as low concavity index and fractal dimension reflect a more irregular shape. For deep WMH, the shape markers that were found appropriate are fractal dimension and eccentricity. Higher eccentricity suggests a more elongated shape. Periventricular/confluent WMH and deep WMH have very different shape appearances which is why different shape markers are used for each of them to most accurately capture their shape. Figure 1.1. Illustration of shapes and different shape markers. Solidity, convexity, concavity index, and fractal dimension are calculated for periventricular/confluent WMH. Eccentricity and fractal dimension are calculated for deep WMH. This Figure is modified from figure 7.3 of Chapter 7.
13 General Introduction 1 Different WMH types (periventricular/confluent and deep) and also different WMH shape patterns seem to be associated with different underlying pathological changes. A more irregular shape of WMH go hand-in-hand with more severe parenchymal changes.21–23 Furthermore, a previous study has shown that a more irregular WMH shape was associated with increased stroke risk and increased mortality in patients with manifest arterial disease.24 I hypothesize that different underlying SVD pathologies result in a different WMH shape that can be quantified by MRIbased WMH shape markers. These shape markers may provide a more detailed characterization of WMH than volume alone. 1.4 AGES-REYKJAVIK STUDY Large longitudinal population-based studies focused on ageing are rare since they are expensive and labor-intensive to carry out. This study type is, however, extremely valuable as it allows investigations with high statistical power. At the same time, they provide high external validity, since the participants of the study come from the general population with minimal inclusion bias. A substantial part of this thesis is focused on data from the Age-Gene/Environment Susceptibility (AGES)-Reykjavik study, a large population-based study.25 The AGES– Reykjavik study originates from the Reykjavik Study, a cohort established in 1967 to prospectively study cardiovascular disease in the general population of Iceland. Included participants were born between 1907 and 1935 and were living in Reykjavik in 1967. For the AGES-Reykjavik study, participants that were still alive were randomly selected for a follow-up between 2002 and 2006 when they underwent amongst other measures, a baseline brain MRI scan and cognitive assessment.26 Another visit took place between 2007 and 2011, where the same brain MRI protocol and the same cognitive assessments were repeated.27 Furthermore, the participants were followed for dementia outcome up to 13.4 years after the first MRI scan session through vital statistics and hospital records, and by the nursing home and homebased resident assessment instrument.28 The AGES-Reykjavik study is an extensive multidisciplinary study and includes besides neuroimaging and neurocognitive data also genetic, cardiovascular and musculoskeletal data.25 In the related chapters of this thesis, the focus will be on neuroimaging and neurocognitive data obtained as part of the AGES study.
14 Chapter 1 1.5 AIM AND OUTLINE OF THIS THESIS The overarching aim of this thesis is to exploit the shape of WMHs to better characterize WMH and thereby to improve the clinical interpretation of WMHs and to investigate whether it could predict clinical outcome. This thesis is mostly based on non-demented and community-dwelling older individuals. Moreover, a study set up focusing on a memory-clinic population will be discussed to get more pathologyfocused insights into the formation of WMH. In Chapter 2, the association of different cardiovascular risk factors with WMH shape was investigated in older adults as included in the Biomarker Development for Postoperative Cognitive Impairment in the Elderly (BIOCOG) study. The study included non-demented older adults scheduled for major elective surgery. Next, in Chapter 3, it was examined whether WMH shape is related to long-term progression of cerebrovascular disease in the AGES Reykjavik dataset. Besides WMH, different types of infarcts, microbleeds and enlarged perivascular spaces and their relationship with WMH shape were evaluated in this chapter. In Chapter 4, the focus was on investigating the association between baseline WMH shape and cognitive decline measured in three different domains (memory, executive function, and processing speed) over 5.2 years in the AGES Reykjavik dataset. In Chapter 5, the association of baseline WMH shape and long-term dementia risk after up to 13.4 years was assessed in the AGES Reykjavik study. In Chapter 6, brain MRI phenotypes were obtained using a hierarchical clustering method in the AGES Reykjavik dataset. In a second step, it was investigated whether these phenotypes are related to long-term risk (10 years) for dementia. In Chapter 7, a novel prospective cross-sectional study is presented applying WMH shape markers and other cutting-edge MRI techniques to further understand processes involved in SVD. This WHIte MAtter hyperintensity Shape and glymphatics (WHIMAS) study is focused on brain MRI determinants of cognitive impairment in geriatric clinic outpatients. Lastly, in Chapter 8, the main findings of this thesis and future directions of research are discussed.
15 General Introduction 1 1.6 REFERENCES 1. World Health Organization. Dementia. Fact Sheet. 2023. https://www.who.int/news-room/ fact-sheets/detail/dementia (accessed 23 Feb 2024). 2. Livingston G, Sommerlad A, Orgeta V, Costafreda SG, Huntley J, Ames D et al. Dementia prevention, intervention, and care. The Lancet 2017; 390: 2673–2734. 3. Wardlaw JM, Smith C, Dichgans M. Small vessel disease: mechanisms and clinical implications. Lancet Neurol 2019; 18: 684–696. 4. Wardlaw JM, Smith C, Dichgans M. Mechanisms of sporadic cerebral small vessel disease: Insights from neuroimaging. Lancet Neurol 2013; 12: 483–497. 5. Pantoni L. Cerebral small vessel disease: from pathogenesis and clinical characteristics to therapeutic challenges. Lancet Neurol 2010; 9: 689–701. 6. Duering M, Biessels GJ, Brodtmann A, Chen C, Cordonnier C, de Leeuw FE et al. Neuroimaging standards for research into small vessel disease—advances since 2013. Lancet Neurol 2023; 22: 602–618. 7. Voigt S, Koemans EA, Rasing I, van Etten ES, Terwindt GM, Baas F et al. Minocycline for sporadic and hereditary cerebral amyloid angiopathy (BATMAN): study protocol for a placebo-controlled randomized double-blind trial. Trials 2023; 24: 1–6. 8. Hajnal J V., de Coene B, Lewis PD, Baudouin CJ, Cowan FM, Pennock JM et al. High signal regions in normal white matter shown by heavily T2-weighted CSF nulled IR sequences. J Comput Assist Tomogr 1992; 16: 506–513. 9. de Bresser J, Portegies MP, Leemans A, Biessels GJ, Kappelle LJ, Viergever MA. A comparison of MR based segmentation methods for measuring brain atrophy progression. Neuroimage 2011; 54: 760–768. 10. Pini L, Pievani M, Bocchetta M, Altomare D, Bosco P, Cavedo E et al. Brain atrophy in Alzheimer’s Disease and aging. Ageing Res Rev 2016; 30: 25–48. 11. Balakrishnan R, Valdés Hernández M del C, Farrall AJ. Automatic segmentation of white matter hyperintensities from brain magnetic resonance images in the era of deep learning and big data – A systematic review. Computerized Medical Imaging and Graphics 2021; 88: 101867. 12. Kuijf HJ, Casamitjana A, Collins DL, Dadar M, Georgiou A, Ghafoorian M et al. Standardized Assessment of Automatic Segmentation of White Matter Hyperintensities and Results of the WMH Segmentation Challenge. IEEE Trans Med Imaging 2019; 38: 2556–2568. 13. Charidimou A, Boulouis G, Frosch MP, Baron JC, Pasi M, Albucher JF et al. The Boston criteria version 2.0 for cerebral amyloid angiopathy: a multicentre, retrospective, MRI– neuropathology diagnostic accuracy study. Lancet Neurol 2022; 21: 714–725. 14. Benveniste H, Nedergaard M. Cerebral small vessel disease: A glymphopathy? Curr Opin Neurobiol 2022; 72: 15–21. 15. van Veluw SJ, Benveniste H, van Osch MJP, Clearance the LFTN of E on B, Bakker ENTP, Carare RO et al. A translational approach towards understanding brain waste clearance in cerebral amyloid angiopathy. Eur Heart J 2024. 16. Peng W, Achariyar TM, Li B, Liao Y, Mestre H, Hitomi E et al. Suppression of glymphatic fluid transport in a mouse model of Alzheimer’s disease. Neurobiol Dis 2016; 93: 215–225. 17. Ringstad G, Valnes LM, Dale AM, Pripp AH, Vatnehol SAS, Emblem KE et al. Brain-wide glymphatic enhancement and clearance in humans assessed with MRI. JCI Insight 2018; 3. 18. Wardlaw JM, Valdés Hernández MC, Muñoz-Maniega S. What are white matter hyperintensities made of? Relevance to vascular cognitive impairment. J Am Heart Assoc 2015; 4: 001140.
16 Chapter 1 19. De Bresser J, Kuijf HJ, Zaanen K, Viergever MA, Hendrikse J, Biessels GJ et al. White matter hyperintensity shape and location feature analysis on brain MRI; Proof of principle study in patients with diabetes. Sci Rep 2018; 8: 1–10. 20. Ghaznawi R, Geerlings MI, Jaarsma-Coes MG, Zwartbol MHT, Kuijf HJ, van der Graaf Y et al. The association between lacunes and white matter hyperintensity features on MRI: The SMART-MR study. Journal of Cerebral Blood Flow and Metabolism 2019; 39: 2486–2496. 21. Alber J, Alladi S, Bae H, Barton DA, Beckett LA, Bell JM et al. White matter hyperintensities in vascular contributions to cognitive impairment and dementia (VCID): Knowledge gaps and opportunities. Alzheimer’s & Dementia: Translational Research & Clinical Interventions 2019; 5: 107–117. 22. Fazekas F, Kleinert R, Offenbacher H, Schmidt R, Kleinert G, Payer F et al. Pathologic correlates of incidental MRI white matter signal hyperintensities. Neurology 1993; 43: 1683–1689. 23. Kim KW, MacFall JR, Payne ME. Classification of White Matter Lesions on Magnetic Resonance Imaging in Elderly Persons. Biol Psychiatry 2008; 64: 273–280. 24. Ghaznawi R, Geerlings M, Jaarsma-Coes M, Hendrikse J, de Bresser J. Association of White Matter Hyperintensity Markers on MRI and Long-term Risk of Mortality and Ischemic Stroke. Neurology 2021. 25. Harris TB, Launer LJ, Eiriksdottir G, Kjartansson O, Jonsson P V., Sigurdsson G et al. Age, gene/environment susceptibility-reykjavik study: Multidisciplinary applied phenomics. Am J Epidemiol 2007; 165: 1076–1087. 26. Saczynski JS, Sigurdsson S, Jonsdottir MK, Eiriksdottir G, Jonsson P V., Garcia ME et al. Cerebral infarcts and cognitive performance: importance of location and number of infarcts. Stroke 2009; 40: 677–682. 27. Sigurdsson S, Aspelund T, Kjartansson O, Gudmundsson EF, Jonsdottir MK, Eiriksdottir G et al. Incidence of Brain Infarcts, Cognitive Change, and Risk of Dementia in the General Population: The AGES-Reykjavik Study (Age Gene/Environment Susceptibility-Reykjavik Study). Stroke 2017; 48: 2353–2360. 28. Morris JN, Hawes C, Fries BE, Phillips CD, Mor V, Katz S et al. Designing the National Resident Assessment Instrument for Nursing Homes. Gerontologist 1990; 30: 293–307.
17 General Introduction 1
CHAPTER 2
DIFFERENT CARDIOVASCULAR RISK FACTORS ARE RELATED TO DISTINCT WHITE MATTER HYPERINTENSITY MRI PHENOTYPES IN OLDER ADULTS Jasmin A. Keller, Ilse M.J. Kant, Arjen J.C. Slooter, Simone J.T. van Montfort, Mark A. van Buchem, Matthias J.P. van Osch, Jeroen Hendrikse, Jeroen de Bresser. Published in: NeuroImage: Clinical 2022, 35, 103131. https://doi.org/10.1016/j.nicl.2022.103131
20 Chapter 2 2.1 ABSTRACT The underlying mechanisms of the association between cardiovascular risk factors and a higher white matter hyperintensity (WMH) burden are unknown. We investigated the association between cardiovascular risk factors and advanced WMH markers in 155 non-demented older adults (mean age: 71 ± 5 years). The association between cardiovascular risk factors and quantitative MRI-based WMH shape and volume markers were examined using linear regression analysis. Presence of hypertension was associated with a more irregular shape of periventricular/confluent WMH (convexity (B (95 % CI)): −0.12 (−0.22–−0.03); concavity index: 0.06 (0.02–0.11)), but not with total WMH volume (0.22 (−0.15–0.59)). Presence of diabetes was associated with deep WMH volume (0.89 (0.15–1.63)). Body mass index or hyperlipidemia showed no association with WMH markers. In conclusion, different cardiovascular risk factors seem to be related to a distinct pattern of WMH shape markers in non-demented older adults. These findings may suggest that different underlying cardiovascular pathological mechanisms lead to different WMH MRI phenotypes, which may be valuable for early detection of individuals at risk for stroke and dementia.
2 21 Cardiovascular risk factors are related to distinct white matter hyperintensity MRI phenotypes 2.2 INTRODUCTION Cerebral small vessel disease (SVD) is associated with the occurrence of dementia and stroke.1 In SVD, different underlying pathological mechanisms (such as small or large vessel atheromas or embolisms) lead to the phenotype of MRI-visible brain changes, namely white matter hyperintensities (WMH), lacunes and microbleeds.2,3 Individual cardiovascular risk factors play an important role in the etiology of SVD as they impact the small vessels via different pathological pathways and hence lead to distinct patterns of SVD related brain changes that may be differentiated.4 An example of this principle is the association between hypertension and deep cerebral microbleeds versus cerebral amyloid angiopathy and lobar microbleeds.5 Although WMH are the key MRI marker of idiopathic SVD, little is known about the association of individual cardiovascular risk factors and distinct patterns of WMH. WMH volume is typically used to study WMH, but this marker fails to fully quantify the complex brain changes related to underlying pathological changes of SVD.4 Moreover, WMH volume alone is insufficient for differentiation of underlying disease mechanisms leading to WMH. Aiming to overcome the limitations of conventional WMH volume markers, in recent studies WMH type and shape were introduced as more advanced WMH markers.6,7 For example, different WMH type (periventricular, deep and confluent) and also different WMH shape is associated with different underlying pathological changes.4,8,9 Furthermore, previous studies on WMH shape show potential diagnostic and prognostic value related to an increased mortality and stroke risk.6,7,10 The investigation of WMH type and shape markers may therefore help in the postulation of potential mechanisms of WMH development. Our hypothesis is that in older adults specific cardiovascular risk factors relate to distinct patterns of WMH, which can be quantified using advanced WMH MRI markers. Studying this hypothesis may aid in the understanding of WMH development and could in the future be used for earlier detection of individuals at risk for stroke or dementia. Therefore, we aimed to investigate the association between cardiovascular risk factors and advanced WMH markers (shape, type, and volume) in non-demented older adults. 2.3 METHODS 2.3.1 Participants We included data from the BioCog consortium study, collected from the University Medical Center Utrecht site.11 Inclusion criteria for the BioCog study were: minimal age of 65 years, a mini-mental state exam (MMSE) score of 24 or higher, and major surgery scheduled of at least 60 min (cardiothoracic (n = 35), gastroenterological (n = 22), gynecologic (n = 6), jaw (n = 11), ear nose throat (13), orthopedic (43), urological
22 Chapter 2 (25)). The MMSE was used as a screening for dementia in order to exclude severely cognitively impaired participants. The study was approved by the medical ethics committee (Medical Ethical Committee Utrecht nr. 14/469). Written informed consent was obtained from all participants. 2.3.2 Procedure and demographics All participants were invited to the hospital before surgery for questionnaires and MRI scanning. Demographic data and medical his- tory questionnaires, and MMSE scores were obtained prior to surgery. Symptoms of anxiety and depression were assessed using the hospital and depression scale (HADS). Scores equal to or above 8 on the depression subscale were considered indicative of depressive symptoms.12 American Society of Anesthesiologists (ASA) classification scores were performed before surgery by anesthesiologists in training. 2.3.3 Cardiovascular risk factors Demographic data and data concerning cardiovascular risk factors was collected using questionnaires and medical history records. Diabetes (type I&II), hypertension and hyperlipidemia were registered in the database if they were known by the subject, and treated. BMI in kg/m3 was collected as an additional cardiovascular risk factor. Obesity was defined as a BMI equal to or above 30 kg/m3. 2.3.4 MRI scans The scans were performed before surgery on a Philips Achieva 3 Tesla MRI system. The standardized MRI scan protocol included a 3D T1-weighted sequence (voxel size = 1.0 × 1.0 × 1.0 mm3; TR/TE = 7.9/4.5 ms) and a 3D fluid-attenuated inversion recovery (FLAIR) sequence (voxel size = 1.11 × 1.11 × 0.56 mm3; TR/TE/TI = 4800/125/1650 ms). 2.3.5 WMH volume, type and shape Two categories of WMH were automatically determined: ‘deep’ WMH, and ‘periventricular/confluent’ WMH. WMH volumes were calculated automatically using validated methods.7 Based on the WMH segmentation data, the mean values per WMH shape marker (solidity, convexity, concavity index, fractal dimension for periventricular/confluent WMH; fractal dimension and eccentricity for deep WMH) were calculated for each participant with an in-house developed method.7 The image processing pipeline is illustrated in Figure 2.1. Statistical parametric mapping (SPM version 12; Wellcome Institute of Neurology University College London, UK, http://www.fil.ion.ucl.ac. uk/spm/doc/) in MATLAB (The MathWorks, Inc., Natick, Massachusetts, United States) was used to register 3D FLAIR images to 3D T1weighted images. Intracranial volume was calculated using the SPM12 with the option
2 23 Cardiovascular risk factors are related to distinct white matter hyperintensity MRI phenotypes ‘tissue volumes’. The lesion prediction algorithm of the lesion segmentation toolbox version 2.0.1513 (https://www.sta tistical-modeling.de/lst.html) for SPM12 was used for WMH segmentation. The quality of the WMH segmentations was visually checked by a trained researcher (IK) under supervision of a neuroradiologist experienced in brain segmentation (JB). Cortical brain infarcts were manually delineated and removed from the WMH probability maps. The lateral ventricles were segmented on the T1-weighted images using the automated lateral ventricle delineation toolbox14 (ALVIN; https://www.nitrc. org/projects/alvin_lv/ in SPM8. A threshold of 0.10 was applied on the probabilistic WMH segmentations. Periventricular WMH were defined as WMH contiguous with the lateral ventricles and extending ≤ 10 mm into the deep white matter. Confluent WMH were defined as WMH contiguous with the lateral ventricles and extending > 10 mm into the deep white matter. Deep WMH were defined as being located > 3 mm from the lateral ventricles and > 5 voxels (3.45 × 10−3 ml). Based on the binary WMH segmentation data, WMH shape markers were calculated.7 For periventricular/confluent WMH, solidity was calculated by dividing lesion volume by the volume of its convex hull. Dividing the convex hull sur- face area by the lesion surface area provides the convexity of the WMH. A lower solidity and a lower convexity indicates a more complex shape.7 The concavity index is a measure of roughness and was calculated from solidity and convexity.7 A higher concavity index indicates a more irregular shape.15 Eccentricity of deep WMH was calculated by dividing the minor axis of the WMH by its major axis. Fractal dimension was calculated as a measure for irregularity of deep, and periventricular/confluent WMH using the box-counting method. Higher fractal dimension and eccentricity values suggest a more complex WMH shape.7 These WMH shape markers were selected because of their ability to capture the expected shape variations of different types of WMH accurately (see Supplementary Table S.2.8.1 for an overview).7 Mean values per WMH shape marker were calculated per patient and used for further analyses. The accuracy of the WMH shape marker calculation is dependent on WMH size. Periventricular/confluent lesions vary in size and especially in earlier stages the lesions are not only smaller, but are in a different range of shape markers. The shape parameters used in our study are more accurate in capturing the WMH shape of bigger periventricular/confluent lesions. Therefore a minimum WMH volume of 4 ml was applied as a threshold for the analyses of periventricular/confluent WMH shape markers to assure our method is robust and accurate.
24 Chapter 2 2.3.6 Statistical analysis Normality of data was assessed by visual inspection of histograms and Q-Q plots. WMH volumes, solidity and concavity index were multiplied by 100 and natural log transformed to correct for non-normal distribution. The statistical analyses performed in this study are exploratory. Associations between cardiovascular risk factors (diabetes, hypertension, BMI, hyperlipidemia) and WMH markers (volume, type, shape) were assessed by linear regression analyses adjusted for age and sex; while WMH volume was additionally adjusted for intracranial volume. In secondary analyses, we performed stepwise linear regression analyses to investigate which of the cardiovascular risk factors accounts for most of the variation of the significantly associated WMH markers. In further secondary analyses, associations between age and sex with the WMH markers were assessed by linear regression analyses; while WMH volumes were adjusted for intracranial volume. Statistical analysis was performed in SPSS version 25. A p value of < 0.05 was considered as statistically significant. Figure 2.1 Schematic illustration of the image processing pipeline. The lateral ventricles were segmented from the T1-weighted MRI images. WMH segmentation was performed using the registered FLAIR images. WMH were subsequently classified into three types (deep, periventricular, and confluent) using two different masks based on the distance from the ventricles. Periventricular WMH were defined as WMH contiguous with the lateral ventricles and extending ≤ 10 mm into the deep white matter. Confluent WMH were defined as WMH contiguous with the lateral ventricles and extending > 10 mm into the deep white matter. Deep WMH were defined as being located > 3 mm from the lateral ventricles and > 5 voxels (3.45 × 10−3 ml). Based on the WMH type, different WMH shape markers were calculated.
2 25 Cardiovascular risk factors are related to distinct white matter hyperintensity MRI phenotypes 2.4 RESULTS A total of 178 participants met the inclusion criteria and were included, of which 23 participants had to be excluded from the current study for the following reasons: WMH segmentation errors (n = 2), over- segmentation of WMH (n = 8), anatomic abnormalities (n = 2), missing scans (n = 3), a large cyst (n = 1), MRI motion artefacts (n = 4) or other artefacts (n = 3). Baseline characteristics of the remaining 155 participants are shown in Table 2.1. A total of 47 % of the participants had hypertension, 34 % had hyperlipidemia, 16 % had diabetes and the mean BMI was 27 ± 4 kg/m2. Presence of hypertension was associated with a more irregular shape of periventricular/confluent WMH (a lower convexity: B (95 % CI): −0.12 (−0.22–−0.03); p = 0.01; a higher concavity index: 0.06 (0.02–0.10); p = 0.01) (see Table 2.2), but not with total WMH volume (0.22 (−0.15–0.59); p = 0.24). Presence of diabetes was associated with a higher deep WMH volume (B (95 % CI): 0.97 (0.25–1.70); p = 0.02) (see Table 2.3). Trends were found for an association between the presence of diabetes and a higher total WMH volume (0.45 (−0.05–0.95); p = 0.07) and a higher perivascular/confluent WMH volume (0.43 (−0.07–0.92); p = 0.09). No associations were found between presence of diabetes and WMH shape markers (see Table 2.2). Neither BMI nor hyperlipidemia were associated with WMH volume (B (95 % CI): 0.97 (0.25–1.70); p = 0.02) (see Table 2.3). Trends were found for an association between the presence of diabetes and a higher total WMH volume (0.45 (−0.05–0.95); p = 0.07) and a higher perivascular/confluent WMH volume (0.43 (−0.07–0.92); p = 0.09). No associations were found between presence of diabetes and WMH shape markers (see Table 2.2). Neither BMI nor hyperlipidemia were associated with WMH volume or shape markers. The mean WMH shape markers stratified for cardiovascular risk factor are shown in Table 2.4 and the mean WMH volumes values stratified for cardiovascular risk factor can be found in Table 2.5. In secondary analyses, we performed a stepwise linear regression to investigate which of the significantly associated cardiovascular risk factors accounts for most of the variation of the WMH marker. The results of these analyses were in line with our primary linear regression analyses (see the Supplementary results). In other secondary analyses, age was associated with periventricular/confluent WMH shape (convexity, concavity index and fractal dimension) and WMH volumes (see Supplementary Tables S.2.8.2 and S.2.8.3). Furthermore, sex was associated with periventricular/confluent WMH shape (solidity).
26 Chapter 2 Table 2.1. Baseline characteristics of the patient population. Total (n=155) Age 71 ± 5 Female sex 50 (32%) MMSE 29 (27,30) ASA score 1 19 (12%) 2 84 (54%) 3 52 (34%) Depressive symptoms 7 (5%) Vascular risk factors Hypertension 72 (47%) Diabetes (type I & II) 24 (16%) BMI (kg/m2) 27 ± 4 Hyperlipidemia 55 (34%) Obesity 30 (19%) Current smoker 13 (8%) Prior TIA or CVA 8 (5%) Data represent n (percentage), mean ± SD or median (interquartile range). MMSE: mini mental state exam; ASA: classification of disease severity for the American Society of Anesthesiologists. BMI: body-mass index. TIA: transient ischemic attack. CVA: cerebrovascular accident. Table 2.2 The association between cardiovascular risk factors and WMH shape markers. Hypertension Diabetes BMI Hyperlipidemia Periventricular/ Confluent WMH† Solidity‡ 0.84 (-0.10–0.27) -0.02 (-0.24–0.20) 0.00 (-0.02–0.02) -0.01 (-0.18–0.19) Convexity -0.12 (-0.22–-0.03)* -0.03 (-0.15–0.09) 0.00 (-0.01–0.01) 0.01 (-0.09–0.11) Concavity index‡ 0.06 (0.02–0.11)* 0.01 (-0.04–0.06) -0.00 (-0.01-0.01) -0.01 (-0.05-0.04) Fractal dimension 0.04 (-0.01–0.10) 0.01 (-0.07–0.07) -0.00 (-0.01–0.01) -0.03 (-0.08–0.03) Deep WMH Eccentricity 0.01 (-0.03–0.06) 0.00 (-0.05–0.06) -0.00 (-0.01–0.01) 0.02 (-0.02–0.06) Fractal dimension 0.01 (-0.10–0.12) -0.03 (-0.17–0.10) -0.01 (-0.02–0.00) -0.08 (-0.17–0.03) The values represent B values (95% confidence interval) of the linear regression analyses adjusted for age and sex. * p<0.05. ‡ Solidity and concavity index were multiplied by 100 and natural log transformed, due to non-normal distribution. † Periventricular/confluent WMH with a volume >4 ml. Periventricular/ confluent WMH: n=73; Deep WMH: n=122.
2 27 Cardiovascular risk factors are related to distinct white matter hyperintensity MRI phenotypes Table 2.3. The association between cardiovascular risk factors and WMH volume. Hypertension Diabetes BMI Hyperlipidemia Total WMH volume 0.22 (-0.15–0.59) 0.45 (-0.05–0.95) 0.00 (-0.04–0.05) 0.14 (-0.24–0.52) Periventricular/confluent WMH volume 0.21 (-0.16–0.57) 0.43 (-0.07–0.92) 0.00 (-0.04–0.05) 0.15 (-0.23–0.53) Deep WMH volume 0.24 (-0.32–0.80) 0.89 (0.15–1.63)* 0.05 (-0.02–0.11) -0.08 (-0.65–0.48) These values represent B values (95% confidence interval) of the linear regression analyses adjusted for age, sex and intracranial volume. WMH volumes were multiplied by 100 and natural log transformed, due to non-normal distribution. * p<0.05. WMH: white matter hyperintensities. BMI: body-mass index. Table 2.4 : Mean WMH shape values per cardiovascular risk factor Total Hypertension Diabetes Obesity Hyperlipidemia Periventricular/confluent WMH* Solidity 0.19 ± 0.08 0.19 ± 0.07 0.19 ± 0.06 0.19 ± 0.07 0.19 ± 0.08 Convexity 1.15 ± 0.21 1.08 ± 0.18 1.15 ± 0.21 1.19 ± 0.12 1.16 ± 0.22 Concavity index 1.19 ± 0.13 1.23 ± 0.14 1.19 ± 0.16 1.15 ± 0.08 1.18 ± 0.13 Fractal dimension 1.83 ± 0.13 1.87 ± 0.14 1.82 ± 0.13 1.80 ± 0.12 1.82 ± 0.13 Deep WMH Eccentricity 0.56 ± 0.12 0.57 ± 0.11 0.56 ± 0.10 0.56 ± 0.15 0.57 ± 0.14 Fractal Dimension 1.83 ± 0.28 1.84 ± 0.32 1.80 ± 0.17 1.76 ± 0.40 1.79 ± 0.34 Data are represented as means ± SD. Obesity was defined as a BMI >30 kg/m2. WMH shape markers are given for the total number of individuals, individuals with diabetes (type I & II), individuals with hypertension, obesity or hyperlipidemia, respectively. *Periventricular/confluent WMH with a volume >4 ml. Periventricular/confluent WMH: total n=73; hypertension: n=36; diabetes: n=16; obesity: n=16; hyperlipidemia: n=31. Deep WMH: total n=122; hypertension: n = 57; diabetes: n=22; obesity: n=25; hyperlipidemia: n=45. Table 2.5: Mean WMH volumes per cardiovascular risk factor Total Hypertension Diabetes Obesity Hyperlipidemia Total WMH volume 7.84 ± 11.37 10.07 ± 13.29 10.14 ± 13.93 6.77 ± 6.74 7.92 ± 10.62 Periventricular/ confluent WMH volume 7.53 ± 11.18 9.67 ± 13.09 9.54 ± 13.56 6.34 ± 6.57 7.61 ± 10.38 Deep WMH volume 0.32 ± 0.59 0.40 ± 0.69 0.61 ± 0.93 0.44 ± 0.74 0.31 ± 0.63 Data are represented as mean volumes (ml) ± SD. WMH volumes are given for the total number of individuals, individuals with diabetes (type I & II), individuals with hypertension, obesity or hyperlipidemia, respectively. Obesity was defined as a BMI >30 kg/m2. Total: n=155; hypertension: n=72; diabetes: n=24; obesity: n=30; hyperlipidemia: n=55. Missing values: hypertension: n=3; BMI: n=5; hyperlipidemia: n=3.
28 Chapter 2 2.5 DISCUSSION The present study aimed to investigate the association of cardiovascular risk factors and novel advanced WMH markers in non-demented older adults. We showed that the presence of hypertension was associated with a more irregular shape of periventricular/confluent WMH, but not with WMH volume. Presence of diabetes was associated with deep WMH volume. BMI or hyperlipidemia showed no association with WMH markers in our study. We found an association between hypertension and a more irregular WMH shape, but did not find an association with WMH volume. An association between hypertension and WMH volume has previously been shown in several large population-based studies focusing on older adults.16,17 Accordingly, hyper- tension is seen as one of the strongest cardiovascular risk factors associated with WMH. In our study—with a smaller sample size than other population-based studies—we did not find such an association. No previous studies have focused on the association between hypertension and WMH shape. The findings of our study might indicate that WMH shape is a more sensitive marker of hypertension-induced brain changes than WMH volume, since WMH shape was associated with hypertension while WMH volume was not. Based on the results of our study, we postulate that hypertension leads to distinct pathological changes within the small vessels of the brain, which manifest as a distinct WMH MRI phenotype. Hypertension has a direct destructive effect on small arteries, arterioles, venules and capillaries in the brain. Progressive pathological changes to the small vessels of the brain, induced by atheromas and micro-embolisms2,18, may manifest as distinct WMH phenotypes on MRI due to the anatomical macrostructure of the small vessels located around the ventricles in the white matter. This highlights the strong vascular component in WMH development already suggested in previous research.3,4 In our study each WMH shape marker that was analyzed represents a different spectrum of shape variations. For example, presence of hypertension is related to the variation in shape of periventricular/confluent WMH represented by a low convexity and high concavity index, but not to variations in shape represented by solidity or fractal dimension. Important to acknowledge in this regard is that convexity and the concavity index are mathematically related to each other. It is difficult to directly translate our described associations on a group-level to an individual patient in a clinical setting. For future translation to clinical practice, artificial intelligencebased models including a combination of several MRI biomarkers are required for more accurate applications in the field of diagnosis or prognosis. At present, the association between diabetes and WMH volume is not entirely understood. Previous studies in community-dwelling older individuals have shown associations of type 2 diabetes mellitus and total WMH volume19,20, while other cross-sectional studies in
2 29 Cardiovascular risk factors are related to distinct white matter hyperintensity MRI phenotypes similar study populations failed to show such an association.21,22 A recent systematic review addressing a possible association between type 2 diabetes and total WMH volume shows an association.23 The review did not only focus on cross-sectional or cohort studies, but also included case-control, and Mendelian randomization studies. Furthermore, a previous case-control study focused on patients with type 2 diabetes showed a higher eccentricity of deep WMH and a larger number of periventricular/ confluent WMH in diabetic patients compared to controls, but no differences in WMH volume.6 In our study in non-demented older adults, we did not find an association of diabetes with WMH shape markers, but found an association of diabetes with deep WMH volume. No previous studies have explored WMH shape in non-demented older adults, therefore our results cannot be directly compared to other studies. Based on these findings, WMH shape may be more sensitive than WMH volume for hypertensioninduced brain changes, but not for diabetes-induced brain changes. The association between hyperlipidemia and WMH remains ambiguous. In a previous study, an association between high cholesterol levels and a larger WMH burden was found in the general population above 65 years of age.24 Participants in this previous study are quite comparable to our study regarding age and cardiovascular risk factor profile. A previous case-control study focusing on stroke patients has shown a potential protective role of hyperlipidemia, as hyperlipidemia was associated with lower WMH volumes in two independent cohorts.25 However, it is unclear if this protective effect is (partially) mediated through the pharmacological treatment of hyperlipidemia with statins.25 This previous study has assessed a specific patient population (stroke patients) with a lower mean age in one of the cohorts compared to our study. No previous study has examined the association between hyper- lipidemia and WMH shape. In our study we did not find an association between hyperlipidemia and WMH volume, nor shape markers. Previous studies conducted using UK-biobank data, with a cross-sectional26, as well as an observational cohort study design27, showed an association of BMI with WMH volume. This effect may be mediated via low-grade systemic inflammation.28 However, no previous study has examined the association between BMI and WMH shape markers. In our study, we found no associations between BMI and WMH volume or shape markers. It should be noted that mean BMI in the UK-biobank study was similar to our study, while mean age was lower compared to our study.26 Little is known about the exact histopathological mechanisms underlying the formation of WMH, mainly due to the small number of pathology studies that have been conducted.29 Interestingly, in a histopathological study, structural markers of vascular dysfunction were found to be associated with WMH.30 More specifically, vascular integrity was shown to be reduced in areas where WMH were present compared to normal appearing white matter.30 Periventricular, and deep WMH were associated with different underlying neuropathological findings in a study
30 Chapter 2 including Alzheimer’s patients, subjects at risk for cerebrovascular disease and healthy controls.31 The study found periventricular WMH volume to be correlated with severity of arteriolosclerosis and breakdown of the ventricular lining.31 Deep WMH volume, however, was correlated with cerebral hemorrhages and microinfarcts, as well as demyelination.31 Depending on the affected cell type (e.g. vascular or parenchymal), the type of pathophysiology and the anatomical structure of the affected areas, WMH shape may be influenced. These differences in WMH etiology may become evident as different WMH shape patterns and locations (for example punctuate versus elongated deep WMH). Some previous studies have also shown a link between WMH shape and underlying histological findings.8,9 Overall, pathological studies confirm the hypothesis that WMH have a heterogenous etiology18—with a strong vascular component. Vascular pathologies may lead to parenchymal changes, subsequently leading to distinct WMH phenotypes on MRI. How different WMH lesion and shape patterns develop related to a certain underlying pathology remains to be investigated in future studies, as in our study it is impossible to link WMH shape patterns to specific underlying mechanisms. Novel advanced MRI markers (such as WMH shape) that are more specific than WMH volume, may help in elucidating the link between histopathology and MRI findings. The strengths of our study are the relatively large sample size, and the application of novel advanced MRI image processing methods. A limitation of our study could be that participants were all scheduled for major elective surgery, and therefore this sample might not be an equivalent of a general population-based group. This could limit the generalizability of our findings to the general population, as this could have led to an underestimation of the associations. Another limitation of our study could be that the cardiovascular risk factors were based on medical records and self-reported rather than objective measurements (e.g. blood pressure, glucose levels and cholesterol/triglyceride levels). Although these risk factors were scored by qualified physicians (anesthesiologists (in training)), we cannot exclude the possibility that some patients may have undiagnosed hypertension, diabetes or hypercholesterolemia. This may partially account for the lack of significant associations between these cardiovascular risk factors and the WMH markers. An additional limitation could be that although most participants had a relatively high MMSE score (median: 29, (IQR: 27,30)), there may have been participants included in this study with mild cognitive impairment. A technical limitation of our method could be that when total WMH volume increases, usually the deep WMH count decreases, as bigger WMH lesions start to overlap with each other. For instance, deep WMH may become part of confluent WMH and only shape markers for periventricular/confluent WMH can be determined. This can partially explain why we found fewer associations with deep WMH shape markers.
2 31 Cardiovascular risk factors are related to distinct white matter hyperintensity MRI phenotypes In conclusion, we showed that different cardiovascular risk factors seem to be related to a distinct pattern of WMH shape markers in non-demented older adults. These findings may suggest that different underlying cardiovascular pathological mechanisms lead to different WMH MRI phenotypes, which may be valuable for early detection of individuals at risk for stroke and dementia. 2.6 ACKNOWLEDGEMENTS The research of Jeroen de Bresser was funded by an Alzheimer Nederland grant (WE.03-2019-08).
32 Chapter 2 2.7 REFERENCES 1. Jellinger KA, Attems J. Challenges of multimorbidity of the aging brain: a critical update. J Neural Transm 2015; 122: 505–521. 2. Wardlaw JM, Smith C, Dichgans M. Mechanisms of sporadic cerebral small vessel disease: Insights from neuroimaging. Lancet Neurol 2013; 12: 483–497. 3. Wardlaw JM, Smith EE, Biessels GJ, Cordonnier C, Fazekas F, Frayne R et al. Neuroimaging standards for research into small vessel disease and its contribution to ageing and neurodegeneration. Lancet Neurol. 2013; 12: 822–838. 4. Alber J, Alladi S, Bae H, Barton DA, Beckett LA, Bell JM et al. White matter hyperintensities in vascular contributions to cognitive impairment and dementia (VCID): Knowledge gaps and opportunities. Alzheimer’s & Dementia: Translational Research & Clinical Interventions 2019; 5: 107–117. 5. Gao Z, Zhai Y, Zhao X, Wang W, Wu W, Wang Z et al. Deep cerebral microbleeds are associated with the severity of lacunar infarcts and hypertension. Medicine 2018; 97: e11031. 6. De Bresser J, Kuijf HJ, Zaanen K, Viergever MA, Hendrikse J, Biessels GJ et al. White matter hyperintensity shape and location feature analysis on brain MRI; Proof of principle study in patients with diabetes. Sci Rep 2018; 8: 1–10. 7. Ghaznawi R, Geerlings MI, Jaarsma-Coes MG, Zwartbol MHT, Kuijf HJ, van der Graaf Y et al. The association between lacunes and white matter hyperintensity features on MRI: The SMART-MR study. Journal of Cerebral Blood Flow and Metabolism 2019; 39: 2486–2496. 8. Fazekas F, Kleinert R, Offenbacher H, Schmidt R, Kleinert G, Payer F et al. Pathologic correlates of incidental MRI white matter signal hyperintensities. Neurology 1993; 43: 1683–1689. 9. Kim KW, MacFall JR, Payne ME. Classification of White Matter Lesions on Magnetic Resonance Imaging in Elderly Persons. Biol Psychiatry 2008; 64: 273–280. 10. Ghaznawi R, Geerlings MI, Jaarsma-Coes M, Hendrikse J, Bresser J de, Group on behalf of the U-SS. Association of White Matter Hyperintensity Markers on MRI and Long-term Risk of Mortality and Ischemic Stroke. Neurology 2021; 96: e2172–e2183. 11. Winterer G, Androsova G, Bender O, Boraschi D, Borchers F, Dschietzig TB et al. Personalized risk prediction of postoperative cognitive impairment – rationale for the EU-funded BioCog project. European Psychiatry 2018; 50: 34–39. 12. Zigmond AS, Snaith RP. The Hospital Anxiety and Depression Scale. Acta Psychiatr Scand 1983; 67: 361–370. 13. Schmidt, P. Bayesian inference for structured additive regression models for large-scale problems with applications to medical imaging. (Maximilians-Universität München, 2017). 14. Kempton MJ, Underwood TSA, Brunton S, Stylios F, Schmechtig A, Ettinger U et al. A comprehensive testing protocol for MRI neuroanatomical segmentation techniques: Evaluation of a novel lateral ventricle segmentation method. Neuroimage 2011; 58: 1051– 1059. 15. Liu EJ, Cashman K V., Rust AC. Optimising shape analysis to quantify volcanic ash morphology. GeoResJ 2015; 8: 14–30. 16. Marcus J, Gardener H, Rundek T, Elkind MSV, Sacco RL, Decarli C et al. Baseline and longitudinal increases in diastolic blood pressure are associated with greater white matter hyperintensity volume: The northern manhattan study. Stroke 2011; 42: 2639–2641. 17. Muller M, Sigurdsson S, Kjartansson O, Aspelund T, Lopez OL, Jonnson P V. et al. Joint effect of mid- and late-life blood pressure on the brain: The AGES-Reykjavik Study. Neurology 2014; 82: 2187–2195.
www.ridderprint.nlRkJQdWJsaXNoZXIy MTk4NDMw