Journal of Alzheimer’s Disease 79 (2021) 163–175 163 DOI 10.3233/JAD-200496 IOS Press White Matter Hyperintensities Are No Major Confounder for Alzheimer’s Disease Cerebrospinal Fluid Biomarkers Linda J.C. van Waalwijk van Doorna,b , Mohsen Ghafoorianc , Esther M.C. van Leijsena , Jurgen A.H.R. Claassend , Andrea Arighie , Marco Bozzalif,g , Jorge Cannash , Enrica Cavedoi,j , Paolo Eusebik , Lucia Farottik , Chiara Fenogliol , Juan Forteam,n , Giovanni B. Frisonii,o , Daniela Galimbertie,l , Viviana Grecop,q , Sanna-Kaisa Herukkar , Yawu Liur , Alberto Lleóm,n , Alexandre de Mendonçah , Flavio M. Nobilis,t , Lucilla Parnettik , Agnese Piccos , Maria Pikkarainenr , Nicola Salvadorik , Elio Scarpinie,l , Hilkka Soininenr , Roberto Tarduccik , Andrea Urbanip,q , Eduard Vilaplanam,n , Olga Meulenbroekd , Bram Platelc , Marcel M. Verbeeka,b,∗,1 and H. Bea Kuiperija,b,1 a Department of Neurology, Donders Institute for Brain, Cognition and Behaviour, Radboud Alzheimer Centre, Radboud University Medical Center, Nijmegen, the Netherlands b Department of Laboratory Medicine, Radboud University Medical Center, Nijmegen, the Netherlands c Department of Radiology and Nuclear Medicine, Radboud University Medical Center, Nijmegen, the Netherlands d Department of Geriatrics, Donders Institute for Brain, Cognition and Behaviour, Radboud Alzheimer Centre, Radboud University Medical Center, Nijmegen, the Netherlands e Fondazione IRCCS Ca’ Granda, Ospedale Policlinico, Milan, Italy f IRCCS Fondazione Santa Lucia, Rome, Italy g Department of Neuroscience, Brighton and Sussex Medical School, University of Sussex, Brighton, United Kingdom h Faculty of Medicine, University of Lisbon, Lisbon, Portugal i Laboratory of Epidemiology, Neuroimaging and Telemedicine, IRCCS San Giovanni di Dio Fatebenefratelli, Brescia, Italy j Sorbonne University, GRC n◦ 21, Alzheimer Precision Medicine (APM), AP-HP, Pitié-Salpêtrière Hospital, Boulevard de l’hôpital, Paris, France; Qynapse, Paris, France k Section of Neurology, Center for Memory Disturbances, University of Perugia, Perugia, Italy l University of Milan, Dino Ferrari Center, Milan, Italy m Sant Pau Memory Unit, Department of Neurology, Hospital de la Santa Creu i Sant Pau, Biomedical Research Institute Sant Pau, Universitat Autònoma de Barcelona, Barcelona, Spain n Center of Biomedical Investigation Network for Neurodegenerative Diseases (CIBERNED), Madrid, Spain o University Hospitals and University of Geneva, Geneva, Switzerland p Fondazione Policlinica Universitario “A. Gemelli” – IRCCS, Rome, Italy q Department of Basic Biotechnological Sciences, Intensivological and Perioperative Clinics, Università Cattolica, Rome, Italy r Department of Neurology, University of Eastern Finland and Kuopio University Hospital, Kuopio, Finland s Department of Neuroscience (DINOGMI), University of Genoa, Genoa, Italy t IRCCS Ospedale Policlinico San Martino, Genoa, Italy 1 These authors contributed equally to this work. P.O. Box 9101, 6500 HB Nijmegen, the Netherlands. Tel.: +31 ∗ Correspondenceto: Dr. Marcel M. Verbeek, TML, r.830, 24 36 14567; Fax: +31 2436 68754; E-mail: Marcel.verbeek@ Department of Neurology, Radboud University Medical Center, radboudumc.nl. ISSN 1387-2877/21/$35.00 © 2021 – The authors. Published by IOS Press. This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial License (CC BY-NC 4.0). 164 L.J.C. van Waalwijk van Doorn et al. / Effect of WMHs on AD CSF Biomarkers Handling Associate Editor: Ralph Martins Accepted 15 October 2020 Abstract. Background: The cerebrospinal fluid (CSF) biomarkers amyloid-␤ 1–42 (A␤42 ), total and phosphorylated tau (t-tau, p-tau) are increasingly used to assist in the clinical diagnosis of Alzheimer’s disease (AD). However, CSF biomarker levels can be affected by confounding factors. Objective: To investigate the association of white matter hyperintensities (WMHs) present in the brain with AD CSF biomarker levels. Methods: We included CSF biomarker and magnetic resonance imaging (MRI) data of 172 subjects (52 controls, 72 mild cognitive impairment (MCI), and 48 AD patients) from 9 European Memory Clinics. A computer aided detection system for standardized automated segmentation of WMHs was used on MRI scans to determine WMH volumes. Association of WMH volume with AD CSF biomarkers was determined using linear regression analysis. Results: A small, negative association of CSF A␤42 , but not p-tau and t-tau, levels with WMH volume was observed in the AD (r2 = 0.084, p = 0.046), but not the MCI and control groups, which was slightly increased when including the distance of WMHs to the ventricles in the analysis (r2 = 0.105, p = 0.025). Three global patterns of WMH distribution, either with 1) a low, 2) a peak close to the ventricles, or 3) a high, broadly-distributed WMH volume could be observed in brains of subjects in each diagnostic group. Conclusion: Despite an association of WMH volume with CSF A␤42 levels in AD patients, the occurrence of WMHs is not accompanied by excess release of cellular proteins in the CSF, suggesting that WMHs are no major confounder for AD CSF biomarker assessment. Keywords: Alzheimer’s disease, amyloid, biomarkers, cerebrospinal fluid, magnetic resonance imaging, tau proteins, white matter hyperintensities, white matter lesions INTRODUCTION fluid-attenuated inversion recovery (FLAIR) T2- weighted magnetic resonance imaging (MRI) scans Alzheimer’s disease (AD) is the most prevalent and are considered biomarkers of SVD [8–13]. form of dementia and is a progressive and irreversible WMHs are caused by destruction of the myelin sheath neurodegenerative disease. Typical neuropatholog- that envelops the axons of neurons. Myelin is the main ical lesions observed postmortem in the brains of component of white matter and facilitates the fast AD patients are the extracellular accumulation of signal transmission along neuronal processes [14]. amyloid-␤ (A␤) protein in both plaques and the cere- Lacunes of presumed vascular origin are cavities bral vasculature, and the intracellular neurofibrillary filled with cerebrospinal fluid (CSF), visible as round tangles containing hyperphosphorylated tau (p-tau) or ovoid subcortical hypointense areas of 3–15 mm protein [1]. on FLAIR T2-weighted MRI scans. Microbleeds are Several lines of evidence suggest that vascular defined as small hypointense areas of 2–10 mm on pathology contributes to the development and clinical T2∗ gradient-recall echo (GRE) or susceptibility- symptoms of AD. Subcortical vascular lesions visu- weighted imaging (SWI) MRI sequences [12]. alized by neuroimaging are collectively referred to as CSF biomarkers (A␤42 , total tau (t-tau), and p- cerebral small vessel disease (SVD). Neuroimaging tau) have been shown to improve the accuracy of AD biomarkers of SVD comprise white matter hyperin- diagnosis to >85% and have therefore recently been tensities (WMHs), lacunar infarcts, and microbleeds included in the diagnostic criteria for AD [15–17]. In [2]. These abnormalities can be observed in brains the CSF of AD patients, the combination of decreased of healthy elderly individuals (typically above the levels of A␤42 and elevated concentrations of both age of 60 years), but are observed in increased num- t-tau and p-tau proteins can be observed, reflecting bers in AD patients, likely contributing to cognitive senile plaques and neurofibrillary tangle pathology, decline [3–7]. WMHs are signal abnormalities in respectively [18] and this combination may predict the white matter that appear as hyperintensities on the progression to AD dementia in patients with mild L.J.C. van Waalwijk van Doorn et al. / Effect of WMHs on AD CSF Biomarkers 165 cognitive impairment (MCI) [19, 20]. In patients with Table 2). We decided to pool the healthy control vascular dementia or stroke (transiently) increased subjects and SMC in one group as neurological con- CSF t-tau and decreased CSF A␤42 levels can be trols. Individuals were labeled as SMC when they observed [21–24]. Therefore, it is possible that SVD presented with cognitive complaints, and results of affects the levels of CSF A␤ and tau proteins and thus clinical investigations were within normal range. affect the diagnostic accuracy of these biomarkers Individuals were labeled as controls if they were for AD. free of neurological disease. SMC are very similar This study aims to identify the relation between to healthy controls in terms of prevalence of amyloid AD CSF biomarkers (A␤42 , t-tau, and p-tau) and the pathology [26]. CSF A␤42 (p = 0.42), t-tau (p = 0.28) presence of WMHs. We hypothesize that the occur- and p-tau levels (p = 0.28) as well as WMH volumes rence of WMHs, especially when localized close to (p = 0.77) were similar between SMC and healthy the ventricles, is accompanied by excess release of control subjects (see Supplementary Table 3). cellular components (e.g., tau proteins) in the extra- We collected data of the subjects who ful- cellular space and as such can modify the patterns of filled the following quality requirements: the time biomarkers that are known to be associated with AD between the lumbar puncture and the T1-weighted pathology. and T2-weighted-FLAIR MRI scans was less than 6 months; both T1-weighted and T2-weighted- MATERIALS AND METHODS FLAIR MRI scans had a maximum voxel size of 2.0 mm × 2.0 mm × 2.0 mm. The T1-weighted MRI Subjects scans were used to examine the brain structure, while the T2-weighted-FLAIR MRI scans were used for T1-weighted structural MRI scans, T2-weighted- the visualization of WMHs. Additionally, informa- FLAIR MRI scans, and AD CSF biomarker data tion about age, sex, center, and scanner type were were acquired from 9 research centers within the recorded. Ventricular volume and total intracranial Biomarkers for Alzheimer’s and Parkinson’s Dis- volume (TIV) of each subject were previously deter- ease (BiomarkAPD) project, a consortium of the mined ([25]; Table 1). In Supplementary Tables 4–6, European initiative Joint Programme for Neurode- information on the MRI scanner type, acquisition generative Diseases Research (JPND). The study was parameters, and whether a center used a specific MRI approved by local ethics committees or the institu- protocol or not, can be found. tional review board of each center. The participants or their legal representatives gave written informed con- Computer aided detection of WMHs sent. Initially, 309 subjects were included for analysis (a sub-set of a previous study on ventricular volumes We pre-processed each MRI scan by a visual check. [25], but for several reasons discussed in the results The MRI scans of unsatisfactory quality in terms section, 137 subjects were excluded from analysis of low transverse resolution, or low structural vis- and therefore n = 172 subjects were finally included ibility due to movement or other sources of noise in this study. were excluded. Thereafter, WMHs were detected by All subjects underwent clinical and neurological a machine learning-based model (computer aided assessment, lumbar puncture, MRI scanning, and detection algorithm) [27] on the T1 and FLAIR CSF analysis at their local laboratory. The patient modalities, resulting in the number of WMH voxels groups comprised: 52 neurological controls (includ- per axial-slice as outcome. ing n = 34 healthy controls, n = 10 subjects for Next, the detection of WMHs by the algorithm whom cognitive disorders were excluded, and n = 8 was visually checked by a trained reader and was subjective memory complainers (SMC), based on adjusted if necessary (post-processing). An example examination of medical records), 72 MCI patients, of a structural T1-weighted MRI scan ran by the com- of which 42 patients were diagnosed with amnestic puter aided detection algorithm can be found in Fig. 1. MCI, and 48 AD patients. Diagnoses were defined Using the number of WMH voxels per axial-slice by clinical and neurological assessment and applica- and information about the voxel size (in mm2 ) and tion of diagnostic criteria (see Supplementary Table 1 distance of axial-slices (in mm), the WMH volume for the MCI and AD diagnostic criteria used in each (expressed in mL) per axial-slice was calculated. The center; the locally applied cut-off values for the use total WMH volume was subsequently calculated by of the CSF biomarkers are shown in Supplementary the sum of the WMH volume of each axial-slice. 166 L.J.C. van Waalwijk van Doorn et al. / Effect of WMHs on AD CSF Biomarkers Table 1 Demographic data and CSF biomarker concentrations across the diagnostic groups Control AD MCI p Sample size: n 52 48 72 Sex: female, n (%) 34 (65) 29 (60) 44 (61) 0.850a Age: y, mean (SD) 61.1 (8.9) 69.0 (8.2) 69.8 (6.8) <0.0001b A␤42 : mean, pg/mL (SD) [A␤42 -positive, %]d 757 (193) [12%] 482 (194) [79%] 706 (334) [36%] <0.0001b t-tau: mean, pg/mL (SD)c [t-tau-positive, %]d 225 (104) [13%] 713 (527) [77%] 492 (294) [57%] <0.0001b p-tau: mean, pg/mL (SD) [p-tau-positive, %]d 42 (15) [4%] 95 (51) [60%] 81 (61) [47%] <0.0001b # of WMH: mean, n (SD) 132 (75) 99 (118) 89 (63) 0.0007b WMH volume: mean, mL (SD) 5.7 (3.1) 9.0 (7.9) 8.7 (7.8) 0.183b WM: mean, mL (SD) 515 (80) 540 (79) 550 (92) 0.056b VV: mean, mL (SD) 23 (15) 42 (23) 40 (22) <0.0001b TIV: mean, mL (SD) 1346 (139) 1363 (118) 1374 (121) 0.364b CSF, cerebrospinal fluid; AD, Alzheimer’s disease; MCI, mild cognitive impairment; SD, standard deviation; A␤42 , amyloid-␤; t-tau, total tau; p-tau, phosphorylated tau; # of WMH, number of white matter hyperintensities; WM, white matter; VV, ventricular volume; TIV, total intracranial volume. a Chi-square test. b Kruskal-Wallis with Dunn’s post hoc test. c Not available for n = 1 AD and n = 3 MCI patients d Based on local cut-off values. Fig. 2. WMH layers in a schematic overview. Brain tissue was divided into 20 layers, from ventricles to the skull, with each Fig. 1. WMH detection on a structural MRI scan. Segmentation layer accounting for 5% of the total distance between ventricles to of the WMHs detected by the computer aided detection system in the skull. To exemplify this division the first three circular lay- the transverse plane projected on a T1-weighted MRI scan of an ers around the ventricles are shown in this schematic picture. AD patient. Periventricular WMHs are indicated in red/light gray. Brain section image was modified from Smart Servier Medical Of note, the (green) circumference, right under ventricle, indicates Art, https://smart.servier.com. a false positive WMH and was removed during post-processing. WMH distribution patterns of WMHs were based on AD patient data and defined as follows: low WMH volume in a layer was defined The region from ventricles to the skull was divided as <20% (<1.32 mL) of the maximum WMH volume in 20 layers, with each layer accounting for 5% of the (i.e., 6.6 mL) measured in any of the layers of the AD total distance between ventricles to the skull. Data on patients. WMH pattern 1 (with low WMH volumes) WMH volumes was available for each of these cir- was defined as low WMH volume (<1.32 mL/layer) cular layers; a schematic overview can be found in in each of the 20 layers of a subjects’ brain. WMH pat- Fig. 2. In order to analyze the distribution of WMHs tern 2 was characterized by a peak in WMH volume across the brain in a standardized manner, we used the close to the ventricles and defined as high WMH vol- following self-defined criteria. Distribution patterns ume (>1.32 mL/layer) in one or more layer(s) of the L.J.C. van Waalwijk van Doorn et al. / Effect of WMHs on AD CSF Biomarkers 167 three layers closest to the ventricles, and low WMH RESULTS volume in other layers. WMH pattern 3 comprised the remaining patterns, generally characterized by a Exclusion of subjects during processing high WMH volume with broad distribution over the Unfortunately, during pre-processing 34 subjects brain, and defined as at least one layer distant from had to be excluded because of movement arte- the ventricles (layers 4–20) with a high WMH volume facts, 19 subjects because of “salt and pepper” noise (i.e., >1.32 mL). effect, with many small white dots dispersed over AD CSF biomarkers the images, and 31 subjects because of insufficient resolution of the MRI scans. Furthermore, after the The ELISAs from Fujirebio (Gent, Belgium) were processing phase, 25 subjects had to be excluded used according to the manufacturers’ protocol for the because no data could be obtained due to an algo- determination of A␤42 (INNOTEST® ␤-AMYLOID rithm error when the MRI scans of these subjects (1–42)), t-tau (INNOTEST® hTAU Ag), and p-tau were run by the computer aided detection system. (INNOTEST® PHOSPHO-TAU (181P)) at each site Moreover, during post-processing, 13 subjects had to separately. Locally applied cut-off values for the use be excluded due to WMH segmentation failures and of AD CSF biomarkers are shown in Supplementary another 13 subjects were excluded due to algorithm Table 2. In all participating laboratories, the CSF sam- failures. A segmentation failure is a failure in the algo- ples were collected in polypropylene tubes [16]. The rithm used for the segmentation of ventricles, while CSF samples were transported directly to the labo- an algorithm failure could be a failure in any of the ratory, centrifuged, and measured or stored at –80◦ C automated steps (e.g., multi-modal registration, brain until use. extraction, bias field correction, intensity standard- ization), which potentially have significant impact on Data and statistical analysis the quality of the WMH detection results. Finally, two more patients were excluded for other reasons; one Statistical analyses were performed using IBM patient due to extreme high WMH volume (116 mL) SPSS Statistics 22 (Armonk, NY, USA) and Graph- and another patient with a non-realistic high p-tau pad Prism 5.03 (La Jolla, CA, USA). Normality value (690 pg/mL). Comparison of available demo- of data distribution was determined using the graphic and CSF biomarker data showed that, apart D’Agostino and Pearson omnibus normality test. For from minor differences in CSF p-tau and t-tau lev- comparison of two groups, a non-parametric Mann- els in control subjects, these data were comparable Whitney test was used. For comparison of three between the groups of excluded versus included groups, the non-parametric Kruskal-Wallis test with control, AD, and MCI subjects. Thus 172 subjects Dunn’s test for multiple comparisons was used. The remained in the study and their demographics are Chi-square test was used to check for sex differences shown in Table 1. In Supplementary Table 7, the between the diagnostic groups. Bivariate correlations detailed overview of included and excluded patients to determine the association of WMH volume with per center can be viewed. the CSF biomarkers were performed using linear regression analysis based on normal distribution. To Clinical validation of AD CSF biomarkers obtain normal distributed data for the linear regres- sion analysis, CSF p-tau, CSF t-tau, WMH/TIV, and The A␤42 CSF concentrations were significantly the arbitrary WMH∗ distance values were log trans- decreased (p < 0.001) in AD patients (mean: 482 ± formed. In the linear regression analysis, age and sex 194 pg/mL) compared to control subjects (mean: were analyzed as covariates, and research center, in 757 ± 193 pg/mL) and MCI patients (mean: combination with or without white matter volume, 706 ± 334 pg/mL) (Fig. 3A). The p-tau (Fig. 3B) were subsequently included as additional covariates. and t-tau (Fig. 3C) concentrations were significantly The WMH∗ distance value is an arbitrary factor that increased (p < 0.001) in AD patients (mean p-tau: combines WMH volume with the relative distance of 95 ± 51 pg/mL, mean t-tau: 713 ± 527 pg/mL) and WMHs to the ventricles, and was calculated using the MCI patients (mean p-tau: 81 ± 61 pg/mL, mean following formula: WMH∗ distance = layer [WMH t-tau: 492 ± 294 pg/mL) compared to control sub- volume (mL) ∗ (100 – relative distance to ventricles jects (mean p-tau: 42 ± 15 pg/mL, mean t-tau: 225 ± (%))]. All tests were two-sided and statistical sig- 104 pg/mL). Table 1 provides an overview of the nificance was defined as p < 0.05. CSF data of all patient groups. 168 L.J.C. van Waalwijk van Doorn et al. / Effect of WMHs on AD CSF Biomarkers Fig. 3. Analysis of AD CSF biomarkers: A␤42 (n = 172), p-tau (n = 172), and t-tau (n = 168) in healthy controls, MCI, and AD patients. Analysis of (A) A␤42 in healthy controls (n = 52), mild cognitive impairment (MCI) (n = 72), and Alzheimer’s disease (AD) patients (n = 48), (B) p-tau in healthy controls (n = 52), MCI (n = 72), and AD patients (n = 48) and (C) t-tau in healthy controls (n = 52), MCI (n = 69), and AD patients (n = 47). Solid bar = median; p-value: ∗∗∗ p < 0.001 (Kruskal-Wallis with Dunn’s post hoc test). Fig. 4. WMH volume per diagnostic group. A) WMH volume per diagnostic group: healthy controls (n = 52), mild cognitive impairment (MCI) (n = 72), and Alzheimer’s disease (AD) patients (n = 48). No significant differences were found between the diagnostic groups (p = 0.18). Solid bar = median. B) WMH volume distribution as a function of the relative distance to the ventricles, per diagnostic group. Healthy controls had a lower WMH volume, but showed a similar distribution pattern as MCI and AD patients. WMH volume and distribution in the diagnostic for the size of the head (TIV). A small, negative groups correlation between CSF A␤42 levels and WMH/ TIV was observed, i.e., higher CSF A␤42 levels No significant differences (p = 0.18) were observed were associated with slightly, but significantly, lower in the mean WMH volume between the different diag- WMH/TIV values in the AD group (r2 = 0.084, nostic groups (controls, MCI, and AD), see Fig. 4A. p = 0.046; Fig. 5A), but not in the MCI (r2 = 0.00, We additionally analyzed the distribution of WMH p = 0.98) or control group (r2 = 0.066, p = 0.065; cor- volume measured in the various layers of the MRI relation data in Table 2). These results for the control scans (Fig. 4B). Overall, a similar distribution of group remained similar after exclusion of SMC WMH volume across the brain was observed for the (n = 8) from this group (r2 = 0.037, p = 0.21). Sub- three diagnostic groups, with a peak WMH volume group analysis showed that in MCI patients positive relatively close to the ventricles (relative distance; at for CSF A␤42 (i.e., with decreased levels based on 5–10% of the distance between ventricles and brain the local cut-off values; Table 1 and Supplemen- skull). tary Table 2), CSF A␤42 levels tended to associate with WMH/TIV values (n = 26, r2 = 0.14, r = –0.38, Association of AD CSF biomarkers with WMH p = 0.056), which was not the case in MCI patients negative for CSF A␤42 (n = 46, r2 = 0.00, r = –0.01, We used linear regression analysis to study the p = 0.97). In contrast, no association of p-tau or t-tau association of AD CSF biomarkers with WMH vol- with WMH/TIV was observed in the AD (Fig. 5B, ume. In these analyses, WMH volume was corrected C), MCI or the control groups (correlation data in L.J.C. van Waalwijk van Doorn et al. / Effect of WMHs on AD CSF Biomarkers 169 Fig. 5. Correlation of WMH volume with CSF A␤42, t-tau and p-tau. White matter hyperintensity volumes corrected for total intracranial volume (WMH/TIV) versus CSF A␤42 (A, D, G), p-tau (B, E, H), and t-tau (C, F, I) concentrations in the Alzheimer’s disease (AD) group (A-C), the mild cognitive impairment (MCI) group (D-F), and the control group (G-I). A significant (p < 0.05), negative correlation was found between (log) WML/TIV and CSF A␤42 concentrations in the AD group (A), but not for (log) p-tau or (log) t-tau in any group or for A␤42 in the MCI and control groups. p and r values derived from linear regression analyses are plotted. Linear regression lines are shown, except for panels A and G where quadratic relations are shown. Table 2 Correlation between WMH volume and AD CSF biomarkers A␤42 p-tau t-tau p r r2 p r r2 p r r2 Control 0.065 –0.258 0.066 0.443 0.109 0.012 0.490 0.098 0.010 MCI 0.975 0.004 0.000 0.250 0.137 0.019 0.530 0.077 0.006 AD 0.046 –0.290 0.084 0.929 –0.013 0.000 0.982 0.003 0.000 Table 2). We sequentially added several covariates to as covariates (overall model: p = 0.045, r2 = 0.230; the analyses, first age and sex, second center and third significance of WMH/TIV in the model: p = 0.031). total white matter volume. The association of A␤42 We repeated the linear regression analysis using levels with WMH/TIV in the AD group remained an arbitrary factor that combines WMH volume with significant when age and sex in combination with the the relative distance of WMHs to the ventricles center and total white matter volume were included instead of WMH/TIV. This arbitrary WMH∗ distance 170 L.J.C. van Waalwijk van Doorn et al. / Effect of WMHs on AD CSF Biomarkers value assigns a higher value to WMH volumes comparisons for A␤42 , t-tau, and p-tau levels in con- closer to the ventricles. Like for the association trols, MCI or AD patients (data not shown). Ventric- with WMH volume, we only observed a significant, ular volumes were, however, significantly increased negative correlation for CSF A␤42 with this arbi- in MCI and AD patients with WMH pattern 2 or 3 trary WMH∗ distance value in the AD group (r2 = versus WMH pattern 1 (MCI: 50.5 ± 24.1 mL versus 0.105, p = 0.025), which remained after including 30.0 ± 13.6 mL, p < 0.001; AD: 49.6 ± 25.6 mL ver- age, sex, center and white matter volume as covariates sus 32.9 ± 13.8 mL, p = 0.009, Fig. 6H), and weak, (overall model: p = 0.025, r2 = 0.252; significance of but significant, correlations between ventricular and WMH∗ distance in the model: p = 0.016). WMH volumes were observed for all three patient groups separately and combined (all: Spearman Association of WMH distribution patterns with r = 0.405, p < 0.001). CSF biomarker levels DISCUSSION To study the relation between WMH distance to the ventricles and CSF biomarker levels in more CSF biomarkers have obtained an increasingly detail, we looked at the WMH distribution patterns prominent position in support of establishing the within a diagnostic group. In the AD group, we early clinical diagnosis of AD, and can also support in could observe three global patterns in the distribu- predicting progression of AD [20]. Establishing and tion of WMH volume within the brain. The most controlling for confounding factors, may further opti- abundant pattern (pattern 1, Fig. 6A) had an over- mize the performance of CSF biomarkers. In a pre- all low WMH volume across the brain. The second vious study we demonstrated that ventricular volume most abundant pattern (pattern 2, Fig. 6B) showed a can be a confounding factor. We showed a slight, but peak in WMH volume close to the ventricles and a significant, improved diagnostic value of A␤42 after low WMH volume comparable to pattern 1 distant correction for ventricular volume [25]. In the current from the ventricles, whereas the last pattern (pattern study, we evaluated WMHs as possible confounding 3, Fig. 6C) showed a high WMH volume distributed factor for AD CSF biomarkers (A␤42 , t-tau, p-tau). over a broader range of the brain. The WMH vol- ume was significantly higher for AD patients with Association of AD CSF biomarkers with WMH pattern 3 as compared to pattern 1 and 2, as well volume as for AD patients with pattern 2 compared to pat- tern 1 (Fig. 6D). The three WMH patterns could also We hypothesized that the occurrence of WMHs is be observed in MCI patients with comparable WMH accompanied by the excess release of cellular com- peak volumes per pattern (Table 3). In control sub- ponents (e.g., tau proteins) in the extracellular space jects, patterns 1 and 2 were also frequently observed, and as a result can modify the patterns of biomarkers in contrast to pattern 3 which was observed only in that are known to be associated with AD pathol- two control subjects (Table 3). ogy. Unlike this hypothesis, we showed that increased We compared patients with WMH pattern 2, i.e., WMH volume, albeit weakly, was significantly asso- WMH close to the ventricles, to patients with WMH ciated with decreased CSF A␤42 concentrations in pattern 3, i.e., WMH both close and distant from the the AD group. This finding suggests that increased ventricles. We observed significantly decreased levels WMH volume is associated with worsened amyloid of A␤42 , but no differences in t-tau or p-tau levels, pathology. We did not observe an association between in AD patients with WMH pattern 3 (mean A␤42 : p-tau or t-tau concentrations and WMH volume in 404 ± 103 pg/mL) as compared to pattern 2 (mean any group, nor between CSF A␤42 levels and WMH A␤42 : 514 ± 146 pg/mL; p = 0.032, Fig. 6E). We nei- volume in the MCI and control groups. Our findings ther observed differences in A␤42 nor in t-tau or p-tau are not in line with a previous study in which it was levels between these patterns 2 and 3 in MCI patients shown that decreased CSF A␤42 and increased neu- (Fig. 6F, G and data not shown). The analysis could rofilament light chain levels are related to increased not be performed for control subjects, since only few WMH volume in a cohort of cognitively intact elderly subjects with WMH pattern 3 were available. individuals [28]. Additionally, two other studies also We also compared patients with WMH pattern reported a significant negative correlation (p < 0.01; 1 to patients with WMH pattern 2 or 3, but we p = 0.013) between CSF A␤42 and WMH volume in did not observe any significant differences in these cognitively normal individuals and MCI patients [29, L.J.C. van Waalwijk van Doorn et al. / Effect of WMHs on AD CSF Biomarkers 171 Fig. 6. WMH distribution patterns in brains of AD patients. WMH in brains of AD patients showed either (A) a distribution of a low WMH volume across the whole brain (pattern 1), or (B) an increased WMH volume (peak) close to the ventricles only (pattern 2), or (C) a high WMH volume distributed across a broader region of the brain (pattern 3). The WMH volume is shown per brain layer for a total of 20 evenly distributed layers per brain (indicated as relative distance from the ventricles in %). The distribution pattern for each patient is shown in grey, with the mean WMH volume per layer for each pattern shown in black. D) Dot plot of the total WMH volumes in the brain of AD patients per WMH pattern. Mean WMH volumes were statistically significant different between the three patterns (Kruskal Wallis test with Dunn’s test for multiple comparisons). A␤42 CSF levels (E), but not t-tau (F), and p-tau levels (G), were significantly decreased in AD patients with WMH pattern 3 compared to pattern 2. H) Ventricular volume was significantly increased in AD patients with WMH pattern 2 or 3 (high WMH volumes) compared to pattern 1 (low WMH volumes). Solid bar = median; p-value: ∗ < 0.05, ∗∗ < 0.01, ∗∗∗ < 0.001. 172 L.J.C. van Waalwijk van Doorn et al. / Effect of WMHs on AD CSF Biomarkers Table 3 WMH pattern characteristics and distribution per patient group All (n) WMH pattern 1 WMH pattern 2 WMH pattern 3 Control n (% of total) 52 30 (58%) 20 (38%) 2 (4%) mean WMH peak volume∗ 0.76 ± 0.25 1.38 ± 0.47 2.65 ± 1.36 MCI n (% of total) 72 36 (50%) 19 (26%) 17 (24%) mean WMH peak volume∗ 0.71 ± 0.36 1.62 ± 0.50 3.00 ± 1.25 AD n (% of total) 48 21 (44%) 16 (33%) 11 (23%) mean WMH peak volume∗ 0.66 ± 0.30 1.59 ± 0.54 2.82 ± 1.79 ∗ Highest (peak) mean WMH volume (in mL ± SD) in a layer per pattern, which was for each pattern and patient group the WMH volume in layer 2. See the Materials and Methods section for a definition of the three WMH patterns. 30]. In all three studies, no AD group was included t-tau levels in CSF if other measures of SVD beyond [28–30]. Recently, another study showed a significant WMHs are included. However, these data were not negative association of CSF A␤42 levels with WMH available for the current study. volume in (healthy) controls and SMCs, and a border- We aimed to study also the relation of the distance significant association in MCI and AD patients [31]. of WMHs to the ventricles with AD CSF biomarkers No associations between vascular risk factors and in addition to the WMH volume. WMHs may rep- CSF A␤42 levels in cognitively normal older indi- resent an independent indicator of cerebrovascular viduals and individuals with abnormal biomarkers impairment in AD patients and may interact with AD were found in another study [32]. The absence of pathology [40]. We expected a greater exchange of an association of CSF A␤42 levels with WMH vol- proteins and influence of WMHs when these are close ume in our study may be due to the heterogeneity to the ventricles, where most of the CSF is located, of our MCI group and thus high variability in A␤42 as opposed to more distant WMHs. Hereto, we used levels observed in these patients. CSF A␤42 levels an arbitrary value that combined both aspects. We in our MCI group were not significantly decreased observed a slightly stronger association with CSF compared to control subjects, which is also reflected A␤42 levels in the AD group when this arbitrary value by a relative low number of MCI patients (36%) was used instead of total WMH volume only. This positive for A␤42 (i.e., with decreased CSF A␤42 suggests that the relative distance of WMH to the level). Indeed, a subgroup in our study diagnosed ventricles contributes to the extent of the associa- with amnestic MCI (who are more likely to con- tion with CSF A␤42 levels. To explore this further, vert to AD) had significantly lower CSF A␤42 levels we analyzed the WMH distribution pattern across the (mean: 680 pg/ml) as compared to the control group brain. We could observe three different distribution (p < 0.05), and had a higher number of patients (48%) patterns. Interestingly, CSF A␤42 levels were lower in positive for A␤42 than the whole MCI group, which AD patients with a broad distribution of WMH across is in line with previous studies [33, 34]. the brain (pattern 3) as compared to AD patients with Our finding of the absence of a correlation between a higher WMH volume only close to the ventricles p-tau or t-tau and WMH volume is consistent with (pattern 2), further suggesting a possible relation of previous studies [28, 29, 31, 35–38]. Tau is mainly WMH distance to ventricles with CSF A␤42 levels. localized in axons, and increased CSF t-tau levels However, the total WMH volume in brains of patients reflect neurodegeneration (formation of neurofibril- with the broad distribution pattern 3 was also higher lary tangles) or neuronal injury and may dominate than for patients with a peak volume close to the ven- the potentially more subtle effects of WMHs on tricles, and therefore the association of either total tau protein release [28, 38]. However, another study WMH volume and distance of WMH lesion from the showed an association between vascular risk burden ventricles could not be separately assessed. measured by the Framingham Stroke Risk Pro- file (including WMHs, silent cerebral infarcts and Limitations microbleeds) and axonal damage, which was more pronounced in patients with AD pathology [39]. The results of this study should be interpreted Thus, it cannot be excluded that SVD may affect with caution for a number of reasons. Firstly, CSF L.J.C. van Waalwijk van Doorn et al. / Effect of WMHs on AD CSF Biomarkers 173 biomarker concentrations were determined in differ- pretation of CSF biomarkers. For this purpose, we ent laboratories and, despite that the same assays used a computer aided detection algorithm for the were used, variation in the pre-analytical and ana- automated measurement of WMHs. We showed that lytical procedures used will influence our data [16, WMH volume, albeit weakly, associated with CSF 41]. Efforts for more standardized methods are A␤42 concentrations in AD, but not in controls and needed to measure CSF biomarkers, and standardized MCI patients. Interestingly, three global patterns in guidelines are developed using uniform reference the distribution of WMH volume within the brain materials within the JPND-BiomarkAPD project could be observed in AD patients. Our data suggests (https://biomarkapd.org/). Secondly, there was nei- that the distance of WMH to the ventricles in addition ther a standardized MRI protocol used for this to WMH volume also slightly contributes to the asso- study nor were the patients scanned using the same ciation with CSF A␤42 levels. In conclusion, since scanner. Therefore, it cannot be excluded that varia- only CSF A␤42 levels weakly and negatively corre- tions in scanning protocol have influenced our data. lated with WMH/TIV, the occurrence of WMHs is Thirdly, we have not evaluated the possible corre- not accompanied by excess release of cellular com- lations between CSF biomarker levels and disease ponents (tau and A␤ proteins) in the CSF. Therefore, progression of AD. Patients at a more advanced stage WMHs are no major confounders in assessing CSF of AD may have both increased WMH volumes and levels of tau and A␤ proteins. decreased CSF A␤42 levels compared to patients at earlier stages of disease. Indeed, a systematic liter- ature review showed a trend in which CSF A␤42 decreases (mean: –0.4 pg/mL/month for 8 studies) ACKNOWLEDGMENTS over time [42]. In contrast, however, in a longitu- dinal study no changes were observed in CSF A␤42 We would like to thank Marcel de Bruijn, BSc, for and p-tau levels over time (mean: 24 ± 13 months) his assistance on the visual quality control of the MRI in MCI and AD patients [43]; therefore, it remains scans. uncertain to what degree the association of WMHs BIOMARKAPD is an EU Joint Programme– Neu- with CSF A␤42 can be explained by differences in rodegenerative Disease Research (JPND) project. disease stage. Fourthly, the MCI group in our study This project was supported through national funding is a mixed group and may include both stable MCI organizations under the aegis of JPND (https://www. patients and MCI patients who will convert to AD at jpnd.eu). These national funding organizations a later stage, which may have affected our findings. included the Netherlands Organisation for Health Furthermore, some laboratories used CSF biomarkers Research and Development (ZonMw), the Nether- as support in the clinical diagnosis of MCI (and AD), lands; the Instituto de Salud Carlos III (ISCIII) (grant while other laboratories did not. It should be noted numbers: PI11/3035, PI11/02425 and PI14/01126), that the proportion of A␤42 -positive cases was com- Spain; the Italian Ministry of Health (grant num- parable when the diagnosis was supported by CSF ber: IG-2014-2354450), Italy; Health Research biomarkers versus when CSF biomarkers were not Council of the Academy of Finland (grant num- used to support the diagnosis. Finally, it would be ber: 263193), Finland; TUBITAK (project number: interesting to study the association of WMHs with 112S360), Turkey; Fundação para a Ciência e a CSF biomarkers in a larger population and to include Tecnologia (FCT) (grant numbers: PIC/IC/83206/ additional CSF biomarkers like A␤40 or covariates 2007 and JPND/0005/2011), Portugal. Addition- like APOE genotype, as well as to develop a model ally, this work was funded by ZonMW and the combining different confounding factors to evaluate Dutch national ‘Deltaplan for Dementia’ (zonmw.nl/ their combined relation with diagnostic accuracy of dementiaresearch), the BIONIC project to Verbeek: CSF biomarkers. We did, unfortunately, not have CSF nr. 733050822), Alzheimer Nederland, Alzheimer’s A␤40 or APOE genotype data available to include in drug discovery foundation, ISAO, the CIBERNED analyses of the current study. program (Program 1, Alzheimer Disease), Fon- doEuropeo de Desarrollo Regional (FEDER), Unión Conclusion Europea, “Una manera de hacer Europa” and a “Marató TV3” grant (grant number: 201412.10), In summary, we studied if the differences in WMH Alzheimer’s Association, IMI Project “PharmaCog”, volume may act as a confounding factor for inter- FP7, H2020, and Swiss National Science Foundation. 174 L.J.C. van Waalwijk van Doorn et al. / Effect of WMHs on AD CSF Biomarkers Authors’ disclosures available online (https:// Pantoni L, Speck O, Stephan BC, Teipel S, Viswanathan A, www.j-alz.com/manuscript-disclosures/20-0496r2). Werring D, Chen C, Smith C, van Buchem M, Norrving B, Gorelick PB, Dichgans M (2013) Neuroimaging standards for research into small vessel disease and its contribution to ageing and neurodegeneration. Lancet Neurol 12, 822-838. SUPPLEMENTARY MATERIAL [13] van Leijsen EMC, de Leeuw FE, Tuladhar AM (2017) Dis- ease progression and regression in sporadic small vessel The supplementary material is available in the disease-insights from neuroimaging. Clin Sci (Lond) 131, 1191-1206. electronic version of this article: https://dx.doi.org/ [14] Schnaar RL (2010) Brain gangliosides in axon-myelin sta- 10.3233/JAD-200496. bility and axon regeneration. FEBS Lett 584, 1741-1747. [15] McKhann GM, Knopman DS, Chertkow H, Hyman BT, Jack CR, Jr., Kawas CH, Klunk WE, Koroshetz WJ, Manly REFERENCES JJ, Mayeux R, Mohs RC, Morris JC, Rossor MN, Schel- tens P, Carrillo MC, Thies B, Weintraub S, Phelps CH (2011) The diagnosis of dementia due to Alzheimer’s dis- [1] Blennow K, de Leon MJ, Zetterberg H (2006) Alzheimer’s ease: Recommendations from the National Institute on disease. Lancet 368, 387-403. Aging-Alzheimer’s Association workgroups on diagnostic [2] Prins ND, Scheltens P (2015) White matter hyperintensities, guidelines for Alzheimer’s disease. Alzheimers Dement 7, cognitive impairment and dementia: An update. Nat Rev 263-269. Neurol 11, 157-165. [16] del Campo M, Mollenhauer B, Bertolotto A, Engelborghs S, [3] Swan GE, DeCarli C, Miller BL, Reed T, Wolf PA, Jack LM, Hampel H, Simonsen AH, Kapaki E, Kruse N, Le Bastard Carmelli D (1998) Association of midlife blood pressure to N, Lehmann S, Molinuevo JL, Parnetti L, Perret-Liaudet A, late-life cognitive decline and brain morphology. Neurology Saez-Valero J, Saka E, Urbani A, Vanmechelen E, Verbeek 51, 986-993. M, Visser PJ, Teunissen C (2012) Recommendations to stan- [4] Bigler ED, Kerr B, Victoroff J, Tate DF, Breitner JC dardize preanalytical confounding factors in Alzheimer’s (2002) White matter lesions, quantitative magnetic reso- and Parkinson’s disease cerebrospinal fluid biomarkers: An nance imaging, and dementia. Alzheimer Dis Assoc Disord update. Biomark Med 6, 419-430. 16, 161-170. [17] Molinuevo JL, Blennow K, Dubois B, Engelborghs S, [5] De Groot JC, De Leeuw FE, Oudkerk M, Van Gijn J, Lewczuk P, Perret-Liaudet A, Teunissen CE, Parnetti L Hofman A, Jolles J, Breteler MM (2002) Periventricu- (2014) The clinical use of cerebrospinal fluid biomarker test- lar cerebral white matter lesions predict rate of cognitive ing for Alzheimer’s disease diagnosis: A consensus paper decline. Ann Neurol 52, 335-341. from the Alzheimer’s Biomarkers Standardization Initiative. [6] Garde E, Lykke Mortensen E, Rostrup E, Paulson OB (2005) Alzheimers Dement 10, 808-817. Decline in intelligence is associated with progression in [18] Clark CM, Xie S, Chittams J, Ewbank D, Peskind E, Galasko white matter hyperintensity volume. J Neurol Neurosurg D, Morris JC, McKeel DW, Jr., Farlow M, Weitlauf SL, Psychiatry 76, 1289-1291. Quinn J, Kaye J, Knopman D, Arai H, Doody RS, DeCarli [7] Prins ND, van Dijk EJ, den Heijer T, Vermeer SE, Jolles C, Leight S, Lee VM, Trojanowski JQ (2003) Cerebrospinal J, Koudstaal PJ, Hofman A, Breteler MM (2005) Cerebral fluid tau and beta-amyloid: How well do these biomarkers small-vessel disease and decline in information processing reflect autopsy-confirmed dementia diagnoses? Arch Neurol speed, executive function and memory. Brain 128, 2034- 60, 1696-1702. 2041. [19] Diniz BS, Pinto Junior JA, Forlenza OV (2008) Do CSF total [8] van Dijk EJ, Prins ND, Vrooman HA, Hofman A, Koud- tau, phosphorylated tau, and beta-amyloid 42 help to predict staal PJ, Breteler MM (2008) Progression of cerebral small progression of mild cognitive impairment to Alzheimer’s vessel disease in relation to risk factors and cognitive con- disease? A systematic review and meta-analysis of the lit- sequences: Rotterdam Scan study. Stroke 39, 2712-2719. erature. World J Biol Psychiatry 9, 172-182. [9] Selnes P, Blennow K, Zetterberg H, Grambaite R, Rosen- gren L, Johnsen L, Stenset V, Fladby T (2010) Effects of [20] Blennow K, Dubois B, Fagan AM, Lewczuk P, de Leon cerebrovascular disease on amyloid precursor protein meta- MJ, Hampel H (2015) Clinical utility of cerebrospinal fluid bolites in cerebrospinal fluid. Cerebrospinal Fluid Res 7, 10. biomarkers in the diagnosis of early Alzheimer’s disease. [10] Schmidt R, Berghold A, Jokinen H, Gouw AA, van der Alzheimers Dement 11, 58-69. Flier WM, Barkhof F, Scheltens P, Petrovic K, Madureira [21] Hesse C, Rosengren L, Andreasen N, Davidsson P, Vander- S, Verdelho A, Ferro JM, Waldemar G, Wallin A, Wahlund stichele H, Vanmechelen E, Blennow K (2001) Transient LO, Poggesi A, Pantoni L, Inzitari D, Fazekas F, Erkinjuntti increase in total tau but not phospho-tau in human cere- T (2012) White matter lesion progression in LADIS: Fre- brospinal fluid after acute stroke. Neurosci Lett 297, quency, clinical effects, and sample size calculations. Stroke 187-190. 43, 2643-2647. [22] de Jong D, Jansen RW, Kremer BP, Verbeek MM (2006) [11] Verhaaren BF, Vernooij MW, de Boer R, Hofman A, Niessen Cerebrospinal fluid amyloid beta42/phosphorylated tau WJ, van der Lugt A, Ikram MA (2013) High blood pressure ratio discriminates between Alzheimer’s disease and vas- and cerebral white matter lesion progression in the general cular dementia. J Gerontol A Biol Sci Med Sci 61, population. Hypertension 61, 1354-1359. 755-758. [12] Wardlaw JM, Smith EE, Biessels GJ, Cordonnier C, Fazekas [23] Koopman K, Le Bastard N, Martin JJ, Nagels G, De F, Frayne R, Lindley RI, O’Brien JT, Barkhof F, Benavente Deyn PP, Engelborghs S (2009) Improved discrimination OR, Black SE, Brayne C, Breteler M, Chabriat H, Decarli of autopsy-confirmed Alzheimer’s disease (AD) from non- C, de Leeuw FE, Doubal F, Duering M, Fox NC, Green- AD dementias using CSF P-tau(181P). Neurochem Int 55, berg S, Hachinski V, Kilimann I, Mok V, Oostenbrugge R, 214-218. L.J.C. van Waalwijk van Doorn et al. / Effect of WMHs on AD CSF Biomarkers 175 [24] Spies PE, Slats D, Sjogren JM, Kremer BP, Verhey FR, [34] Hampel H, Teipel SJ, Fuchsberger T, Andreasen N, Wiltfang Rikkert MG, Verbeek MM (2010) The cerebrospinal fluid J, Otto M, Shen Y, Dodel R, Du Y, Farlow M, Möller HJ, amyloid beta42/40 ratio in the differentiation of Alzheimer’s Blennow K, Buerger K (2004) Value of CSF beta-amyloid1- disease from non-Alzheimer’s dementia. Curr Alzheimer 42 and tau as predictors of Alzheimer’s disease in patients Res 7, 470-476. with mild cognitive impairment. Mol Psychiatry 9, 705-710. [25] van Waalwijk van Doorn LJ, Gispert JD, Kuiperij HB, [35] Guzman VA, Carmichael OT, Schwarz C, Tosto G, Claassen JA, Arighi A, Baldeiras I, Blennow K, Bozzali Zimmerman ME, Brickman AM (2013) White matter hyper- M, Castelo-Branco M, Cavedo E, Emek-Savas DD, Eren E, intensities and amyloid are independently associated with Eusebi P, Farotti L, Fenoglio C, Ormaechea JF, Freund-Levi entorhinal cortex volume among individuals with mild cog- Y, Frisoni GB, Galimberti D, Genc S, Greco V, Hampel H, nitive impairment. Alzheimers Dement 9, S124-131. Herukka SK, Liu Y, Llado A, Lleo A, Nobili FM, Oguz [36] Gold BT, Zhu Z, Brown CA, Andersen AH, LaDu MJ, KK, Parnetti L, Pereira J, Picco A, Pikkarainen M, de Tai L, Jicha GA, Kryscio RJ, Estus S, Nelson PT, Scheff Oliveira CR, Saka E, Salvadori N, Sanchez-Valle R, Santana SW, Abner E, Schmitt FA, Van Eldik LJ, Smith CD (2014) I, Scarpini E, Scheltens P, Soininen H, Tarducci R, Teunis- White matter integrity is associated with cerebrospinal fluid sen C, Tsolaki M, Urbani A, Vilaplana E, Visser PJ, Wallin markers of Alzheimer’s disease in normal adults. Neurobiol AK, Yener G, Molinuevo JL, Meulenbroek O, Verbeek MM Aging 35, 2263-2271. (2017) Improved cerebrospinal fluid-based discrimination [37] Kester MI, Goos JD, Teunissen CE, Benedictus MR, Bouw- between Alzheimer’s disease patients and controls after man FH, Wattjes MP, Barkhof F, Scheltens P, van der Flier correction for ventricular volumes. J Alzheimers Dis 56, WM (2014) Associations between cerebral small-vessel 543-555. disease and Alzheimer disease pathology as measured by [26] Jansen WJ, Ossenkoppele R, Knol DL, Tijms BM, Schel- cerebrospinal fluid biomarkers. JAMA Neurol 71, 855-862. tens P, Verhey FR, Visser PJ, Aalten P, Aarsland D, Alcolea [38] Soldan A, Pettigrew C, Zhu Y, Wang MC, Moghekar A, D, et al. (2015) Prevalence of cerebral amyloid pathology Gottesman RF, Singh B, Martinez O, Fletcher E, DeCarli in persons without dementia: A meta-analysis. JAMA 313, C, Albert M (2020) White matter hyperintensities and CSF 1924-1938. Alzheimer disease biomarkers in preclinical Alzheimer dis- [27] Ghafoorian M, Karssemeijer N, van Uden IW, de Leeuw FE, ease. Neurology 94, e950-e960. Heskes T, Marchiori E, Platel B (2016) Automated detection [39] Osborn KE, Alverio JM, Dumitrescu L, Pechman KR, Gif- of white matter hyperintensities of all sizes in cerebral small ford KA, Hohman TJ, Blennow K, Zetterberg H, Jefferson vessel disease. Med Phys 43, 6246. AL (2019) Adverse vascular risk relates to cerebrospinal [28] Osborn KE, Liu D, Samuels LR, Moore EE, Cambronero fluid biomarker evidence of axonal injury in the presence FE, Acosta LMY, Bell SP, Babicz MA, Gordon EA, Pech- of Alzheimer’s disease pathology. J Alzheimers Dis 71, man KR, Davis LT, Gifford KA, Hohman TJ, Blennow K, 281-290. Zetterberg H, Jefferson AL (2018) Cerebrospinal fluid beta- [40] Brickman AM (2013) Contemplating Alzheimer’s disease amyloid42 and neurofilament light relate to white matter and the contribution of white matter hyperintensities. Curr hyperintensities. Neurobiol Aging 68, 18-25. Neurol Neurosci Rep 13, 415. [29] Wei K, Tran T, Chu K, Borzage MT, Braskie MN, Harring- [41] Bjerke M, Portelius E, Minthon L, Wallin A, Anckarsater ton MG, King KS (2019) White matter hypointensities and H, Anckarsater R, Andreasen N, Zetterberg H, Andreas- hyperintensities have equivalent correlations with age and son U, Blennow K (2010) Confounding factors influencing CSF beta-amyloid in the nondemented elderly. Brain Behav amyloid Beta concentration in cerebrospinal fluid. Int J 9, e01457. Alzheimers Dis 2010, 986310. [30] Marnane M, Al-Jawadi OO, Mortazavi S, Pogorzelec KJ, [42] Zhou B, Teramukai S, Yoshimura K, Fukushima M (2009) Wang BW, Feldman HH, Hsiung GY (2016) Periventric- Validity of cerebrospinal fluid biomarkers as endpoints ular hyperintensities are associated with elevated cerebral in early-phase clinical trials for Alzheimer’s disease. J amyloid. Neurology 86, 535-543. Alzheimers Dis 18, 89-102. [31] Walsh P, Sudre CH, Fiford CM, Ryan NS, Lashley T, Frost [43] Kester MI, Scheffer PG, Koel-Simmelink MJ, Twaalfhoven C, Barnes J (2020) CSF amyloid is a consistent predictor of H, Verwey NA, Veerhuis R, Twisk JW, Bouwman FH, white matter hyperintensities across the disease course from Blankenstein MA, Scheltens P, Teunissen C, van der Flier aging to Alzheimer’s disease. Neurobiol Aging 91, 5-14. WM (2012) Serial CSF sampling in Alzheimer’s disease: [32] Bos I, Vos SJB, Schindler SE, Hassenstab J, Xiong C, Grant Specific versus non-specific markers. Neurobiol Aging 33, E, Verhey F, Morris JC, Visser PJ, Fagan AM (2019) Vascu- 1591-1598. lar risk factors are associated with longitudinal changes in cerebrospinal fluid tau markers and cognition in preclinical Alzheimer’s disease. Alzheimers Dement 15, 1149-1159. [33] Mendes T, Cardoso S, Guerreiro M, Maroco J, Silva D, Alves L, Schmand B, Gerardo B, Lima M, Santana I, de Mendonça A (2019) Can subjective memory complaints identify A␤ positive and A␤ negative amnestic mild cogni- tive impairment patients? J Alzheimers Dis 70, 1103-1111.