Patient Population

COMPASS-ND is a Canada-wide multi-center prospective, longitudinal cohort study. The overall study design and methods have been published previously. ^{Reference Chertkow, Borrie and Whitehead1} The study is registered on clinicaltrials.gov (NCT03402919). In brief, the objective is to create a platform for research on neurodegenerative and vascular causes of mild cognitive impairment (MCI) and dementia. Recruitment began in 2016 and is expected to complete in approximately 2023. The study is conducted with ethical approval by institutional review boards at each site, with written informed consent of the participants. In this manuscript we report the methods for assessing vascular contributions to neurodegeneration, with interim findings.

In total, 2310 participants will be recruited in these groups: cognitively intact elderly (n = 660), subjective cognitive impairment (SCI; 300), MCI (400), subcortical ischemic vascular MCI (V-MCI; 200), mild dementia due to clinically probable Alzheimer’s disease (AD; 150), mild dementia of mixed etiology (200), Parkinson’s disease (PD) and Lewy body dementias (LBDs; 200), and frontotemporal dementia (200). Full selection criteria have been published previously. ^{Reference Chertkow, Borrie and Whitehead1} Participants are being recruited, generally from specialty cognitive clinics, at 31 sites across Canada.

Diagnoses are made by the site investigator based on all available clinical and research information available at the time of assessment. A screening diagnosis is made based on the information from the referral and screening along with referral information including clinical neuroimaging, and then a clinical diagnosis is made additionally incorporating information collected at the research clinical and physical assessment but prior to research MRI or other biomarker information. Group analyses in this paper are based on the clinical diagnosis groupings.

Based on a priori scientific interest, two groups of patients with vascular cognitive impairment were identified using specific selection criteria: subcortical ischemic V-MCI, and mixed etiology dementia with a vascular component. Criteria for V-MCI were derived from consensus criteria from the American Heart Association ^{Reference Gorelick, Scuteri and Black2} and International Society for Vascular Cognitive and Behavioral Disorders (Vas-Cog). ^{Reference Sachdev, Kalaria and O’Brien3} V-MCI participants were required to be age ≥60, have MCI according to National Institute on Aging-Alzheimer’s Association criteria (NIA-AA), ^{Reference Albert, DeKosky and Dickson4} not have a prior history of clinical stroke, and to have evidence of cerebrovascular disease on brain computed tomography (CT) or MRI defined as two or more supratentorial infarcts (i.e., excluding brainstem or cerebellar infarcts) or beginning confluent or confluent white matter hyperintensity (WMH) according to the Age-Related White Matter Change (ARWMC) scale ^{Reference Wahlund, Barkhof and Fazekas5} on any slice. For the V-MCI group, participants with history of symptomatic stroke were excluded so that the focus is on V-MCI caused by covert cerebrovascular disease, predominantly due to cerebral small vessel disease. Criteria for mixed dementia were adapted from NIA-AA criteria for dementia due to AD ^{Reference McKhann, Knopman and Chertkow6} and required that a non-AD cause of dementia should additionally be present. Although the presence of vascular disease was not required for the mixed dementia group, in practice most cases of mixed dementia have a vascular component. ^{Reference Rabinovici, Carrillo and Forman7} Persons with symptomatic stroke within the last year were excluded. Full details of the study inclusion and exclusion criteria are provided in the online-only supplement (online Table I).

Measurements and Outcomes

Data on age, sex, past medical and surgical history, medications, tobacco and alcohol intake, hearing, and vision are collected. Cognitive and neurobehavioral measurements include the Montreal Cognitive Assessment, ^{Reference Nasreddine, Phillips and Bedirian8} Clinical Dementia Rating Scale, ^{Reference Morris9} Mild Behavioral Impairment Checklist ^{Reference Ismail, Aguera-Ortiz and Brodaty10} and Apathy questionnaire. The Hachinski Ischemic Score ^{Reference Hachinski, Iliff and Zilhka11} is collected. A two-hour neuropsychological test battery is administered. Full details of these assessments have been published previously. ^{Reference Chertkow, Borrie and Whitehead1} Participants will be followed for two years with a repeat of these assessments at two years. At this time, repeat brain imaging is not planned.

Physiological measurements include sitting and orthostatic blood pressure, height and weight (to calculate body mass index), and waist and hip circumference. Blood measurements include complete blood count, creatinine, hemoglobin A1c, insulin, homocysteine, C-reactive protein, high-density and low-density lipoprotein (HDL and LDL), and apolipoprotein B. Lumbar puncture is offered as a voluntary option and will be analyzed for beta-amyloid, total tau, and phosphorylated tau. Additional measurements potentially relevant to defining vascular contributions to neurodegeneration are shown in Table 1.

Table 1: COMPASS-ND assessments of interest for vascular cognitive impairment

Full details of the neuropsychological test battery have been published previously. ^{Reference Chertkow, Borrie and Whitehead1}

Brain MRI are acquired using the previously validated Canadian Dementia Imaging Protocol (www.cdip-pcid.ca). ^{Reference Duchesne, Chouinard and Potvin12,Reference Potvin, Chouinard and Dieumegarde13} The protocol is summarized in Table 2. Compliant with the Standards for Reporting Vascular Changes on Imaging (STRIVE), ^{Reference Wardlaw, Smith and Biessels14} it was harmonized on three major vendor platforms and includes a common procedure for qualification, quality control, and quality assurance.

Table 2: Canadian Dementia Imaging Protocol ^{Reference Duchesne, Chouinard and Potvin12}

FLAIR = fluid attenuated inversion recovery; GRE = gradient-recalled echo; BOLD = blood oxygen level dependent.

A Vascular Core Lab for visual review of brain MRI is distributed between the University of Calgary and Sunnybrook Hospital (University of Toronto). Each MRI is visually reviewed for evidence of vascular brain injury by one of two trained readers qualified by the Vascular Core Lab. Readers are required to be neurologists or radiologists with more than five years of clinical experience with cerebrovascular lesions on MRI, to complete a training module, and to exhibit good inter-rater reliability on a training set (kappa ≥ 0.60) for common vascular lesions (presence of brain infarcts, presence of microbleeds, and Fazekas WMH rating scale ^{Reference Fazekas, Chawluk, Alavi, Hurtig and Zimmerman15} ). MRI scans are reviewed, blinded to clinical information, using a standardized case report form. Infarcts, microbleeds, and WMH are defined according to STRIVE. ^{Reference Wardlaw, Smith and Biessels14} Infarcts are classified by size, number, location, and vascular territory, which allows their subdivision as lacunar versus non-lacunar type. Microbleeds are classified by number and location as in the Brain Observer Microbleeds (BOMBS) scale. ^{Reference Cordonnier, Potter and Jackson16} If microbleeds or cortical superficial siderosis are present their pattern is classified as probably, possibly, or not consistent with cerebral amyloid angiopathy as defined by revised Boston criteria. ^{Reference Linn, Halpin and Demaerel17} When lobar microbleeds or cortical superficial siderosis were accompanied by non-lobar microbleeds, the pattern is described as “mixed”. ^{Reference Pasi, Charidimou and Boulouis18} WMH are visually rated using the Fazekas Scale. ^{Reference Fazekas, Chawluk, Alavi, Hurtig and Zimmerman15} Each scan is visually reviewed for quality. Standard operating procedures have been developed and are stored in a study binder, following FDA guidance for imaging endpoints and guidelines for multi-center brain imaging studies. ^{Reference Wiseman, Meijboom and Valdes Hernandez19}

Neuroimaging evidence of a potential vascular cause or contribution to cognitive decline is defined, a priori, as either the presence of two or more infarcts or beginning confluent or confluent WMH according to the ARWMC scale on any MRI or CT slice. The ARWMC scale was chosen because it is simple for sites to apply and can be used with either CT or MRI, facilitating site identification of participants for the V-MCI group. Additionally, the MRI-based Fazekas scale will be recorded. The imaging criteria for V-MCI are based on Vas-Cog criteria for pure vascular cognitive impairment, ^{Reference Sachdev, Kalaria and O’Brien3} but to increase sensitivity the criteria were revised to require only two infarcts instead of three and to include beginning confluent WMH instead of confluent WMH, and to increase specificity supratentorial infarcts are required. Post-hoc analyses will test the validity of these criteria for identifying patients at risk of continued cognitive decline. Strategic locations of infarcts were not part of the imaging criteria because the required locations and sizes are not precisely defined in the literature.

A visual assessment of medial temporal atrophy is made using the Scheltens scale, ^{Reference Korf, Wahlund, Visser and Scheltens20} which is rated from 0 to 4 on each side and was averaged for this analysis. Quantitative segmentation of the hippocampus using high-resolution T1-weighted is planned.

Quantitative analyses of brain images will be batched and performed near the end of the study recruitment period according to the best available methods at that time. The Canadian Dementia Imaging Protocol acquires images suitable for analysis of global and regional brain volumes including cortical thickness; DTI for analysis of white matter connectivity, and multiple BOLD volumes suitable for analyzing resting state functional connectivity.

The data from COMPASS-ND used in this paper, including MRI Digital Communication in Medicine (DICOM) is stored and curated on the Longitudinal Online Research Information System (LORIS). ^{Reference Mohaddes, Das and Abou-Haidar21,Reference Das, Zijdenbos, Harlap, Vins and Evans22}

Statistical Analyses, Sample Size, and Data Access Policies

The study is designed to be an open access platform for secondary analyses of the risk factors, predictors, and cognitive outcomes of common neurodegenerative diseases. Therefore, there is no pre-defined primary outcome.

The V-MCI group sample size of 200 was based on having 80% power to detect factors that would double the risk of progression to dementia at two years (from 20% to 40%), assuming 20% loss to follow-up at two years, alpha of 0.05, and a two-tailed p value.

Access to COMPASS-ND data will be provided to the worldwide research community according to the policies of the Canadian Consortium on Neurodegeneration in Aging (CCNA; https://ccna-ccnv.ca/policies/). Project proposals should be submitted to the publications and Data Access Committee (PDAC). After study recruitment is completed, data will be provided to the Global Alzheimer’s Association Interactive Network (GAAIN) for access through their portal.

To explore the feasibility of recruitment and study measurements, we present descriptive statistics of the distribution of vascular risk factors and vascular MRI findings in the first 200 recruited participants according to clinical diagnosis. V-MCI and mixed dementia are grouped for analysis. To compare prevalence across groups the chi-square test is used, and to compare medial temporal atrophy scores analysis of variance (ANOVA) is used. The kappa statistic is used to test the reliability of the two readers on 50 COMPASS-ND cases (7 with infarcts and 10 with microbleeds), with values ≥0.60 indicating substantial agreement. Agreement was based on the presence or absence of any lesions of that type; there were too few participants with multiple lesions to test the reliability of lesion counts. Because these are interim, exploratory analyses without the full planned sample size, multivariable adjustments are not done and there is no adjustment for multiple comparisons. Analyses were done using SAS version 9.4 (Cary, NC). A p value of <0.05 is considered significant.