All patients with a first-ever TIA or stroke, without known cognitive decline pre-stroke, admitted to the stroke unit at Bærum Hospital between February 2007 and July 2009 were invited to participate in the CAST study. Patients with previous stroke or TIA, subarachnoid hemorrhage, life expectancy of less than one year, known cognitive decline as indicated by a score of ≥ 3.44 on The Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE)(43), or patients who did not speak Norwegian were excluded. Stroke survivors were invited to participate in a follow-up study from February 2014 to July 2016. Details can be found in previous published papers (40, 44).
Examinations and Assessments
At baseline, stroke etiology was classified according to the Trial of ORG 10172 (TOAST) classification (45), and patients underwent neuroimaging with computed tomography (CT) and/or magnetic resonance imaging (MRI).
In short, at baseline, at one and at seven-year post-stroke, vascular risk factors, and current medication were recorded. Fasting blood samples and an electrocardiography (ECG), were obtained, and body mass index (BMI) was calculated. Neurological impairment was assessed using the National Institutes of Health Stroke Scale (NIHSS) (46). Cognitive evaluation included the Mini Mental State Examination (MMSE) (47), the clock drawing test (48), the Trail making test part A and B (TMT-A and B ) (49), and the 10-word memory test (50) . Additional tests at seven year post-stroke were the Montreal Cognitive Assessment (MoCA) (51) and Controlled Oral Word Association Test (COWAT) (52). Global functional outcome was assessed by the modified Rankin Scale (mRS) (53) and personal activities of daily living (p-ADL) by the Barthel Activities of Daily Living Index (53).
Supplementary investigations at one and seven years included MRI of the brain, carotid ultrasound, and when possible, lumbar puncture was performed. Cerebrospinal fluid (CSF) was collected in polypropylene tubes and immediately transported to the local laboratory, in accordance with the manufacturers’ instructions. Biomarkers for neurodegenerative disease (amyloid-β peptide (Aβ₄₂) levels, total tau (T-tau) and phosphorylated tau (P-tau)) were quantified with commercially available ELISAs (Fujirebio Europe, Gent, Belgium). The laboratory recommended a cut-off value of Aβ₄₂≤ 550ng/L for abnormality, modified from Sjögren et al (54).
At one- and seven-years follow-up, MRI scans were acquired on a Philips Intera system 1.5 tesla (Philips Medical Systems, Best, The Netherlands). The MRI study protocol consisted of 3D-T1, axial T2, 3D-FLAIR, DWI and SWI sequences.
MRI investigations were evaluated for focal vascular lesions, medial temporal lobe atrophy (MTLA), white matter lesions (WML), and global cortical atrophy (GCA), by two radiologists, blinded to the clinical information. Any discrepancies were resolved by consensus. MTLA was graded from 0 to 4; with MTLA grade 0 = no atrophy, MTLA 4 = highest degree of atrophy. MTLA 0–1 is considered normal (55). WML was rated using the visual rating scale proposed by Fazekas, scores ranging from 0 to 3 (56). GCA was rated using the visual rating scale known as Pasquier scale, ranging from 0 to 3 (57).
MRI Segmentations and Analyses
Cortical reconstruction and volumetric segmentation were performed with the FreeSurfer image analysis suite version 6.0.0 (http://surfer.nmr.mgh. harvard.edu/). This includes segmentation of the subcortical white-matter and deep gray-matter volumetric structures (58) and parcellation of the cortical surface (59). This labels cortical sulci and gyri, and thickness values are calculated in the regions of interest. All segmentations were manually inspected.
18F-Flutemetamol PET CT acquisition
Seven years post-stroke, patients were examined with a Siemens Biograph40 mCT scanner (Siemens Healthineers, Erlangen, Germany). All patients received an intravenous injection of approximately 185 MBq 18F-Flutemetamol (mean 188 MBq, range 165-218 MBq). Image acquisition started approximately 90 minutes after injection (mean 91 minutes, range 78 – 108) with a low dose CT followed by PET acquisition for 20 minutes (four frames of five minutes). 3D dynamic emission data were obtained with a resolution recovery algorithm with time of flight (TrueX with two iterations, 21 subsets and a Gaussian filter with FWHM of 2 mm, matrix size 400x400). Reconstructed images had a slice thickness of 2 mm and a voxel size of 2 x 2 x 2 mm3.
Qualitative classification of 18F-Flutemetamol PET
18F-Flut PET images were visually classified as positive (18F-Flut-PET (+)) or negative (18F-Flut-PET (-)) by at least two nuclear medicine physicians with experience in 18F-Flut PET, and recorded in patients’ medical records in line with standard clinical practice between 2015-2017. In addition, one nuclear medicine physician with experience in 18F-Flut PET repeated classification in 2019, and in case of discrepancy between the classifications, an additional expert was consulted. All image classifications were performed according to the validated electronic reader program (60) and as described previously (15).
Quantitative classification of 18F-Flutemetamol PET
Motion correction of the 18F-Flut PET was performed using frame-by-frame rigid registration, then the frames were summed to a single time-frame image and rigidly registered to the anatomical MRI volume using a 6-parameter rigid registration as implemented in the Statistical Parametrical Mapping (SPM 12, Wellcome Trust Centre for Neuroimaging, UCL, UK) toolbox. 18F-Flut-PET standardized uptake value ratios (SUVr) were obtained by normalization to the brainstem. Both the cerebellar cortex, pons and brainstem are widely used Flutemetamol reference regions, acquiring amyloid build plaques not until the fifth and final phase of amyloid deposition (61). Our choice of reference region was influenced by local tradition, a feeling that the structural masks of the cerebellar cortex were less accurate, and a long series of publications utilizing the brainstem or pons as reference region. Prior to normalization, the brainstem mask was eroded by 1mm to avoid partial volume effects, inaccurate segmentation or co-registration. 18F-Flut-PET uptake was analyzed in five pre-selected cortical regions of interest (ROIs), as defined by Desikan et al. (62) and implemented in FreeSurfer as described above, known to hold substantial amyloid plaques in AD: the precuneus and posterior cingulate combined, anterior cingulate, prefrontal, inferior parietal, and lateral temporal cortex (63). The 18F-Flut-PET uptake for each region was averaged across the hemispheres. We further calculated a composite SUVr by averaging the uptake in the above-mentioned regions. SUVr were not calculated to categorize scans as positive or negative but used as continuous variables in correlation analyses.
Outcomes and Diagnosis of cognitive function
The criteria outlined by Winblad et al. (64) were used for MCI, and the International Classification of Diseases 10th revision (ICD-10) criteria (2) for dementia diagnoses at both one and seven years post-stroke. The diagnoses were made in consensus meetings by two senior neurologists (B.F and B.T) and one senior geriatrician (A.R.Ø). Details of the novel method for sub-classification have previously been reported (41), with six potential subgroups; degenerative MCI or degenerative dementia, vascular MCI (MCI VaD) or vascular dementia (dementia VaD) or mixed degenerative and vascular MCI or dementia. The evaluations were based on the results from the medical history, cognitive assessments, the IQCODE and information regarding daily functioning. Sub-classification for proposed underlying etiology was based on MRI findings of vascular or degenerative brain changes, biomarkers in the CSF, the patient´s vascular risk factors and clinical cognitive profile. Patients were classified with possible vascular disease when the radiological findings revealed WMLs without MTLA, while MTLA without WMLs was interpreted as degenerative etiology. Patients with combined pathologies were classified with mixed vascular and degenerative disease.
Categorical variables were compared with Pearson's chi-squared test and continuous variables with independent Student t-test. The relationship between death as dependent variable and age-adjusted CSF as independent variables were assessed with logistic regression analyses. Correlation analyses of 18F-Flut-PET SUVr and measures of neurodegeneration and cognition, were performed assessing the Pearson correlation coefficient in continuous variables (MTA, CSF Aβ₄₂ levels, GCA, TMT-A), Spearman rho in ranked data (MMSE). Linear regression analyses were performed with 18F-Flut-PET SUVr as explanatory variable. Each regression model was adjusted for age. Statistical analyses were performed using SPSS Statistic version 23.