Abstract
INTRODUCTION
Cerebral small-vessel disease (SVD) is a leading cause for cognitive impairment and dementia [1]. Although small vessels cannot be detected on conventional imaging [2], some of their parenchymal manifestations are visible on magnetic resonance imaging (MRI). The clinical consequences of SVD, such as cognitive decline, only weakly correlate with each presumable marker of SVD [3, 4]. It is unclear whether these parenchymal alternations are indeed key players in the pathogenesis of cognitive decline, or only markers for yet unknown processes. In a previous study, we showed that white matter hyperintensities (WMH) predict cognitive performances following first-ever mild to moderate stroke or transient ischemic attack (TIA) [5]. However, WMH is only one among other radiological markers of SVD. In 2013, an international working group developed standards for reporting vascular changes on neuroimaging (STRIVE) [6]. These standard definitions of MRI SVD markers include WMH, lacunes, lobar and deep cerebral microbleeds (CMB), and enlarged perivascular spaces (PVS).
So far only a few studies have combined these SVD features to capture total SVD burden and its clinical applications in stroke survivors, and none have investigated its association with post-stroke cognitive performance [7, 9].
In this study we sought to determine whether adding other SVD markers to WMH using STRIVE, as well as their total score by a previously suggested scale [8], improves prediction of post-stroke cognitive performances.
METHODS
Ethics statement
The study was approved by the Tel Aviv Sourasky Medical Center ethics committee. Written informed consent was obtained from all participants.
Study population and data collection
We analyzed data of stroke or TIA patients obtained from participants enrolled in the Tel Aviv Brain Acute Stroke Cohort (TABASCO) study [10], an ongoing prospective study of first-ever mild to moderate (National Institutes of Health Stroke Scale (NIHSS) <17) ischemic stroke or TIA patients aiming to identify predictors for post-stroke cognitive impairment (ClinicalTrials.gov # NCT01926691).
TIA was defined by clinical symptoms lasting less than 24 hours with no lesion on diffusion weighted imaging (DWI). Stroke patients were defined based on their clinical presentation and/or the presence of hyperintense lesion on DWI. For stroke patients with no DWI lesion, stroke diagnosis was confirmed by a stroke clinician (J.M).
From April 2008 to December 2014, a total of 575 consecutive eligible patients according to study inclusion and exclusion criteria [10] were enrolled to the TABASCO study. From the present analysis, patients were excluded if: (1) they did not have baseline MRI (n = 114 patients) or one or more of the sequences essential for the calculation of SVD score were missing (n = 37); (2) their cognitive data at one year were missing (n = 123); or (3) patients deceased before follow-up (n = 35). This left a total of 266 patients (see Fig. 1).
For each patient, demographic data, medical history, and comorbidities were collected as previously described [10]. Vascular risk factors were assessed according to the Framingham Stroke Risk Profile (FSRP) score [11].
MRI acquisition
All images were acquired within 7 days of stroke onset on a 3T GE scanner (GE Signa EXCITE, Milwaukee, WI, USA) using an 8–channel head coil. Imaging parameters were previously described [10].
MRI analyses
Results are reported in accordance with STRIVE [6]. We used the previously described SVD burden score [7, 8], an ordinal scale (0 to 4) counting the presence of each of the four MRI markers for SVD. In more details, SVD burden score was composed of (1) Chronic lacunar infarcts: presence of one or more lacunes was defined as sharply demarcated hypointense lesions sized between 3 mm and about 15 mm in diameter on T1-weighted images with corresponding hypointense lesions with hyperintense rim on FLAIR. (2) White matter hyperintensities (WMH) were graded using the Fazekas score [12, 13]. If confluent WMH (Fazekas score 2 and 3) were present, one point was awarded. (3) Cerebral microbleeds (CMB) were defined as round hypointense lesions on T2-weighted gradient echo-images with a diameter <10 mm. CMBs were then divided to lobar versus deep [14]. If ≥1 deep or lobar CMB were present one point was awarded. (4) Enlarged perivascular spaces (PVS) were defined as smooth margin, round, oval, or linear-shaped lesions, sized up to 3 mm, with signal intensity equal to cerebrospinal fluid (CSF) on T1-weighted images. Enlarged perivascular spaces at the level of the basal ganglia as well as at the level of centrum semiovale were counted [9, 16]. We counted enlarged perivascular spaces in the most affected hemisphere. One point was awarded if 30 or more enlarged perivascular spaces were present at any of the locations. Based on FreeSurfer V5.1 image analysis suite (http://surfer.nmr.mgh.harvard.edu/) [17, 18], ventricular CSF volume was calculated and adjusted to intracranial volume to represent a measure of cerebral atrophy.
For the first 50 patients, all images were independently rated by three raters (E.A., J.M., and E.K.) for the presence of chronic lacunar infarcts, WMH, CMB, and PVS. Total Kappa was 0.84. Kappas for the presence of asymptomatic lacunar infarcts and CMB were 0.67 and 0.70, respectively. Weighted kappas for WMH and PVS in the basal ganglia were 1.0 for each. In case of unclear lesions, SVD burden score was ascertained byconsensus.
Ischemic infarct type
Presence of acute ischemic infarcts was assessed by a senior neuroradiologist (O.A.) based on DWI. DWI lesions were categorized into 4 groups: no lesions, cortical, sub-cortical, or subtentorial infarct [19] (Table 1).
Cognitive assessments
NeuroTraxTM (NeuroTrax Corp., Bellaire, TX) computerized cognitive testing was performed individually by each patient 12 months following stroke/TIA. This is a computerized battery of neuropsychological tests that is used for reliable detection of cognitive state in cognitively healthy, mild cognitive impairment and mild dementia subjects [20]. It provides an overall measure of cognitive function (global cognitive score) as well as evaluation of specific cognitive domains (executive function, memory, visual spatial, verbal function and attention). Each domain score is normalized to fit a standardized scale (mean 100; Standard deviation 15). Patients were defined as having cognitive impairment if their Neurotrax global cognitive score was more than 1.5 SD below the mean after one year.
Statistical analysis
Statistical analysis was conducted using SPSS for Windows software (version 19.0). Between group comparisons were done using the independent sample t test for continues variables. The χ2, Mann-Whitney, and the Kruskal–Wallis tests were used to assess categorical variables. Pearson (for interval scaled variables) or Spearman (for ordinal variables) correlations were used to assess relationships between age, cognitive domain scores and SVD burden. Analysis of variance (ANOVA) was used to compare between lesion type, SVD score and cognitive performance. Bonferroni correction was used if needed.
In order to assess the relationships between admission WMH score (according to Fazekas [13]) or SVD burden and cognitive performance in the different cognitive domains, two sets of linear regressions were carried out: unadjusted regressions (Table 2, model 1) and multivariate regressions (Table 2, model 2) controlling for age, gender, years of education, admission NIHSS, history of hypertension, and evaluation of cerebral atrophy via normalized ventricular CSF volume.
Further, univariate logistic regressions were performed to ascertain the effects of age, years of education, male gender, stroke or TIA, lesion side, vascular risk factors, admission NIHSS, SVD features, SVD score, and WMH score on the likelihood that patients will develop cognitive impairment one year after the event.
RESULTS
Patients’ characteristics
Of the 266 patients, 217 (81.6%) were diagnosed with ischemic stroke and 49 (18.4%) with TIA. Stroke etiologies (based on TOAST criteria [21]) were as follows: 109 (50.2%) lacunar stroke, 31 (14.3%) cardioembolic stroke, 23 (10.6%) large-artery atherosclerotic stroke, 54 (24.9%) stroke of other or undetermined etiology. No differences in SVD scores or cognitive performances were observed between stroke and TIA patients, and they were therefore grouped together for further analyses. Mean age upon admission was 66.4±9.4 years and 60.9% were males. Other demographic, clinical, and cognitive characteristics are summarized in Table 1. Patients excluded from the present analyses (Fig. 1) were less educated (12.7±3.7 versus 13.6±3.8, p = 0.004) and older (68.5±10.3 versus 66.4±9.4, p = 0.013), with similar neurological deficit measured by NIHSS upon admission (2.8±2.9 versus 2.6±2.8, p = 0.473) and similar SVD burden (Mann-Whitney U test U = 16515,p = 0.913).
SVD burden association with age, education, vascular risk factors, and sex
Significant associations were found between SVD burden score and age (r = 0.265, p = 0.001) and FSRP (r = 0.321, p = 0.001), but not with years of education. In addition, male versus female comparison revealed higher SVD burden score for male group(p = 0.038)
Associations between SVD features, WMH score, SVD score, and cognition
No associations were found between cognition and lacunes count, PVS score, or microbleeds count (Supplementary Table 1). WMH score based on the Fazekas scale [13] showed significant negative associations with all cognitive domains except for verbal function. Association remained significant following an adjustment for age, education, during admission NIHSS, gender, history of hypertension, and evaluation of cerebral atrophy using normalized ventricular CSF volume (Table 2). Negative correlations between total SVD burden score and cognitive scores were observed for global cognitive, memory, and visual spatial scores only (all p < 0.05). However, following adjustment for confounders no associations remained significant (Table 2). Adding each of STRIVE biomarkers (lacunes, lobar and deep CMBs, basal ganglia and centrum semiovale PVS) to WMH did not considerably alter the effect size of association between WMH and cognitive performances in multi-variable model.
Univariate predictors of cognitive impairment
During the one-year follow-up period, 27 (10.2%) patients developed cognitive impairment (see Methods for definition). Univariate logistic regressions revealed that years of education was associated with a reduction in the likelihood to develop cognitive impairment (OR 0.79 95% CI 0.70–0.89), while WMH score was associated with increased likelihood (OR 1.517 95% CI 1.006–2.290). All other parameters did not emerge as predictors of cognitive impairment in this cohort (Supplementary Table 2).
Infarction type and cognitive score
For patients defined as stroke, ANOVA analysis comparing different lesion type (no DWI lesions, cortical, sub-cortical, and sub-tentorial) revealed no significant differences in global cognitive score or in any of the cognitive sub-domains. Similar results were obtained when stroke patients without DWI lesions were excluded. Further analysis comparing different stroke types according to the TOAST criteria and also lesion lateralization (right versus left) revealed no significant differences in global cognitive score or in any of the cognitive sub-domains as well.
DISCUSSION
In the present study, we found that adding other SVD markers to WMH does not improve predication of post-stroke cognitive performance and WMH remains the sole predictor for post-stroke cognitive performance, among first-ever mild to moderate ischemic stroke or TIA patients.
The association found between WMH based on Fazekas scale [13] alone and cognitive performance is in line with our previous paper [5] and others [22–24]. No associations were found between cognitive performances and other SVD markers. Lacunes have been associated with cognitive impairment in some SVD studies [25, 26], but not in others [27]. Notably, lacunes are often mentioned in relation to processing speed measure [25, 26]. This domain, however, was not assessed in our cognitive battery. Studies looking at association between CMB and cognitive decline have produced inconsistent results. In a recent study, a threshold effect was suggested for this association in patients with high CMB count [28]. Of note, according to total SVD score6 used in the present study, one point was awarded if one CMB or more were present. A quantitative score, therefore, may be more predictive. Perivascular spaces were a very frequent finding in our cohort, and there was almost no variance between patients, hence no association was found.
It is possible that the ischemic event triggers and hastens cognitive impairment processes through multiple mechanisms such as enhanced inflammatory response or compensatory processes. WMH may be the best indicator of these processes, even though the pathological basis of the later is not fully understood. Previous studies found different pathological alternations underlying WMH, such as demyelination, axonal loss, thickening of blood vessels, and valerian degeneration [29, 30]. The lack of association between other SVD markers and post-stroke cognitive performances may imply that these changes are indeed only markers for not yet determined process, not detected on conventional imaging. The use of ultra-high resolution MRI [31] may improve our understanding of the pathophysiology underlying post-stroke dementia. It should also be noted that the suggested SVD burden score does not encompass other SVD parameters such as cerebral microinfarcts [32] and cortical atrophy [33], which may add further information on overall brain “health” and affect cognitive outcome. In an exploratory post hoc analysis, adding atrophy as confounder did not alter results.
Strengths of this study include the relatively large, prospective, single center, first-ever stroke or TIA cohort, the identical 3T MRI protocol used for all patients, and the comprehensive cognitive assessment in multiple domains. In addition, MRI features of SVD were rated according to the neuroimaging standards for research of SVD [6].
Several limitations should be considered. First, this study could have benefited from follow-up MRIs which could provide information on the dynamics of SVD features and their relation to cognition. Second, patients were included in this study only if they were reported to be free of cognitive decline (IQCODE, Informant Questionnaire on Cognitive Decline in the Elderly) prior to the event. Yet, our patients’ cognitive performance before the ischemic event was not systematically evaluated. Third, SVD features were analyzed as dichotomous and not continuous variables. Fourth, one of the main problems in stroke research is the heterogeneity of lesions in term of location, etiology, and size. For that reason, we chose to assess the effect of lesion type (cortical, sub-cortical, and sub-tentorial) on cognitive outcome. Fifth, we acknowledge the high dropout rate of patients excluded from the present analyses, yet our results relate to 266 patients with complete cognitive and radiological evaluation.
Summary
To the best of our knowledge, this study is the first to examine the associations between SVD burden during admission and cognitive performance after one year in first-ever stroke or TIA patients. WMH score was associated with cognitive performance, while adding more SVD markers or total SVD burden score were not.
