The Norwegian Cognitive impairment after stroke study (Nor-COAST): study protocol of a multicentre, prospective cohort study

Reported rates for PCI varies considerably between studies. Many studies reports dementia but not MCI, and in studies that include MCI the operational definition has a large impact on the reported rate [8]. Pooled dementia rates suggest that 10% of first ever stroke survivors develop incident dementia within the first year, and 30% after recurrent stroke [9]. A Norwegian study reported an incidence of MCI and dementia one year after first ever stroke of 37.5 and 19.6%, respectively [10]. Heterogenity in reported rates of cognitive impairment following stroke can be explained by casemix differences resulting from population selection and attrition bias [6] and by varying assessment time and tools, operational definitions and diagnostic criteria [8]. Further there are few studies on the course of PCI, especially of late-onset PCI, and it is hypothesised that the course of PCI strongly depends on the underlying mechanisms [11‐14]. To quantify the overall burden of PCI and explore distinct trajectories and their determinants, larger inclusive studies with longer follow-up are warranted.

A complex interplay of vascular pathology, neurodegenerative and inflammatory processes contributes to the progression of PCI [11, 14‐16]. Cognitive impairment predisposes to stroke [15, 16] and patients with dementia get more severe strokes [17], supporting that the two conditions are related. Several imaging characteristics associated with PCI have been identified. Some imaging findings are more extensively studied, especially the relationship between white matter hyperintensities and cognitive impairment [18]. However, recently there has been more attention on the association between degenerative and vascular changes in the brain [2]. Chronic inflammation seems to be involved in the pathogenesis of both cerebrovascular disorders and Alzheimer’s disease. Immune activation and immune depression are present in the acute, sub-acute and chronic stage after a stroke and diminished or impaired inflammatory mechanisms are likely to be important factors in the pathways leading to dementia [19]. Cytokines are potential drivers in addition to neurovascular dysfunction, endothelial activation [20, 21] and genetic risk including Apolipoprotein Eε4 (APOE ε4) carrier status [22]. Several clinical markers are predictors of PCI, and the risk appears to increase with increasing number of predictors. Predictors include age, pre-stroke cognitive impairment, pre-stroke functioning and frailty as well as characteristics of the stroke lesion. In addition, complications in the acute and sub-acute phase as delirium, infections, falls and seizures seem to increase the risk [6].

Early and late-onset PCI is probably driven by somewhat different mechanisms, with brain resilience and characteristics of the stroke lesion being the dominant driver for early-onset PCI, while small vessel disease, Alzheimer pathology and recurrent strokes are more pronounced in late-onset PCI [11, 14, 23, 24].

Prevention of recurrent strokes is important in preventing PCI [6] and can probably be achieved in up to 80% by optimal secondary prevention [25]. A multi domain intervention including vascular risk monitoring and physical activity has shown to prevent cognitive decline in at-risk stroke free elderly people suggesting that secondary prevention could have an additional and independent effect on cognition and overall brain health [26]. However, guidelines for secondary prevention are rather complex, involving secondary preventive medications and healthy lifestyle including smoking cessation, moderate alcohol consumption, healthy diet and regular physical activity [27]. Adherence to both medication and lifestyle interventions have been reported to be suboptimal in many patients [28], and adherence to secondary prevention may be problematic in patients with cognitive impairment after stroke.

There is evidence that physical activity enhances cognitive function after stroke [29], and especially executive function in individuals with MCI [30]. A recent meta-analysis, representing data from 736 participants included in 14 randomized controlled trials, demonstrated a significant positive effect of physical activity on cognition post-stroke [31]. Possible mechanisms involved are reduced levels of inflammation [32], improved brain circulation, neurogenesis and increased size of hippocampus [33, 34]. On the other hand, sedentary behaviour has been identified as a risk factor for vascular disease, and hypothesized to increase the risk of cognitive impairment after stroke [35].

The aim of the Norwegian Cognitive Impairment After Stroke study (Nor-COAST), is to quantify and measure levels of cognitive impairments in a Norwegian general stroke cohort and to identify biomarkers and clinical markers associated with overall prognosis for early and late onset PCI.

The Nor-COAST study is organised within five work packages (WP). The aim of WP 1 is to describe incidence and distinct trajectories of PCI. In WP 2, the aim is to increase knowledge concerning the pathogenesis of early and late-onset PCI by analysing brain imaging, APOE ε4 typing and blood markers, and to explore associations between neurodegenerative disease, underlying vascular disease, inflammation and stroke characteristics. In WP 3, we aim to identify predictors for PCI in order to develop a novel risk score consisting of a few clinical markers that will enable planning of individualized prevention and treatment. The aim of WP 4 is to study the impact of physical activity and sedentary behaviour on PCI in the short and long term, and to explore associations between motor function measured by performance based tests and progression of PCI. WP 5 aims to explore the impact of cognitive function on adherence to secondary prevention and evaluate whether optimal secondary prophylaxis is protective against decline in cognitive function over time Fig. 1.

Based on the Norwegian stroke registry we estimated that about 1000 stroke patients in total would be available for inclusion per year in the participating hospitals. Experience from earlier studies at the same hospitals involved [10, 38] indicated that about half of eligible participants would be included and that 20–25% would be lost to follow-up during the data collection period of 18 months. With 750 patients available for follow-up, we will be able to detect an association corresponding to a correlation of 0.11 or higher (power 86% at significance level 5%). Inclusion of approximately 1000 participants would therefore be sufficient to answer the main research questions within each work package. Based on these assumptions we decided on a preplanned inclusion stop at 1000 participants or a maximum of two years inclusion period.

Eligibility criteria are: 1) Admittance to one of the five centres within seven days after symptom debut. 2) Acute stroke diagnosed according to the World Health Organisation (WHO) criteria or with findings of acute infarction or intra-cerebral haemorrhage on MRI. 3) Scandinavian speaking 4) over > 18 years and 5) living in the catchment area of the recruiting hospitals. Exclusion criteria were expected survival less than three months.

Assessments during the hospital stay are performed at discharge or the seventh day of the stay for participants with longer hospital stay. Study related follow-ups are performed at the outpatient clinic three and 18 months, and three years’ post-stroke. Participants unable to attend the outpatient clinic are assessed by telephone interview and/or proxy information is collected. Assessments and interviews are performed by trained research assistants, using a standardised paper Case Report Form (CRF), and then plotted in web-based CRF developed by the Unit for Applied Clinical Research, NTNU. Table 1 provides an overview of data collected by time and data source.

Demographics and participant characteristics will be presented using descriptive statistics. Prevalence and incidence of dementia will be presented by crosstabs. Statistical analyses for trajectories will be performed with group-based trajectory modelling, which uses maximum likelihood to identify groups of individuals with statistically similar trajectories Two primary outputs of the estimations in group-based trajectory modelling are 1) shapes of trajectories of each group as specified by group-specific polynomial functions of time and 2) estimated percentages of the population following each identified trajectory [70]. Both results from visual and volumetric brain analyses will be used to identify MRI predictors of early or delayed of cognitive impairment after stroke. Brain MRI analyses will be combined with machine learning techniques to explore eventual clusters in the study population and to build predictive models. Network analysis with the “NetworkX” software package [71]. is used to discover patient clusters. Predictive models are build with classification algorithms like decision trees with “scikit-learn” [72]. and gradient boosting with “XGBoost” [73]. Linear regression, logistic regression and other statistical methods will be used according to the specific research questions within each work package. A statistician is part of the project managing group and will be involved in developing analysis plans for the main publications and quality insurance of statistical approach in each work package.