Predicting post-stroke cognitive impairment using acute stroke neuroimaging and other biomarkers

Abstract

BACKGROUND: Cognitive impairment is common following a stroke. Structural neuroimaging and blood tests are a routine part of acute stroke management, and may identify those at risk of post-stroke cognitive problems. This thesis focuses on identifying clinically relevant risk factors, present at the time of stroke, that are associated with post-stroke cognitive impairment, including dementia.

METHODS AND RESULTS: I first performed a systematic review to identify neuroimaging features, present on acute stroke computed tomography (CT) brain scans, that are associated with post-stroke cognitive problems. I found that white matter hyperintensities (WMH), cerebral atrophy, and pre-existing stroke lesions were associated with a two to threefold increased risk of post-stroke dementia. WMH were also associated with a threefold increase in risk of post-stroke cognitive impairment.

Since structural magnetic resonance imaging (MRI) is becoming more common in clinical practice, I performed a second review focusing on associations between features seen on acute stroke MRI and cognitive outcome. Severity of WMH (3 studies, N=704, adjusted OR=1.26, 95% CI=1.06-1.49), presence of cerebral atrophy (3 studies, N=453, unadjusted OR=2.48, 95% CI=1.15-4.62), cerebral microbleeds (2 studies, N=9151, adjusted OR=1.36, 95% CI=1.08-1.70), and worse cerebral small vessel disease (2 studies, N=499, OR=1.34, 95% CI=1.12-1.61) were associated with post-stroke cognitive problems. There were no clear associations with other features.

Findings from the systematic reviews informed the neuroimaging features that I looked at in my third project, a Scottish data linkage study. I identified two existing research datasets of stroke patients (N=185) which contained coded reports of acute stroke CT brain scans. The Lothian Research Safe Haven used the participants’ Community Health Index (CHI) numbers to link to electronic health records (Hospital Admissions, Death Records, and Prescribed Drugs) and identify dementia diagnoses. Twenty-one participants (11%) developed post-stroke dementia. Moderate-to-severe WMH and presence of cerebral atrophy were associated with an increased risk of post-stroke dementia.

To address my research question on a larger scale, I designed a second data linkage study using National Swedish Registry data. This project aimed to identify blood biomarkers associated with post-stroke dementia. AMORIS is a Swedish cohort of 812,073 people who had blood samples collected during routine health check-ups, or as an outpatient, between 1985 and 1996. AMORIS has already been linked to multiple National Patient Registers within Sweden, including stroke and dementia registers. Through linkage of Swedish data, I am able to identify people with a reported blood sample in the AMORIS cohort who had a stroke, and have subsequently developed dementia. I planned to access these data remotely from Edinburgh; however, there were complex legal challenges when gaining approvals for this international project during the pandemic. I therefore discuss the process of setting up this collaborative project and the challenges of conducting cross-border research.

If we could predict who is at risk of post-stroke cognitive problems, is it feasible to discuss risk of dementia at the time of stroke? To address this question I conducted an online survey of sixty UK healthcare professionals who care for patients with stroke. In summary, healthcare professionals were aware of the risk factors for post-stroke dementia (over 90% of respondents selected previous stroke, older age, and global atrophy as risk factors). A majority of healthcare professionals (57%; N=34/60) thought it would be beneficial to inform stroke survivors about risk of dementia, and 75% (N=45/60) thought it would be beneficial for carers to know. However, of the respondents who cared for patients with acute stroke in the past year, the majority (89%; N=47/53) said they rarely/never discussed risk of dementia with their patients.

CONCLUSIONS: In conjunction with other demographic, lifestyle, and vascular risk factors, acute stroke brain scans may assist healthcare professionals in identifying people at risk of post-stroke cognitive problems. Whether stroke survivors would want to know this prognostic information and how to clearly communicate this information requires further research. Electronic health records are a large-scale valuable resource for identifying clinically relevant risk factors of post-stroke dementia.