Predicting post-stroke cognitive impairment using acute CT neuroimaging: A systematic review and meta-analysis

Background Identifying whether acute stroke patients are at risk of cognitive decline could improve prognostic discussions and management. Structural computed tomography neuroimaging is routine in acute stroke, and may identify those at risk of post-stroke dementia or post-stroke cognitive impairment (PSCI). Aim To systematically review the literature to identify which stroke or pre-stroke features on brain computed tomography scans, performed at the time of stroke, are associated with post-stroke dementia or PSCI. Summary of review We searched electronic databases to December 2020. We included studies reporting acute stroke brain computed tomography, and later diagnosis of a cognitive syndrome. We created summary estimates of size of unadjusted association between computed tomography features and cognition. Of 9536 citations, 28 studies (41 papers) were eligible (N = 7078, mean age 59.8-78.6 years). Cognitive outcomes were post-stroke dementia (10 studies), PSCI (17 studies), and one study analyzed both. Fifteen studies (N = 2952) reported data suitable for meta-analyses. White matter lesions (WML) (six studies, N = 1054, OR = 2.46, 95% CI = 1.25-4.84), cerebral atrophy (four studies, N = 558, OR = 2.80, 95% CI = 1.21-6.51), and pre-existing stroke lesions (three studies, N = 352, OR = 2.38, 95% CI = 1.06-5.32) were associated with post-stroke dementia. WML (four studies, N = 473, OR = 3.46, 95% CI = 2.17-5.52) were associated with PSCI. Other computed tomography features were either not associated with cognitive outcome, or there were insufficient data. Conclusions Cognitive impairment following stroke is of great concern to patients and carers. Features seen on visual assessment of acute stroke computed tomography brain scans are strongly associated with cognitive outcomes. Clinicians should consider when and how this information should be discussed with stroke survivors.