Post-stroke cognitive impairment: More than a lesion-symptom model.

Theoretical neuropsychology has traditionally investigated brain-behaviour relationships by measuring the extent to which individual cognitive impairments map onto distinct lesion correlates. However, individual post-stroke cognitive impairments (PSCI) rarely occur in isolation. In the current study, we employ multivariate analysis techniques to explore the extent to which individual, multi-domain patterns of PSCI can be distinguished into distinct cognitive profiles and the extent to which these profiles are associated with lesion anatomy. Latent Class Analysis (LCA) was conducted on domain-specific cognitive screening data from a representative stroke cohort (n = 2172). Voxel-level and network-level lesion mapping was used to identify lesion correlates of identified PSCI profiles. In addition, the association of the PSCI profiles with general brain health (e.g., atrophy severity, white matter integrity), and demographic characteristics was investigated. LCA identified two viable cognitive class solutions: a 5-class model and a 13-class model. The 5-class solution distinguished classical lateralised stroke deficits (e.g., aphasia, neglect) alongside a minimal impairment and non-lateralised global impairment profile. In contrast, the 13-class solution provided finer-grained differentiation, particularly for non-lateralised cognitive profiles which were more strongly associated with premorbid health and education level. Importantly, lesion anatomy alone could not fully account for class distinctions. While lesion location was predictive, particularly, in hyper-acute stages, profiles for patients tested 2 weeks post-stroke revealed less influence of lesion location and more of lesion volume. These findings provide a novel, multivariate conception of PSCI, which establishes the theoretical groundwork necessary to support future translational research aimed at improving clinical care and predicting cognitive trajectories.