Serum lipidomic alterations associated with post-stroke cognitive impairment: an exploratory prospective study

Post-stroke cognitive impairment (PSCI) is a common complication of stroke that adversely affects long-term functional outcomes and quality of life among survivors. However, reliable biomarkers for early risk prediction remain limited. Dysregulation of lipid metabolism has been implicated in ischemic stroke and may contribute to PSCI development.

This study investigated early serum lipidomic alterations associated with PSCI in patients with acute ischemic stroke and evaluated whether integrating lipidomic features with clinical indicators improves predictive performance.

In this prospective cohort study, 123 patients with acute ischemic stroke were enrolled. Serum samples were obtained within 24 h of stroke onset, and untargeted lipidomic profiling was performed using ultra-high-performance liquid chromatography-high-resolution mass spectrometry. Cognitive function was assessed at 3 months using the Telephone Montreal Cognitive Assessment (T-MoCA). Patients were classified into PSCI and post-stroke cognitively normal (PSCN) groups. Candidate lipids were identified through univariate analysis, multivariable logistic regression, pathway enrichment analysis, and least absolute shrinkage and selection operator (LASSO) regression. Clinical, lipidomic, and combined prediction models were developed and internally validated.

Among the 123 patients, 62 developed PSCI and 61 remained cognitively normal. A total of 1,584 lipid annotations meeting MSI Level 2 criteria were retained for subsequent analyses. Although 87 lipids differed nominally between the PSCI and PSCN groups, none remained significant after false discovery rate correction. After adjustment for clinical confounders, 65 lipids remained nominally associated with PSCI. Pathway enrichment analysis implicated glycerophospholipid metabolism and choline-related phospholipid metabolism. LASSO regression identified LPC 22:6/0:0 as a candidate lipidomic feature. In the internal validation set, the clinical, lipidomic, and combined models yielded areas under the curve (AUCs) of 0.746, 0.623, and 0.763, respectively. The combined model demonstrated favorable discrimination and calibration, as well as potential clinical net benefit.

These findings suggest early serum lipid metabolic alterations in patients who developed PSCI. LPC 22:6/0:0 may be associated with post-stroke cognitive outcomes, and a model integrating clinical and lipidomic features may support early PSCI risk stratification.