Objective To systematically collate and evaluate metabolomics-based biomarkers of ischemic stroke (IS) to guide clinical diagnosis and treatment. Methods Comprehensive literature searches were conducted in PubMed, Embase, and Web of Science using “IS” and “metabolomics” as core keywords, covering publications up through February 2024. Any original metabolomic research related to IS was selected. Key information such as study demographics, study type, objectives, metabolomic analysis methods, and main findings were extracted and analyzed. Frequently mentioned metabolites were subjected to enrichment analysis using the MetaboAnalyst 6.0 platform. Results A total of 51 studies were included. Quality assessment revealed that 54.8% of the diagnostic studies and 69.2% of the prognostic studies were high-quality, with most controlling for confounding factors. Metabolite analysis revealed associations between decreased proline, isoleucine, valine, and alanine levels with IS. Increased tyrosine, glutamine, phenylalanine, sphingomyelin, glutamate, lactate and glucose, and decreased LysoPC (18:2), histidine, and methionine levels were linked to IS onset. Specific metabolite combinations, such as serine, isoleucine, betaine, PC (5:0/5:0), and LysoPE (18:2), showed high precision in predicting acute ischemic stroke (AIS) (training set AUC = 0.988, test set AUC = 0.971). Glycine-serine-threonine and valine-leucine-isoleucine pathways were significant in diagnosing IS and AIS, and in differentiating ischemic and hemorrhagic strokes, as well as identifying post-stroke depression and cognitive impairment. Conclusion This study confirms the potential diagnostic and prognostic value of changes in amino acids and lipids, as well as other metabolites and metabolic pathways, in IS. These findings highlight the promise of metabolomics in IS diagnosis, differential diagnosis, risk assessment, and complication identification. However, further validation is needed due to the varying quality of the included studies. Systematic review registration https://www.crd.york.ac.uk/PROSPERO/#myprospero , identifier CRD42022335505.
Ding et al. (2025) studied this question.