Abstract
BACKGROUND: Limited understanding of the pathophysiology of cerebral small vessel disease (CSVD) has hampered the development of effective treatments. Lipidomics offers a promising approach for identifying molecular signatures, clarifying underlying pathogenic mechanisms, and predicting disease severity and progression. METHODS: A total of 1161 participants with lipidomic data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database were included and matched to each neuroimaging marker of CSVD separately, including cerebral microbleeds (CMBs, n = 578), white matter hyperintensities (WMHs, n = 650), lacunes (n = 1125), and CSVD burden (n = 546). Three complementary classification strategies (class-based grouping, model-based clustering, and individual lipid species) were employed to investigate lipid signatures across various CSVD markers. A multimodel regression framework followed by a series of sensitivity analyses was used to further identify lipid species showing robust associations with each CSVD marker. RESULTS: A total of 46 lipid classes, 25 lipid clusters, and 749 lipid species were quantified. Individuals with CSVD presented broadly reduced plasma lipid levels, particularly those of glycerophospholipids and glycerolipids. A multimodel regression framework initially screened 32 lipid species associated with the presence or progression of CSVD. Subsequent sensitivity analyses narrowed these to 13 robust species, including phosphatidylcholine, triacylglycerol, phosphatidylethanolamine, alkenyl-phosphatidylethanolamine, and sphingomyelin. Among these, elevated levels of PC(36:4)[+ OH] and TG(56:6)[NL-20:4] were associated with lower odds of CMBs and reduced WMH volumes, respectively, whereas higher levels of PC(15:0_20:3) and SM(d16:1/19:0) were linked to lower odds of lacunes, with all associations consistently observed across baseline, year 1, and year 2. CONCLUSION: This study revealed dysregulated lipid metabolism across distinct magnetic resonance imaging phenotypes of CSVD and revealed multiple lipid species that are consistently associated with the presence and progression of these phenotypes, underscoring the potential of lipidomics for the earlier identification and prevention of CSVD and informing future diagnostic development and mechanistic studies. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1186/s12944-026-02891-9.