Cerebrospinal fluid lipid profiles as exploratory biomarkers for pediatric meningitis: a proof-of-concept case series

To investigate the cerebrospinal fluid (CSF) lipid metabolite profiles of pediatric patients with purulent meningitis (PM) or viral meningitis (VM) and to explore their differential diagnostic potential.

In this proof-of-concept case series, 13 CSF samples from 10 pediatric patients were analyzed by lipidomics, distributed across four sample-level analytical groups: acute-phase PM (PM_A, n = 3) and recovery-phase PM (PM_R, n = 3) from 3 PM patients with paired sampling, acute-phase VM (VM_A, n = 3) from 3 VM patients, and non-meningitic controls (N, n = 4). Lipid separation and detection were performed by UPLC-MS/MS on a Q Exactive high-resolution mass spectrometer, and data were processed using LipidSearch v.4.1.

Compared with the N group, the PM_A group displayed exploratory lipid alterations with 30 increased and 45 decreased metabolites, predominantly involving sphingolipid (SPH), dihexosylceramide (Hex2Cer), monohexosylceramide (Hex1Cer), lysophosphatidylcholine (LPC), and acylcarnitine (AcCa) species. Elevated Hex2Cer combined with reduced SPH represented a distinctive exploratory PM_A pattern. Receiver operating characteristic (ROC) analysis identified LPC(20:3) and Hex2Cer(d42:3) as exploratory candidate features for distinguishing PM_A from N (apparent AUC = 0.833 for each marker; leave-one-out jackknife AUC range 0.750-1.000, mean 0.833). Multi-feature composite AUCs were not reported because, at n = 7, any composite would be highly susceptible to overfitting. The VM_A group showed limited deviation from N (4 increased, 6 decreased metabolites) but differed from PM_A in 32 increased and 15 decreased metabolites, primarily within SPH, phosphatidylserine (PS), MePC, LPC, and AcCa subclasses. SM(d35:2) emerged as an exploratory candidate distinguishing VM_A from both N and PM_A, but at n = 3 vs. 4 the apparent AUC = 1.000 is statistically expected for some of the 344 screened metabolites by chance alone, and the candidate is therefore hypothesis-generating only.

Cerebrospinal fluid lipidomics revealed disease-stage-specific alterations in pediatric PM and VM. These findings should be regarded as hypothesis-generating in this small proof-of-concept cohort and require validation in larger pediatric studies.