Untargeted plasma lipidomics reveals alterations in steroid and lipid metabolism in fibromyalgia

Abstract Fibromyalgia (FM) is a chronic pain condition affecting approximately 2–4% of the general population, characterized by a marked predominance in females; however, no single etiologic target or mechanism has been clearly established. In this study, an untargeted lipidomics approach was applied to comprehensively profile plasma lipid species in FM patients and healthy controls (C). A total of 54 participants were enrolled (31 with FM and 23 C), of whom 52 (29 FM and 23 C) were included in the final lipidomic analysis following the loss of two FM samples during experimental procedures. Plasma lipid profiles were acquired using an untargeted ultra high-performance liquid chromatography–ion mobility–quadrupole time of flight-tandem mass spectrometry (UHPLC–ion mobility–QTOF-MS/MS) platform in both positive and negative ionization modes. The dataset was subsequently normalized and explored by principal component analysis. Discriminant features were identified using recursive feature elimination with logistic regression (LR) as the base estimator, embedded within a repeated nested cross-validation framework to minimize overfitting and information leakage. Fourteen informative lipid species were identified through this procedure, and the optimized LR pipeline subsequently yielded a parsimonious five-lipid panel consisting of TG(18:1/18:2/20:4), SM(d18:2/24:1), SM 41:2, hexosylceramide (HexCer)(d18:2/24:1), and pregnanetriol sulfate. Repeated nested cross-validation demonstrated moderate and reproducible classification performance, with mean outer-fold AUC values of 0.709 for LR, 0.709 for SVM, and 0.699 for gradient boosting. Participant-level aggregated out-of-fold predictions yielded AUCs of 0.759 (95% CI 0.624–0.876) for LR, 0.823 (95% CI 0.706–0.929) for SVM, and 0.765 (95% CI 0.623–0.892) for gradient boosting. Following Benjamini–Hochberg correction across all 227 detected lipid species, pregnanetriol sulfate was the only lipid that remained statistically significant (FDR q = 0.0455), showing lower abundance in FM patients than in controls. In addition, several lipid species, including PE(16:1/20:4), SM 41:2, SM(d18:2/24:1), PC(18:1/18:2), and PC(18:0/20:2), showed nominal differences between groups but did not remain statistically significant after FDR correction. Collectively, our findings suggest subtle alterations in steroid and lipid metabolism in fibromyalgia. Lower plasma pregnanetriol sulfate, the only lipid remaining significant after FDR correction, may reflect alterations in steroid metabolism, while nominal changes in selected phospholipid and sphingolipid species may indicate differences in membrane lipid metabolism. Although validation in larger independent cohorts is required, the identified lipid panel provides candidate markers for further investigation and molecular characterization of fibromyalgia.