A Cohort Study of Environmental Exposures and Metabolic Alterations in Early Pregnancy Loss Using Non-Targeted Analysis

Early pregnancy loss (EPL) is a common reproductive outcome with multifactorial etiologies that remain incompletely understood. To characterize metabolic alterations and potential environmental chemical features associated with EPL, we conducted non-targeted analysis (NTA) using Liquid Chromatography (LC) with electrospray ionization (ESI) coupled to a Quadrupole Time-Of-Flight (QTOF) High-Resolution Mass Spectrometer (HRMS) on maternal plasma samples collected between 6 and10 weeks of gestation (45 EPL, 23 ongoing pregnancy controls). In total, 1,997 ESI + features and 1,031 ESIfeatures differed significantly between EPL and controls (BH-adjusted p < 0.01), and unsupervised analyses revealed clear separation between pregnancy loss and control metabolomic profiles. Recurrent pregnancy loss (RPL, n = 33) samples exhibited largely overlapping metabolic patterns with EPL, indicating shared biochemical perturbations across pregnancy loss subtypes. Among annotated endogenous metabolites, EPL and RPL were characterized by reduced abundances of progesterone and other steroid intermediates, along with elevated medium-chain acylcarnitines and altered glycerophospholipids, consistent with disrupted steroidogenesis and mitochondrial fatty acid β-oxidation. Metabolite correlation networks and pathway enrichment analyses converged on lipid-and energy-related pathways, highlighting coordinated dysregulation of acylcarnitines, oxidized fatty acids, and central energy metabolites. In parallel, confirmed environmental contaminants, including the plasticizer acetyl tributyl citrate and the herbicide metabolite deethylatrazine, were differentially abundant in pregnancy loss samples, and dibutyl phthalate showed biologically plausible correlations with endogenous lipid and steroid metabolites, suggesting potential interactions between environmental exposures and maternal metabolic pathways. XGBoost machine learning models effectively discriminated between EPL and control samples, with feature importance analysis identifying reduced progesterone and lipid-derived metabolites, including 13-oxo-9,11tridecadienoic acid, as key contributors to EPL classification.