O-089 Maternal plasma cell-free RNA as a liquid biopsy for first-trimester screening of early and late-onset preeclampsia

Abstract

Study question Could liquid biopsy using maternal plasma cell-free RNA (cfRNA) serve as early screening for early-onset preeclampsia (EOPE) and late-onset preeclampsia (LOPE)? Summary answer Distinct cfRNA profiles identified early in pregnancy allow the accurate risk-prediction of both EOPE and LOPE months before their clinical diagnosis. What is known already Screening efforts to predict preeclampsia have traditionally focused on placental biomarkers, such as sFLT1 and PlGF, which are effective for short-term prediction when preeclampsia is clinically suspected, but have significant limitations in accuracy and broad application. Liquid biopsy-based cfRNA studies offer valuable insights for early prediction, though these findings remain unimplemented clinically. Recently, we conducted a multi-omic analysis of decidualization resistance (DR)—a defect in endometrial cell differentiation—in patients with prior severe preeclampsia, identifying biomarkers that could enable earlier diagnosis1. 1Nat Med (2025). https://doi.org/10.1038/s41591-024-03407-7 Study design, size, duration This prospective longitudinal study involves 9,586 pregnant women recruited from 14 hospitals across Spain between September 2021 and June 2024 (ClinicalTrials.gov: NCT04990141). Participants provided 20 mL of peripheral blood at three time points: T1: 9–14 weeks, T2: 18–28 weeks, T3: >28 weeks or at the time of preeclampsia diagnosis using ACOG/FIGO guidelines. Plasma cfRNA was analyzed in EOPE (n = 42) and LOPE (n = 43) cases compared to normotensive controls that had uncomplicated pregnancies (n = 75). Participants/materials, setting, methods cfRNA was isolated using the MiRNeasy Serum/Plasma Advanced Kit from plasma samples (n = 457) collected throughout pregnancy from the selected participants (n = 160) and sequenced using Illumina technology. Participants were divided into discovery (70%) and validation (30%) sets to develop and validate the predictive models. Lasso regression was used to select the relevant cfRNAs and algorithms were evaluated using leave-one-out cross-validation. The algorithm with the best F1-score was selected and validated in the hold-out set. Main results and the role of chance We developed a first-trimester predictive model for EOPE using 36 transcripts, achieving 83% sensitivity, 88% specificity and an AUC of 0.85 in the validation sample set, and detecting risk 18.0 weeks before onset. Of these, 47.2% were associated with DR signature (e.g. PAEP, MMP7). For the second trimester, the most effective EOPE model used 87 cfRNA transcripts achieving 87% sensitivity, 84% specificity, and an AUC of 0.85 in the validation set, approximately 8.5 weeks before onset. The 36.8% of transcripts included in the model were associated with the DR signature (e.g. IGF1, WNT5A). For LOPE, the best model included 92 cfRNAs, achieving 86% sensitivity, 89% specificity and an AUC of 0.88 in the hold-out set, detecting risk 14.9 weeks before onset and just included two cfRNAs associated with DR (HES4 and SPEF1). Thus, models were enriched with DR-associated cfRNAs, while LOPE models reflected broader systemic contributions, highlighting distinct biology between subtypes. Additionally, at diagnosis, 24,336 transcripts were significantly altered in EOPE (FDR < 0.05) and 11,859 in LOPE (FDR < 0.05) compared to controls, revealing enriched processes linked to fetal and maternal organ-specific damage, with an exacerbated proinflammatory state defining EOPE. Limitations, reasons for caution To ensure the generalizability of the predictive models presented before their clinical application, this study requires external validation, which is currently underway (ClinicalTrials.gov: NCT06716242) Wider implications of the findings We provide a reliable, standardized screening tool based on liquid biopsy that improves preeclampsia risk screening accuracy and efficiency. DR emerges as a key feature in EOPE prediction, distinguishing its pathophysiology from LOPE. These insights enable improved subtype differentiation and tailored interventions to reduce preeclampsia risks during pregnancy. Trial registration number Yes