Fetal Fraction in Pregnancy Loss Reflects Maternal-Fetal Physiologic Integration–a prospective cohort study in the Copenhagen Pregnancy Loss Cohort

STUDY QUESTION Can maternal biomarkers of maternal-fetal physiologic integration predict the fetal fraction of cell-free DNA in pregnancy loss? SUMMARY ANSWER Maternal organ system biomarkers, including cardiometabolic, hepatic, renal and thyroid markers, substantially predict fetal fraction in pregnancy loss (Area Under the Receiver Operating Characteristics Curve (AUC-ROC) 0.77–0.80), indicating that fetal fraction may reflect integrated maternal-fetal physiology. WHAT IS KNOWN ALREADY Fetal fraction, which is the proportion of cell-free DNA in maternal plasma derived from the placenta, increases with gestational age and decreases with higher maternal BMI. Existing prediction models on fetal fraction rely mainly on demographics and placental hormones, and the contribution of broader maternal organ-system biomarkers remains unclear. STUDY DESIGN, SIZE, DURATION Prospective multi-center cohort study within the Copenhagen Pregnancy Loss study with development cohort (n = 992) from a tertiary university hospital and external validation cohorts from two secondary university hospitals (n = 275 and n = 241). Enrollment was from November 2020 to February 2023. PARTICIPANTS/MATERIALS, SETTING, METHODS Women ≥18 years with confirmed intrauterine pregnancy loss before 22 weeks at three Danish university hospitals were included. Fetal fraction was estimated from cfDNA sequencing using SeqFF. Six clinical characteristics and 36 biochemistry measurements, selected to reflect physiological domains implicated in placentation including thyroid, renal, lipid, and inflammatory markers, were analyzed using Bayesian Hurdle lognormal regression and gradient boosting models with external validation. MAIN RESULTS AND THE ROLE OF CHANCE In adjusted hurdle models, higher total thyroid hormone levels were associated with adequate fetal fraction (e.g., total triiodothyronine (TT3) Odds Ratio (OR) 2.76 (95% CI 1.31–5.87); total thyroxine (TT4) OR 2.42 (95% CI 1.43–4.29)), while higher levels of creatinine and uric acid were inversely associated (e.g., creatinine OR 0.39 (95% CI 0.17–0.91); uric acid OR 0.47 (95% CI 0.26–0.90)). The magnitude of fetal fraction was additionally associated with HDL cholesterol/total cholesterol ratio, uric acid and creatinine. A gradient-boosting model achieved AUC-ROC 0.77 in internal cross-validation and 0.80/0.78 in external validation (two secondary university hospitals), with aspartate transaminase (AST)/alanine transaminase (ALT) ratio among the strongest predictors alongside beta subunit of human chorionic gonadotropin (βhCG) and gestational age. LIMITATIONS, REASONS FOR CAUTION The cross-sectional design and pregnancy loss focus limit inference regarding the directionality between maternal biomarkers, placental health and fetal fraction. SeqFF-based fetal fraction estimation adds measurement noise versus genotype methods, while our pre-specified biomarker panel limits scope to investigated parameters. Our findings require validation in ongoing pregnancies. WIDER IMPLICATIONS OF THE FINDINGS The findings suggest fetal fraction may reflect maternal-fetal physiologic integration rather than merely a technical quality-control parameter. The associations between fetal fraction and cardiometabolic, thyroid, hepatic and renal markers suggest that fetal fraction may offer broader insights into maternal physiology and maternal organ system adaptation during pregnancy, influencing cell-free DNA dynamics (reflecting placental shedding, maternal clearance and dilution). The model enables personalized clinical decision-making regarding cell-free fetal DNA test timing in pregnancy loss and thereby improving the efficiency and success rate of cell-free fetal DNA testing. FUNDING The preparation of our cohort was funded by the Ole Kirks Foundation, while the study itself received support from the BioInnovation Institute Foundation (BII21SG1020554, NNF20SA0066125, and NNF15OC0016662), the Novo Nordisk Foundation (NNF23OC0087269, NNF22OC0077221), the A.P. Moller Foundation, the Independent Research Fund Denmark (6215-00037B), and the Pioneer Center for Statistical and Computational Methods for Advanced Research to Transform Biomedicine (SMARTbiomed), DNRF grant number P4. DISCLOSURES HSN has received scientific grants paid to her institution from Freya Biosciences, Ferring Pharmaceuticals, BioInnovation Institute, Ministry of Education, Novo Nordisk Foundation, Augustinus Fonden, Oda og Hans Svenningsens Fond, Demant Fonden, and Ole Kirks Fond. HSN has received personal payments or honoraria for lectures and presentations from Ferring Pharmaceuticals, Merck, AstraZeneca, Cook Medical, Gedeon Richter, IBSA Nordic, Novo Nordisk, and Bessins. SB declares speaker’s fees from Novo Nordisk and Extempore. All other authors declare no conflict of interests. This article is the subject of a patent (application number WO2025181337A1) filed by the Capital Region of Denmark, where TSH, DW, and HSN are named inventors. TRIAL REGISTRATION NUMBER N/A