Abstract Objective We aimed to establish a high‐sensitivity, multimetabolite rule‐out model for the development of preeclampsia (PE), prioritizing minimizing false negatives to exclude low‐risk individuals from intensive surveillance confidently. Methods In this prospective, nested case–control study, maternal serum samples were collected between 20 +0 and 25 +6 weeks of gestation, before the diagnosis of PE, from the BRISA (Brazilian Ribeirão Preto and São Luís) prenatal cohort. Participants were prospectively followed until delivery, and PE cases were diagnosed during follow‐up. Among 1400 enrolled women, 940 delivered at our institution; 30 were later diagnosed with PE, and 28 were included in the metabolomics analysis. The control group encompassed 28 healthy women, matched for maternal age and body mass index. A targeted metabolomics approach analyzed a 41‐metabolite panel using liquid chromatography coupled with tandem mass spectrometry. Eight metabolites below the detection limit in ≥15% of samples were excluded. Results Compared to the control group, univariate analysis revealed lower methionine and glutamine concentrations and elevated threonine levels in the case group. Classification models based on the remaining 33 metabolites, using leave‐one‐out cross‐validation, achieved 100% sensitivity but had limited specificity (21%). A reduced four‐metabolite model maintained 100% sensitivity with 50% specificity, while our refined, best‐performing six‐metabolite model achieved 100% sensitivity and 61% specificity with a negative predictive value of 100%. Conclusion Our models demonstrated considerable potential as diagnostic tools for ruling out PE, thereby reinforcing their future applicability in screening and monitoring strategies during gestation to improve clinical decision‐making and alleviate the burden on healthcare systems.
Martins et al. (Tue,) studied this question.