Prediction of gestational age using urinary metabolites in term and preterm pregnancies

Abstract

Assessment of gestational age (GA) is key to provide optimal care during pregnancy. However, its accurate determination remains challenging in low- and middle-income countries, where access to obstetric ultrasound is limited. Hence, there is an urgent need to develop clinical approaches that allow accurate and inexpensive estimations of GA. We investigated the ability of urinary metabolites to predict GA at time of collection in a diverse multi-site cohort of healthy and pathological pregnancies (n = 99) using a broad-spectrum liquid chromatography coupled with mass spectrometry (LC-MS) platform. Our approach detected a myriad of steroid hormones and their derivatives including estrogens, progesterones, corticosteroids, and androgens which were associated with pregnancy progression. We developed a restricted model that predicted GA with high accuracy using three metabolites (rho = 0.87, RMSE = 1.58 weeks) that was validated in an independent cohort (n = 20). The predictions were more robust in pregnancies that went to term in comparison to pregnancies that ended prematurely. Overall, we demonstrated the feasibility of implementing urine metabolomics analysis in large-scale multi-site studies and report a predictive model of GA with a potential clinical value.

Figure 1

Study design and cohort characteristics. (a) Urine samples from 99 pregnant women were collected across 5 sites and analyzed using a broad-spectrum metabolomics LC–MS platform. The sources of the images are described in the Methods section. (b) GA at collection and at delivery across the collection sites. Urine samples were collected early in pregnancy 8–19 weeks and 49 women delivered at term (> 37 weeks’ GA) and 50 women delivered preterm (≤ 37 weeks’ GA). (c) Structural categorization of detected urine metabolites according to the “Superclass level” of the ClassyFire classification system.

Figure 2

Prediction of gestational age at time of collection and associated biological processes. (a) Performance of the restricted RF prediction model of GA that uses three metabolites (C19H26O7S, C24H30O9 and estriol glucuronide) and all the samples in the study (n = 99). The model was validated in an independent cohort (n = 20). The blue area represents the 95% confidence interval. (b) Principal component analysis using predictive metabolites (P-value < 0.05). PC1 and PC4 were chosen because they associated the most strongly with GA. (c) KEGG metabolic pathway enrichment analysis. (d) Volcano plot of annotated significant metabolites (P-value < 0.05). Beta coefficients were calculated using a linear modeling and P-values were calculated from Spearman correlations. (e) Top 6 metabolites in the predictive model and LOESS fit across all the samples. The grey area represents the 95% confidence interval.

Figure 3

Prediction of gestational age at time of collection in term and preterm pregnancies. (a) Distribution of GA at sample collection in term (n = 49) and preterm pregnancies (n = 50). (b) Performance of the RF prediction models of GA in term and preterm deliveries. (c) P-values of selected metabolites in term and preterm RF models. Metabolites that are most predictive tend to be significant in both models. The top 10 metabolites are represented in red. (d) Coefficient of variation of the top 10 metabolites across GA ranges.

Figure 4

Biological processes associated with term and preterm prediction models. (a) KEGG metabolic pathway enrichment analysis using metabolites selected in the term and preterm RF models (P-value < 0.05). (b) Venn diagram of validated metabolites predictive of GA (P-value < 0.05) in term and preterm models. (c) Pairwise spearman correlation network. Nodes were color-coded by model significance and their size represents the betweenness centrality.