Novel non-invasive model may help predict preeclampsia before symptoms

In a recent study published in the journal Nature Medicine, researchers developed a novel clinical diagnostic screening test to identify pregnancies at risk of developing preeclampsia (PE), a potentially life-threatening complication characterized by hypertension.

Cell-free DNA methylome analysis for early preeclampsia prediction
Study: Cell-free DNA methylome analysis for early preeclampsia prediction. Image Credit: Petrovich Nataliya/

They used methylation differences in plasma-derived cell-free DNA (cfDNA) to stratify PE risk and diagnose the condition at 12 weeks of gestation, significantly earlier than conventional diagnostic tools. A model developed using this approach was able to predict early-onset PE with 72% accuracy and 80% specificity, making it a promising tool to combat this peripartal condition in the future.

Preeclampsia and the need for early accurate diagnosis

Preeclampsia (PE) is a common and potentially life-threatening peripartal complication, identified as a leading cause of morbidity and mortality during pregnancy. The World Health Organization (WHO) estimates that 2–10% of pregnancies globally suffer from PE, with incidences higher in developing countries than their developed counterparts.

Two main types of PE have been identified – early-onset PE, which manifests before 34 weeks of gestation, and late-onset PE, which presents after the 34-week period. Both PE types display overlapping symptoms, but early-onset PE is far more dangerous, with an eightfold higher risk of morbidity and fetal mortality than late-onset PE.

Identifying PE early in the gestation period is thus imperative for initiating clinical interventions to treat the condition during the critical period (before 16 weeks of gestation).

Unfortunately, conventional diagnostic procedures are insensitive or rely on operator experience, wherein sensitivity depends on sonographers’ skill. While PE’s etiopathogenesis (causal factors) remain poorly understood, the placenta has been identified as playing a vital role in disease pathophysiology, especially for early-onset PE.

Chorionic villus (tiny projections of placental tissue) samples from pregnant women suggest that PE pathogenesis precedes observable symptoms and occurs during the first trimester. Chorionic villus sample, while effective for PE diagnosis, is invasive and has been associated with miscarriage risk, making its use suboptimal.

Research has explored cell-free DNA (cfDNA) profiling as a minimally invasive alternative to Chorionic villus sampling. Obtained from the maternal plasma, placenta-derived cfDNA at 10-12 weeks of gestation comprises about 10% of maternal cfDNA. Maternal hematopoietic cells are the primary source of cfDNA. Notably, cfDNA depicts both placental genotypes and epigenetic information, including DNA methylation (DNAme) and nucleosome positions, which studies suggest can be used for PE diagnosis and risk profiling during the first and second trimesters.

About the study

The present study aims to investigate if cfDNA methylation (cfDNAme) can be used to diagnose and assess PE risk in pregnant women during the critical initial 16 weeks of gestation, enabling potential clinical intervention. Researchers began by collecting cfDNA from conventionally diagnosed PE patients (case-cohort; n = 64) and gestation-matched controls (control-cohort; n = 38).

This cfDNA was analyzed using target-enrichment bisulfite sequencing to identify regions of interest depicting low DNAme in blood. Analysis revealed 35,411 regions, of which 28,898 were retained following quality filtering. These regions were subject to differential methylation analysis to identify hypo- and hypermethylation. To understand methylation sources, cfDNAme from peripheral blood samples was analyzed.

Researchers then built an unsupervised model to hierarchically cluster differentially methylated regions from both cfDNA and DNA from blood and placentas. Using an elastic net regularized regression analysis, the model was trained on cfDNAme data to separate cases and controls.

To evaluate whether DNAme could be observed before PE symptom manifestation, researchers collected and analyzed cfDNA from women in their first trimester (n = 199), 75 of which manifested early-onset PE symptoms later.

Clinical parameters, including gestational age at sampling, fetal fraction, maternal age, body mass index (BMI), parity, method of conception, ethnicity, diabetes and smoking, were matched between case and control groups.

De Borre et al. (2023)

Researchers then build a tenfold cross-validated model to predict a PE risk metric (score) using cfDNAme data for 1,542 regions of interest obtained before or after 12 weeks of gestation. To identify sources of signals predicting PE, hierarchical model clustering was employed.

Finally, model validation was done using an external dataset comprising 197 randomized, unselected pregnancies. In tandem with logistic regression analysis, the extended prior risk model (ePRM) was used to assess model prediction sensitivity.

Study findings

Of the 28,898 differentially methylated regions retained after quality filtering, 2,068 and 1,547 regions depicted hypo- and hypermethylation. Of these, 132 and 99 regions, respectively, passed multiple testing corrections and were used for model training.

Peripheral blood DNA analyses revealed that cfDNAme changes were associated more strongly with placenta-derived than blood cell-derived DNA, with changes occurring predominantly near development-related genes (placental blood vessel formation, trophoblast differentiation, and placental maturation).

The unsupervised model was found to cluster case and control cohorts separately using both cfDNA and placental- or blood cell-derived DNA.

A tenfold cross-validation analysis revealed robust performance (area under the curve (AUC) = 0.979, 95% confidence interval (CI): 0.956–1.000), indicating that cfDNAme profiling can confirm a PE diagnosis. Notably, 54 of 64 patients with PE received steroids to accelerate fetal lung maturation before cfDNA sampling, but this did not affect PE-associated cfDNAme changes.

De Borre et al. (2023)

The final trained model used 1,392 regions to classify blood and placenta-derived DNA, with results suggesting that placental DNA was a more accurate predictor of early-onset PE (AUC = 0.678) than blood (AUC = 0.527). This implies that the model can identify and use PE-associated cfDNAme signals from the placenta to diagnose PE in pregnancies.

Trained model analyses of cfDNA samples obtained from women before (n = 38), during (n = 112), or after (n = 49) 12 weeks of pregnancy revealed that, while all cohorts showed higher cfDNAme scores for cases compared to controls, results from week 12 onwards were significant and capable of identifying early-onset PE (AUC = 0.792 for week 12 versus 0.629 before week 12).

This suggested that the trained model can be used to predict and diagnose PE in women pre-symptomatically by week 12 of gestation, earlier than conventional diagnostic approaches.

External model validation using ePRM and logistic regression elucidated that the trained model had 72% accuracy and 80% specificity, more than sufficient for clinical diagnostic use.


Preeclampsia, despite being a common and severe condition, is hard to study because it only occurs spontaneously in primates. Conventional in vitro methodologies and in vivo murine models cannot be recruited to test PE hypotheses, forcing researchers to focus on etiopathogenetic PE studies conducted on human PE patients.

Due to the scarcity of research into PE, conventional diagnostic procedures are highly invasive with miscarriage risk or subject to high sonographer experience bias. Furthermore, most traditional diagnostic approaches are of low sensitivity, especially preceding PE symptomatic presentation.

Clinical evidence shows that prophylaxis can reduce the incidence of preterm PE by 62% and of early-onset PE by 82%, but risk estimates should be available well before 16 weeks of gestation because prophylaxis initiated after this time window is ineffective.

​​​​​​​De Borre et al. (2023)

In the present study, researchers employed differential methylation screening of cfDNA obtained from the plasma of pregnant women, a minimally invasive procedure, to build a model capable of accurately predicting PE and evaluating PE risk as early as 12 weeks of gestation.

The model was found to have an accuracy of 72% and a specificity of 80%, significantly higher than present diagnostic approaches and without the demerit of operator bias.

Our study thus paves the way for prospective cohort studies to accurately assess performance. As such, the predictive power of the cfDNAme profiling we describe holds promise for improving detection of PE—and potentially other disorders—and may thus enable comprehensive monitoring of the health of mother and child during pregnancy.

De Borre et al. (2023)

Journal reference:
Hugo Francisco de Souza

Written by

Hugo Francisco de Souza

Hugo Francisco de Souza is a scientific writer based in Bangalore, Karnataka, India. His academic passions lie in biogeography, evolutionary biology, and herpetology. He is currently pursuing his Ph.D. from the Centre for Ecological Sciences, Indian Institute of Science, where he studies the origins, dispersal, and speciation of wetland-associated snakes. Hugo has received, amongst others, the DST-INSPIRE fellowship for his doctoral research and the Gold Medal from Pondicherry University for academic excellence during his Masters. His research has been published in high-impact peer-reviewed journals, including PLOS Neglected Tropical Diseases and Systematic Biology. When not working or writing, Hugo can be found consuming copious amounts of anime and manga, composing and making music with his bass guitar, shredding trails on his MTB, playing video games (he prefers the term ‘gaming’), or tinkering with all things tech.


Please use one of the following formats to cite this article in your essay, paper or report:

  • APA

    Francisco de Souza, Hugo. (2023, August 31). Novel non-invasive model may help predict preeclampsia before symptoms. News-Medical. Retrieved on December 10, 2023 from

  • MLA

    Francisco de Souza, Hugo. "Novel non-invasive model may help predict preeclampsia before symptoms". News-Medical. 10 December 2023. <>.

  • Chicago

    Francisco de Souza, Hugo. "Novel non-invasive model may help predict preeclampsia before symptoms". News-Medical. (accessed December 10, 2023).

  • Harvard

    Francisco de Souza, Hugo. 2023. Novel non-invasive model may help predict preeclampsia before symptoms. News-Medical, viewed 10 December 2023,


The opinions expressed here are the views of the writer and do not necessarily reflect the views and opinions of News Medical.
Post a new comment
You might also like...
Ohio State Wexner Medical Center wins $14 million award for aspirin dose study