Preeclampsia remains one of the most feared complications of pregnancy, a condition that can escalate from mild hypertension to life-threatening seizures within days, and it strikes without warning in women who seemed perfectly healthy at their first prenatal visit. Now a team of researchers in China has reported a new way to peer into the earliest stages of the disease by combining three-dimensional ultrasound imaging of the placenta with maternal clinical data, producing prediction models that they say can identify which pregnancies are likely to develop early-onset and late-onset preeclampsia months before any symptom appears. The study, published in Reproductive Sciences, followed more than 1,600 pregnant women prospectively and converted the results into practical scoring charts, known as nomograms, that clinicians could eventually use at the bedside.
The research, led by Ming Liu and Huifeng Wang of the Affiliated Taian City Central Hospital of Qingdao University, together with colleagues from Taian Hospital of Traditional Chinese Medicine, rests on a simple but powerful biological insight. Preeclampsia is widely understood to originate in the placental bed, the zone where the placenta anchors itself into the wall of the uterus. In healthy pregnancies, specialized cells from the placenta remodel the spiral arteries that feed it, transforming narrow, muscular vessels into wide, low-resistance channels that can deliver the enormous blood flow a growing fetus demands. When this remodeling fails, the placenta is starved of perfusion, becomes stressed, and releases factors into the maternal circulation that damage the mother’s blood vessels, producing the hypertension, protein in the urine, and organ injury that define the disease.
Because the roots of preeclampsia lie in this early vascular failure, the investigators reasoned that the best warning signs should be visible in the first and early second trimesters, long before the placenta’s distress spills over into maternal symptoms. Their study enrolled 1,602 women carrying singleton pregnancies between October 2020 and May 2025. Each participant underwent detailed assessments at two windows of early pregnancy, between 11 and 14 weeks and again between 15 and 18 weeks of gestation. The team collected standard clinical information, measured the pulsatility index of the uterine arteries, a Doppler marker of resistance in the vessels supplying the womb, and acquired three-dimensional power Doppler volumes of the placenta itself.
Those three-dimensional scans yielded a set of quantitative parameters that go far beyond what a conventional two-dimensional image can offer. The researchers measured placental volume, a direct gauge of how much placental tissue has grown; the placental quotient, which relates placental size to gestational age; and the placental blood flow vascularization index, a power Doppler metric that captures the density of blood vessels and the intensity of flow within the placental tissue. By comparing measurements taken at the two time points, they also calculated a placental vascular growth rate, a dynamic measure of how rapidly the placenta’s blood supply was expanding during the critical window when spiral artery remodeling should be under way.
As the pregnancies progressed, the participants were followed until delivery, and the outcomes were sorted according to the gestational age at which preeclampsia developed. Twenty-nine women developed early-onset preeclampsia, the form that appears before 34 weeks and carries the greatest risk to both mother and baby. Fifty-eight developed late-onset disease, which emerges at or after 34 weeks and is more common but generally less severe. The remaining 1,515 women served as normal controls. This trimester-specific framing matters because many researchers now view early-onset and late-onset preeclampsia as related but distinct entities, with early-onset disease more tightly linked to failed placental implantation and late-onset disease involving a greater contribution of maternal cardiovascular factors and placental aging.
To build their prediction models, the team used univariable and multivariable Firth’s logistic regression, a statistical approach designed to perform reliably even when the outcome of interest is rare, as it inevitably is with a condition like early-onset preeclampsia in a general obstetric population. The analysis identified chronic hypertension as an independent risk factor for both forms of the disease, confirming the well-established link between pre-existing maternal cardiovascular disease and preeclampsia. In contrast, higher values of the placental perfusion parameters, including placental volume, placental quotient, and the vascularization index, acted protectively, consistent with the idea that a well-perfused, vigorously growing placenta is far less likely to trigger the maternal syndrome.
The performance of the resulting models was striking, particularly for the early-onset form. The combined model, which integrated maternal clinical characteristics with uterine artery pulsatility index and the three-dimensional placental parameters, achieved an area under the receiver operating characteristic curve of 0.980 for early-onset preeclampsia, with a 95 percent confidence interval of 0.968 to 0.992. An area under the curve of that magnitude indicates near-perfect discrimination within the study cohort, meaning the model separated women who developed early-onset disease from those who did not with remarkable accuracy. For late-onset preeclampsia, the combined model achieved an area under the curve of 0.827, with a confidence interval of 0.765 to 0.889, a level of discrimination that is strong for a condition known to be harder to predict and more heterogeneous in its origins.
Discrimination alone does not make a prediction model clinically useful, so the researchers subjected their models to a battery of validation tests. Bootstrap-based internal validation, in which the model is repeatedly refitted on resampled versions of the same cohort, confirmed acceptable calibration, meaning the predicted probabilities tracked reasonably well with the observed rates of disease. Decision curve analysis, a technique that quantifies the net benefit of acting on a model’s predictions across a range of risk thresholds, demonstrated positive net clinical benefit, suggesting that using the model to guide decisions would help more patients than it would harm compared with treating everyone or no one. From the optimal combined models, the team constructed static nomograms, graphical scoring tools that allow a clinician to plot an individual patient’s values for each predictor and read off a personalized probability of developing early-onset or late-onset preeclampsia.
The practical implications could be substantial. Preeclampsia affects a significant share of pregnancies worldwide and imposes enormous costs on health systems, both through acute maternal and neonatal complications and through the elevated long-term cardiovascular risk that affected women carry for the rest of their lives. Low-dose aspirin started early in pregnancy can reduce the risk of preterm preeclampsia, but only if the women who need it can be identified in time. Current first-trimester screening approaches, which combine maternal history with blood pressure, uterine artery Doppler, and serum biomarkers, perform well in some settings but have shown inconsistent results across populations, and many published prediction models have never been validated outside the cohorts in which they were built. A tool based on three-dimensional placental imaging offers a potentially direct window into the pathophysiology itself, measuring the health of the placental bed rather than relying solely on downstream markers.
The authors are careful to note the limits of their work. The study was conducted at a single center, and although internal validation by bootstrap resampling supported the models’ apparent performance, the generalizability of the results to other populations, scanners, and clinical settings remains unclear. External validation in independent cohorts is required before the nomograms can be deployed in routine antenatal screening. The relatively small number of early-onset cases, twenty-nine in total, also means the estimates carry uncertainty, even with the use of regression methods suited to rare outcomes. Still, the study represents one of the most complete attempts to date to fuse dynamic three-dimensional placental perfusion imaging with trimester-specific modeling of both early-onset and late-onset disease. If external validation confirms what this cohort suggests, the humble ultrasound machine, already present in nearly every prenatal clinic on earth, could become a far sharper instrument for spotting, and ultimately preventing, one of pregnancy’s most dangerous complications.
Subject of Research: Trimester-specific prediction of early- and late-onset preeclampsia using 3D placental bed perfusion ultrasound parameters and maternal clinical characteristics
Article Title: 3D Placental Bed Perfusion Ultrasound: Development and Internal Validation of Trimester-Specific Prediction Models for Early- and Late-Onset Preeclampsia
Article References: Liu, M., Zhou, Z., Li, J., Wang, H., Yu, L., Qi, C., & Wang, H. (2026). 3D Placental Bed Perfusion Ultrasound: Development and Internal Validation of Trimester-Specific Prediction Models for Early- and Late-Onset Preeclampsia. Reproductive Sciences. https://doi.org/10.1007/s43032-026-02234-0
Image Credits: AI Generated
DOI: 10.1007/s43032-026-02234-0
Keywords: preeclampsia, early-onset preeclampsia, late-onset preeclampsia, 3D ultrasound, placental perfusion, uterine artery pulsatility index, placental volume, prediction model, nomogram, pregnancy, obstetrics, Doppler imaging
News Source: Ophelia Keating. (October 6, 2026). 3D Placental Ultrasound Models Predict Preeclampsia Weeks Before Symptoms Appear. Scienmag.
