Bayesian analysis clarifies preeclampsia prevention timing

A post-hoc Bayesian analysis of the PREVENT-PE randomised trial has provided new insights into the effectiveness of risk-stratified timed birth for preventing term preeclampsia. Published in eClinicalMedicine in November 2026, the study examined data from 4,734 participants recruited at King’s College Hospital in London. The analysis focused on how the timing of planned birth, determined by preeclampsia risk assessments at 35 to 36 weeks of gestation, influenced outcomes across different risk strata. While the parent trial demonstrated an overall reduction in preeclampsia, this exploratory analysis sought to quantify the uncertainty and potential benefits or harms within specific delivery windows.

Preeclampsia is a serious complication of pregnancy associated with approximately 46,000 maternal deaths and 500,000 perinatal deaths globally each year. Although preterm preeclampsia carries higher risks of adverse events, term preeclampsia is approximately four times more common and accounts for the majority of maternal complications and a high proportion of perinatal complications. The PREVENT-PE trial was the first randomised controlled trial to investigate the impact of screening for preeclampsia risk at 35 to 36 weeks, followed by stratification of participants with a risk of at least 1 in 50 to timed birth at 37, 38, 39, or 40 weeks. This intervention strategy aimed to prevent the onset of term preeclampsia by delivering the baby before the condition could develop.

The current analysis was restricted to participants from King’s College Hospital, a single site within the broader trial. It included all 2,871 intervention participants and 1,863 control participants who had stored serum samples and haemodynamic measurements necessary to calculate their 35 to 36-week preeclampsia risk. Of the 2,881 women randomised to the control group at this site, eight were lost to follow-up, and 1,010 were excluded because the required measurements were unavailable. The final sample comprised 4,734 participants, with baseline characteristics broadly comparable to the full trial population. The median age of participants was 34.2 years, and the most commonly self-reported ethnic group was White, accounting for 71.3% of the cohort.

Risk stratification in the intervention group was based on the Fetal Medicine Foundation competing-risks model, which uses maternal characteristics, medical history, mean arterial pressure, and serum concentrations of placental growth factor and soluble fms-like tyrosine kinase-1. Participants were assigned to specific delivery windows based on their calculated risk: those with a risk of at least 1 in 2 were offered birth at 37 weeks, while those with lower risks were scheduled for 38, 39, or 40 weeks. Women with a risk below 1 in 50 and all control participants received usual care, with the timing of birth determined by spontaneous labour, clinical indications, or maternal request. For control participants, the same risk strata were reconstructed retrospectively using stored samples, allowing for within-stratum comparisons of benefit and harm.

The primary outcome of the analysis was birth with preeclampsia, defined by the 2021 International Society for the Study of Hypertension in Pregnancy criteria. Overall, birth with preeclampsia occurred in 117 (4.1%) of 2,871 intervention participants and 106 (5.7%) of 1,863 controls. The posterior median incidence ratio was 0.72, with a 95% credible interval of 0.56 to 0.93, indicating a 28% reduction in incidence in the intervention group. The posterior probability of benefit was 0.995, suggesting a high likelihood that the intervention reduced the risk of preeclampsia. This estimate was consistent in direction and magnitude with the adjusted frequentist estimate from the full PREVENT-PE trial, which reported an approximately 30% reduction in the primary outcome.

When examining specific risk strata, the results revealed varying levels of precision and certainty. At 39 and 40 weeks, the estimates favoured the intervention and were relatively precise. These strata comprised 846 (17.9%) of the participants and accounted for 41.3% of all preeclampsia cases. The posterior median incidence ratios were 0.60 at 39 weeks and 0.45 at 40 weeks, with posterior probabilities of benefit of 0.985 and 0.992, respectively. These findings suggest that timed birth at these later gestational ages is associated with a substantial reduction in preeclampsia incidence.

In contrast, the 37-week and 38-week strata, which represented the highest-risk groups, showed substantially more uncertainty. These strata comprised only 182 (3.8%) of the participants, with small control-group sizes of 26 and 51 participants, respectively. The posterior probabilities of benefit were 0.463 at 37 weeks and 0.622 at 38 weeks, and the credible intervals included both potentially important benefit and harm. The intervention resulted in materially earlier delivery in these strata, with 93.0% of intervention pregnancies at 37 weeks delivered at or before 37+6 weeks, compared with 38.5% in the control group. However, the imprecise estimates mean that the balance between preventing preeclampsia and the consequences of early-term delivery remains unclear for these highest-risk women.

Secondary outcomes included emergency caesarean section and neonatal unit admission for at least 48 hours. Overall, there was no significant difference in emergency caesarean rates between the intervention and control groups, with a posterior median incidence ratio of 1.03. In the 39-week stratum, the posterior probability of harm was 0.857, although the credible interval was wide, indicating uncertainty. For neonatal unit admission, the overall posterior median incidence ratio was 0.97, with a posterior probability of benefit of 0.591. In the 37-week and 38-week strata, the posterior probabilities of harm were 0.746 and 0.825, respectively, but the estimates were imprecise and did not establish a definitive increase in neonatal risk. The stratum-specific data were insufficient to establish neonatal benefit or harm in any individual stratum.

The authors highlight several limitations of the study. As a post hoc exploratory analysis, it was not prespecified in the trial protocol, and the numbers of participants and events were small in some strata, resulting in wide credible intervals. The analysis was also restricted to a single study site, and risk could only be calculated for a subset of control participants due to missing data. This means the intervention and control groups in this analysis do not represent the complete randomised groups, and selection bias cannot be excluded. Differences in self-reported ethnicity between included and excluded controls suggest that residual within-stratum differences may exist, although ethnicity is a sociocultural rather than biological characteristic.

The study concludes that while randomised evidence supports risk assessment with risk-stratified timed birth as an overall strategy to reduce term preeclampsia, this exploratory analysis does not establish different treatment effects across risk strata. The higher posterior probabilities of benefit at 39 and 40 weeks should not be interpreted as evidence that the treatment effect was greater in these strata, as posterior probabilities depend on precision and baseline risk. For the highest-risk strata at 37 and 38 weeks, the evidence remains limited regarding the balance of benefit and harm of routine planned birth. The findings suggest that alternative strategies, such as more intensive surveillance, might be considered for these women, although such approaches were not evaluated in PREVENT-PE.

The Fetal Medicine Foundation competing-risks model, used to assign participants to specific delivery windows, estimates the distribution of gestational age at delivery with preeclampsia. This calculation integrates maternal characteristics, medical history, mean arterial pressure, and serum concentrations of placental growth factor and soluble fms-like tyrosine kinase-1. For control participants, who did not undergo this assessment as part of the trial intervention, the research team reconstructed their 35 to 36-week risk using stored serum samples and haemodynamic measurements. This retrospective assignment allowed for within-stratum comparisons that were not available in the parent trial analysis, ensuring that later outcomes did not contribute to the initial risk calculation.

Screening and the communication of the 35 to 36-week preeclampsia risk were integral components of the intervention pathway, occurring exclusively in the intervention group. Consequently, the comparisons made at 39 and 40 weeks estimate the effect of the intervention strategy as implemented, rather than isolating the causal effect of timed birth alone. These co-interventions, which include the screening process and subsequent clinical responses to the risk information, mean that the observed benefits cannot be attributed solely to the timing of delivery. The 41-week stratum, where protocol-specified timing of birth was identical in both groups, serves as a negative control for the timing-of-birth component, though the data cannot distinguish between chance, residual selection, or effects of other intervention elements.

The outcomes examined in this analysis provide only a partial assessment of the potential benefits and harms of timed birth. While birth with preeclampsia was the primary outcome, and emergency caesarean section and neonatal unit admission for at least 48 hours were key secondary outcomes, these metrics do not capture the full spectrum of clinically important maternal and neonatal morbidity. Specifically, the analysis does not account for the severity of preeclampsia or fetal growth restriction. Therefore, the present analysis cannot establish the overall clinical benefit-harm balance within individual strata, as it lacks data on these broader clinical dimensions.

One possible explanation for the uncertain effect of timed birth in the 37-week and 38-week strata is that preeclampsia may already be developing by the time of screening at 35 to 36 weeks. As preeclampsia develops along a continuum and may become clinically apparent before or after delivery, earlier delivery might alter the timing of its clinical presentation rather than prevent its occurrence. This speculative explanation is consistent with the earlier delivery and substantial overlap in cumulative preeclampsia incidence observed in these highest-risk strata, where the intervention resulted in materially earlier delivery than usual care.

Subject of Research: Post-hoc Bayesian analysis of the PREVENT-PE trial on risk-stratified timed birth for preeclampsia prevention

Article Title: Preeclampsia prevention by risk-stratified timed birth: a post-hoc Bayesian analysis of the PREVENT-PE randomised trial

Article References: Wright, D., Singh, L., Syngelaki, A., Magee, L. A., von Dadelszen, P., Wright, A., & Nicolaides, K. H. (2026). Preeclampsia prevention by risk-stratified timed birth: a post-hoc Bayesian analysis of the PREVENT-PE randomised trial. eClinicalMedicine, 101, Article 104309. https://doi.org/10.1016/j.eclinm.2026.104309

Image Credits: AI Generated

DOI: 10.1016/j.eclinm.2026.104309

Keywords: preeclampsia, Bayesian analysis, PREVENT-PE trial, risk stratification, timed birth, obstetrics, randomised trial, maternal health, neonatal outcomes, clinical trial, prevention, risk-stratified