Performance of a first-trimester combined screening for preterm preeclampsia in the United States population using the Fetal Medicine Foundation competing risks model.

Abstract

Background: Preeclampsia complicates 2-8% of pregnancies and is a leading cause of maternal and perinatal morbidity and mortality. In the United States, current guidelines recommend low-dose aspirin based on clinical risk factors, which identify only a minority of those at risk for preterm preeclampsia. The Fetal Medicine Foundation (FMF) competing risks algorithm improves prediction by combining maternal characteristics, biophysical parameters, and biochemical markers. While validated internationally, its performance in the U.S. population remains unclear.

Objective: To evaluate the performance of the FMF first-trimester combined screening algorithm for predicting preterm preeclampsia in a diverse U.S. nulliparous population.

Study design: We performed a retrospective cohort study using data from the Nulliparous Pregnancy Outcomes Study: Monitoring Mothers-to-Be (nuMoM2b). We excluded those with delivery before 20 weeks, missing outcome data, or aspirin use. The FMF model incorporated maternal factors, mean arterial pressure, placental growth factor, pregnancy-associated plasma protein-A, and uterine artery pulsatility index (adjusted from second-trimester values). The primary outcome was preterm preeclampsia, defined as preeclampsia requiring delivery <37 weeks. Model performance was assessed via area under the receiver-operating characteristics curve (AUROC) with 95% confidence interval (CI), with optimal cut-off determined using Liu's method. Sensitivity, specificity, positive and negative predictive values (PPV and NPV), likelihood ratios (LHR), and odds ratios (OR) were calculated using the optimal cut-off. Goodness of fit was evaluated using a calibration plot and decision curve analysis.

Results: Of 8,664 nulliparous individuals, 189 (2.2%) had preterm preeclampsia. The FMF model revealed an AUC of 0.75 (95% CI 0.71-0.78), indicating moderate discriminative ability. Using a 0.7% cut-off, sensitivity was 64.0%, specificity 72.5%, PPV 4.9%, and NPV 98.9%. Positive LHR was 2.3, negative LHR 0.5, and OR 4.7. The calibration plot showed the model underestimated risk, while recalibration improved alignment. Decision curve analysis demonstrated net clinical benefit across commonly used thresholds (1-12%).

Conclusion: The FMF first-trimester screening algorithm demonstrated moderate predictive performance. Recalibration enhanced risk estimation, and decision curve analysis supported clinical utility, highlighting the need for population-specific adjustment.