Externally validated prediction models for pre-eclampsia: systematic review and meta-analysis.

Area of Science:

• Obstetrics and Gynecology
• Clinical Prediction Models
• Perinatal Health

Background:

• Pre-eclampsia (PE) poses significant risks to maternal and infant health.
• Accurate prediction models are crucial for timely intervention and improved outcomes.
• Existing models require rigorous external validation to assess their real-world performance.

Purpose of the Study:

• To systematically review and meta-analyze the performance of externally validated prediction models for various types of pre-eclampsia (PE).
• To identify the most effective prediction models for any-onset, early-onset, late-onset, and preterm PE.
• To evaluate the consistency of predictors and the generalizability of these models.

Main Methods:

• Systematic search of five databases and Google Scholar until May 2023, adhering to PICOTS criteria.
• Data extraction using the CHARMS checklist and risk of bias assessment with the PROBAST tool.
• Meta-analysis of discrimination (AUC) and calibration performance for included models.

Main Results:

• Twenty-three studies yielded 52 externally validated PE prediction models; no two models shared identical predictors.
• The Fetal Medicine Foundation (FMF) model demonstrated superior performance for preterm PE prediction (AUC 0.90), with good calibration.
• Most other models showed limited discrimination (median AUC 0.66) and were prone to overfitting; only two maternal-characteristics-based models for any-onset PE performed reasonably well.

Conclusions:

• Externally validated models for any-, early-, and late-onset PE generally exhibit suboptimal discrimination and calibration, with inconsistent predictor variables.
• The FMF triple-test model shows excellent performance for preterm PE prediction across diverse settings.
• The feasibility of implementing the FMF model outside high-resource settings may be limited by its reliance on specialized biomarkers.