RESUMO
OBJECTIVE: This systematic review and meta-analysis aimed to evaluate the performance of existing externally validated prediction models for pre-eclampsia (specifically for any- early- late-onset and preterm pre-eclampsia). METHODS: A systematic search was conducted in five databases (MEDLINE, Embase, Emcare, CINAHL, and Maternity and Infant Care Database) to identify studies based on Population, Index model, Comparator, Outcome, Timing, and Setting (PICOTS) approach until May 20, 2023. We extracted data using the CHARMS checklist and appraised risk of bias using PROBAST tool. Discrimination and calibration performance were meta-analysed when appropriate. RESULTS: Twenty-three publications reported 52 externally validated prediction models on pre-eclampsia (twenty any-onset, seventeen early-onset, fourteen late-onset, and one preterm pre-eclampsia). No model had the same set of predictors. Fifteen, two, and three any-onset pre-eclampsia models were externally validated once, twice, and thrice, respectively, and the Fetal Medicine Foundation (FMF) preterm model was widely validated in sixteen different settings. The most common predictors were maternal characteristics (pre-pregnancy BMI, prior pre-eclampsia, family history of pre-eclampsia, chronic medical conditions, and ethnicity) and biomarkers (uterine artery pulsatility index and pregnancy-associated plasma protein-A). The model for preterm pre-eclampsia (triple test FMF) had the best performances with a pooled area under the receiver operating characteristics curve (AUROC) of 0.90 (95% prediction interval (PI) 0.76 - 0.96) and was well-calibrated. The other models generally had poor to fair discrimination performance (AUROC median 0.66, range 0.53 to 0.77) and were overfitted in calibration after external validation. Apart from the FMF model, only the two most validated models in any-onset pre-eclampsia using isolated maternal characteristics, produced reasonable pooled AUROCs of 0.71 (95% PI 0.66 - 0.76) and 0.73 (0.55 - 0.86). CONCLUSION: Existing externally validated prediction models for any-, early-, and late-onset pre-eclampsia have limited discrimination and calibration performance with inconsistent input variables. The triple test FMF model had excellent discrimination performance in predicting preterm pre-eclampsia in numerous settings, but the inclusion of specialised biomarkers may limit feasibility and implementation outside of high-resource settings. This article is protected by copyright. All rights reserved.