ABSTRACT
OBJECTIVE: The prediction of preeclampsia in pregnancy has resulted in a plethora of prognostic models. Yet, very few make it past the development stage and most fail to influence clinical practice. The timely identification of high-risk pregnant women could deliver a tailored antenatal care regimen, particularly in low-resource settings. This study externally validated and calibrated previously published models that predicted the risk of preeclampsia, based on blood pressure (BP) at multiple time points in pregnancy, in a geographically diverse population. METHODS: The prospective INTERBIO-21st Fetal Study included 3,391 singleton pregnancies from Brazil, Kenya, Pakistan, South Africa, Thailand and the UK, 2012-2018. Preeclampsia prediction was based on baseline characteristics, BP and deviation from the expected BP trajectory at multiple time points in pregnancy. The prediction rules from the Avon Longitudinal Study of Parents and Children (ALSPAC) were implemented in the INTERBIO-21st cohort. RESULTS: Model discrimination was similar to the development cohort. Performance was best with baseline characteristics and a BP measurement at 34 weeks' gestation (AUC 0.85, 95 % CI 0.80-0.90). The ALSPAC models largely overestimated the true risk of preeclampsia incidence in the INTERBIO-21st cohort. CONCLUSIONS: After recalibration, these prediction models could potentially serve as a risk stratifying tool to help identify women who might benefit from increased surveillance during pregnancy.