Blood pressure at early pregnancy and gestational hypertensive disorders: a prospective cohort study in China

Abstract

BackgroundGestational hypertension and pre-eclampsia are major causes of perinatal mortality. Prediction of gestational hypertension and pre-eclampsia is of great interest because it enables early intervention, thus improving prognosis. Most existing prediction models consist of biomarkers, which might be unavailable in low-resourced countries. We aimed to establish a prediction model of gestational hypertension and pre-eclampsia using data at early pregnancy. MethodsWe studied women with singleton delivery from Born in Guangzhou Cohort Study (BIGCS), China. Predictors included maternal age, educational level, income level, prepregnancy weight, height, passive smoking, and blood pressure collected at the first antenatal-care visit (around 16 weeks' gestation). Information on diagnosis of gestational hypertension or pre-eclampsia was extracted from medical records using international classification of disease code (ICD-10). We used logistic regression to develop prediction models. Discrimination and calibration were assessed with receiver operation characteristics (ROC) and calibration plot, respectively. FindingsBetween Feb 1, 2012, and Jan 1, 2016, we recruited 12 915 women, of which 326 (2·52%) women were diagnosed with gestational hypertension and 82 (0·66%) had pre-eclampsia. The prediction model for gestational hypertension with maternal characteristics alone had an area under the ROC-curve of 0·67 (95% CI 0·62–0·72). Maternal mean arterial pressure (MAP) had an area under the curve (AUC) of 0·74 (95% CI 0·70–0·79), whereas the AUC of the model with MAP and maternal characteristic combined was 0·76 (0·72–0·81), which was slightly better than for MAP alone (p=0·03). Results for prediction of pre-eclampsia were very similar to those of gestational hypertension. Calibration plots showed that the prediction model with MAP had good fit. InterpretationOur findings show that MAP has acceptable predictive ability of gestational hypertension and pre-eclampsia and can be used to triage further care. Our relatively large sample size ensured stronger statistical power. Model validation need to be performed in a separate population. FundingNational Natural Science Foundation of China (81673181), Guangzhou Science and Technology Bureau, Guangzhou, China (2011Y2-00025, 201508030037)