The Neonatal and Obstetrics Risks Assessment (NORA) study was a prospective cohort conducted at the KK Women’s and Children’s Hospital in Singapore between September 2010 and October 2014.15 Singleton pregnancies less than 14 weeks of gestation were recruited during the study period. The exclusion criteria were multiple gestations, chronic medical conditions such as renal disease or systemic lupus erythematosus and pregnancies complicated by aneuploidy or fetal anomaly. Gestational age was confirmed from the fetal crown-rump length. Detailed interviews, biophysical measurements, ultrasound scans and blood sample collections were performed at recruitment (11–14 weeks), 18–22 weeks, 28–32 weeks and 34 weeks onward, respectively. Information on pregnancy complications, labour and delivery and neonatal outcomes was collected through medical chart review. A written informed consent was obtained from all participating women.

A total of 1013 women were enrolled initially and 934 of them completed all 4 antenatal visits. We excluded 8 participants without pregnancy outcomes, leaving 926 participants for the final analysis (online supplemental figure 1).

Preeclampsia was defined according to the guidelines of International Society for the Study of Hypertension in Pregnancy16: systolic blood pressure (SBP) ≥140 mm Hg and/or diastolic blood pressure (DBP) ≥90 mm Hg on at least two occasions 4 hours apart after 20 weeks of gestation in a previously normotensive women, and proteinuria: urinary albumin ≥300 mg/24 hours urine collection or ≥1+ dipstick. For women with chronic hypertension, superimposed preeclampsia was defined as a significant increase in blood pressure compared with baseline (30 mm Hg systolic, 15 mm Hg diastolic) in association with new-onset proteinuria. If proteinuria was present at baseline, superimposed preeclampsia was diagnosed if there was doubling of urinary protein excretion after 20 weeks of gestation in association with a significant increase in blood pressure. Superimposed preeclampsia was also diagnosed if blood pressure was elevated and there were elevated liver enzymes (two times baseline) and a low platelet count (<100×10⁹/L). Preeclampsia or superimposed preeclampsia was subdivided according to gestational age at delivery into term (≥37 weeks) and preterm term (<37 weeks). Gestational hypertension was defined as newly onset hypertension after 20 weeks of gestation without proteinuria.

Based on the US Preventive Services Task Force,3 we classified women as high-risk, moderate-risk and low-risk groups by maternal demographic characteristics and medical history. High-risk group included women with history of preeclampsia, chronic hypertension, diabetes mellitus or autoimmune disease; moderate-risk group included women with more than one of the following factors: obesity (body mass index ≥30 kg/m²), maternal age over 35 years, nulliparity, family history of preeclampsia and previous adverse outcome, including preterm birth and low birth weight; the rest of the participants were considered as the low-risk group. As several studies showed that in women with preeclampsia, especially in those requiring early delivery, uterine artery pulsatility index (UtA-PI) was increased from early pregnancy.17–19 Women with UtA-PI (11–14 weeks) over the 95th percentile for gestation was also included in the high-risk group.

Maternal venous blood was collected by venipuncture using 10 mL non-heparinised tubes at each antenatal visit. Serum was isolated by centrifugation at 2000 revolutions/min for 15 min and stored at −80℃ for subsequent analysis. Samples were measured for soluble fms-like tyrosine kinase 1 (sFlt-1) and PlGF by means of the fully automated Elecsys assays on an electrochemiluminescence immunoassay platform (cobas e411 analyzers, Roche Diagnostics). The detection limit was approximately 6 pg/mL for sFlt-1 and <2 pg/mL for PlGF.

Doppler ultrasound examinations were carried out transabdominally. At 11–14 weeks of gestation, a sagittal section of the uterus was obtained and colour flow mapping was used to identify each uterine artery along the side of the cervix and uterus at the level of the internal os.20 At 18–22 weeks, 28–32 weeks and 34 weeks onward, colour Doppler was used to identify each uterine artery at the apparent crossover with the external iliac arteries.21 Pulsed-wave Doppler was used with the sampling gate set at 2 mm to cover the whole vessel. When three similar waveforms were obtained consecutively, the pulsatility index (PI) and the resistance index (RI) were measured and mean PI and mean RI of the left and right arteries were calculated, respectively.

Blood pressure was taken by validated automated devices (Omron HEM 705 LP, Omron Healthcare) which were calibrated periodically. Women were in a sitting position and their arms were supported. A correct cuff size was used and the middle of cuff was positioned on woman’s upper arm at the level of the right atrium.22 Either a small (<22 cm), normal (22–32 cm) or large (33–42 cm) adult cuff was used depending on the mid-arm circumference. After a rest for 5 min, blood pressure was measured by trained nurses and three recordings were made at 1 min intervals. We calculated SBP and DBP as the average of the three measurements. MAP was calculated from SBP and DBP measures using the following formula: MAP=DBP+1/3×(SBP-DBP).

Normality of continuous variables was assessed by the Kolmogorov-Smirnov test. Numeric data were expressed as mean (SD) or as median (IQR) for normally and non-normally distributed data, respectively. Maternal characteristics and pregnancy outcomes were compared between women with preeclampsia and without preeclampsia using Wilcoxon tests for continuous variables and χ² analysis for categorical variables. The distributions of sFlt-1, PlGF and sFlt-1/PlGF ratio were transformed logarithmically to approximate Gaussian distribution. Covariance analysis was used to compare differences in logarithm-transformed angiogenic factors values, blood pressure levels and uterine artery Doppler values between women with preeclampsia and without preeclampsia adjusted for covariants. Geometric means and 95% CIs were calculated by taking the exponent of the logarithm transformed mean. If variables did not fit the sphericity tests, repeated measure analysis of variance was performed to examine the differences of variables at four time points between women with preeclampsia and without preeclampsia. The performance of sFlt-1, PlGF, sFlt-1/PlGF ratio, MAP and UtA-PI for the prediction of preeclampsia was assessed by the area under the receiver operating characteristic curve (AUC). Areas were compared using the method of DeLong et al.23 Cut-off values were determined by the maximising the sum of sensitivity and specificity. We used SAS V.9.4 for all statistical analyses.

Patients and the public were not directly involved in the design, conduct or reporting in our study.

Our large prospective cohort study in Asian population suggested that compared with angiogenic factors and UtA-PI levels, MAP had significantly higher AUCs at 11–14, 18–22 and 28–32 weeks of gestation for prediction of preeclampsia, especially for the term preeclampsia. For the prediction of preterm preeclampsia, MAP and PlGF were similar. Moreover, in high-risk women, MAP, PlGF and UtA-PI had equivalent predictive values at 11–14 weeks of gestation; in moderate-risk and low-risk women, MAP predicted better than PlGF and UtA-PI in early pregnancy. Our study also confirmed that women who developed preeclampsia, compared with women without preeclampsia, had significantly lower levels of PlGF, higher levels of sFlt-1/PlGF ratio and MAP throughout pregnancy.

A previous meta-analysis including more than 60 000 women suggested that MAP yielded an AUC of 0.79 in the first trimester and 0.76 in the second trimester. But most of the studies did not report single or multiple measurements of blood pressure.11 Poon et al showed that MAP at 11–13⁺⁶ weeks alone had an AUC of 0.734 for prediction of preeclampsia. Considering the interarm blood pressure differences, they measured blood pressure in both arms simultaneously and recorded the last two stable measurements.24 25 Previous study suggested that the overall screening performance using MAP determined from either the left or right arm, or the average of both arms was not significantly different. But when the average of two or three recordings were used for calculating MAP, the performance of screening tended to improve with an increasing number of recordings.26 Although we measured blood pressure in one arm, three readings were recorded at 1-min intervals. Our results showed that MAP in Asian women had moderate to high performance for the prediction of preeclampsia, with AUCs of 0.86 (95% CI 0.78 to 0.95), 0.87 (95% CI 0.80 to 0.95) and 0.91 (95% CI 0.85 to 0.98) at 11–14, 18–22 and 28–32 weeks, respectively.

Meanwhile, in our Asian population, angiogenic factors showed low to moderate values for the prediction of preeclampsia, with AUCs for PlGF 0.67 (95% CI 0.56 to 0.79), 0.77 (95% CI 0.66 to 0.88), and 0.78 (95% CI 0.65 to 0.91) at 11–14, 18–22 and 28–32 weeks, respectively. Our results were similar to those in a longitudinal study which showed AUCs of 0.53 (95% CI 0.48 to 0.58), 0.57 (95% CI 0.51 to 0.63), 0.66 (95% CI 0.61 to 0.71) and 0.72 (95% CI 0.67 to 0.77) at 10, 17, 24 and 35 weeks of gestation, respectively, for PlGF.27 PlGF/sFlt-1 ratio also showed similar performance as PlGF in the mixed population.27 28 Thus, individual angiogenic factors did not perform well in predicting preeclampsia.

The Fetal Medicine Foundation proposed a combined screening test, including maternal factors, MAP, PlGF and UtA-PI, which was found to predict 75% of preterm preeclampsia and 47% of term preeclampsia at a false-positive rate of 10% in European population.6 However, this algorithm was less predictive in Asian population.8 9 Except for the under-measurement of UtA-PI, the lower incidence of preterm preeclampsia and prevalence of maternal risk factors in Asian women might reduce the effectiveness of the algorithm.9 Studies that clarify the causes of the racial differences in preeclampsia are few. Difference in metabolic profiles during 26–28 gestational weeks between White European and South Asian women suggested that White European women had higher levels of lipoproteins, cholesterol, glycerides, phospholipids and monosaturated fatty acids than South Asian women.29 A systematic review, examining the association between the endothelial nitric oxide synthase polymorphism and preeclampsia risk stratified by ethnicity, identified a higher risk of preeclampsia in the white or mixed populations with the 894 G>T polymorphism, but not in those of Asian or African descent.30 Another systematic review reported that IL10 −1082 G>A polymorphism significantly increased the risk of preeclampsia in Asian women, but not in White population.31 Besides impaired placentation, immunological factors, endothelial function, and inflammation have all been identified as contributors to the pathophysiology of preeclampsia.32 Although circulating PlGF and MAP mainly reflect placentation and maternal endothelial function, respectively, racial disparities on pathophysiology might influence the effectiveness of the predictors. As our study showed that while MAP performed better than PlGF for the prediction of preeclampsia and term preeclampsia throughout gestation, for the prediction of preterm preeclampsia, these predictors were not significantly different.

Blood pressure measurement has been used as a routine screening and diagnostic tool in antenatal care for decades. Although the American Heart Association statement on blood pressure measurement has been commonly accepted worldwide, which arm to measure and how to record the measurements were not specified in the recommendation.22 Our results, along with other studies, suggested that multiple recordings of blood pressure could improve the accuracy for estimating risk of hypertensive disorders. Although we measured blood pressure on one arm, three continuous recordings at 1-min interval could produce an AUC of 0.86 (95% CI 0.78 to 0.95) with 13.0% positive predictive value at 11–14 weeks of gestation. This performance in Asian women was better than that in European where blood pressure was measured on both arms with three recordings (AUC 0.773, 95% CI 0.768 to 0.778). Thus, except for a proper position, appropriately sized cuff and regularly checked devices, multiple recordings should also be emphasised in future clinical practice.

As to the differences between MAP and PlGF in predicting preeclampsia, the predictive values for preterm preeclampsia were not significantly different between MAP and PlGF in the Asian population. In addition, MAP performed better in the prediction of term preeclampsia. Considering the costs and performance of circulating PlGF level for the prediction of preeclampsia, our results support the pragmatic guide for the first-trimester screening of preeclampsia by the International Federation of Gynecology and Obstetrics, in which the baseline screening method should be a combination of maternal risk factors with MAP.

Our study compared predictive values of MAP, angiogenic factors and UtA-PI for predicting preeclampsia in a large Asian population. The NORA cohort was a well-performed prospective study, with standard protocols on measurements of blood pressure, circulating angiogenic factors and UtA-PI. Thus, our results were reliable and comparable with other studies. Prophylactic use of low-dose aspirin was not routinely recommended during study period and three women reported using antihypertensive medication during pregnancy. The incidence of preterm preeclampsia and term preeclampsia was 0.9% and 1.3% in the cohort. Thus, our results were less likely affected by the medication issue. However, the exclusion of women with multiple gestations and chronic medical conditions, and lower incidence of preeclampsia in Asian women per se, might have impacts on the relatively low occurrence of preeclampsia in the NORA cohort. On the other hand, our cohort measured blood pressure on one arm and we could not directly compare the performance of MAP from both arms and from one arm. Besides, predictive values of biomarkers at 28–32 weeks might be overestimated, as women who had already been diagnosed as preeclampsia might not be excluded due to unknown time of diagnosis in our study. The potential presence of white coat hypertension might impact on the performance of blood pressure in the prediction of preeclampsia.

MAP is a good predictor for preeclampsia, especially term preeclampsia, in Asian women. For prediction of preterm preeclampsia, the performance of MAP and PlGF was similar. Our experience indicates that multiple recordings in antenatal blood pressure measurement are encouraged to achieve better accuracy of MAP.