Early prediction of preeclampsia and small-for-gestational-age via multi-marker model in Chinese pregnancies: a prospective screening study

Preeclampsia (PE) and small -for -gestational -age (SGA) are two of the main contributors to maternal and perinatal morbidity and mortality worldwide [1‐3], which could impose life-long negative effects on an individual [4, 5]. PE is a disorder of pregnancy with typical characteristics of hypertension and proteinuria, which affects multiple physiological systems after 20 weeks’ gestation [6]; SGA newborns are generally defined as those with weight below the 10th percentile for the gestational age [7]. Early identification of PE and SGA will enable clinicians to take preventative steps and offer intensive monitoring to the high-risk pregnancy group, thereby decreasing the incidence of serious diseases in both mothers and their offspring.

According to the guidelines of the UK National Institute for Health and Clinical Excellence (NICE) and the American Congress of Obstetricians and Gynecologists (ACOG), the traditional approach to evaluate women with high risk of developing PE is to derive risk factors from their medical histories and maternal demographic features. This kind of screening method remains recommended for clinical use [8, 9]. Previous studies have demonstrated that maternal serum pregnancy associated plasma protein A (PAPP-A) and placental growth factor (PLGF) MoM levels at 11 to 13 weeks’ gestation were lowered in pregnant women who subsequently developed PE and also in those who delivered SGA infants [10, 11]. In recent years, the Fetal Medicine Foundation (FMF) has developed a new algorithm for screening high-risk PE pregnancies at 11–13 weeks’ gestation, combining a number of maternal factors such as mean arterial pressure (MAP), uterine artery pulsatility index (UTPI), serum PLGF and PAPP-A MoM values [12, 13]. It outperforms the existing criteria recommended by NICE and ACOG [14, 15]. The algorithm has also shown potential to predict delivery of SGA [16], since the two placenta-related pregnancy complications (PE and SGA) may share the common pathophysiological mechanism. Moreover, the FMF screening algorithm has already been widely validated in Europe [17‐19].

Heretofore, most of the combined model studies on prediction of PE and SGA have been focused on the Caucasian and African-American populations, whereas the data of the Asians are insufficient. Our study adopted the algorithm in Pre-Eclampsia Predictor™ by Perkin Elmer (hereafter named “PREDICTOR”) which has similar performance to the FMF algorithm in predicting PE [20]. It is a standalone software application designed for calculating prior and posterior risks of early and late PE in first trimester, and performs risk calculation by using the Bayesian algorithm developed by Professor Howard Cuckle [21]. No previous study has investigated the applicability of PREDICTOR algorithm to predicting PE and SGA by combining maternal factors and biomarkers in first trimester in Mainland China. More importantly, an effective screen method for PE and SGA will allow more accurate research on prophylactic intervention. In light of these considerations, it is significant to demonstrate the effectiveness of the PREDICTOR algorithm in prediction of PE and SGA in Mainland Chinese pregnancies.

This study aims to, firstly, estimate the predictive performance of screening model for PE in Chinese pregnant women in first trimester through a combination of maternal characteristics, MAP, serum PLGF and PAPP-A, by applying PREDICTOR algorithm; secondly, to investigate potential predictive value for SGA through the same screening protocol.

This prospective study demonstrated that our screening model at 11–13 weeks’ gestation could identify pregnancies developing early PE efficiently; furthermore, it performed better in detection of PE with SGA than PE with AGA. Nevertheless, by using the same screening method, the performance of prediction of isolated early and late SGA was evidently lower. Thus, we supposed that the prediction of SGA may depend heavily on the presence of PE. Moreover, this study found that women with PE or those delivered SGA had decreased serum PLGF and PAPP-A. And previous studies have proposed that the low concentrations of serum markers in the first trimester are likely to be the consequence of impaired trophoblastic invasion of the spiral arteries [26]. Although the pathophysiological mechanisms behind PE and SGA are still not fully understood, impaired placenta indeed plays a central role in the development of these two conditions.

In this study, we used the PREDITCOR algorithm developed by Professor Howard Cuckle to calculate early and late PE risks. The prior risk is a product of a combination of multiple demographic-based likelihoods. In the aspect of diagnostic accuracy, it is less effective than using posterior risk (combining prior likelihoods and likelihoods of MAP, PLGF and PAPP-A). Our data provide further support for the hypothesis that screening performance for PE at 11–13 weeks’ gestation could be improved significantly by adding MAP, serum PLGF and PAPP-A. Furthermore, according to the aforesaid software default setting, the theoretic posterior risk cutoff (1:20) was less effective in early screening of PE and apparently not suitable for the population in our research, whereas from our calculation, the suitable posterior risk cutoffs for early and late PE were 1:45 and 1:151, individually. Therefore, we can draw a conclusion that laboratories should adjust the cutoff values in accordance with the screening population and their lab conditions in order to achieve the best screening performance to define high-risk PE group in the most appropriate fashion.

The past decade has seen extensive studies being conducted to screen for PE based on the FMF algorithm which combines maternal characteristics with different biochemical and biophysical markers at 11–13 weeks’ gestation [13, 17, 19]. Yet only a very few medical units applied the PREDICTOR algorithm to test its validity in prediction of PE and SGA in first trimester [20]. Moreover, in reviewing literature, few data has been found relating to predicting PE and SGA risks in Chinese pregnancies at 11–13 weeks’ gestation by using the PREDICTOR algorithm.

It is noteworthy that the findings in this report indicate that Perkin Elmer’s PREDICTOR algorithm is an effective method of detecting early PE in Chinese pregnancies. According to Table 4, the DRs of early PE, under 5 and 10% FPRs, were both 87.5%, which were similar to those of previous researches conducted in western countries. For instance, a prospective observational nonintervention cohort study in a U.S. population screened for PE through a combination of maternal characteristics, MAP, UTPI and serum biomarkers (PLGF, PAPP-A, alphafetoprotein), in which the DRs of early PE, at either 5% or 10% FPR, were both 85% [27]. Also, another previous study showed that among 8775 pregnancies, the DRs of delivery with PE < 32 weeks’ gestation screened by maternal factors combining MAP, PAPP-A and PLGF, at FPRs of 5 and 10%, were 76 and 88%, respectively [17]. Additionally, O’Gorman et al. screened 35,948 singleton pregnancies at 11 to 13 weeks among whom they found that by combining maternal factors with MAP, serum PLGF and PAPP-A in screening, the DRs of early PE delivery before 32 weeks’ gestation, at 5 and 10% FPRs, were 76 and 85%, respectively [28]. In our study, we concluded that DRs for SGA differed substantially with and without PE. A similar conclusion was made by a previous study, which assessed the effectiveness of first-trimester screening for SGA neonates using a combination of maternal characteristics, MAP, serum biochemical markers and UTPI [29]. The DR for early SGA in their study was higher and the DR for late SGA was lower than the data in our report (60% versus 41.67% in early, and 23% versus 28% in late, at FPR of 10%).

This study carries following strengths. First, it is a prospective study with a complete follow-up in Chinese population; second, the implication of Chinese population in prediction of PE has not been closely examined by the PREDICTOR algorithm before; third, it supplements the data of combined multi-marker model to estimate patient-specific risks and investigates the performance of screening for two placental-related conditions; fourth, our screening approach effectively tests the feasibility of its future large-scale implementation in China and involves health economic considerations. As shown above in section of Methods in this report, maternal history recording and blood pressure measurement are a part of the normal routine pregnancy care. Also, maternal serum PAPP-A is widely used in screening for Down syndrome, and serum PLGF measurement can be further conducted on the same automated machines with same sample which can provide results quickly, with merely an additional cost for the reagents. This combining method for the prediction of PE can be incorporated into the present-day screening method for fetal aneuploidies, but whether or not this first-trimester screening method for PE could be readily available to the entire Chinese population requires further investigation.

Nonetheless, there are two important limitations of the study. One is that despite comparatively large sample size (over 3000), there were only eight cases of early PE due to the relatively low incidence of PE in southern China. But we evaluated the sample size by R package, and the results showed that for both early or late PE, the sample sizes were adequate and had enough power (> 99%). The prevalence of PE varies among different ethnic groups globally. Compared with the European and African American women, the incidence of PE in our research was relatively low [30]. The other limitation is that we did not perform the testing of Doppler ultrasound detection of UTPI. Our testing model did not include this marker (UTPI) into risk calculation, even though it has shown its promising value in predicting PE and SGA [31]. The reason is that measurement of UTPI requires specific training of sonographers to achieve uniformity of results among different operators. Due to large volume of outpatients, it is extremely difficult to ensure the quality of UTPI and to attain accurate results of every pregnancy as part of the current routine ultrasound scans in China. Presumably, the performance of our multivariate screening may be further improved by adding accurate UTPI, which is expected to be included in our next-step study. Yet it need be stressed that our screening model is still feasible in detection of PE with delivery < 34 weeks’ gestation.

The conventional approach to PE diagnosis is based on the occurrence of clinical symptoms discovered during routine obstetrics clinical visit in the second or third trimester of pregnancy, however, a new pyramid of pregnancy care having been proposed in 2011 states that effective early identification of the high-risk groups of subsequently developing PE and other pregnancy complications in the first trimester could reduce the incidence of these complications through pharmacological intervention with low-dose aspirin (LDA) [32, 33]. Further evidences in European populations have suggested that the risk of PE could be predicted and greatly reduced by the prophylactic application of LDA [34]. The aforementioned findings have demonstrated the vital role the prediction of PE plays in its clinical diagnosis, prevention and incidence reduction. What’s more, the ASPRE trial (Combined Multi-Marker Screening and Randomized Patient Treatment with Aspirin for Evidence-Based Preeclampsia Prevention) has shown that 150 mg/day of aspirin taken from 11 to 14 weeks to 36 weeks of gestation had reduced the incidence of early and preterm PE [35]. A secondary analysis of data from the ASPRE trial reported that in pregnancies at high risk of PE, treatment with LDA has shorten the length of stay in the neonatal intensive care unit [36]. Another research deriving data from SPREE (Screening programme for preeclampsia) and ASPRE concluded that the combined screening test of first-trimester pregnancies for PE could also identify a high proportion of cases of preterm SGA which can be prophylactically prevented by the use of aspirin [37]. But only a few small-sample studies in China have reported that pregnant women with high risk of PE obtained a better maternal and child prognosis after preventively taking aspirin [38]. Li et al. has launched a multicenter randomized controlled trial (RCT), low-dose Aspirin in the Prevention of Pre-Eclampsia in China (APPEC study), to evaluate the efficacy of LDA for PE prevention on high-risk pregnancies in Chinese [39]. Nevertheless, to achieve this, researches of accurate PE prediction at 11–13 weeks in Chinese are pivotal to future RCTs in the determination of the role that LDA plays in reducing the incidence of PE.

Our first-trimester integrated screening is valid in terms of predicting early PE, but late-onset PE (i.e. ≥34 weeks) accounts for a large proportion of all PE, which remains a significant challenge of effective early screening. Furthermore, the performance of prediction for isolated SGA is not as good as the one for SGA with PE. Consequently, screening strategy for late PE and isolated SGA demands additional potential markers and more effective screening model.

To conclude, this study shows that our screening model could be effective for predicting early PE in Chinese pregnancies even without requiring ultrasound UTPI, but it is relatively less valuable for late PE and SGA neonates in the absence of PE. Furthermore, the performance of SGA screening by same protocol is strongly associated with PE.