Re: Prediction of pre-eclampsia: review of reviews.

Abstract

The meta-analysis of previous meta-analyses by Townsend et al.1 does not benefit but rather hinders research and clinical implementation of strategies for the prediction and prevention of pre-eclampsia (PE). It is inevitable that there is limited value in performing such umbrella meta-analyses and that misleading conclusions can be drawn when combining data from a series of heterogeneous studies in terms of size and type of populations examined, methodology used, timing of biomarker measurement, adjustment or not for confounders, use of different models combining risk factors from maternal demographic characteristics and medical history with biomarkers, as well as use of a plethora of PE definitions resulting in delivery at different gestational ages. Whilst we agree with the conclusion of Townsend et al.1 that well-designed prospective studies of predictive markers and randomized intervention studies are necessary to develop and validate new prediction models before their introduction into clinical practice, we strongly disagree that these have yet to be performed. In the last three decades, many well-designed prospective studies involving tens of thousands of patients have established all the necessary requirements for effective clinical implementation of routine screening for PE2. We present here some of the evidence generated by our group; however, many other researchers have contributed through their data to the following conclusions. First, it is now established that the best approach for estimation of individual patient-specific risk of delivery with PE before any specified gestational age is by a combination of maternal factors and biomarkers, obtained either individually or in combination at any stage in pregnancy, using the competing-risks model. This model for estimation of risk based on maternal factors was originally derived from a prospective study of 58 884 singleton pregnancies3 and was subsequently updated based on the study of 120 492 pregnancies4. Second, many potential biomarkers have been extensively investigated and it has been established that, at present, useful biomarkers for the prediction of PE are: uterine artery pulsatility index (UtA-PI), mean arterial pressure (MAP) and serum placental growth factor (PlGF) measured at 11–13 and 19–24 weeks' gestation; UtA-PI, MAP, PlGF and serum soluble fms-like tyrosine kinase-1 (sFlt-1) measured in the early third trimester; and MAP, PlGF and sFlt-1 evaluated in the late third trimester5-28. The methodology for obtaining appropriate measurements of biomarkers and auditing of results has also been established. Third, it is established that biomarker levels depend on gestational age, maternal weight and race, method of conception, medical conditions and elements of obstetric history, and they are also affected by the instrument used for their measurement. Consequently, appropriate evaluation and application of biomarkers in screening requires prior standardization by expressing the measured values as multiples of the median (MoM)28-31. In pregnancies that develop PE, MoM values of MAP, UtA-PI and sFlt-1 tend to be higher and PlGF tends to be lower than in normal pregnancies; the effect size increases with increasing severity of the disease, quantified by the gestational age at delivery5-8. Fourth, the competing-risks model has been successfully applied for assessment of risk for PE and stratification of pregnancy care by a combination of maternal factors and biomarkers in the first, second and third trimesters of pregnancy32-44. In the first trimester, the competing-risks approach utilizing maternal factors, MAP, UtA-PI and PlGF was used to identify women at high risk of developing preterm PE; at a 10% screen-positive rate, 90% of early-PE cases and 75% of those with preterm PE were predicted in both a training dataset of 35 948 singleton pregnancies and in two independent, non-intervention, multicenter studies involving 8775 and 16 451 singleton pregnancies, respectively32, 45-48. We have also shown, in a study involving 66 964 pregnancies, that the predictive performance of this approach is similar irrespective of whether PE is defined by the traditional criteria of the International Society for the Study of Hypertension in Pregnancy (ISSHP), which requires the presence of both hypertension and proteinuria, or the new criteria of ISSHP and the American College of Obstetricians and Gynecologists which include cases without proteinuria but with evidence of renal, hepatic or hematological dysfunction49-52. Fifth, a randomized controlled trial found that aspirin administration after 23 weeks' gestation in a high-risk group identified by second-trimester screening was not effective in preventing PE53. In contrast, a more recent randomized trial demonstrated that, in women at high risk of PE, administration of aspirin (150 mg/day) from 11–14 until 36 weeks' gestation reduces the risk of early PE and preterm PE by about 90% and 60%, respectively, and the length of stay in the neonatal intensive care unit by about 70%54, 55. There is no justification for further delay in the implementation of strategies for prediction and prevention of PE in routine clinical practice.