Maternal Serum Placental Growth Factor, Soluble Fms-Like Tyrosine Kinase-1, and Soluble Endoglin in Twin Gestations and the Risk of Preeclampsia—A Systematic Review

Multiple gestation is one of the key risk factors for the occurrence of preeclampsia (PE). Soluble fms-like tyrosine kinase-1, placental growth factor, and soluble endoglin are molecules involved in the process of angiogenesis with a proven role in the pathogenesis of PE. The aim of the review was to summarize available data on maternal serum levels of the above-mentioned factors and their usefulness in predicting PE in twin pregnancies. Only original research articles written in English were considered eligible. Reviews, chapters, case studies, conference papers, experts’ opinions, editorials, and letters were excluded from the analysis. No publication date limitations were imposed. The systematic literature search using PubMed/MEDLINE, Scopus, Embase, and Cochrane Library databases identified 338 articles, 10 of which were included in the final qualitative analyses. The included studies showed significant differences in maternal serum levels of the discussed factors between women with twin pregnancies with PE and those who did not develop PE, and their promising performance in predicting PE, alone or in combination with other factors. The identification of the most effective algorithms, their prompt introduction to the clinical practice, and further assessment of the real-life performance should become a priority.


Introduction
According to the definitions of the American College of Obstetricians and Gynecologists and the International Society for the Study of Hypertension in Pregnancy, preeclampsia (PE) is defined as the new onset of hypertension after 20 weeks of gestation, accompanied with proteinuria or, in the absence of proteinuria, thrombocytopenia, renal insufficiency, impaired liver function, or pulmonary edema [1,2]. It affects about 5% of pregnancies worldwide with great regional diversity [3,4]. Together with eclampsia, it remains one of the leading causes of maternal mortality worldwide [4,5]. Although risk factors are well established, the majority of PE cases occur in otherwise healthy pregnant women. One of the independent risk factors is a multiple gestation, which was the subject of this review [2]. The occurrence of PE in multiple gestations is higher than in singleton gestations. Laine et al. recently The article was written in accordance with the principles contained in preferred reporting items for systematic reviews and meta-analyses (PRISMA) statement [31]. The systematic literature search for articles concerning PlGF, sFlt-1, and sEng in predicting PE in twin pregnancies was performed using four databases: PubMed/MEDLINE, Scopus, Embase, and Cochrane Central Register of Controlled Trials (CENTRAL). The last search was performed on October 23, 2019. We did not contact the authors of the papers to obtain additional information. The search strategy was suited to the specific database (details provided in Table 1).
Only original research articles written in English were considered eligible, whereas reviews, chapters, case reports or case series, conference papers and abstracts, experts' opinions, editorials, and letters to the editor were excluded from the analysis. No publication date limitations were imposed. Titles, abstracts, and keywords of research works obtained via the search process, described above, were screened independently by all study authors. The next step, after rejecting papers that visibly did not meet the criteria, involved reviewing full-text publications by two authors. A customized data extraction sheet was used for the collection of the following information: type of study, population demographics, inclusion and exclusion criteria, methodology, diagnostic tools used, and results. The risk of bias was assessed using the Newcastle-Ottawa Quality Assessment Scale modified by authors for this work [32]. Additionally, other potential additional sources of bias, not included in the scale, have been described in the subsection "Risk of bias assessment" in the Results section. Any disagreements were resolved through consensus with all study authors.

Characteristics of Retrieved Studies
The implemented systematic literature search identified 338 articles. After adjusting for duplicates with the use of EndNote X9 automatic duplicate search followed by manual verification, 220 studies remained, and 10 of which finally met the inclusion criteria. Details on the selection process are presented in a customized PRISMA flow chart in Supplementary Materials Figure S1. The basic characteristics of the studies included in the review are summarized in Table 2. To investigate if differences in sFlt1, sEng, and PlGF in high-risk patients would identify women who later developed PE in a manner similar to low-risk women.
Secondary analysis of samples obtained during a multicenter RCT of low-dose aspirin in the prevention of PE conducted between 1991 and 1995 [34]. To assess levels of sFlt1, sEng, and PlGF in maternal serum in the 1st trimester of TGs and establish if the mode of conception influences the angiogenic status.

MG
A prospective study on women with TGs or SGs who attended the first-trimester screening visit, conducted between 2008 and 2010.

Boucoiran et al. 2013 [38]
To determine the accuracy of PlGF, sFlt-1, and inhibin A in SGs and MGs for predicting PE and SGA.
A prospective multicenter cohort study nested in an RCT of antioxidant supplementation for the prevention of PE conducted between 2004 and 2006 [39].

MG: 69 TG + 5 triplets
LoD; in the whole study group up to 90% Caucasian To determine if early pregnancy serum markers in high-risk women who develop PE vary depending on the risk factor.
Secondary analysis of samples obtained during a multicenter RCT of low-dose aspirin for the prevention of PE conducted between 1991 and 1995 [34].

Serum Concentrations of sFlt-1, PlGF, sFlt1:PlGF Ratio, and sEng in PE and non-PE Twin Pregnancies
The general trends in serum concentrations of sFlt-1, PlGF, sFlt1:PlGF ratio, and sEng in PE and non-PE twin pregnancies reported in the selected studies are summarized in Table 3. Table 3. Serum concentrations of sFlt-1, PlGF, sFlt1:PlGF ratio, and sEng in women in twin pregnancies who developed PE compared to women in twin pregnancies who did not develop PE.

Accuracy of sFlt-1, PlGF, sFlt1:PlGF Ratio, and sEng in Predicting PE in Twin Gestations
Six out of ten selected studies attempted to determine the accuracy of the examined angiogenic factors in predicting PE in twin pregnancies. Detailed data on this subject are collected in Table 4.

Risk of Bias Assessment and Limitations
In the individual studies, the risk of bias was assessed using the Newcastle-Ottawa Quality Assessment Scale modified by authors for this work ( [32]; See Supplementary Materials Table S1). Although most of the studies revealed low or medium bias risk, it is noteworthy that half of the studies in our review were secondary analyses. This, with a high probability, indicated a substantially higher bias risk than the original study. Another possible source of bias for this review could be the fact that two of the included studies defined the study group as "multiple gestation", without specifying whether they were only twin pregnancies, and, in addition, the study by Boucoiran et al. covered also five triplets [33,38,40]. Nevertheless, the authors assumed that this should not be a source of a major bias for this review. Another factor potentially interfering with the collective interpretation of studies' results was the variety of methods used. Firstly, the studies were conducted on populations, which differed in terms of factors that might presumably affect the initial risk of PE like race and ethnicity, maternal age, body mass index, or gestational age. Similarly, the diversity of assays, machines, as well as specimens used, might also be the source of potential bias. Abbreviations: AUC-area under the receiver operating characteristic curve; DR-detection rate; FPR-false-positive rate; GW-gestational week; MAP-mean arterial pressure; PAPP-A-pregnancy-associated protein A; PE-preeclampsia, PlGF-placental growth factor; PP13-placental protein 13; SBP-systolic blood pressure; sFlt-1-soluble fms-like tyrosine kinase-1; UTPI-uterine artery pulsatility index.

Differences in Serum Concentrations of sFlt-1, PlGF, and sEng Between Singleton and Twin Pregnancies
The studied biomarkers are largely of placental origin. Hence, differences in their concentrations between singleton and twin gestations were hypothesized. Seven out of ten papers included in this systematic review reported such differences [33,37,38,41,44,46,47]. Two other studies did not include singleton pregnancies [35,45], and, in the third one, no such comparisons were made [40]. sFlt-1 maternal serum levels were reported to be statistically higher in women with twin pregnancies in comparison with singleton gestations [33,37,38,41,47]. Saleh et al. found a similar relationship only in a group unaffected by PE, while, in the PE group, no significant differences were observed [47]. PlGF concentrations were also higher in twins compared to singletons [33,37,38,41,44,46,47]. Dröge et al. and Saleh et al. reported PlGF levels to be higher in the group of preeclamptic women with a twin gestation in comparison with women with a singleton pregnancy [41,47], whereas Francisco et al. observed higher PlGF concentrations only in dichorionic twins [46]. Only two studies examined differences in sEng concentrations. One showed its higher level in twin pregnancies [33], and the other revealed no significant differences between singletons and twins [37].
Several other articles reporting differences in PlGF, sFlt-1, and sEng concentrations between singletons and twins have been published to date. Since they did not consider preeclamptic twin pregnancies, they were not included in the review. According to Maynard et al., the maternal sFlt-1 level was higher in multiple gestations compared to high-risk singletons, and PlGF was significantly higher in multiples before 31 weeks of gestation [49]. Furthermore, Bdolah et al. found higher sFlt-1 in twins but similar PlGF concentrations in twins and singletons [50]. Faupel-Badger et al. reported higher sFlt-1 and sEng. However, lower PlGF levels were noted in twin gestations [51].

Serum Concentrations of sFlt-1, PlGF, and sEng in Monochorionic and Dichorionic Twin Gestations
Regarding the studies included in the systematic review, only three investigated the correlation between chorionicity and angiogenic factor concentrations. In two, no significant differences in PlGF, sFlt-1, nor sEng levels between MC and DC twin gestations were found [37,44]. Francisco et al. reported higher PlGF in DC twin pregnancies that did not develop PE, in comparison to singleton pregnancies, while no such differences in MC twins [46]. Nevertheless, other reports of such a correlation are available in the literature. Faupel-Badger et al. found the concentrations of sFlt-1 and sEng to be higher in monochorionic than in dichorionic twin gestations after adjustment for gestational age [51]. Cowans and Spencer reported PlGF concentrations to be 41% higher in DC, but only 16% higher in MC compared to singleton pregnancies [52].

Differences in Serum Concentrations of sFlt-1, PlGF, and sEng between Non-Preeclamptic Twin Pregnancies and Those Who Developed PE
Five out of six studies, which investigated differences between sFlt-1 levels in PE and non-PE twin gestations, reported significantly higher maternal serum levels in those mothers who developed PE [33,35,37,38,41]. The above studies included women in all trimesters of pregnancy -the first [37], the second [33,38], and the third [35,41]. Interestingly, Sanchez et al. found statistical differences only in PE and non-PE twins conceived with assisted reproductive technologies, without significance in spontaneously conceived ones [37]. Conversely, Boucoiran et al. found significant differences only between 24 and 26 weeks of pregnancy, whereas they were absent between 12 and 18 weeks of gestation [38].
Nine out of ten studies included in the review investigated PlGF levels in PE and non-PE women. Seven papers reported lower PlGF concentrations in women with a twin gestation who developed PE compared to non-PE multiples [33,38,40,41,[44][45][46]. Alike for sFlt-1, differences were shown for different gestational ages ranging from late first to the early third trimester. Two other studies did not report such differences [35,47]. Both of those studies concerned women in the third trimester of twin pregnancies with suspicion or symptoms of PE forming the control groups, which, to some extent, maybe the reason for obtaining different results.
One of the selected articles investigated sEng concentrations in women with twin pregnancies with and without PE [33]. sEng was significantly higher between 31 to 35 weeks of gestation in subjects who later developed preeclampsia.
Only Dröge et al. analyzed the relationship between biomarkers and the severity of PE [41]. Authors found significantly higher sFlt-1, lower PlGF plasma levels, and, consequently, higher sFlt-1/PlGF ratio compared to healthy controls in both mild and severe groups. Among the included articles, we did not find direct comparisons of biomarkers' concentrations between early and late-onset PE in twin gestations.

The Usefulness of Selected Angiogenic Factors in the Prediction of PE in Twin Gestations
A practical aspect of differences in the concentration of biomarkers allowed the creation of algorithms for long-or short-term prediction of PE. Currently, PlGF, sFlt-1:PlGF ratio, and The Fetal Medicine Foundation calculators are widely used [26,[53][54][55].
Six out of ten studies included in this review attempted to determine the usefulness of selected biomarkers in the prediction or the diagnosis of PE in twin pregnancies [35,38,41,[44][45][46][47]. A vast majority of the algorithms proposed by authors were characterized by promising parameters.
Rana et al. built an algorithm based on the highest systolic blood pressure, proteinuria, gestational age, and sFlt-1:PlGF ratio for the prediction of PE-related adverse outcomes in twins within the next 2 weeks and received area under the receiver operating characteristic curve (AUC) of 0.85 for a 75% false-positive rate (FPR). The authors also noted that AUC was slightly higher (0.87) when only women <34 weeks of gestation were included in the analysis. Moreover, they proposed sFlt1:PlGF >75 as more suitable for the diagnosis of PE in this group, instead of sFlt1:PlGF >85 that had been validated as the optimal cut-off for singletons [36,56]. Dröge et al. suggested a cut-off point of ≥53 as the most optimal for twins; however, for both, women <34 and ≥34 gestational weeks [41]. Moreover, Saleh et al. recently published a paper, which evaluated the usefulness of sFlt-1:PlGF ratio of ≤38 in the prediction of the short-term absence of PE in late second and third-trimester twin pregnancies [47]. Five out of thirteen preeclamptic twin gestations had sFlt-1:PlGF ratio >38, and four out of eight non-preeclamptic twin gestations had sFlt-1:PlGF ratio ≤38. Thus, the authors concluded that such a ratio is not applicable for ruling out PE in twin pregnancies.
Boucoiran et al. showed that the prediction of PE development in current pregnancy was possible with high accuracy (AUC 0.81, 10% FPR) at the early stages of pregnancy, between 12 and 18 weeks, only with the use of PlGF serum levels [38]. Furthermore, complex algorithms were published, using various combinations of maternal factors (history, mean arterial pressure, uterine artery pulsatility index) and biochemical markers (PlGF, serum pregnancy-associated plasma protein-A, placental protein 13, free β-human chorionic gonadotropin, and α-fetoprotein) [44][45][46]. The authors of research, which was conducted in the largest group of patients from all papers included in our review, created an algorithm with the use of maternal risk factors, PlGF, uterine artery pulsatility index (UTPI), and mean arterial pressure (MAP) in the first trimester of pregnancy. In a mixed population (singletons and twins) with the risk cut-off of 1 in 75 for PE at <37 gestational weeks, the detection rate of PE at <32, <37, and <42 weeks in singletons was accounted for 91%, 77%, and 57% with screen-positive rate (SPR) of 13%. The analogous values for twins were 100%, 99%, and 97%, but with a high SPR of 75% [46]. Moreover, the AUC values for this algorithm were decreasing with the increasing gestational age at the delivery of the twin pregnancies complicated by PE, as shown in Table 4. Only one paper investigated the possible accuracy of sEng in predicting PE. The adjusted odds ratio (aOR) of developing PE for a twofold increase in sEng was accounted for 2.98 (95% confidence interval (CI) 1.44-6.36). As regards other biomarkers, the results were as follows: sFlt-1 aOR = 2.07 (95% CI 1.15-3.89), PlGF aOR = 0.50 (95% CI 0.30-0.83), and sFlt1+sEng/PlGF aOR = 2.18 (95% CI 1.46-3.32) [33].

Conclusions
This systematic review is the most recent summary of available knowledge about maternal serum levels of PlGF, sFlt-1, and sEng and the risk of PE in twin pregnancies. Most of the studies included in the review reported statistical differences in maternal serum levels of discussed biomarkers between singleton and twin gestations and between PE and non-PE ones. Several proposed algorithms for the prediction and diagnosis of PE seem promising. However, according to that current knowledge, determination of their usefulness in diagnosing or ruling out the PE in all twin pregnancies is not possible. Moreover, the reference ranges of analyzed biomarkers in uncomplicated twin pregnancies are also not available. Large prospective studies with repeatable measurements at different weeks of pregnancy, as well as comparisons of maternal characteristics, chorionicity, onset, and severity, are needed for improvement of the algorithms. Subsequently, their prompt introduction into the clinical practice and further assessment of the real-life performance could help improve the quality of care for women with twin pregnancies.