Clinical Performance of The sFlt-1/PlGF Test For The Prediction of Preeclampsia Among Asymptomatic and Symptomatic Pregnant Women in Estonia: A Nested Case-Control Study


 Background: Pre-eclampsia (PE) is a pregnancy complication manifesting as new-onset hypertension and other maternal organ dysfunction after 20th gestational weeks. The study aimed to evaluate the applicability and limitations of the maternal serum soluble fms-like tyrosine kinase-1/placental growth factor (sFlt-1/PlGF) test in a clinical setting for the prediction of PE among symptomatic and asymptomatic pregnant women. There is limited knowledge on the performance of this test in asymptomatic women and thus, its value for screening purposes to predict PE is not confidently settled.Methods: The study group comprised of 215 patients developing either PE (n=29) or gestational hypertension/proteinuria (n=22) or representing controls (n=164). Patients had been sampled within 180-291 gestational days in the presence (symptomatic, n=31) or absence (asymptomatic, n=216) of PE-alerting symptoms. Serum samples collected during prospective cohort study ’Happy Pregnancy’ at the Women’s Clinic, Tartu University Hospital, Estonia, and they were analyzed using the BRAHMS sFlt‑1 Kryptor/BRAHMS PlGF plus Kryptor PE ratio test (Thermo Fisher Scientific, Henningdorf, Germany). The results were interpreted using recommendations by Stepan et al 2015 Ultrasound Obstet Gynecol. Results: The assignment of ’Rule-out PE’ (sFlt-1/PlGF ratio <38) had a negative predictive value >99% for four weeks for both asymptomatic and symptomatic women. Among 29 sera assigned to the ‘Rule-in PE’(sFlt-1/PlGF >85/110), only 18 pregnancies (62.1%) eventually developed PE. For four weeks period, the overall PE detection rate was 83% for asymptomatic and 50% for symptomatic pregnancies. Pregnancies receiving false predictions of PE risk based on sFlt-1/PlGF ratio represented either cases with an isolated small-for-gestational age fetus or blood sampling after 34 gestational weeks.Conclusions: The first study in Estonian patients confirmed high reliability of the proposed cut-off value sFlt-1/PlGF <38 as a “Rule-out PE” cut-off value for two weeks in both symptomatic and asymptomatic pregnancies. The test’s limitation in our clinical setting appeared to be high false positive rate in pregnancies with other placental pathologies than PE or due to physiological increase in sFlt-1/PlGF in late gestation. For correct prediction of PE, more specific recommendations are urgently needed for the application and interpretation of the test in routine clinical practice.


Background
Pre-eclampsia (PE) is a pregnancy disorder de ned by the new-onset of hypertension after 20 gestational weeks accompanied by proteinuria or other maternal organ dysfunction. In extreme cases it may rapidly progress to the form of life-threatening eclampsia.
Additionally to severe damage to maternal health, preeclamptic pregnancies also cause fetal complications such as growth restriction or intrauterine death. PE affects 2 to 5% of pregnancies worldwide and is one of the major causes of maternal and perinatal morbidity and mortality. 1,2,3 Current clinical predictive and diagnostic toolset for PE includes routine monitoring of blood pressure and urinary protein levels. 4 However, women with the developing PE may also miss any alerting symptoms. Extensive research has been devoted to identify maternal serum biomarkers for early PE prediction, facilitating timely intervention and preventive measures.
The pathogenesis of PE is associated with the imbalance of pro-and antiangiogenic factors such as sFlt-1 (soluble fms-like tyrosine kinase 1) and PlGF (placental growth factor). In symptomatic women the sFlt-1/PlGF test can help to distinguish between PE development and a transient elevation of blood pressure. 5 As increased sFlt-1/PlGF ratio in maternal serum can be detected up to 5 weeks before the onset of PE symptoms, these biomarkers represent potential screening tool among asymptomatic women. 6 Although currently no effective third trimester prevention for preeclampsia is available, early detection of the high risk for developing PE helps to select women who need more frequent follow-up visits for timely diagnosis and referral to specialized centres. The application of glucocorticoids and neuroprotection help to alleviate the problems caused by PE-induced preterm birth. Recently, manifestation of pravastatin since 36 th g. week has attracted attention as a potential preventative or therapeutic candidate for late onset PE, supported by pilot clinical trials 7 . Commercially offered solutions for the detection of sFlt-1/PlGF biomarker ratios to predict PE development have been rapidly introduced and rule-in and rule-out thresholds for the PE prediction have been established. 5,8 However, international community has concluded that before use in daily clinical practice further studies on asymptomatic and symptomatic pregnant women are needed. 9 This study aimed to retrospectively evaluate the applicability and limitations of the maternal serum sFlt-1/PlGF test in a routine clinical setting in Estonia to predict PE development during the second half of pregnancy. The potential of the sFlt-1/PlGF test as a PE screening tool was evaluated by comparing the outcomes obtained for the asymptomatic women and patients presenting at least one PE symptom at sampling. The cases with false-positive, false-negative and inconclusive test results were critically assessed in the context of their detailed clinical history.

Methods 'Happy Pregnancy' cohort of pregnant women
The analyzed patients were recruited, and the respective clinical data and biomaterials were collected during a monocentric prospective 'Happy Pregnancy' project (full name 'Development of novel non-invasive biomarkers for fertility and healthy pregnancy': PI: M.L.). The project has been approved by the Ethics Committee of Human Research of the University Clinic of Tartu, Estonia (permissions no. 221/T-6, 17.12.2012 and 286/M- 18, 15.10.2018) and was carried out in compliance with the Helsinki Declaration. A written informed consent to participate in the study was obtained from each individual prior to recruitment. All participants were of white European ancestry and living in Estonia.
During March 2013 -August 2015, in total 2,334 unselected pregnant women had been enrolled at their rst antenatal visit at the Women´s Clinic, Tartu University Hospital, Estonia (clinical PI: K.R.). The pregnancy follow-up was based on the national guidelines approved at 2011 by the Estonian Gynaecologists` Society. 10 The collected clinical and epidemiological data, as well as applied clinical criteria are speci ed in Supplementary Methods S1, Additional File 1. Among the nal cohort of 2,257 eligible women 61 (2.7%) had eventually developed preeclampsia (PE), 52 (2.3%) had been diagnosed with gestational hypertension (GH) and 17 (0.8%) with proteinuria (Table 1). In total 138 pregnancies (5.9%) resulted in a preterm birth (PTB), a delivery before 37 gestational weeks (< 259 gestational days).
Serum sampling for research purposes was performed across gestation in parallel with regular clinical visits and blood-draw based tests according to routine pregnancy monitoring procedures ( Figure S1, Additional File 2). At each blood draw, the symptoms alerting to PE were assessed and documented. The diagnosis of PE followed the international guidelines at the time of recruitment (ISSHP, 2014), simultanous co-occurence of a new-onset hypertension (HTN; ≥140/ ≥90 mmHg) after 20 gestational weeks and proteinuria or other signs of maternal organ dysfunction. 1 According to the absence or presence of the signs alerting to PE at blood draw or up to four days later, serum samples collected from the pregnant women were classi ed as asymptomatic or symptomatic (GH, proteinuria, PE).
The sample set and patient group analyzed in the current study Prospectively drawn 252 serum samples from the 'Happy Pregnancy' cohort participants were retrospectively subjected to the commercially offered sFlt-1/PlGF test (Thermo Fisher Scienti c). The samples had been collected from 215 pregnant women with different pregnancy outcome (PE, n=29; isolated GH or proteinuria, n=22; controls, n=164) 1-69 days before delivery and covered 180 -291 gestational days ( Table 1). The formation of the study group centered around the available serum samples drawn from PE cases (n=42 samples) and isolated GH/proteinuria cases (n=30 samples) after 25 th gestational weeks until term. Respective gestational and maternal age matched control serum samples (n=180) representing uncomplicated pregnancies were selected from the cohort biobank, aiming to match two to three control serums per each included sample of a PE or isolated GH/proteinuria case ( Figure S2, Additional le 3; Figure S3, Additional le 4). The study group included 35/215 pregnancies with the representation of several consecutive blood samples, enabling to assess also gestational dynamics of biomarkers (Table S1, Additional le 5). From among 29 women, who eventually developed PE, 11 had been sampled twice and one patient thrice. Among 22 women with a later diagnosis of GH/proteinuria, six had been subjected to blood draw two and one patient three times. Control group included 16/164 pregnancies that had been sampled twice.

Implementation of sFlt-1/PlGF test for the cryopreserved serum samples
After blood draw, serum samples had been immediately separated and stored at -80°C for maximum 1.5 years (Supplementary Methods S2, Additional le 6). After thawing the samples were aliquoted and sent on dry ice to the Synlab Germany service laboratory (Leinfelden, Germany) without no information regarding the pregnancy course and outcome. Freeze-thaw stability of serum sFlt-1 and PlGF concentrations has been previously reported. 11,12 Concentrations of serum sFlt-1 and PlGF were analyzed using the BRAHMS sFlt-1 Kryptor/BRAHMS PlGF plus Kryptor PE ratio test (Thermo Fisher Scienti c, Henningdorf, Germany). Reportable measurement values of both biomarkers were derived from the service provider for 247/252 (98%) serum samples (failure rate 2%; Table S2, Additional le 7). Failed samples were excluded from further analysis.
Prediction of the PE development was based on the estimated sFlt-1/PlGF ratio. As no validated cut-offs are published specially for Kryptor, Elecsys® assay validated cut-offs were used as recommended by Stepan et al. 13 Serum samples were classi ed as 'Rule-out PE at least within 1 week' when the calculated sFlt-1/PlGF ratio was < 38. 'Rule-in PE' decision alerting to the disease was applied for the samples exhibiting sFlt-1/PlGF ratio >85 when drawn before 34 gestational weeks; and > 110 when drawn from 34 gestational weeks onwards. Values of the sFlt-1/PlGF within these thresholds (38-85/110) were considered as inconclusive. 13 Statistical analysis Summary estimates of the data (median, minimum -maximum; mean ± standard deviation) were calculated and all statistical tests were implemented using the STATA software ver. 13.1 (StataCorp TX, USA). For the correct assignment of small-for-gestational-age (SGA) diagnosis, the fetal growth calculator based on INTERGROWTH-21 st Project was applied to convert the newborn birth weight into gestational age and sex-adjusted z-scores. 14 A newborn was categorized as SGA in case the z-score for sex-and gestational age adjusted birthweight was lower than -1. To compare groups Wilcoxon rank-sum test was used for continuous variables and Chi 2 test for categorical variables. Statistical signi cance was de ned as P <0.05.

Results
Characteristics of the sample set selected from the 'Happy Pregnancy' cohort The study group comprised of 215 pregnancy cases (9.5%) recruited to the 'Happy Pregnancy' cohort, including 29 PE, 22 isolated GH/proteinuria and 164 uncomplicated gestations. To maximize the number of analyzed cases of each pregnancy scenario, the study group was enriched for the patients diagnosed with PE (13.5 vs 2.7% compared to the full cohort and GH/proteinuria during the index pregnancy (10.2% vs 3.1%; Table 1). In the analyzed sample set, compared to control group, PE subgroup included a higher proportion of women with the rst pregnancy (79.3% versus 50.6%, P = 3.0×10 -3 ) and multiparous women with PE/GH in their previous pregnancy (17% vs 14.6%, P = 1.5×10 -2 ). Patients with GH/proteinuria exhibited higher BMI (mean 28.6 vs 24.6 kg/m 2 , P =1.2 ×10 -3 ) and were diagnosed with gestational diabetes more often (22.7% vs 7.3% P =1.5 ×10 -2 ). There was no statististical in maternal age and the proportion of pregnancies concieved by in vitro fertilization procedure between the study subgroups.
The serum samples with reported sFlt-1/PlGF result (n=247) were categorized as either 'symptomatic' (n=31) or 'asymptomatic' (n=216) depending on the presence or absence of PE symptoms at the blood draw or up to 4 days ( Table 2). Statistically signi cant differences (p<0.05) between the subgroups are indicated as following: *Rule-out vs Inconclusive; † Rule-out vs Rule-in; ‡ Inconclusive vs Rule-in. § Two consecutive serum samples from the same patient.
For the assignement of small-for-gestational-age (SGA) diagnosis, the fetal growth calculator based on INTERGROWTH-21 st Project was applied to convert the newborn birth weight into gestational age and sex-adjusted z-scores. 14 Newborn was categorized as SGA in case the Z-score for sex-and gestational age adjusted birthweight was lower than -1.

Prognostic yield of sFlt-1/PlGF test in symptomatic and asymptomatic pregnancies
The pro le of the sFlt-1/PlGF test outcome differed between the samples drawn from symptomatic and asymptomatic patients ( Figure  1, Table 2). Although 'Rule-out PE' was the most prevalent sFlt-1/PlGF test outcome in both groups, it was far more frequent assignment among asymptomatic patients (79 vs 42%, P = 8.7×10 -6 ), whereas outcome category 'Rule-in PE' was enriched among symptomatic cases (9% vs 29%, P =4.0×10 -3 ). The 'Rule-out' cut off sFlt-1/PlGF ratio <38 resulted in the estimated negative predictive value (NPV) >99% for up to four weeks for both, symptomatic and asymptomatic patients (Table 3).  Figure 1, Table 2). False positive 'Rule-in PE' was assigned to 11 cases, 2/31 (6.5%) and 9/216 (4.2%) samples from symptomatic and asymptomatic cases, respectively (Table 3). In false-positive cases the serum samples were taken at 210-291 gestational days, and the patients delivered within 1-54 days after sampling either vaginally (n=7) or by C-section (n=4) ( Table S4, Additional le 9). The estimated positive predictive value (PPV) for 'Rulein PE' in next 28 days for asymptomatic cases was 50% and until term 55%, whereas for symptomatic cases it was 67% and 78%, respectively. The overall detection rate (DR) to predict the onset PE during four weeks was 83% for asymptomatic and 50% for symptomatic pregnancies. When extending the period until delivery, the DRs were 41% and 47%, respectively (Table 3).
'Inconclusive' test results (38≤sFlt-1/PlGF≤85/110) were assigned to nine of 31 samples (29%) drawn from symptomatic patients compared to 12% (25/221) in asymptomatic cases (P = 8.0×10 -3 ) (Figure 1, Table 2). Among the nine symptomatic patients, six cases (67%) either already had the clinical diagnosis (n=2) or developed PE 1-29 days later (n=4). Only 20% (25/221) of the 'Inconclusive' category samples representing asymptomatic cases had been drawn from pregnancies with a later onset of PE. When exploring the application of sFlt-1/PlGF ratio ≥38 to predict manifestation of PE for up to 4 weeks, the DR was nearly maximum, 100% for women with and 92% without PE-alerting symptoms (Table 3). However, introduction of this non-stringent cut-off would result in a high false positive rate (FPR). Among symptomatic pregnancies, the estimated FPR for the PE prediction during the next week was 41% and for the next month 32% (Table 3). Respective FPRs among asymptomatic women were 20% and 17%.
The test performance depends on the dynamics of sFlt-1 and PlGF across gestation With advancing gestational age, a general trend towards increased serum sFlt-1, decreased PlGF and consequently, higher sFlt-1/PlGF estimates was observed in all investigated pregnancy outcomes (Figure 2A-C). A rise in sFlt-1/PlGF ratio was detected in nearly all individual cases with two or more available consecutive serum samples, irrespective of the pregnancy scenario (n=35; Figure 2D). The most prominent increase in sFlt-1/PlGF ratio was detected in normal gestation near delivery (preterm vs term samples: 4.0 vs 37.8; P = 2.9×10 -10 ; Table S5, Additional le 10). At term, there were no statistical differences in the distributions of sFlt-1 measurements between controls, preeclampsia and GH/proteinuric pregnancies. In total 11 of 38 (29%) term pregnancy samples exhibited sFlt-1/PlGF ratio >110, but only 4 of them developed PE ( Figure 2C), underlying the limited value of this test near term.
Despite overall similar trends, the gestational dynamics of sFlt-1 and PlGF differed among the three studied pregnancy scenarios (Table   4). Already during 25-33 rd gestational weeks, the future PE cases exhibited signi cantly higher sFlt-1 measurements compared to both, control pregnancies (median 2788 vs 1178 pg/ml; P = 7.9x10 -7 ) and those diagnosed later with GH/proteinuria (2788 vs 1290; P = 2.9x10 -3 ). However, there was no difference between the two non-PE groups. Interestingly, reduced serum PlGF levels were observed not only among future PE cases compared to controls (median PlGF 70 vs 311 pg/ml; P = 1.9x10 -8 ), but also in pregnancies developing isolated GH/proteinuria (156 vs 311 pg/ml; P=9.7x10 -3 ). It can be speculated that whereas low concentration of circulating PlGF represents a general hallmark of suboptimal pregnancy physiology, only increased sFlt-1 serum levels and consequently high sFlt-1/PlGF ratio drives the development of PE. Data are given as median, minimum-maximum. Parameter distributions between groups were compared using Wilcoxon rank-sum test.
sFlt-1, soluble fmf-like tyrosine kinase-1; PlGF, placental growth factor; GH, gestational hypertension; PTN, proteinuria Challenges in applying sFlt-1/PlGF test outcome in clinical decision-making Although in several clinical cases the sFlt-1/PlGF test has a clear bene t in clinical decision making and prediction of PE (Cases A-C, Table 5), there is a challenge to handle inconclusive and false-positive test results in clinical routine. In one hand, an inconclusive test outcome may assist to identify the patients needing careful monitoring for early detection of PE (Case D), but it could also cause unreasonable anxiety and unnecessary visits (Case E). The most common causes behind false positive sFlt-1/PlGF test results in asymptomatic patients are isolated SGA (Case G) and closeness of delivery (Case H), especially when sampling occurred after 37 th gestational weeks. Complications other than PE were noted in patients with sFlt-1/PlGF >110 near or at term such as uterine rupture (FP-4; Table S4, Additional le 9), fetal distress leading to emergency C-section (FP-7). High sFlt-1/PlGF ratio without development to PE could lead to earlier unnecessary interventions such as induced preterm delivery that may be not optimal for this pregnancy (Case I, Table 5). Consistent with previous reports, the study con rmed a high NPV (>99%) for the sFlt-1/PlGF<38 estimate to rule-out PE for up to two weeks in the presence or absence of PE symptoms. 5,15,16,17 In our study, 13 samples resulting sFlt-1/PlGF<38 but drawn from cases who later developed PE, the period between sampling and PE onset exceeded 16 days.
Elevated estimate of sFlt-1/PlGF often resulted in PE in symptomatic women, but at the expense of high FPR. Although cut-off ≥38 could identify correctly all cases with the PE onset in four weeks, almost every second or third woman with isolated HTN or proteinuria would be falsely classi ed as a possible PE pregnancy. However, when limiting PE diagnosis only with values exceeding the diagnostic cut-off (sFlt-1/PlGF>85/110), based on out data, FPR can be reduced to 17% and 14% in one and four weeks, respectively, but may fail to con rm the clinical diagnoses of PE in a third of the patients.
All false positive samples (sFlt-1/PlGF >85/110) from asymptomatic patients observed in our study were taken after 33 g. weeks, most of them at term. The study data showed that a normal rise in the sFlt-1/PlGF ratio with advancing gestation limits the diagnostic accuracy of the test already after 33-34 gestational weeks and especially close to delivery. Increasing sFlt-1/PlGF ratio towards the end of pregnancy has been also reported before. 6,18 In our study, only four of 11 term pregnancy samples exhibiting sFlt-1/PlGF ratio >85/110, developed further to PE. As most PE cases manifest at term, a clearly interpretable diagnostic/predictive test for this time window is especially warranted. 19 As a solution, inclusion of additional biomarkers has been suggested to improve the prediction accuracy, especially close to term. 20,21 Another scenario behind increased sFlt-1/PlGF ratio in asymptomatic women is fetal growth restriction (FGR) and other placental syndromes, such as premature birth, intrauterine fetal death, abruption of placenta etc. 22,23 Remarkable proportion of adverse maternal and neonatal outcomes among patients with false positive test for PE taken at mid-pregnancy has also been pointed out by Black et al. 24 In concordance, more than third of false positive cases in our study suffered from SGA or fetal distress.
Apart from PE prediction sFlt/PlGF ratio test could be also useful pointing out the pregnant women increased risk for other placenta related adverse outcome. However, the main challenge for sFlt/PlGF ratio test as a predictive tool remains the timing of the test since the prediction time for PE is limited up to 4 weeks and the usefulness of the test is starting to decrease after 34 weeks due to physiological increase in sFlt/PlGF ratio.
It has been proposed that sFlt-1 alters normal endothelial cells by hindering the angiogenetic effect of PlGF. 6,25 We observed low PlGF both among patients with isolated GH/proteinuria and PE, but the level of sFlt-1 before 34 gestational weeks was higher only in cases destined to develop PE (Table 4). It can be speculated that lower sFlt-1 level prevents from generalized endothelial damage leading to manifestation of PE. Our nding is in line with several other studies. 26,27,28 However, opposite results have been demonstrated. 29,30 The physiological increase of sFlt-1 from 34 g. weeks onwards but constantly low PlGF due to placental stress shift the sFlt-1/PlGF ratio among patients with isolated GH/proteinuria and the discrimination these cases from PE may be more complicated.
Our study had also some limitations. The applied methodological approach is restricted by the the lack of platform-speci c validated clinical cut-off values for the BRAHMS sFlt-1 Kryptor/BRAHMS PlGF plus Kryptor PE ratio test. The currently applied cut-off values have been validated for the Elecsys ® assay (Roche Diagnostics, Mannheim, Germany). Although several studies have shown highly comparable results in clinical performance of Kryptor and Elescys ® , a recent study has raised a concern that the Kryptor test measures PlGF concentrations ~20% lower that Elescys ® assay. 31,32,33 Consequently, this results in a higher sFlt-1/PlGF estimates and, therefore may increase FPR for PE prediction when applying the Elescys ® -platform based cut-off values.

Conclusions
Despite a relatively small number of samples analysed our study, we could replicate the major ndings from larger studies. 5,15,16,17 sFlt-1/PlGF ratio estimated for the third trimester serum measurements is highly reliable for the exclusion of PE for at least two weeks. However, elevated sFlt-1/PlGF level could not be currently considered as an essential part for the diagnosis nor prediction of PE. Although the recommended diagnostic cut-off often predicted PE development, the presence of PE alerting symptom(s), SGA or additional placental pathologies as well as advanced gestational age at sampling must be considered in clinical decision making. Clearer guidelines for the clinical practice, e.g. timing of the test, further management, clear cut-offs for different platforms, are needed.
Abbreviations FPR -false positive rate