Plasma concentrations of angiogenic/anti-angiogenic factors have prognostic value in women presenting with suspected preeclampsia to the obstetrical triage area: a prospective study

Abstract

Objective: To prospectively determine the prognostic value of maternal plasma concentrations of placental growth factor (PlGF), soluble endoglin (sEng) and soluble vascular endothelial growth factor receptors-1 and -2 (sVEGFR-1 and -2) in identifying patients with suspected preeclampsia (PE), who require preterm delivery (PTD) or develop adverse outcomes.

Study design: This prospective cohort study included 85 consecutive patients who presented to the obstetrical triage area at 20–36 weeks with a diagnosis of “rule out PE.” Patients were classified as: 1) those who remained stable until term (n = 37); and 2) those who developed severe PE and required PTD (n = 48). Plasma concentrations of PlGF, sEng and sVEGFR-1 and -2 were determined by ELISA.

Results: Patients with PlGF/sVEGFR-1 ≤0.05 multiples of the median (MoM) or PlGF/sEng ≤0.07 MoM were more likely to deliver preterm due to PE [adjusted odd ratio (aOR) 7.4 and 8.8], and to develop maternal (aOR 3.7 and 2.4) or neonatal complications (aOR 10.0 and 10.1). Among patients who presented <34 weeks of gestation, PlGF/sVEGFR-1 ≤ 0.035 MoM or PlGF/sEng ≤0.05 MoM had a sensitivity of 89% (16/18), specificity of 96% (24/25) and likelihood ratio for a positive test of 22 to identify patients who delivered within 2 weeks. The addition of the PlGF/sVEGFR-1 ratio to standard clinical tests improved the sensitivity at a fixed false-positive rate of 3% (p = 0.004) for the identification of patients who were delivered due to PE within 2 weeks. Among patients who had a plasma concentration of PlGF/sVEGFR-1 ratio ≤0.035 MoM, 0.036-0.34 MoM and ≥0.35 MoM, the rates of PTD <34 weeks were 94%, 27% and 7%, respectively.

Conclusions: The determination of angiogenic/anti-angiogenic factors has prognostic value in patients presenting to the obstetrical triage area with suspected PE for the identification of those requiring preterm delivery and at risk for adverse maternal/neonatal outcomes.

• Maternal outcome
• neonatal outcome
• placental growth factor
• pregnancy
• soluble endoglin
• soluble vascular endothelial growth factor receptor-1

Introduction

Preeclampsia (PE), one of the “great obstetrical syndromes” [1–4], is characterized by systemic intravascular inflammation [5–7], endothelial cell dysfunction [8–11], excessive thrombin generation [12–18], an anti-angiogenic state [19–40] and is often associated with multiple organ involvement [41–60]. However, PE is fundamentally a placental disease [37,61–63] which manifests itself, in most cases, by involvement of the vascular (i.e. hypertension) and renal systems (i.e. proteinuria). The diagnosis of PE typically requires the presence of hypertension and proteinuria [64]. Sometimes, patients who are subsequently diagnosed with PE present initially with either hypertension [65–68] or proteinuria [69–73], and may have some signs or symptoms indicating yet, they do not meet the full diagnostic criteria for PE [74]. Indeed, even in some cases of HELLP syndrome, patients present without hypertension and proteinuria, and only develop these two abnormalities if pregnancy is allowed to continue [43,75]. In addition, when PE involves organs other than the vascular system and the kidney, it becomes the “great imitator” [76]. Such cases are frequently misdiagnosed and initially thought to have disorders unrelated to pregnancy. Hence, variability in clinical presentation presents a management challenge.

While approximately 1 in 10 pregnant women exhibit some signs and/or symptoms associated with PE, only 20–25% of these patients will eventually be diagnosed as having the disease [77–79]. Women with signs and/or symptoms associated with PE (i.e. elevated blood pressure, headache, abdominal pain, edema, etc.) are frequently referred to an obstetrical triage area for assessment of multiple maternal organ systems and fetal involvement [80,81]. The standard work-up includes blood pressure determination, urine analysis for protein, a platelet count, determination of liver enzymes in peripheral blood, and a blood smear to detect schistocytes [64,79]. The prognostic performance of these tests in determining which patients will develop PE, require preterm delivery (PTD), or have maternal and/or neonatal morbidity is poor [82–97]. Consequently, many patients with signs and/or symptoms associated with PE are hospitalized for observation. Those diagnosed with PE in preterm gestation often undergo long-term hospitalization or frequent monitoring as outpatients [98,99]. Thus, the lack of adequate biomarkers to predict disease progression and adverse outcomes in patients with suspected PE [82,84,100] results in substantial financial burden to the health care system (i.e. frequent visits, hospitalizations, intensive laboratory surveillance, serial ultrasound examination and antepartum testing) [101–105].

To address the need for biomarkers with high predictive value, we previously reported the results of a retrospective study which indicated that plasma concentrations of angiogenic/anti-angiogenic factors [placental growth factor (PlGF), soluble vascular endothelial growth factor receptors (sVEGFR) -1, soluble endoglin (sEng)] have prognostic value in patients with “suspected PE” at preterm gestations who present to the obstetrical triage area [106]. We found that, based on the results of these analytes, it is possible to stratify patients into those at high, moderate and low risk of requiring PTD within 2 weeks [106].

The purpose of this study was to determine whether the results of the retrospective study could be replicated prospectively. Namely, to examine whether maternal plasma concentrations of angiogenic/anti-angiogenic factors have prognostic value in identifying patients presenting to the obstetrical triage area with the diagnosis “rule out PE” and at risk for PTD due to PE or develop other complications within 2 weeks of presentation.

Materials and methods

Study design

This prospective cohort study included singleton pregnancies who presented to the obstetrical triage unit of Hutzel Women’s Hospital, Detroit, MI, for the diagnosis of “suspected PE” from July 2010 to March 2011. Inclusion criteria were: (1) singleton pregnancy; (2) no known major fetal anomaly; (3) gestational age 20–36 6/7 weeks; and (4) signs/symptoms of PE (elevated blood pressure, proteinuria, headache, blurred vision, epigastric or right upper quadrant pain and edema). All participants provided written informed consent and donated a blood sample for research purposes.

Based on our prior study [106], we estimated that approximately 40% of patients presenting with suspected PE prior to 37 weeks of gestation would have an abnormal biomarker ratio using thresholds selected based on ROC curve analysis. Accordingly, we estimated that to have 80% power to detect a 2-fold difference in the rate of preterm delivery due to severe PE (between patients with abnormal biomarker ratio compared to those with normal biomarker ratio) with a 5% limit on Type I error probability, 85 patients would be required. Considering patients to be excluded for loss to follow-up (10%), we aimed to enroll 93 patients into this study.

Demographic data, medical, surgical and obstetrical history were recorded. Delivery outcomes were reviewed and used to classify patients into two groups: patients who remained stable until delivery at ≥37 weeks (Group I) and patients who developed severe PE requiring PTD (Group II). Maternal and neonatal complications within two weeks after presentation to the triage area were recorded as well. The collection and use of samples for research purposes was conducted under protocols approved by the Institutional Review Boards of Wayne State University and the Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Department of Health and Human Services (NICHD/NIH/DHHS).

Clinical definitions

PE was defined as the new onset of hypertension and proteinuria that developed after 20 weeks of gestation. Hypertension was defined as systolic ≥140 or diastolic blood pressure ≥90 mm Hg, measured on two occasions, 4 h to 1 week apart. Proteinuria was defined as a urine protein of ≥300 mg in a 24-h urine collection, or two random urine specimens obtained 4 h to 1 week apart containing ≥1+ by dipstick [64].

“Maternal morbidity” was considered to be present if one or more of the following complications was observed: (1) eclampsia; (2) Hemolysis elevated liver enzymes low platelet count syndrome; (3) pulmonary edema; (4) oliguria (<400 mL/day); (5) renal insufficiency (creatinine > 1.2 mg/dL); (6) abruptio placentae (defined as clinical vaginal bleeding and uterine tenderness with or without placental pathologic findings); (7) elevated liver enzymes or thrombocytopenia; (8) antepartum admission to the maternal intensive care unit (MICU); or (9) fetal death. “Adverse neonatal outcome” was defined as one or more of the following complications: (1) respiratory distress syndrome [107,108]; (2) intraventricular hemorrhage [109]; (3) bronchopulmonary dysplasia [110]; or (4) neonatal death.

Patients were managed by attending physicians blinded to the results of the analytes under study. The approach in our institution is that patients with mild PE diagnosed prior to 37 weeks of gestation are managed expectantly, – with weekly platelet counts and liver function tests in addition to non-stress tests. Patients diagnosed with severe PE are hospitalized to assess maternal/fetal conditions and the need for delivery. Individually, patients are managed at the discretion of the attending physician.

Sample collection and immunoassays

Venipuncture was performed and blood was collected into tubes containing EDTA. Samples were centrifuged and stored at −70 °C. Maternal plasma concentrations of PlGF, sVEGFR-1, sEng and sVEGFR-2 were determined by sensitive and specific immunoassays (R&D Systems, Minneapolis, MN) as previously described [106]. The inter- and intra-assay coefficients of variation (CV) obtained were as follows: PlGF, 6.0 and 4.8%, respectively; sEng, 2.3 and 4.6%, respectively; sVEGFR-1, 1.4 and 3.9%, respectively; and sVEGFR-2, 2.0 and 4.0%, respectively. The sensitivity of the assays were as follows: PlGF, 9.52 pg/ml; sVEGFR-1, 16.97 pg/ml; sEng, 0.08 ng/ml; and sVEGFR-2, 19.01 pg/ml.

Statistical analysis

Normality of data was assessed by inspection of histograms and the Kolmogorov–Smirnov test. Continuous variables were compared using the Mann-Whitney U test or unpaired Student's t-test depending upon the distribution of the data. Contingency tables and χ² tests were used to compare proportions. Methods used to construct maternal plasma angiogenic/anti-angiogenic factors and ratio concentration multiples of the median (MoM) have been previously reported [106]. Briefly, quantile regression was used to calculate the median concentration of a particular analyte and the ratio conditional upon gestational age in 180 participants in a longitudinal study at Hutzel Women’s Hospital who had uncomplicated pregnancies and delivered at term. MoMs were calculated by dividing the observed analyte concentrations by the previously determined expected median ratio conditional on gestational age. Dichotomization thresholds were selected in a prior study [106] based on inspection of receiver operating characteristic (ROC) curves.

Logistic regression was used to assess the magnitude of association between biomarkers and selected pregnancy outcomes. Multiple logistic regression models were constructed to account for potentially confounding factors selected based on clinical information (gestational age at presentation, average mean arterial blood pressure, chronic hypertension, combined parity and/or history of PE, and tobacco use). Potentially confounding factors were entered with respect to time-order into a full model. Model reduction was performed based on the plausibility of regression coefficients, association between independent and dependent variables, and the magnitude of change in the main effect parameter estimates; independent variables whose relationship with the dependent variable yielded a p-value >0.10 were excluded from the final models. Diagnostic performance metrics were calculated and correlated sample non-parametric statistical techniques were used to compare area under the ROC curves (AUC) [111]. Differences in sensitivity at a fixed false positive rate were tested using McNemar’s tests, and exact p-values are reported.

Survival analysis and Cox proportional hazard models were utilized to examine the relationship between sampling-to-delivery interval and biomarker MoMs, adjusting for the above-mentioned potential confounders. Patients who delivered preterm due to causes other than PE had the interval between triage and delivery treated as a censored observation. Analysis was conducted with SPSS V.15 (SPSS Inc., Chicago, IL) and SAS version 9·3 (Carry, NC, U.S.A). A p value <0.05 was considered significant.

Results

During the study period, 122 visits to the obstetrical triage unit for the diagnosis of “suspected PE” met the inclusion criteria. After exclusion of repeated visits (n = 19), those who were lost to follow up (n = 8), and those who declined to participate (n = 10), 85 patients met the inclusion criteria. Among them, 37 remained stable until term (Group I) and 48 developed severe PE requiring PTD (Group II).

Demographic, clinical characteristics, pregnancy outcome and plasma concentrations of angiogenic/anti-angiogenic factors

Demographic/clinical characteristics, maternal morbidity and adverse neonatal outcomes are presented by group in Table 1. Patients who developed severe PE and subsequently required a PTD had a higher mean systolic and diastolic blood pressure and a higher frequency of maternal morbidity and/or adverse neonatal outcomes than those who delivered at term (each p < 0.001).

Table 1. Demographic/clinical characteristics, maternal morbidity and adverse neonatal outcome by study groups.

	Stable until delivery at term (n = 37)	Developed PE requiring preterm delivery (n = 48)	p
Age (years)	26 (19.5–32.5)	26.5 (21.0*33.8)	0.6
African American	34 (91.9%)	44 (91.7%)	0.9
Nulliparous	16 (43.2%)	20 (41.7%)	0.8
Previous preeclampsia	7/21 (33.3%)	12/28 (42.9%)	0.5
Smoking	6 (16.2%)	5 (10.4%)	0.4
Body mass index (kg/m^2)	32.0 ± 7.7	30.9 ± 8.2^α	0.5
Chronic hypertension	18 (48.6%)	22(45.8%)	0.8
Pregestational diabetes	2 (5.4%)	3 (6.3%)	0.9
Chief complaint in triage area
Elevated blood pressure	31 (84%)	47 (97.9%)	0.04
Proteinuria	8 (21.6%)	10 (20.8%)	0.9
Edema	2 (5.4%)	7 (14.6%)	0.3
Headache	15 (40.5%)	22 (45.8%)	0.6
GA at venipuncture (weeks)	35.0 (32.3–35.9)	32.9 (30.1–35.4)	0.1
Average blood pressure in triage area
Systolic (mmHg)	135 ± 11	158 ± 16	<0.001
Diastolic (mmHg)	82 ± 8	97 ± 11	<0.001
Mean arterial (mmHg)	100 ± 8	117 ± 11	<0.001
Proteinuria in triage area	11 (29.7%)	37 (77.1%)	<0.001
Interval to delivery (days)	25 (13–46)	3 (1–9)	<0.001
GA at delivery (weeks)	38.3 (37.1–39.0)	34.7 (32.1–36.1)	<0.001
Birthweight (g)	3056 ± 483	1866 ± 625	<0.001
<10%	6 (16.2%)	19 (39.6%)	0.02
Adverse maternal and neonatal outcome
Indicated preterm delivery <34 weeks	–	20 (41.7%)	<0.001
Fetal death	1 (2.1%)	–	0.4
Maternal morbidity
Eclampsia (at term)	1 (2.7%)	–	0.4
HELLP	–	1 (2.1%)	1.0
Pulmonary edema	–	2 (4.2%)	0.5
Oliguria	–	6 (12.5%)	0.03
Serum Creatinine >1.2 mg/dL	–	5 (10.4%)	0.06
Abruptio placentae	2 (5.4%)	5 (10.4%)	0.5
Elevated AST	3 (8.1%)	7 (14.6%)	0.5
Thrombocytopenia	–	3 (6.3%)	0.3
Antepartum admission to ICU	–	7 (14.6%)	0.02
Composite adverse maternal outcome	5 (13.5%)	24 (50%)	<0.001
Neonatal complications
Fetal or neonatal death	1 (2.1%)	–	0.4
RDS	3 (8.1%)	8 (16.7%)	0.3
BPD	–	3 (6.3%)	0.2
IVH	2 (5.4%)	4 (8.3%)	0.5
Adverse neonatal outcome	4 (10.8%)	10 (20.8%)	0.2
Composite maternal morbidity  and/or adverse neonatal outcome	8 (21.6%)	29 (60.4%)	<0.001

α: n = 47; value expressed as median (interquartile), mean ± SD or number (%).

PE, preeclampsia; GA, gestational age; AST, aspartate amniotransferase; ICU, intensive care unit; RDS, respiratory distress syndrome; BPD, bronchopulmonary dysplasia; IVH, intraventricular hemorrhage.

Three patients in Group I delivered preterm: two due to spontaneous preterm delivery, and another one due to anti-D isoimmunization. One patient in Group I was diagnosed with chronic hypertension and superimposed preeclampsia at 35 weeks and 4 days. She was discharged home on the same day and had placental abruption with a fetal death at 37 6/7 weeks.

The median MoM plasma concentrations of angiogenic/anti-angiogenic factors differed significantly by patient groups (each p < 0.005; Figure 1). A plasma concentration of PlGF/sVEGFR-1 of ≤ 0.05 MoM and that of PlGF/sEng ≤0.07 MoM had the highest likelihood ratio for a positive test (8.2). In contrast, a plasma concentration of PlGF ≤0.4 MoM had the lowest likelihood ratio for a negative test, 0.13 for the identification of patients who developed severe PE requiring a PTD (Table 2).

Figure 1. Median maternal plasma Multiple of Median (MoM) concentrations of soluble vacular endothelial growth factor receptor (sVEGFR)-1, soluble endoglin (sEng), sVEGFR-2, placental growth factor (PlGF), PlGF/sVEGFR-1 ratio and PlGF/sEng ratio in patients who subsequently developed severe PE requiring preterm delivery and those who remained stable until term [sVEGFR-1: median 4.46 MoM, interquartile range (IQR) 2.5–9.8 MoM vs. 1.5 MoM, IQR 0.9–2.4 MoM; p < 0.001; sEng: median 3.3 MoM, IQR 2.1–5.6 MoM vs. 1.2 MoM, IQR 0.9–1.7 MoM; p < 0.001; sVEGFR-2: median 0.69 MoM, IQR 0.5–0.9 MoM vs. 0.9 MoM, IQR 0.7–1.1 MoM; p < 0.001; PlGF: median 0.15 MoM, IQR 0.08–0.23 MoM vs. 0.47 MoM, IQR 0.3–1.1 MoM; p < 0.001; PlGF/sVEGFR-1: median 0.03 MoM, IQR 0.008–0.07 MoM vs. 0.4 MoM, IQR 0.2-0.9 MoM; p < 0.001; PlGF/sEng: median 0.04 MoM, IQR 0.01–0.09 MoM vs. 0.4 MoM, IQR 0.2–0.9 MoM; p < 0.001].

Table 2. Diagnostic performance of plasma concentrations of angiogenic/anti-angiogenic factors for the identification of patients who subsequently developed severe preeclampsia requiring preterm delivery.

	Cut-off (MoM) from ROC curves in previous publication	Sensitivity (%)	Specificity (%)	LR for a positive test (95% CI)	Posterior probability (95% CI)	LR of a negative test (95% CI)	Posterior probability (95% CI)
sVEGFR-1	≥3.46	58.3 (28/48)	86.5 (32/37)	4.3 (1.9–10)	85% (71–93)	0.48 (0.3–0.7)	38% (31–47)
sEng	≥2.7	64.6 (31/48)	91.9 (34/37)	7.9 (2.6–24)	91% (77–97)	0.39 (0.3–0.6)	34% (25–43)
PlGF	≤0.4	91.7 (44/48)	62.2 (23/37)	2.4 (1.6–3.7)	76% (67–83)	0.13 (0.05–0.4)	14% (6–31)
sVEGFR-2	≤0.74	56.3 (27/48)	67.6 (25/37)	1.7 (1.02–2.9)	69% (57–79)	0.65 (0.4–0.9)	46% (36–55)
PlGF/sVEGFR-1	≤0.05	66.7 (32/48)	91.9 (34/37)	8.2 (2.7–25)	91% (78–97)	0.36 (0.2–0.6)	32% (24–42)
PlGF/sEng	≤0.07	66.7 (32/48)	91.9 (34/37)	8.2 (2.7–25)	91% (78–97)	0.36 (0.2–0.6)	32% (24–42)

Prevalence of severe preeclampsia requiring preterm delivery = 56.5% (48/85).

Value expressed as % (n/N).

MoM, multiples of the median; ROC, receiver-operating characteristic; LR, Likelihood Ratio; CI, confidence interval; PlGF, Placental Growth Factor, sEng, soluble endoglin, sVEGFR-1, soluble vascular endothelial growth factor receptor-1, sVEGFR-2, soluble vascular endothelial growth factor receptor-2.

Plasma concentrations of PlGF/sVEGFR-1 ≤0.05 MoM and PlGF/sEng ≤0.07 MoM were significantly associated with indicated PTD due to PE, maternal morbidity, composite adverse neonatal outcome, and composite maternal morbidity and/or neonatal outcome after adjustment for potential confounders (Table 3). The ratio of angiogenic/anti-angiogenic factors performed better than the mean arterial pressure or laboratory tests in identifying patients who required delivery within 2 weeks when the analysis was restricted to patients who presented prior to 34 weeks of gestation (Figure 2A; n = 43) compared to analysis of the full cohort (Figure 2B; n = 85). Therefore, subsequent analysis was restricted to patients who presented before 34 weeks of gestation.

Figure 2. Area under the Receiver-Operating Characteristic (ROC) curves for angiogenic/anti-angiogenic factor ratio Multiples of the Median (MoM) compared to mean arterial pressure and combined standard tests (mean arterial pressure, serum creatinine, serum uric acid, aspartate aminotransferase, platelet count) for the identification of patients who delivered within 2 weeks (A) Restricted to patients who presented before 34 weeks of gestation; and (B) Included all patients in the cohort.

Table 3. Likelihood (unadjusted and adjusted) of each outcome according to angiogenic/antiangiogenic factors ratio.

	Outcome				Unadjusted			Adjusted-I (Full)			Adjusted-II (Reduced)
	No		Yes
Outcome - Predictor	%	n/N	%	n/N	OR	95% CI		OR	95% CI		OR	95% CI
Indicated Preterm Delivery due to PE
PlGF/sVEGFR-1 MoM ≤0.05	8.1	3/37	66.7	32/48	22.7	6.0	85.2	7.1	1.3	39.7	7.4*	1.6	34.1
PlGF/sEng MoM ≤0.07	8.1	3/37	66.7	32/48	22.7	6.0	85.2	8.0	1.5	42.9	8.8*	1.9	40.2
Maternal morbidity
PlGF/sVEGFR-1 MoM ≤0.05	32.1	18/56	58.6	17/29	3.0	1.2	7.6	2.2	0.7	7.3	3.7^α	1.4	10.0
PlGF/sEng MoM ≤0.07	32.1	18/56	58.6	17/29	3.0	1.2	7.6	2.3	0.7	7.1	2.4^β	0.8	7.3
Adverse neonatal outcome
PlGF/sVEGFR-1 MoM ≤0.05	33.8	24/71	78.6	11/14	7.2	1.8	28.2	6.4	0.9	44.9	10.0^γ	2.0	48.9
PlGF/sEng MoM ≤0.07	33.8	24/71	78.6	11/14	7.2	1.8	28.2	5.9	0.9	37.7	10.1^γ	2.1	49.8
Composite maternal morbidity and/or  adverse neonatal outcome
PlGF/sVEGFR-1 MoM ≤0.05	17.4	4/23	65.0	13/20	8.8	2.1	36.4	8.6	0.8	90.7	2.8*	0.9	8.8
PlGF/sEng MoM ≤0.07	17.4	4/23	65.0	13/20	8.8	2.1	36.4	5.9	0.8	45.7	3.1*	1.0	9.5

OR, odds ratio, CI, confidence interval; PE, preeclampsia; PlGF, placental growth factor, sEng, soluble endoglin; sVEGFR-1, soluble vascular endothelial growth factor receptor-1; ORs represent the likelihood of outcome among subjects exhibiting analyte ratio concentrations below centile cutoff relative to the likelihood of outcome among subjects at or above cutoff; Adjusted-I (Full) models included age, African American ethnicity, chronic hypertension, mean arterial pressure, and combined nulliparity/previous preeclampsia; Adjusted- II (Reduced) models include final multivariable adjustment after excluding factors that were not significantly associated with the dependent variable from the Full model- variables included are indicated as follows:

*adjusted for average mean arterial pressure.

α: adjusted for nulliparity, previous preeclampsia.

β: adjusted for nulliparity, previous preeclampsia, average mean arterial pressure.

γ: adjusted for gestational age at blood sampling.

Findings restricted to patients presenting prior to 34 weeks of gestation

Among patients who presented to the obstetrical triage area before 34 weeks, 18 (42%) delivered due to PE and 13 (30.2%) developed adverse maternal/neonatal outcomes within 2 weeks (Table 4). A ratio of PlGF/sVEGFR-1 ≤0.035 MoM and a ratio of PlGF/sEng of ≤0.05 MoM achieved a sensitivity of 89% (16/18), a specificity of 96% (24/25) and a likelihood ratio for a positive and negative test of 22 and 0.12, respectively, in identifying patients who required delivery within 2 weeks (Table 5). Table 6 displays the magnitude of association between biomarker MoMs and selected outcomes. The ratios of PlGF/sVEGFR-1 ≤0.035 MoM and PlGF/sEng ≤0.05 MoM were significantly associated with indicated PTD within 2 weeks (aOR 53 and 58, respectively) and composite maternal morbidity and/or adverse neonatal outcome within 2 weeks (aOR 19.5) after adjustment for potential confounders.

Table 4. Demographic/clinical characteristics and adverse maternal and neonatal outcomes of the patients who presented to the triage area at <34 weeks according to interval to delivery within and more than 2 weeks.

	Delivery more than 2 weeks (n = 25)	Delivery within 2 weeks (n = 18)	p
Age (years)	29 (20–36)	24.5 (19–31)	0.1
African American	23 (92.0%)	16 (88.9%)	1.0
Nulliparous	11 (44.0%)	9 (50.0%)	0.8
Previous preeclampsia	4/14 (28.6%)	5/9 (55.6%)	0.4
Smoking	3 (12.0%)	1 (5.6%)	0.6
Body mass index (kg/m^2)	34.0 ± 7.1	31.2 ± 9.5	0.4
Chronic hypertension	16 (64.0%)	7 (38.9%)	0.1
Pre-gestational diabetes	1 (4.0%)	0	1.0
Chief complaint in triage area
Elevated blood pressure	22 (88%)	18 (100%)	0.3
Proteinuria	7 (28.0%)	4 (22.2%)	0.7
Edema	2 (8.0%)	1 (5.6%)	1.0
Headache	10 (40.0%)	7 (38.9%)	0.9
GA at blood sampling (weeks)	30.6 (27.6–33.0)	31.5 (29.0–32.3)	0.9
Average blood pressure in triage area
Systolic (mmHg)	145 ± 17	160 ± 18	0.005
Diastolic (mmHg)	85 ± 9	102 ± 14	<0.001
Mean arterial (mmHg)	105 ± 8	117 ± 11	<0.001
Proteinuria in triage area	3 (20%)	19 (67.9%)	0.003
Platelet count	237 (190–278)	192 (172–235)	0.09
AST	19 (14–34)	23 (18–34)	0.1
Serum creatinine (mg/dL)	0.6 (0.6–0.7)	0.7 (0.6–0.7)	0.08
Serum uric acid (mg/dL)	3.4 (3.0–3.8)α	4.9 (3.9–5.5)β	0.001
Hospitalized	15 (60%)	18 (100%)	0.002
Interval to delivery (days)	41 (22–65)	3 (2–6)	<0.001
GA at delivery (weeks)	37 (35.3–38.9)	32 (29.7–32.7)	<0.001
Birthweight (g)	2767 ± 730	1283 ± 384	<0.001
Adverse maternal and neonatal outcomes
Indicated preterm delivery <34 weeks	2 (8%)	18 (100%)	<0.001
Maternal morbidity at 2 weeks
Eclampsia	–	–
HELLP	–	–
Pulmonary edema	–	–
Oliguria	–	3 (16.7%)	0.06
Serum Creatinine >1.2 mg/dL	–	1 (5.6%)	0.4
Abruptio placentae	–	1 (5.6%)	0.4
Elevated AST	–	4 (22.2%)	0.025
Thrombocytopenia	–	–	–
Antepartum admission to ICU	–	3 (16.7%)	0.06
Fetal Death	–	–
Composite adverse maternal outcome	–	9 (50%)	<0.001
Neonatal complications at 2 weeks
Neonatal death	–	–	–
RDS	–	6 (33.3%)	0.003
BPD	–	3 (16.7%)	0.06
IVH	–	3 (16.7%)	0.06
Adverse neonatal outcome	–	8 (44.4%)	<0.001
Composite maternal morbidity and/or  adverse neonatal outcome	–	13 (72.2%)	<0.001

Value expressed as median (interquartile), mean ± SD or number (%); GA, gestational age; AST, aspartate amniotransferase; ICU, intensive care unit; RDS, respiratory distress syndrome; BPD, bronchopulmonary dysplasia; IVH, intraventricular hemorrhage.

Table 5. Diagnostic performance of plasma concentrations of angiogenic/anti-angiogenic factors among patients who presented at <34 weeks of gestation for the identification of patients who subsequently delivered within 2 weeks.

	Cut-off (MoM) derived from ROC curves in previous publication	Sensitivity (%)	Specificity (%)	LR for a positive test (95% CI)	Posterior probability (95% CI)	LR of a negative test (95% CI)	Posterior probability (95% CI)
sVEGFR-1	≥4.0	78 (14/18)	96 (24/25)	19 (2.8–135)	93% (67–99)	0.23 (0.1–0.6)	14% (7–28)
sEng	≥3.0	83 (15/18)	96 (24/25)	21 (3.0–144)	94% (69–99)	0.17 (0.06–0.5)	11% (4–26)
PlGF	≤0.15	72 (13/18)	92 (23/25)	9 (2.3–35)	87% (63–96)	0.3 (0.1–0.6)	18% (9–32)
sVEGFR-2	≤0.74	72 (13/18)	72 (18/25)	2.6 (1.3–5.2)	65% (48–79)	0.39 (0.2–0.8)	22% (11–38)
PlGF/sVEGFR-1	≤0.035	89 (16/18)	96 (24/25)	22 (3.2–153)	94% (70–99)	0.12 (0.03–0.4)	8% (2–24)
PlGF/sEng	≤0.05	89 (16/18)	96 (24/25)	22 (3.0–153)	94% (70–99)	0.12 (0.03–0.4)	8% (2–24)

Prevalence of patients who subsequently delivered within 14 days = 41.9% (18/43).

Value expressed as % (n/N); MoM, multiples of the median; ROC, receiver-operating characteristic; LR, Likelihood Ratio; CI, confidence interval; PlGF, Placental Growth Factor, sEng, soluble endoglin, sVEGFR-1, soluble vascular endothelial growth factor receptor-1; sVEGFR-2, soluble vascular endothelial growth factor receptor-2.

Table 6. Likelihood (unadjusted and adjusted) of selected pregnancy outcomes according to angiogenic/antiangiogenic factor ratio in patients who presented to the triage area at <34 weeks of gestation.

	Outcome				Unadjusted			Adjusted
	No		Yes
Outcome - Predictor	%	n/N	%	n/N	OR	95% CI		OR	95% CI
Delivery within 2 weeks
PlGF/sVEGFR-1 MoM ≤0.035	4	1/25	88.9	16/18	192.0	16.0	>999.999	53.7	5.9	485.8
PlGF/sEng MoM ≤0.05	4	1/25	88.9	16/18	192.0	16.0	>999.999	58.3	6.3	538.5
Week 2 maternal morbidity
PlGF/sVEGFR-1 MoM ≤0.035	26.5	9/34	88.9	8/9	22.2	2.4	203.3	10.1	1.3	80.8
PlGF/sEng MoM ≤0.05	26.5	9/34	88.9	8/9	22.2	2.4	203.3	9.4	1.2	72.0
Week 2 adverse neonatal outcome					Insufficient variation to model
PlGF/sVEGFR-1 MoM ≤0.035	25.7	9/35	100	8/8
PlGF/sEng MoM ≤0.05	25.7	9/35	100	8/8
Week 2 composite maternal morbidity  and/or adverse neonatal outcome
PlGF/sVEGFR-1 MoM ≤0.035	16.7	5/30	92.3	12/13	60.0	6.3	571.9	19.5	2.5	150.1
PlGF/sEng MoM ≤0.05	16.7	5/30	92.3	12/13	60.0	6.3	571.9	19.5	2.6	144.1

OR, odds ratio, CI, confidence interval; PlGF, placental growth factor, sEng, soluble endoglin, sVEGFR-1, soluble vascular endothelial growth factor receptor-1. ORs represent the likelihood of outcome among subjects exhibiting analyte ratio concentrations below centile cutoff relative to the likelihood of outcome among subjects at or above cutoff; Adjusted models included accounted for the association between mean arterial pressure and the outcome variable, Firth’s penalized maximum likelihood estimation was also used due to small sample size.

The biomarkers under study achieved a greater AUC (0.94) compared to combined standard clinical and laboratory tests (AUC 0.91) for the identification of patients who required PTD within 2 weeks. Table 7 displays the sensitivity for the identification of patients who required delivery within 2 weeks or those who developed composite adverse maternal and/or neonatal outcomes within 2 weeks by pre-specified false-positive rate (3%, 5% and 10%) for biomarker MoMs, blood pressure and standard laboratory tests for PE. The addition of the PlGF/sVEGFR-1 ratio to standard tests significantly improved the sensitivity at a fixed false-positive rate of 3% (p = 0.004) and marginally improved sensitivity at a fixed false-positive rate of 5% (p = 0.06) for the identification of patients who would be delivered due to PE within 2 weeks (Figure 3). Therefore, angiogenic/anti-angiogenic factor MoMs outperformed the standard tests used in the triage area for the prediction of these adverse outcomes, given that they had a higher sensitivity than standard laboratory tests.

Figure 3. Addition of the PlGF/sVEGFR-1 ratio to standard tests significantly improved the sensitivity at a fixed false-positive rate of 3% (p = 0.004) and marginally improved sensitivity at a fixed false-positive rate of 5% (p = 0.06) for the identification of patients who would be delivered due to preeclampsia within 2 weeks.

Table 7. Sensitivities of each outcome according to angiogenic/antiangiogenic factors ratio in patients who presented to the triage area at <34 weeks of gestation while fixing different false positive rates (FP).

	Sensitivities (%) at fixed false positive rates (%)
Outcome - Predictor	FP: 3%	95% CI		FP: 5%	95% CI		FP: 10%	95% CI
Delivery with 14 Days
PlGF/sVEGFR-1	72.2	46.5	90.3	88.9	65.3	98.6	88.9	65.3	98.6
PlGF/sEng	77.8	52.4	93.6	88.9	65.3	98.6	88.9	65.3	98.6
Systolic blood pressure (mmHg)	22.2	6.4	47.6	44.4	21.5	69.2	44.4	21.5	69.2
Serum AST	16.7	3.6	41.4	16.7	3.6	41.4	16.7	3.6	41.4
Serum creatinine (mg/dL)	11.1	1.4	34.7	22.2	6.4	47.6	55.6	30.8	78.5
Mean arterial pressure (mmHg)	5.6	0.1	27.3	44.4	21.5	69.2	44.4	21.5	69.2
Diastolic blood pressure (mmHg)	5.6	0.1	27.3	44.4	21.5	69.2	55.6	30.8	78.5
Serum uric acid (mg/dL)	0.0			11.8	1.5	36.4	70.6	44.0	89.7
Platelet counts	0.0			0.0			5.6	0.1	27.3
Proteinuria	0.0			0.0			0.0
Week 2 composite maternal morbidity  and/or adverse neonatal outcome
PlGF/sVEGFR-1	46.2	19.2	74.9	92.3	64.0	99.8	92.3	64.0	99.8
PlGF/sEng	61.5	31.6	86.1	92.3	64.0	99.8	92.3	64.0	99.8
Serum uric acid (mg/dL)	15.4	1.9	45.4	23.1	5.0	53.8	46.2	19.2	74.9
Systolic blood pressure (mmHg)	30.8	9.1	61.4	53.8	25.1	80.8	61.5	31.6	86.1
Diastolic blood pressure (mmHg)	53.8	25.1	80.8	53.8	25.1	80.8	61.5	31.6	86.1
Mean arterial pressure (mmHg)	15.4	1.9	45.4	53.8	25.1	80.8	61.5	31.6	86.1
Serum creatinine (mg/dL)	30.8	9.1	61.4	53.8	25.1	80.8	53.8	25.1	80.8
Serum AST	23.1	5.0	53.8	23.1	5.0	53.8	23.1	5.0	53.8
Platelet counts	0.0			7.7	0.2	36.0	7.7	0.2	36.0
Proteinuria	0.0			0.0			0.0

PlGF, placental growth factor; sEng, soluble endoglin; sVEGFR-1, soluble vascular endothelial growth factor receptor-1; AST, aspartate aminotransferase.

Time-to-event analysis confirmed that a low maternal PlGF/sVEGFR-1 ratio (≤0.035 MoM) or PlGF/sEng ratio (≤0.05 MoM) was associated with a shorter interval-to-delivery after adjusting for potential confounders – gestational age at presentation, average mean arterial blood pressure, chronic hypertension, combined parity and history of PE, and tobacco use [hazard ratio = 25 (95% CI 6.4–104.0) and 29.4 (95% CI 6.9–125.1); respectively; Figure 4].

Figure 4. Kaplan–Meier survival curves by angiogenic/anti-angiogenic ratio Multiple of Median (MoM) cut-offs of patients who delivered prior to 37 weeks of gestation (right censored). Patients who delivered preterm due to causes other than PE had the interval between blood draw and delivery treated as a censored observation. (A = PlGF/sVEGFR-1; B = PlGF/sEng).

Table 8 displays the disposition (hospitalization versus discharge home) of the patient after visiting the triage area, rate of PTD and adverse perinatal outcomes according to the previously proposed grading of severity criteria based on plasma MoM concentrations of the PlGF/sVEGFR-1 ratio and the PlGF/sEng ratio [106]. The cut-off values for patients at high-risk were based on ROC curves for the identification of patients who delivered within 2 weeks and those for patients at low-risk were based on ROC curves for the identification of patients who remained stable until term [106]. Almost 40% of patients presenting prior to 34 weeks had analyte ratio MoM concentrations classified as Zone 3 (high-risk), while 35–42% were characterized as Zone 1 (low-risk) and 20–25% were considered Zone 2 (moderate-risk). Thirty-three (77%) patients who presented to the triage area with “suspected PE” at less than 34 weeks were hospitalized. Among them, one-third (11/33) were classified as Zone 1 (low-risk). Nine of these eleven patients did not have hypertension in the severe range, and five complained of headache in the triage area. A single patient within Zone 1 (6.7%) delivered within 2 weeks and had an abnormal liver function test. This patient had chronic hypertension and was referred to the triage area at 31 4/7 weeks because of an elevated blood pressure (161/86 mmHg).

Table 8. Disposition of the patient after visiting the triage area (hospitalization versus discharge home), rate of preterm delivery and adverse perinatal outcomes according to the previously proposed grading of severity criteria based on plasma MoM concentrations of the PlGF/sVEGFR-1 ratio and the PlGF/sEng ratio.

	PlGF/sVEGFR-1 Ratio (MoM)			PlGF/sEng Ratio (MoM)
	≥0.35, Zone 1 (n = 15)	0.0351-0.349 Zone 2 (n = 11)	≤0.035 Zone 3 (n = 17)	≥0.30 Zone 1 (n = 18)	0.051-0.299 Zone 2 (n = 8)	≤0.05 Zone 3 (n = 17)
Home (Column %)	4 (26.7%)	5 (45.5%)	1 (5.9%)	6 (33.3%)	4 (50.0%)	0
Hospitalized (Column %)	11 (73.3%)	6 (54.5%)	16 (94.1%)	12 (66.7%)	4 (50.0%)	17 (100%)
Risk category Number (Row %)	Low	Intermediate	High	Low	Intermediate	High
	15 (34.9%)	11 (25.6%)	17 (39.5%)	18 (41.9%)	8 (18.6%)	17 (39.5%)
Delivery <34 weeks	6.7% (1/15)	27.3% (3/11)	94.1% (16/17)	5.6% (1/18)	25.0% (2/8)	100% (17/17)
Delivery within 14 days	6.7% (1/15)	9.1% (1/11)	94.1% (16/17)	5.6% (1/18)	12.5% (1/8)	94.1% (16/17)
Delivery within 7 days	6.7% (1/15)	– (0/11)	88.2% (15/17)	5.6% (1/18)	12.5% (1/8)	82.4% (14/17)
Maternal morbidity (within 2 weeks)	6.7% (1/15)	0	47.1% (8/17)	5.6% (1/18)	0 (0/8)	47.1% (8/17)
Adverse neonatal outcome (within 2 weeks)	0	0	47.1% (8/17)	0	0	47.1% (8/17)
Composite maternal morbidity and/or  adverse neonatal outcome (within 2 weeks)	6.7% (1/15)	0	70.6% (12/17)	6.7% (1/15)	0	70.6% (12/17)

Value expressed as % (n/N) PlGF, Placental Growth Factor, sEng, soluble endoglin, sVEGFR-1, soluble vascular endothelial growth factor receptor-1.

In contrast, among patients classified as high-risk (Zone 3) by either a PlGF/sVEGFR-1 ratio ≤0.035 MoM or a PlGF/sEng ratio ≤0.05 MoM, the rate of PTD before 34 weeks was 94% (16/17) and 100% (17/17), respectively; the rate of PTD within 2 weeks was 94% (16/17) for each ratio, and the rate of composite adverse maternal and/or neonatal outcome within 2 weeks of visiting the triage area was 70% (12/17) for each ratio (Table 8).

Discussion

Principal findings of the study

1) Patients who presented to the obstetrical triage area with “suspected PE” and subsequently required a PTD had a significantly lower mean plasma MoM concentration of angiogenic/anti-angiogenic ratio than those who remained stable until term; 2) low plasma concentrations of the PlGF/sVEGFR-1 ratio (≤0.05 MoM) or the PlGF/sEng ratio (≤0.07 MoM) were significantly associated with indicated PTD and composite maternal morbidity and/or adverse neonatal outcome in patients presenting at <37 weeks of gestation; however, the ratio of angiogenic/anti-angiogenic factors performed better than standard laboratory tests in identifying patients who required delivery within 2 weeks when the analysis was restricted to patients who presented prior to 34 weeks of gestation; and 3) a PlGF/sVEGFR-1 ratio ≤0.035 MoM and a PlGF/sEng ratio ≤0.05 MoM were significantly associated with composite maternal morbidity and/or adverse neonatal outcome, and a shorter interval-to-delivery than those above these cutoffs among patients who presented <34 weeks of gestation.

Meaning of the findings

The results of this study add further support to the concept that it is possible to classify patients presenting to the obstetrical triage area before 34 weeks of gestation with a suspected diagnosis of PE, based upon the results of PlGF, sVEGFR-1 and sEng [106,112]. We had previously proposed a three-tier classification of patients based on the results of these analytes and their ratios [106]. This classification yielded encouraging results, but needed replication. This is important because data-driven techniques for prediction generally perform well on the data set used for the development of a particular algorithm, but they generally underperform when applied to an independent sample. This is a key clinical challenge because algorithms are developed for application to new patients, not in patients from whom the models were derived [113]. The current study was designed to test the value of the proposed classification system constructed based on retrospective data. This method classified patients into three groups: those at high, moderate and low risk for indicated PTD and/or adverse maternal or neonatal outcome.

The results of the current study independently confirm that our classification system has prognostic value in identifying patients at risk for PTD or adverse maternal/neonatal outcome. Among patients presenting prior to 34 weeks of gestation, those classified as “high risk” (Zone 3) had a 94% probability of PTD and 70% developed maternal and/or neonatal complications within 2 weeks. On the other hand, patients classified as “low risk” (Zone 1) were unlikely to deliver within 2 weeks (6.7%) or develop maternal and/or neonatal complications (6.7%). We propose that such patients can be discharged home after the managing physicians have performed a thorough evaluation and considered the likelihood of compliance with a regimen for monitoring and follow-up. For example, in this cohort, 9 of 11 patients who presented to the triage area before 34 weeks of gestation had maternal plasma concentrations of angiogenic/anti-angiogenic factors in Zone 1, and did not have blood pressure in the severe range or abnormal laboratory tests, but were hospitalized. All delivered more than 2 weeks (range 20–111 days) after the initial presentation.

The management of patients at intermediate risk (Zone 2) remains a challenge: 25–27% delivered prior to 34 weeks of gestation, although few (9–12%) developed maternal complications. Additional biomarkers will be required to improve the prognostic assessment of these patients. Alternatively, serial testing may prove to be informative.

Recently, a model to predict maternal complications in patients admitted with the diagnosis of PE has been proposed [114]. The current study had a different objective since we aimed to identify patients who required PTD and those who developed maternal and/or neonatal complications utilizing information in the triage area prior to admission.

Strengths and limitations of the study

The strengths of this study are its prospective nature and the independent replication of the previous findings [106]. Other than our first retrospective study of angiogenic/anti-angiogenic factors in the obstetrical triage area [106], three studies reported an association between maternal plasma angiogenic/anti-angiogenic factor concentrations in women who presented to the obstetrical triage area with suspected PE and adverse maternal and neonatal outcomes [112,115,116]. However, indicated preterm birth was included in the composite maternal and/or neonatal complications in these studies. We reported this outcome separately from other complications since this is an observational study and institutional management policies may influence this outcome (i.e. policy that required delivery of all patients with severe PE at 34 weeks of gestation). Moreover, we have gone beyond reporting an association between the results of biomarker determination and outcome by proposing an algorithm for follow up and management.

The limitations of this study include the lack of statistically significant differences in AUCs observed in this study when comparing the performance of standard clinical parameters and the additional inclusion of angiogenic/anti-angiogenic factors. However, it is important to note that comparisons of the AUCs are insensitive measures of biomarker utility [117]. In other clinical fields (such as cardiovascular disease), the limitations of ROC curves as a tool for model evaluation have been recently highlighted, and the comparison of the AUC is considered an insensitive measure of improvement in model accuracy [117–119]. Second, this study utilized the conventional ELISA assay method rather than a rapid automated assay system whose results could be available for clinical decision-making in the obstetrical triage area. However, previous studies have demonstrated that measures obtained by rapid point-of-care and conventional ELISA assay systems are highly correlated [120–123]. Third, this is an observational study, and further investigation is required to determine whether implementation of management policies based upon the risk assessment would prove to be cost-effective and achieve the clinical goals.

Conclusion

Maternal plasma concentrations of angiogenic/anti-angiogenic factors have prognostic value for patients who present with suspected PE before 34 weeks of gestation. We propose that these biomarkers allow prospective stratification of patients whose illness will progress sufficiently to require a preterm delivery or develop adverse maternal and/or neonatal outcomes.

Declaration of interest

This research was supported, in part, by the Perinatology Research Branch, Division of Intramural Research, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Department of Health and Human Services (NICHD/NIH); and, in part, with Federal funds from NICHD, NIH under Contract No. HHSN275201300006C.

Notes

¹Presented at the 58th Annual Meeting of the Society for Gynecologic Investigation, March 21–24, 2012, San Diego, CA.