Study design and patients
We present a prospective, observational study which was conducted at the University Hospitals in Basel and Geneva, Switzerland [19]. The Competent Ethics Committee of Northwestern Switzerland and Geneva (IRB approval numbers EKNZ PB_2016_02490 and GE 14–216) approved the study protocol, and written informed consent was obtained from all participants. Women who were > 18 years of age with a singleton pregnancy were included if they had at least one PE risk factor: nulliparous overweight or obese women with body mass index (BMI) > 26.1 kg/m2, nulliparous women > 40 years of age, pre-existing diabetes, essential hypertension or renal disease, pregnancy-induced hypertension, gestational diabetes (defined by at least one pathological value of fasting glucose (> 5.1 mmol/l) or at one (> 10.0 mmol/l) or two hours (> 8.5 mmol/l) after a 75-g glucose load, utero-placental dysfunction (defined by abnormal uterine perfusion with mean pulsatility index >95th percentile in the second trimester and/or bilateral uterine artery notching), previous PE, eclampsia, or HELLP, thrombophilia with high risk for PE (homozygous factor V Leiden or methylenetetrahydrofolate reductase (MTHFR) C677T defects, or the combination of heterozygous factor II G20210A and heterozygous factor V Leiden defects diagnosed in a DNA analysis prior pregnancy), antiphospholipid antibodies, or family history of PE, eclampsia, or HELLP in first-degree relatives. Additionally, women who had symptoms suspicious of PE (two combined findings of clinical symptoms like headache and/or scotoma and/or epigastric pain and/or excessive edema and/or new onset proteinuria (> 1+ in dipstick)) were asked to participate. Exclusion criteria included diagnosis of PE at sample collection, chromosomal aberrations, fetal malformations, abortion, or stillbirth at < 22 weeks of gestation. All eligible women were followed regularly with recording of demographic characteristics, medical history, clinical examinations, and blood draws for biomarker analysis (GlyFn, PAPPA2, PlGF, and sFlt-1). High-risk women with suggestive clinical findings and symptomatic women were treated expectantly, depending on their clinical condition, until delivery. The results of the biomarker analysis were not available until the end of study and did not, therefore, influence management decisions.
Diagnostic criteria for hypertensive diseases in pregnancy
Pre-existing hypertension was defined as systolic blood pressure > 140 mmHg and/or diastolic blood pressure > 90 mmHg diagnosed before conception or < 20 weeks of gestation. Gestational hypertension was determined as new onset of hypertension developing > 20 weeks of gestation without proteinuria. The following criteria for PE were used to establish the diagnosis: New-onset systolic blood pressure > 140 mmHg and/or diastolic blood pressure > 90 mmHg measured on two occasions at least 6 h apart but within one week and new-onset proteinuria with > 30 mg/24-h urine protein collection or > 2+ in dipstick or spot urine (> 3 mg/dL or protein/creatinine ratio > 3 mg protein/mmol creatinine) > 20 weeks of gestation. Eclampsia was defined as new onset of tonic-clonic seizures associated with PE, which could not be assigned to any other cause. HELLP syndrome was considered when haemolysis (lactic acid dehydrogenase > 600 IU/L, and/or lowered haptoglobin), elevated liver enzymes (aspartate amino transferase exceeding 70 IU/L) and low platelets (platelet counts < 100,000/μL) occurred.
Diagnostic criteria for intrauterine growth restriction (IUGR)
IUGR was defined as an estimated fetal weight < 10th percentile (adjusted for gender and ethnicity according to charts routinely used by both sites [20]) plus pathological finding(s) in Doppler indices (cerebro-placental ratio < 5th percentile and/or a uterine artery pulsatility index >95th percentile in the second trimester) or a birth weight < 3rd percentile [21].
Assessment of GlyFn, PAPPA2, PlGF, and sFlt-1
All maternal serum samples were aliquoted and stored at − 80 °C until analysis. Commercial immunoassay kits for sFlt-1 and PlGF (R&D systems, Minneapolis, MN, USA), PAPPA2 (Ansh Labs, Webster, TX, USA), and GlyFn (DiabetOmics, Inc., Hillsboro, OR, USA) were used according to manufacturer’s instructions. Inter-assay coefficients of variation for these commercial kits ranged from 1.89–6.65% and the intra-assay coefficients ranged from 2.1–4.5%. Biomarker thresholds for PlGF and sFlt-1 were chosen based on published literature using R&D immunoassays [22]; abnormal PlGF levels are those < 100 pg/ml and abnormal sFlt-1 levels are those > 7000 ng/ml. The threshold for PAPPA2 > 200 ng/ml was determined from prior biomarker studies (unpublished data). GlyFn threshold > 315 μg/ml were derived from the current dataset which best discriminated cases from non-cases and require additional validation in future studies.
Point-of-care test (Lumella™ test system)
A prototype GlyFn POC test strip was previously described that employed a fluorescently labeled fibronectin polyclonal antibody as both the detection and capture antibody, with the signal from maternal serum measured using a commercial automated cassette reader [18]. In the current study, serum samples were analyzed for GlyFn using the second-generation Lumella™ PE test (DiabetOmics, Inc.) according to the manufacturer’s instructions. Test strips were configured with monoclonal antibodies against GlyFn labeled with gold particles for quantification using a hand-held Lumella™ reader system. Briefly, 5 μl of serum is diluted 1:350 in running buffer and 120 μl of diluted serum is added to the test strip and inserted into the reader. The GlyFn concentration is displayed on the reader at the end of 10 min. Calibration information is supplied by the manufacturer as a lot-specific radiofrequency identification (RFID) tag on each test kit. The measurable range of the Lumella™ assay is 100 ng/mL to 800 μg/mL vs 10–2000 μg/mL for the prototype version [16]. The intra-/interassay coefficients of variation at mean concentrations of 50–800 μg/mL were 8.6/10.4 and 9.2/10.2%, respectively.
Participant/sample selection
From a prospective cohort, 226 unique samples were collected. Fifty-seven samples were excluded as we restricted the current investigation to samples derived > 20 and < 37 weeks of gestation and to women who developed clinical PE within 40 days of sample collection or did not develop PE but had a sample collected at similar gestational age. High-risk samples were chosen based on matching for gestational age (within 1 week). No exclusions were needed because of the matching of high-risk women. Women with a diagnosis of PE prior to sample collection were excluded from the analysis. Analyses were restricted to one sample per woman and the earliest sample were chosen from women in the PE group who had multiple samples collected to better represent early prediction. Thereby another 18 samples were excluded because of multiple measurement within 40 days period. Finally, 151 women with samples were included in the current analysis.
Statistical analyses
Baseline maternal characteristics were stratified for women within these groups. The nonparametric, two-sided Wilcoxon rank sum test was used to compare differences between groups for continuous variables, as they are more robust than non-normal distributions, as well as outlying observations. The χ2 test was used for categorical variables. We also compared co-morbidities, pre-existing renal disease, pre-existing diabetes, pre-existing hypertension, and gestational hypertension. Biomarker distributions for women with and without development of clinical PE were calculated and compared, and medians and interquartile ranges (IQR) of the original scales are reported. Non-parametric test equivalent to receiver-operating characteristic (ROC) curve were used to inferentially compare biomarker distributions. Confirmed delivery outcomes were also compared between groups, including gestational age at delivery, neonatal birth weight, Apgar scores, cesarean sections, preterm births, IUGR, and SGA.
ROC curves, the area under the curve (AUC), along with corresponding 95% confidence intervals (CIs) for PE diagnosis were generated using predicted probabilities from simple logistic regression models [23]. We estimated and compared the operating characteristics (sensitivity, specificity) using thresholds described previously (> 315 U/mL for GlyFn, > 200 ng/mL for PAPPA-2, < 100 pg/mL for PlGF, and ≥ 7000 ng/mL for sFlt-1) for detection of PE. We evaluated the ability of the various biomarkers to predict the onset of PE within four weeks of sample collection. Predicted probabilities from simple logistic regression were used to create ROC curves, AUCs, and 95% CI’s [23]. A comparison of the GlyFn plate immunoassay to the GlyFn POC test was performed on samples assayed by both methods. Pearson correlation coefficient was calculated to compare the methods. ROC curves were generated for each method to ascertain classification accuracy. All statistical analyses were performed using R (3.2.2) via Rstudio software version 1.0.136 (https://www.rstudio.com/products/RStudio/). ROC curves were created using the pROC package [24].