Clinical Performance of a Delphi Consensus and the sFlt-1/PlGF-ratio in Guiding Delivery Decisions in Early-Onset Preeclampsia Before 32 Weeks

Abstract

Objective

A recent Delphi consensus defined organ-system-based parameters intended to reflect disease severity and delivery urgency in preeclampsia (PE) ≤ 32 weeks. This study evaluates the applicability of these parameters in a real-world cohort.

Methods

In a retrospective cohort of 1928 women with suspected PE, 148 presented at ≤ 32 weeks, of whom 84 (56.8%) delivered at ≤ 32 weeks. Delivery indications documented in medical records were compared with organ-based Delphi criteria, and Likert scores were calculated. The association between the Delphi Score and the risk of PE-related delivery, as well as Delphi organ-system scores, mean maximal arterial blood pressure and the sFlt-1/PlGF-ratio were examined using regression analysis. ROC analyses and likelihood ratio tests were used to assess the Delphi score, sFlt-1/PlGF-ratio, and the combination of both to predict PE-related delivery.

Results

The Delphi score was associated with an increased risk of PE-related deliveries ≤ 32+0 weeks (OR 1.27 per point, 95% CI 1.12–1.45), p < 0.001) but demonstrated limited predictive accuracy (AUC 0.699 (95% CI 0.612–0.787). Adding the sFlt-1/PlGF-ratio markedly improved model performance (χ ² from 16.3 to 72.4, p < 0.001; AUC 0.891, 95% CI 0.839–0.943). In the organ subscores, the uteroplacental and fetal Delphi criteria showed the strongest association (44% delivering ≤ 32 weeks vs. 9.4% delivering > 32 weeks). Severe hypertension (≥ 160/100 mmHg), not reflected in the Delphi criteria, led to 60.7% of deliveries.

Conclusion

In this retrospective real-world cohort, the Delphi score showed limited applicability for guiding delivery decisions in PE ≤ 32 weeks. Incorporating broader hypertension thresholds and biomarkers may enhance decision-making approaches.

Background

Preeclampsia (PE) remains a major cause of maternal and perinatal morbidity and mortality worldwide, while specific therapeutic options to improve outcomes are limited 1 2 3 . As no causal therapy currently exists, management focuses on maternal and fetal surveillance, antihypertensive treatment, and timely iatrogenic delivery when indicated 4 5 . However, indications for iatrogenic delivery are not precisely defined, nor is the severity of the condition, and they vary between guidelines and the policies of individual hospitals. Once PE is diagnosed, international guidelines recommend either expectant management or delivery depending on gestational age. Between 24 and 34 weeks of gestation, when the fetus is viable but at high risk of prematurity-related complications, the German–Austrian–Swiss guideline advises prolongation of pregnancy provided that maternal and fetal conditions remain stable 1 2 . Immediate delivery is recommended in the presence of severe complications, including therapy-refractory severe hypertension, progressive renal insufficiency or acute renal failure, maternal oxygen saturation < 90%, cardiac decompensation, acute pulmonary edema, disseminated intravascular coagulation, progressive thrombocytopenia (platelets < 50 G/l), placental abruption, intrauterine fetal death, persistent severe upper abdominal pain, new-onset neurological symptoms, or eclampsia. Between 34 and 37 weeks, recommendations are less definitive, and risks and benefits must be assessed individually. After 37 weeks, delivery is recommended in all cases of PE 5 .

Due to the absence of uniformly classified severity grades in PE and the lack of clearly defined thresholds for rising blood pressure that does not respond to therapy or organ dysfunction, in the German–Austrian–Swiss guidelines, decisions regarding delivery timing in early-onset PE remain largely individualized. Furthermore, factors such as non-PE-related maternal circumstances, the expertise of the treatment team, and the risk tolerance of both patients and clinicians, as well as institutional policies and standard operating procedures, influence the decision-making process. Although decisions regarding delivery must be individualized and depend on the infrastructure of the treating center as well as the clinical circumstances of the pregnant woman, standardized criteria for an objective assessment of disease severity – independent of these factors – are needed to enable comparability across clinical studies. This is particularly relevant for interventional studies, where the primary endpoint is defined as the prolongation of pregnancy until an indication for delivery is reached 3 4 5 6 .

From 2019 to 2020, a Delphi consensus study was conducted among fifty German, Austrian and Swiss senior obstetricians with the aim of reaching agreement on parameters that could serve as criteria for describing the severity of PE. The basic idea was that PE represents a dynamic and evolving process in which newly emerging clinical features vary in their individual significance but may cumulatively constitute an indication for delivery, even when considered insufficient in isolation. A case scenario was presented involving a woman who developed PE < 32 weeks of gestation and had completed antenatal corticosteroid therapy for fetal lung maturation. Within this framework, individual symptoms and clinical as well as laboratory findings were submitted to the Delphi consortium for evaluation. Each parameter was to be assessed independently on a Likert scale ranging from 0 to 4 (0 indicates no indication for delivery and 4 indicates definite immediate delivery), according to its significance in determining the indication for delivery 7 .

Building on this work, the present study aimed to evaluate the applicability of these criteria in a retrospective cohort of patients with PE who presented to our perinatal center before 32 weeks’ gestation. We reviewed the indications for delivery of those delivering ≤ 32 weeks and > 32 weeks documented in the patients’ electronic medical records and analyzed whether they aligned with the indications and the perceived need for delivery according to the Likert-scale scores assigned by the experts in the Delphi consensus.

Finally, although the ratio of the angiogenic biomarkers sFlt-1 (soluble Fms-like tyrosine kinase-1) to PlGF (placental growth factor) was evaluated in the Delphi process, the panel did not rate it as an indication for delivery at ≤ 32 weeks’ gestation (Likert score: 0). We therefore additionally investigated its association with PE-related delivery at ≤ 32 weeks’ gestation.

Methods

Study design

The patients of the present cohort were derived from a previous retrospective real-world study investigating the utility of the sFlt-1/PlGF-ratio to predict PE-related pregnancy complications in women presenting to our tertiary perinatal center with signs and symptoms of PE and/or fetal growth restriction (FGR) between August 2010 and December 2019 8 9 . This study examines the indications for delivery in women who were finally diagnosed with pre-eclampsia (PE), according to the criteria of hypertension (≥ 140/90 mmHg) and proteinuria (≥ 300 mg/d). The inclusion process is shown in Supplementary Fig. S1 online . Eligible patients were ≥ 18 years of age with singleton pregnancies ≥ 22 weeks of gestation. Inclusion criteria were new-onset hypertension, proteinuria, headache, visual disturbances, epigastric pain, progressive oedema, weight gain ≥ 1 kg/week, FGR, abnormal maternal or fetal Doppler findings. Exclusion criteria were fetal malformations or chromosomal abnormalities, pregnancy termination, prior administration of antenatal corticosteroids before PE-related presentation in our clinic, intrauterine fetal death at first presentation, and incomplete outcome data.

No study protocol with predefined criteria for delivery was in place. Treating physicians were aware of the laboratory results, and decisions regarding delivery were made by the responsible senior physician in consultation with the patient following informed consent.

We assessed all delivery indications documented in the electronic medical records and determined whether they corresponded to the organ-based delivery criteria of the Delphi consensus proposed by Brodowski et al. 7 . The Delphi criteria, along with the corresponding Likert-scale values for each variable, are shown in Table 1 using the color scheme from the original publication. Other delivery indications obtained from the clinical records are presented below for each organ system.

Table 1 Indications for delivery in routine PE cohorts with delivery at ≤ 32 and > 32 weeks’ gestation, including Delphi Consensus indications and the corresponding Likert scale, as well as additional indications documented in the patients’ medical records.

	Score	PE-related delivery ≤ 32 weeks n = 84	PE-related delivery > 32 weeks n = 64
Abbreviations: AKI = acute kidney injury; aPTT = activated partial thromboplastin time; cCT = cranial computed tomography; CPR = cerebral-placental-ratio; CTG = cardiotocography; ECG = electrocardiogram; HELLP = hemolysis, elevated liver enzymes, low platelets; MCA = middle cerebral artery; MRI = magnetic resonance imaging.
Cardiovascular system		2 (2.4%)	4 (6.3%)
Therapy resistant hypertension of 220/140 mmHg	4	1 (1.2%)	2 (3.1%)
Any pathological cardiac event	4	1 (1.2%)	2 (3.1%)
Retrosternal pressure/pain/thoracic pain with pathological ECG and/or echocardiography	4	1 (1.2%)	–
Retrosternal pressure/pain/thoracic pain without pathological ECG and/or echocardiography	2	–	–
Therapy resistant hypertension as an indication to delivery		32 (38.1%)	30 (46.9%)
Severe hypertension ≥ 160/110 mmHg		51 (60.7%)	35 (54.7%)
Severe hypertension ≥ 200/110 mmHg		12 (14.3%)	7 (10.9%)
Sensory system		3 (3.6%)	3 (4.7%)
Papilla congestion	4	–	–
Visual complaints	3	3 (3.6%)	3 (4.7%)
Hearing problems	2	–	–
Central nervous system		21 (25.0%)	19 (29.7%)
Eclampsia	4	1 (1.2%)	–
Headache with pathological cCT/MRI findings	4	–	–
Somnolence	4	1 (1.2%)	1 (1.6%)
Hyperreflexia/widened reflex zones	3	5 (6.0%)	–
Dizziness and/or balance problems	2	3 (3.6%)	–
Headache without pathological cCT/MRI findings	1	14 (16.7%)	18 (28.1%)
Respiratory system		5 (6.0%)	3 (4.7%)
Oxygen saturation < 90%	3	–	1 (1.6%)
Pulmonary edema	3	–	1 (1.6%)
Pleural effusion	3	4 (4.8%)	1 (1.6%)
Dyspnea	2	5 (6.0%)	3 (4.7%)
Hematological system		12 (14.3%)	6 (9.4%)
Disseminated intravascular coagulation	4	–	–
Platelets < 50000 cells/dl or 50 Gpt/l	4	3 (3.6%)	1 (1.6%)
Platelets < 100000 cells/dl or 100 Gpt/l	2	9 (10.7%)	5 (7.8%)
Serum sodium value < 120 mmol/l	2	–	–
Isolated decreasing platelets		–	2 (3.1%)
Hyponatremia < 126 mmol/l			1 (1.6%)
Renal and excretion system		5 (6.0%)	3 (4.7%)
Acute severe renal insufficiency (AKI stage 3)	3	–	1 (1.6%)
Moderately severe or severe oliguria to anuria (AKI stage 2 or 3)	3	2 (2.4%)	–
Acute mild or moderately severe renal insufficiency (AKI stage 1 or 2)	2	3 (3.6%)	2 (3.1%)
Worsening pre-existing nephropathy			1 (1.6%)
Metabolic system		25 (29.8%)	14 (21.9%)
Acute liver failure (transaminase value > 500/l + coagulation disorder (detectable by INR and/or aPTT)	4	–	–
Acute liver failure (e.g. very rapid within 24 hours, increase of transaminases and bilirubin + collapse in coagulation (detectable by INR and/or aPTT)	4	–	–
Liver hematoma, liver capsule rupture	4	–	–
Acute liver dysfunction (transaminase value > 70 and < 500 U/l)	2	17 (20.2%)	6 (9.4%)
Nausea	2	5 (6.0%)	2 (3.1%)
Vomiting	2	2 (2.4%)	2 (3.1%)
Pain in the upper abdomen/epigastrium	2	12 (14.3%)	9 (14.1%)
Laboratory changes indicating HELLP syndrome without fulfilling the Delphi Criteria (at least elevated transaminases, decreasing or decreased platelet count)		1 (1.2%)	–
Uteroplacental and fetal system		37 (44.0%)	6 (9.4%)
CTG pathology evaluated according to FIGO classification	4	32 (38.1%)	6 (9.4%)
Vaginal bleeding with abnormal sonography	4	4 (4.8%)	2 (3.1%)
Short-time variability < 2.6 ms (26+0–28+6 weeks of gestation) or < 3.0 ms (29+0–31+6 weeks of gestation)	3	–	–
Ductus venosus: Increased pulsatility (PIV > 95th percentile)	3	–	–
Ductus venosus: absent a-wave or reverse flow of the a-wave	3	4 (4.8%)	–
A. umbilicalis: reverse end-diastolic flow	3	7 (8.3%)	–
Vaginal bleeding with normal sonography	2	–	–
Any pathological fetal Doppler findings		24 (28.6%)	4 (6.3%)
A. umbilicalis > 95th percentile		18 (21.4%)	4 (6.3%)
CPR < 1 or		11 (13.1%)	3 (4.7%)
MCA < 5th percentile		6 (7.1%)	2 (3.1%)
End-diastolic zero-flow of the umbilical artery		12 (14.3%)	2 (3.1%)
Intrauterine fetal death		3 (3.6%)	–
Placental abruption without bleeding		1 (1.2%)	–
Patients fulfilling at least one criterium with a Likert Scale of 4 (absolute indication to delivery)		38 (45.2%)	12 (18.8%)
Patients with overall Score < 4 as a maximum value		24 (28.6%)	43 (67.2%)
Patients not classified according to the Delphi Score criteria		22 (26.2%)	23 (35.9%)
sFlt-1/PlGF-ratio		336.0 (167.5–577.0)	23.5 (7.0–95.0)
sFlt-1 ng/ml		8607.0 (5529.0–13468.3)	2787.5 (1549.5–6309.0)
PlGF ng/ml		27.0 (16.3–40.0)	120.0 (60.5–262.5)

The respective scores were calculated according to Likert scale from 0 to 4 (0 = no indication for delivery; 4 = absolute indication for immediate delivery) 7 . The predictive performance of the Delphi Score and the sFlt-1/PlGF-ratio, as well as the combination of both, to predict PE-related deliveries at ≤ 32 weeks was tested.

The influence of relevant parameters associated with these endpoints – including the individual Delphi-based organ impairment scores, mean maximal arterial blood pressure, and the sFlt-1/PlGF-ratio – was assessed using regression analyses performed both with and without adjustment for gestational age at diagnosis of preeclampsia as a potential confounder. In addition, correlations between the sFlt-1/PlGF-ratio, the total Delphi score, and the individual organ-system subscores were evaluated. We examined correlations between the sFlt-1/PlGF-ratio and both the total Delphi score and the individual organ-system subscores.This study was approved by the institutional review board (IRB) of Charité – Universitätsmedizin Berlin (EA1/252/19).

Biochemical assessment

Serum sFlt-1 and PlGF concentrations were measured when patients presented with signs of a PE or FGR. We used commercial assays on the Cobas platform (Roche Diagnostics GmbH, Mannheim, Germany), as part of standard clinical care. Patients and the attending physicians were aware of the sFlt-1/PlGF-ratio. Other laboratory markers used to indicate delivery, such as sodium level, platelet count, transaminases and creatinine, were considered if they were measured two to zero days before delivery. Doppler investigations (uterine, umbilical and middle cerebral artery, as well as the ductus venosus) were performed according to Fetal Medicine Foundation standards measurement, the indicated Doppler investigations show the examination results of the days at delivery 10 .

Statistical methods

The patient’s data were collected from medical records in our data systems Viewpoint by GE (Solingen, Germany) or SAP (Walldorf, Germany). Analyses and visualizations were performed with the IBM Statistical Package for Social Sciences, Version 31.0 (SPSS, IBM Corp., Armonk, NY, USA). Descriptives statistics were performed calculating median and interquartile ranges (IQR) as indicated for continuous-, and numbers and frequencies (%) for categorical variables.

To evaluate the influence of the Delphi Score and the sFlt-1/PlGF-ratio, we performed uni- and multivariate regression analyses. Subsequently, more detailed multivariable regression models were used to assess the influence of the individual Delphi-defined organ-system scores, mean maximal arterial blood pressure, and the sFlt-1/PlGF-ratio on PE-related delivery at ≤ 32 weeks. ROC analyses for single and combined parameter were performed. The likelihood ratio test was applied to determine whether the discriminatory performance and the prediction probability are enhanced by a combined model (Delphi score plus sFlt-1/PlGF-ratio) in comparison to a single prediction parameter (Delphi score). Additionally, we examined Spearman’s rho correlations between the sFlt-1/PlGF-ratio, the total Delphi score and the individual organ-system subscores.

Results

Patients characeristics and pregnancy outcomes

Of 1928 eligible patients with suspected PE and FGR, 148 patients met the inclusion criteria and 84 (56.8%) exhibited a PE-related delivery ≤ 32 weeks. The most common presenting symptoms were new-onset hypertension (50.0% of those delivering ≤ 32 weeks, 31.3% > 32 weeks); suspected FGR (45.2% ≤ 32 weeks, 10.9% > 32 weeks) new-onset proteinuria (34.5% ≤ 32 weeks, 12.5% > 32 weeks); elevated liver enzymes 15.5% ≤ 32 weeks and 7.8% > 32 weeks), and headache (6.0% ≤ 32 weeks, 9.4% > 32 weeks). Patient baseline characteristics including the gestational age at time of diagnosis of PE are presented in Table 2 . The median sFlt-1/PlGF-ratio of the patients who delivered ≤ 32 weeks was 336.0 (IQR 167.5–577.0), patients with a delivery > 32 weeks had a median sFlt-1/PlGF-ratio of 23.5 (IQR 7.0–95.0).

Table 2 Baseline characteristics of the routine PE-cohorts with delivery at ≤ 32 and > 32 weeks’ gestation and symptoms at first presentation with suspected PE.

	Delivery ≤ 32 weeks n = 84	Delivery > 32 weeks n = 64
Abbrevations: AL = alanin-aminotransferase; AST = aspartat-aminotransferase; FGR = fetal growth restriction; PE = preeclampsia; UA pH = umbilical artery pH.
Gestational age at presentation, weeks	26+1 (23+6–28+6)	28+6 (26+4–31+2)
Gestational age at diagnosis of PE	27+5 (24+6–29+6)	35+2 (32+6–36+5)
Maternal age, years	34.0 (30.0–37.8)	32.5 (28.0–39.0)
Maternal age ≥ 40 years	13 (15.5%)	8 (12.5%)
Body mass index, kg/m 2	25.0 (22.0–30.3)	27.4 (23.0–32.8)
Smoking	1 (1.2%)	1 (1.6%)
Ethnicity
Caucasian	80 (95.2%)	57 (89.1%)
African/Afro-American	1 (1.2%)	4 (6.3%)
Asian	–	3 (4.7%)
Other/unknown	3 (3.6%)	–
Nulliparity	41 (48.8%)	23 (35.9%)
Autoimmune disease	12 (14.3%)	7 (10.9%)
Familiar history of PE	5 (6.0%)	10 (15.6%)
History of maternal hypertension	27 (32.1%)	16 (25%)
Maternal antiphospholipide syndrome	–	3 (4.7%)
Pre-existing maternal kidney disease	3 (3.6%)	4 (6.3%)
Maternal history of PE	23 (27.4%)	20 (31.3%)
Gestational diabetes	5 (6.0%)	5 (7.8%)
New onset of hypertension	42 (50.0%)	20 (31.3%)
Systolic blood pressure, mmHg	150.0 (140.0–170.0)	140.0 (129.0–153.0)
Diastolic blood pressure, mmHg	90 (80.8–100.0)	90.0 (80.0–99.0)
New proteinuria	29 (34.5%)	8 (12.5%)
Weight gain ≥ 1 kg/week	1 (1.2%)	1 (1.6%)
Severe edemas	6 (7.1%)	3 (4.7%)
Elevated liver enzymes (ALT < 31 U/l), AST < 35 U/l)	13 (15.5%)	5 (7.8%)
Platelet count ≤ 150/nl	8 (9.5%)	3 (4.7%)
Visual disturbances	2 (2.4%)	2 (3.1%)
Headache	5 (6.0%)	6 (9.4%)
Epigastric pain	5 (6.0%)	4 (6.3%)
Ascites/Pleural effusion	1 (1.2%)	–
FGR suspected	38 (45.2%)	7 (10.9%)
Oligohydramnio	7 (8.3%)	3 (4.7%)
Umbilical artery pulsatility index	1.38 (1.16–1.71) n = 77	1.22 (0.98–1.40)
Mean uterine pulsatility index	1.78 (1.42–2.01) n = 73	1.22 (1.04–1.54)

Pregnancy outcomes are summarized in Table 3 .

Table 3 Pregnancy outcomes of the PE routine cohorts with delivery at ≤ 32 and > 32 weeks’ gestation.

GA at delivery (weeks)	28+3 (26+2–30+5)	35+4 (33+5–37+1)
Abbreviations: AL = alanin-aminotransferase; AST = aspartat-aminotransferase; FGR = fetal growth restriction; PE = preeclampsia.
Delivery within 48 h	20 (23.8%)	1(1.6%)
Delivery within one week	37 (44.0%)	8 (12.5%)
ACS completed	67 (79.8%)	20 (31.3%)
ACS at least one administration	80 (95.2%)	20 (31.3%)
Days until delivery	9.50 (3.0–21.0)	50.0 (30.3–68.0)
Fetal birthweight (g)	917.5 (582.5–1118.8)	2247.5 (1752.5–2812.5)
Percentile of fetal birthweight	10.0 (3.0–22.5)	19.5 (5.0–33.5)
Birth mode
Spontaneous	3 (3.6%)	11 (17.2%)
Cesarean	81 (96.4%)	53 (82.8%)
Emergency cesarean	7 (8.3%)	3 (4.7%)
Vacuum/forceps	–	–
5 minute Apgar < 7	13(15.5%)	2 (3.1%)
UA pH	7.26 (7.21–7.29)	7.25 (7.21–7.29)
FGR + PE	62 (73.8%)	18 (28.1%)
PE/FGR-related delivery	84 (100%)	62 (96.9%)
HELLP syndrome	20 (23.8%)	8 (12.5%)
Renal failure	4 (4.8%)	1 (1.6%)
Maternal cerebral bleeding	–	–
Neonatal Respiratory distress syndrome	72 (85.7%)	5 (7.8%)
Necrotizing enterocolitis	2 (2.2%)	–
Fetal/neonatal death	7 (8.3%)	–
Intrauterine fetal death	3 (3.6%)	–
Neonatal death	4 (4.8%)	–
Abruption of placenta	5 (6.9%)	3 (4.7%)
Neonatal intraventricular hemorrhage	5 (6.0%)	–
Maximal blood pressure before delivery (mmHg)
Systolic	160.0 (150.0–180.0)	160.0 (150.0–170.0)
Diastolic	100.0 (91.3–11.0)	100.0 (90.0–110.0)
Antihypertensive medication in pregnancy	64 (76.2%)	48 (75.0%)
Methyldopa	57 (67.9%)	47 (73.4%)
Metoprolol	24 (28.6%)	16 (25.0%)
Nifedipin retard	4 (4.8%)	4 (6.3%)
Amlodipin	4 (4.8%)	2 (3.1%)
Magnesium	30 (35.7%)	11 (17.2%)
Urapidil	28 (33.3%)	5 (7.8%)

Delphi consensus criteria in relation to real-world delivery indications

Fig. 1 shows the prevalence of the Delphi consensus criteria by organ system, as well as the most common indications for delivery documented in the medical records of women with PE-related deliveries at ≤ 32 weeks. An overview of all documented indications for delivery and their frequencies in the cohort of patients with PE-related delivery ≤ 32 weeks is shown in Table 1 .

Fig. 1.

Indications for delivery according to the Delphi Consensus Score and the most common indications for delivery in clinical routine in patients with PE-related delivery at ≤ 32 and > 32 weeks’ gestation. The figure is adapted from the Delphi Consensus publication by Brodowski et al.

The severity of cardiovascular impairment in patients delivering ≤ 32 weeks is reflected by the high proportion of severe hypertension ≥ 160/110 mmHg as an indication for delivery (60.7% of women delivering ≤ 32 weeks). Moreover, 14.3% exhibited hypertension ≥ 200/110 mmHg. In contrast, only 2.4.% of patients delivering ≤ 32 weeks fulfilled any cardiovascular criterion of the Delphi consensus, and only one patient (1.2%) showed severe hypertension ≥ 220/140 mmHg, which would constitute an absolute indication for delivery according to the Delphi criteria.

Overall, delivery indications related to the uteroplacental and fetal system showed close alignment with the prevalent indications in the present real-world cohort. Pathological cardiotocography was observed in 44% of women delivering ≤ 32 weeks. Signs of severe FGR, such as an absent or negative a-wave in the ductus venosus, were less frequent (8.3%), a vaginal bleeding suspicious for placental abruption occurred in 4.8%. Notably, umbilical artery diastolic zero-flow and umbilical artery pulsatility index > 95 ^th percentile – both markers of severe fetal impairment in later stages of pregnancy beyond 32 weeks of gestation are not included in the Delphi consensus criteria – were documented as delivery criteria in 21.4% of patients delivering ≤ 32 weeks.

Unexpected intrauterine fetal death occurred in three pregnancies (3.6% ≤ 32 weeks), and one patient in the cohort delivering ≤ 32 weeks (1.2%) experienced placental abruption without vaginal bleeding, which was not captured by the Delphi consensus definition of placental abruption. However, this patient also showed a pathological CTG, indicating the need for immediate delivery.

Findings related to the metabolic system within the Delphi consensus framework served as delivery indications in 29.8% of patients delivering ≤ 32 weeks. Only one patient (1.2%) delivering ≤ 32 weeks underwent delivery due to signs of HELLP syndrome, without meeting the Delphi-defined threshold for impaired liver function (transaminases > 70 U/l).

Central nervous system (CNS) symptoms as delivery indications occurred in 25% of patients delivering ≤ 32 weeks, most commonly headache without abnormal MRI/CT findings; headache with suspicious neuroimaging did not occur, but one patient experienced eclamptic seizures (1.2% ≤ 32 weeks).

Contingency tables detailing the sensitivities and specificities of the Delphi score and squared Delphi score for the prediction of delivery at ≤ 32 weeks and delivery within 48 hours are presented in Supplementary Material 3a–d, online .

Regression analysis and comparative predictive performance of the Delphi score and sFlt-1/PlGF-ratio

Univariate analysis showed that a higher Delphi score was associated with an increased risk of PE-related deliveries at ≤ 32+0 weeks of gestation (OR: 1.27 per point, 95% CI 1.12–1.45), p < 0.001). However, the model explained only a limited proportion of the variance (Nagelkerke R ² = 0.16).

ROC analyses demonstrated AUCs for the Delphi Score of 0.699 (95% CI 0.612–0.787) for ≤ 32+0 weeks and corresponding AUCs of 0.895 (95% CI 0.844–0.946) for the sFlt-1/PlGF-ratio ( Fig. 2 ).

Fig. 2.

ROC-curves indicating the predictive accuracies of a PE-related delivery ≤ 32 weeks with the Delphi Score, a combined model of the Delphi Score and the sFlt-1/PlGF-ratio, and the sFlt-1/PlGF-ratio.

Adding the sFlt-1/PlGF-ratio to the Delphi score markedly improved model performance for PE-related deliveries ≤ 32+0 weeks (increase in χ ² from 16.3 to 72.4, p < 0.001), consistent with a significant likelihood ratio test. The combined model showed good discrimination (AUC 0.891, 95% CI 0.839–0.943), with the sFlt-1/PlGF-ratio remaining the only independent predictor and the Delphi score adding no further predictive value ( Fig. 2 ).

Multivariable logistic regression analysis including the Delphi score and the sFlt-1/PlGF-ratio as independent predictors showed that the Delphi score was not significantly associated with early delivery at ≤ 32+0 weeks (OR 1.12, 95% CI 0.96–1.29), whereas a 100-unit increase in the sFlt-1/PlGF-ratio was significantly associated (OR 2.35, 95% CI 1.68–3.29). Further analyses integrating individual organ-based Delphi criteria, mean maximal arterial blood pressure, and the sFlt-1/PlGF-ratio are presented in Supplementary Material 2a and Suppl. Fig. S2 online. The sFlt-1/PlGF-ratio remained significantly associated with delivery at ≤ 32+0 weeks (OR 2.31 per 100 units, 95% CI 1.62–3.31; p < 0.001). Additionally, a higher uteroplacental–fetal system score was associated with increased risk (OR 1.33, 95% CI 1.04–1.71; p = 0.024). When gestational age at diagnosis of preeclampsia was included as a potential confounder in the multivariable logistic regression analysis, higher gestational age at diagnosis emerged as a protective factor against delivery ≤ 32+0 weeks (OR 0.80, 95% CI 0.72–0.90). In this model, neither the sFlt-1/PlGF-ratio nor the uteroplacental system score remained significantly associated with the outcome (Supplementary Material 1b and 2c/d online) .

The sFlt-1/PlGF-ratio correlated moderately with the Delphi score (Spearman’s ρ = 0.401, p < 0.001). Among the individual organ systems, significant correlations with the sFlt-1/PlGF-ratio were observed only for the uteroplacental system (Spearman’s ρ = 0.416, p < 0.001) and the metabolic system (Spearman’s ρ = 0.180, p = 0.028). No significant correlations were identified for the remaining systems (Supplementary Material 4 online).

Discussion

Main findings

In this retrospective real-world cohort of patients with PE, 56.8% experienced a PE-related delivery ≤ 32 weeks. Criteria related to the metabolic, hematological, and central nervous systems showed the closest alignment with the actual delivery indications of the cohort. Moreover, lower blood pressure thresholds such as ≥ 160/110 mmHg were commonly used as indications for delivery. The uteroplacental and fetal criteria of the Delphi panel demonstrated strong concordance with real-world indications; however, severe maternal or fetal complications, such as intrauterine fetal death or suspected placental abruption without preceding bleeding, were not captured by the Delphi framework. Regression and ROC analyses showed that the sFlt-1/PlGF-ratio, as well as its combination with the Delphi Score, significantly predicted PE-related delivery at ≤ 32 weeks, whereas the Delphi Score alone demonstrated lower predictive performance.

Study population

The present cohort with early signs of PE and FGR < 32 weeks does not fully correspond to the originally intended population with PE < 32 weeks, as some patients showed signs of PE and FGR at presentation but only later fulfilled the former gold-standard diagnostic criteria of hypertension and proteinuria.

However, patients in the cohort delivering ≤ 32 weeks fulfilled these criteria at < 32 weeks, with a median gestational age of 27+5 weeks, and all patients had suspected PE, reflecting the typical clinical scenario of suspected early-onset PE, in which decisions regarding hospital admission, antenatal corticosteroid administration, and potential iatrogenic delivery must be made 1 8 9 . With regard to the fetal criteria of the Delphi consensus, this suggests that criteria from the TRUFFLE study were adopted, including reversed end-diastolic flow in the umbilical artery, a negative a-wave in the ductus venosus, and/or pathological fetal heart rate tracings 11 12 . In contrast, our cohort showed that less severe Doppler abnormalities, such as zero end-diastolic flow (14.3%) and an umbilical artery pulsatility index > 95 ^th percentile (21.4%), were already associated with delivery ≤ 32 weeks. This could be explained by the fact that TRUFFLE was published in 2015, and our cohort was treated between 2010 and 2019. Clear recommendations on timing of delivery in cases of FGR with concomitant PE are however still lacking in current guidelines.

Concordance between Delphi delivery criteria and real-world indications in early-onset PE

Several Delphi criteria – particularly those related to the metabolic, sensory, and central nervous systems, as well as most uteroplacental and fetal indicators – showed good concordance with the delivery indications documented in our retrospective cohort. However, the retrospective design does not allow for conclusions about the relative weighting of these criteria in real-world clinical decision-making.

One of the most frequently documented indications for delivery in this cohort was blood pressure ≥ 160/100 mmHg 1 . This level is already associated with severe cardiovascular and neurological complications in PE. The complications – such as cardiac events, retrosternal pain with pathological echocardiographic or ECG findings, or headache with abnormal neuroimaging – are however Delphi consensus criteria for delivery 1 13 14 . The threshold for elevated blood pressure of 220/140 mmHg proposed by the Delphi consensus, as well as the inclusion of time-consuming prenatal cardiac or neurological examinations outside the delivery room in the indication list, may in practice worsen maternal and fetal perinatal outcomes.

However, neither a blood pressure threshold of 220/140 mmHg nor the lower thresholds frequently applied in our real-world cohort have yet been prospectively evaluated with regard to their benefits or potential disadvantages for maternal and fetal outcomes. This highlights the need for future prospective studies before refined or adapted blood pressure thresholds can be incorporated into clinical guidelines as delivery indications. Furthermore, consideration of antihypertensive treatment and magnesium sulfate administration may help to better characterize the clinical conditions under which delivery decisions are made, thereby enhancing the applicability of such thresholds and improving their alignment with real-world clinical practice.

In addition, consideration may be given to including severe complications such as intrauterine fetal death and suspected placental abruption – even when occurring without preceding vaginal bleeding – since these events were not encompassed by the current Delphi criteria yet remain relevant for decision-making in early-onset PE.

Influence of the Delphi score as a predictor of PE-related delivery ≤ 32 Weeks

The Delphi score was originally developed for prospective use, assuming that once a clinical parameter occurs, it remains present, and additional findings accumulate to increase the score. Because our study retrospectively assessed delivery indications in patients with PE, direct comparison with its intended prospective application is limited.

The requirement that only hypertension ≥ 240/120 mmHg contributes to the cardiovascular component of the Delphi score may partly explain its low predictive accuracy for delivery ≤ 32 weeks in our cohort. Moreover, since the study included women with early signs of PE regardless of their eventual delivery timing, the discriminatory capacity between those delivering ≤ 32 versus > 32 is inherently reduced, as patients with later delivery also exhibited severe PE features. Prospective evaluation of dynamic changes in the Delphi score among women who remain pregnant versus those who require early delivery would help clarify this issue. A total of 28.6% of patients who delivered at ≤ 32 weeks had a Delphi Score < 4. This finding indicates that a considerable proportion of early deliveries would not have been correctly identified by the Delphi Score alone.

Consistent with this, the predictive performance of the Delphi Score in ROC analysis was comparatively low, with AUCs of 0.699 for delivery ≤ 32 weeks. Additional markers – such as the sFlt-1/PlGF-ratio, particularly when combined with the Delphi Score – may therefore be useful to improve predictive accuracy in the future.

Influence of the sFlt-1/PlGF-ratio on PE-related delivery ≤ 32 weeks

In this cohort, the sFlt-1/PlGF-ratio was more than tenfold higher in women delivering ≤ 32 weeks compared with those delivering > 32 weeks. Numerous prospective studies have already demonstrated that markedly elevated sFlt-1/PlGF-ratios are strongly associated with imminent delivery and rapid disease progression in early-onset PE 15 16 17 18 19 . Although the ratio was unblinded in our real-world cohort and may therefore have influenced clinical decision-making, the moderate yet significant correlations between the biomarker and both the overall Delphi score and the metabolic, uteroplacental and fetal subscores – those most consistent with actual delivery indications – support its value as an indicator of clinically relevant disease severity, but also suggests that the angiogenic profile only partially reflects the disease burden represented by the clinical organ dysfunction score.

Although inclusion of gestational age at diagnosis of preeclampsia in the regression analyses attenuated the associations of both the Delphi score and the sFlt-1/PlGF-ratio with delivery ≤ 32 weeks, we propose that these parameters – the pathophysiology-based sFlt-1/PlGF-ratio and the clinically derived symptom-based scores – may provide complementary information for decision-making regarding early delivery.

It should also be considered that the historical gold-standard definition of preeclampsia applied in our cohort before 2019, namely hypertension in combination with proteinuria, differs from the broader clinical suspicion of preeclampsia present in all patients at the time of study inclusion, when the question of delivery indication arose. Future prospective studies are needed to clarify the influence of gestational age at diagnosis when applying the updated diagnostic criteria for preeclampsia, which define the disease as hypertension in combination with any preeclampsia-specific symptom 1 14 .

Limitations

This study has several important limitations. Its retrospective design inherently restricts the interpretation of the findings.

In the absence of a formal sample size calculation, the analyses – particularly multivariable regression and ROC analyses – should be interpreted as exploratory. In addition, the relatively small number of events may increase the risk of overfitting and limit the generalizability of the findings.

All clinical information was extracted from electronic medical records and was therefore dependent on the completeness and accuracy of routine documentation. Consequently, relevant clinical information required for the full application of the Delphi criteria may have been missing, leading to a potential underestimation of the calculated Delphi severity scores. This may partly explain why only approximately half of the patients who underwent delivery ≤ 32 weeks reached the proposed threshold score of 4, despite clear clinical indications for delivery.

Furthermore, it remains to be determined whether delivery indications derived from the Delphi consensus or those observed in our retrospective single-center cohort result in improved maternal and fetal outcomes; this question should be addressed in future prospective studies.

Interpretation of the sFlt-1/PlGF-ratio as a predictor of delivery is further limited by the clinical context in which the biomarker was measured. In our cohort, testing was performed only in cases with clinical suspicion of preeclampsia, resulting in a population inherently enriched for elevated ratios. This indication-based testing may limit the ability to assess the independent predictive value of the sFlt-1/PlGF-ratio for indication of delivery.

Finally, as a single-center cohort, the generalizability of the findings may be limited. Local clinical practices, thresholds for intervention, and documentation standards may differ from those at other institutions, potentially influencing the observed concordance between clinical decisions and the Delphi-based scoring system.

Despite these limitations, the study provides valuable insights into the real-world applicability of the Delphi consensus and highlights critical aspects for refinement of future severity scoring systems in preeclampsia.

Conclusion

The organ-based severity criteria of the Delphi consensus largely reflected the delivery indications observed in this real-world cohort. Nonetheless, incorporating lower blood pressure thresholds and the sFlt-1/PlGF-ratio as an additional prognostic marker may allow for a more precise assessment of disease severity and better alignment with clinically relevant PE complications. In an expanded prediction model incorporating both the total Delphi Score and the sFlt-1/PlGF-ratio, the sFlt-1/PlGF-ratio remained a strong independent predictor of early delivery, whereas the Delphi Score contributed only marginally to the predictive performance.

The ongoing multicenter AsSeDePre32 study aims to prospectively define additional clinical and laboratory parameters and translate them into a scoring system reflecting the urgency of delivery. Further refinement – particularly through clearer laboratory and blood pressure thresholds – may support the development of a clinically applicable algorithm to guide decision-making in pregnancies complicated by PE and fetal growth restriction.