Analysis of the Longitudinal Behavior of Serum Levels of Soluble Flt1 and Placental Growth Factor in Pregnant Patients With Systemic Lupus Erythematosus

Abstract

Objective

This study analyzed longitudinal trajectories of soluble Flt1 (sFlt1) levels, placenta growth factor (PlGF) levels, and sFlt1:PlGF ratios in a cohort of pregnant patients with systemic lupus erythematosus (SLE).

Methods

Blood samples were collected (14–18, 24–26, 30–32, 34–36, and 38–40 weeks), stored at −80°C, and evaluated for serum levels of sFlt1, PlGF, and sFlt1:PlGF ratios. Patients were classified as inactive SLE (Systemic Lupus Erythematosus Pregnancy Disease Activity Index [SLEPDAI] <4), active disease (SLEPDAI ≥4), or preeclampsia (PE). Medians and interquartile ranges were calculated for each group, and linear models with random effects were used.

Results

A total of 527 samples were obtained from 163 patients, and all patients were subsequently classified as having inactive disease (109 patients [66.9%]), active disease (33 patients [20.2%]), and inactive disease with PE (21 patients [12.9%]). In exploratory analysis, patients with PE had higher mean serum levels of sFlt1 and sFlt1:PlGF ratios and lower PlGF levels than patients with inactive and active SLE (P = 0.01 to P < 0.001). Using linear models with random effects, there was no significant differences in mean serum levels of these angiogenic markers comparing inactive and active disease. Patients with PE showed a marked increase in sFlt1 levels from the 24th week, constantly low PlGF levels from the 14th week, and progressive increase of sFlt1:PlGF ratio during pregnancy. All these differences were statistically significant compared to the groups without PE.

Conclusion

Pregnant patients with SLE who developed PE had higher sFlt1 levels and sFlt1:PlGF ratios and lower PlGF levels, and these last two changes were detected at the beginning of second trimester, before clinical manifestation. SLE activity did not interfere with longitudinal behavior of these angiogenic markers.

INTRODUCTION

Pregnancy in patients with systemic lupus erythematosus (SLE) has an increased risk of adverse events including miscarriages, infections, premature delivery, intrauterine growth restriction, and preeclampsia (PE) compared to the general population. ¹ There is also an increased risk of lupus flares that contribute to the increased frequency of some of these events and require specific immunosuppressive treatments. ²

SIGNIFICANCE & INNOVATIONS.

• Pregnant patients with systemic lupus erythematosus (SLE) who developed preeclampsia (PE) had higher soluble Flt1 (sFlt1) levels and sFlt1:placenta growth factor (PlGF) ratios and lower PlGF levels compared to patients with inactive and active SLE.

• These changes could be detected as early as the second trimester, several weeks before clinical manifestation of PE.

• SLE activity did not interfere with longitudinal behavior of these angiogenic markers, presenting a pattern similar to that observed in pregnant patients with inactive SLE and significantly different from that found in patients with inactive SLE prone to developing PE.

• Analysis of sFlt1 levels, PlGF levels, and sFlt1:PlGF ratio may help establish an early and accurate differential diagnosis between active disease and PE and also predict adverse events in pregnancy.

The first studies reporting the involvement of angiogenic and antiangiogenic angiogenic markers in the pathogenesis of PE were published in the late 1990s. Torry et al ³ analyzed the serum levels of placenta growth factor (PlGF) in healthy pregnant people without SLE and found statistically significant lower levels in those who developed PE. Additionally, the study by Maynard et al ⁴ showed that this reduction in circulating levels of PlGF, and vascular endothelial growth factor (VEGF) was associated with an increase in serum soluble Flt1 (sFlt1) levels. The authors also demonstrated that the excess of sFlt1 in vitro resulted in endothelial dysfunction, and that this imbalance could be reversed by exogenous VEGF and PlGF. ⁴ Based on these initial observations, over the last three decades, there has been great interest in studying the behavior of these angiogenic markers in normal pregnancies, in those that develop PE, and those with associated comorbidities. ⁵ , ⁶ , ⁷ , ⁸ , ⁹

The classic clinical manifestations of PE, including hypertension and proteinuria, are very similar to those of active lupus nephritis (LN). Therefore, differentiating between these two conditions is always a challenge and virtually impossible with traditional laboratory assessments. A cross‐sectional study of patients with SLE demonstrated that the analysis of serum levels of angiogenic (PlGF) and antiangiogenic (sFlt1) angiogenic markers may allow accurate differentiation between these two conditions. ¹⁰ Moreover, there is also evidence that the pathologic changes observed in PlGF and sFlt1 serum levels occur early in pregnancy. ¹¹

However, the longitudinal trajectory of these two angiogenic markers levels throughout pregnancy in patients with SLE, and whether disease activity, including LN, influences these levels in the long term have not yet been well elucidated. This study aimed to analyze the longitudinal behavior of sFlt1 levels, PlGF levels, and sFlt1:PlGF ratio in a cohort of pregnant patients with SLE and to correlate these changes with adverse pregnancy outcomes.

PATIENTS AND METHODS

Patients

This prospective study included patients with SLE who had single pregnancies between May 2015 and May 2023. All enrolled patients met the revised diagnostic criteria for SLE stated by the American College of Rheumatology ¹² and were followed up at a single center, the high‐risk prenatal clinic of a tertiary health unit (Prenatal Care of Autoimmune Diseases and Thrombophilia, Universidade do Estado do Rio de Janeiro, Brazil). All patients with SLE attending the high‐risk prenatal clinic were invited to participate in the present study and all agreed. Inclusion was performed sequentially according to routine prenatal appointment scheduling. Clarifications regarding the research protocol were provided and the informed consent form was signed before the research interview. Participants’ race was self‐reported according to the classification of the Brazilian Institute of Geography and Statistics. Disease activity was assessed for all patients in the study group using a semistructured questionnaire, a complete physical examination, and laboratory tests. Patients with malformed fetuses, twin pregnancies, and no data on delivery and those who did not sign the informed consent form were excluded from the study. All patients who developed PE superimposed on SLE activity were excluded from the main groups to prevent this association from acting as a confounding factor. This study was approved by the Hospital Universitario Pedro Ernesto Research Ethics Committee (registration number: 2866/2011‐CEP/HUPE/CAAE: 0017.0.228.000‐11). All clinical manifestations, laboratory changes specific to SLE, and permanent damage accrual as defined by the Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index ¹³ were obtained through a review of medical records and direct interviews with each patient, at which point a specific clinical form was filled out for this study.

Screening and follow‐up visits

At the initial prenatal visit, in addition to medical and obstetric history and a complete physical examination, blood samples were obtained for a complete blood counts and measurement of serum aspartate aminotransferase, alanine aminotransferase, and creatinine, research of anti‐Ro, anti‐La, anti‐RNP, anti‐DNA, anticardiolipin, and antibeta2glycoprotein I antibodies, and lupus anticoagulant. Urinalysis, 24‐hour proteinuria, and/or urinary protein:creatinine ratio were also requested. Monthly clinical care was provided by a multidisciplinary team of obstetricians and rheumatologists. Consultations and complementary examinations were performed according to the clinical needs. Gestational outcomes were assessed by reviewing the medical records and analyzing the following parameters: route of delivery, gestational age at delivery, birth weight, Apgar score, and presence of adverse events (premature delivery, premature rupture of membranes, PE, gestational hypertension, gestational diabetes, placental abruption, fetal death, and neonatal death).

SLE activity and diagnosis of PE

Clinical assessment of disease activity was performed at each prenatal visit, and laboratory tests for this assessment were routinely requested every trimester or earlier whenever a flare was suspected. SLE activity was classified according to the Systemic Lupus Erythematosus Disease Activity Index score adapted to pregnancy (Systemic Lupus Erythematosus Pregnancy Disease Activity Index [SLEPDAI]) ¹⁴ at the moment of blood collection. Analysis of serum parameters for the SLEPDAI score (complement, anti‐DNA, thrombocytopenia, and leukopenia) was based on data available in the patient's medical records at the time of consultation. Urinalysis was performed using phase microscopy and included detection of cellular casts, dysmorphic hematuria, and pyuria to calculate the SLEPDAI score. Proteinuria was measured in a 24‐hour urine sample or estimated using the urinary protein:creatinine ratio in an isolated urine sample when the former was not available. A SLEPDAI score greater than or equal to 4 was assumed to indicate moderately active disease, whereas patients with a SLEPDAI score less than 4 were classified as having mild activity or remission. This cutoff to define activity during pregnancy has been previously used by our group ¹⁰ and other authors. ¹⁴

PE was diagnosed according to the criteria provided by the American College of Obstetricians and Gynecologists. ¹⁵ For patients who already had proteinuria at the start of prenatal care, the worsening of this condition was considered in the context of differential diagnosis between PE and SLE renal activity.

Maternal angiogenic factor assay

Peripheral venous blood samples were collected at the following gestational age intervals after enrollment: 14 to 18, 24 to 26, 30 to 32, 34 to 36, and 38 to 40 weeks (when possible, ie, when labor had not yet occurred). The samples were centrifuged for 15 minutes at 2,500 revolutions per minute, and the serum was stored at –80°C until cytokine level measurement, according to the manufacturer's recommendations. The serum levels of PlGF and sFlt1 (picograms per milliliter) were assessed after delivery, with the researchers being anonymized to the pregnancy outcomes, using the Elecsys PlGF and Elecsys sFlt1 automated electrochemiluminescence immunoassay system (Cobas 8000 e701 module, Roche Diagnostics).

Analysis

To obtain homogeneous groups in relation to disease activity and PE, all patients were classified into the three groups only after delivery. Those who remained quiescent throughout pregnancy were classified as “inactive” (group 1). Those who developed disease activity (with SLEPDAI ≥ 4) at any time during pregnancy were all classified as active SLE (group 2), and the patients who developed PE (all with quiescent SLE) constituted group 3. Definitive outcome groups were assigned after a careful review of each case by a rheumatologist and an obstetrician specialized in high‐risk pregnancies. Descriptive analysis included measures of central tendency and dispersion (mean, SD, median, and interquartile range [IQR]) for numerical data and frequency and percentage for categorical data. The comparison of delivery and newborn characteristics among the groups (inactive, active disease, and PE) was made using Student's t‐test for independent samples and the chi‐square test or Fisher's exact test for categorical data.

Over different gestational periods, the medians and IQRs of serum sFlt1 levels, PlGF levels, and sFLt1:PlGF ratios were calculated for each group. The Kruskal‐Wallis test was used to compare the distribution of these cytokine levels in each group. If the hypothesis of equality between the groups was rejected, a post hoc two‐by‐two comparison was performed using the Wilcoxon rank sum test. We opted for nonparametric tests because the distribution of the outcome variables did not meet the assumption of data normality.

Linear models with random effects (or mixed‐effects models) were used, with serum sFlt1 levels, PlGF levels, and the sFLt1:PlGF ratio as dependent variables. This choice considered the inherent correlation between measurements collected from the same patient at different points in time. Moreover, mixed‐effects models allow for the inclusion of individuals in the analysis even if they miss some measurements or have a different number of repeated measurements than the remaining individuals. Logarithmic transformation was applied to each of the three variables to better fit the assumption of data normality. Thus, the exponential of the adjusted coefficients indicated an average multiplicative increase in the levels of these angiogenic markers between the groups. Group 1 (pregnant patients with inactive SLE and no PE) was used as a reference for these comparisons.

The intercept variance (between groups) indicates how much the mean values of groups vary around the overall intercept. It reflects the differences between groups (inactive disease, active SLE, and PE) in the variables (sFlt1 levels, PlGF levels, and sFlt1:PlGF ratio). A high intercept variance suggests significant differences across groups. The residual variance (within groups) represents the unexplained noise in the data after accounting for the main factors under investigation. It indicates the variation among individuals within the same group that the model cannot explain. Essentially, it captures the individual differences that persist within the group. Finally, the intraclass correlation (ICC) measures how much individuals within the same group resemble each other compared to those in different groups. It is used to assess the reliability of measurements or the degree of similarity within groups. The ICC ranges from 0 to 1, in which a value closer to 1 indicates that individuals in the same group are very similar (indicating significant within‐group correlation), whereas a value closer to 0 suggests little similarity within groups.

Moreover, during the analysis, we implemented models adjusted for gestational period, exposure, and the interaction between these variables. No adjustment was made by the SLEPDAI as a nominal variable because this variable was used in the construction of the three groups to be compared. Its introduction into the multivariate model would be a source of collinearity, leading to problems in model fitting. We tested the statistical significance of the interaction terms. In addition, we evaluated the graphical representation of these effects by displaying the predicted values of each outcome variable on the original scale at different gestational ages. A significance level of 5% was considered in all analyses, which were conducted using R software (version 4.1.2).

RESULTS

Among all pregnant patients followed up at the high‐risk prenatal clinic, 185 patients met the eligibility criteria and were included in this study. Of these, three patients (1.6 %) did not complete the study due to miscarriage (pregnancy interrupted before the 20th week), and nine patients did not complete the study (4.8 %) due to loss of follow‐up. Five patients were excluded because of twin pregnancies, two patients with malformed fetuses were excluded, and three patients with both active disease and PE were excluded to avoid defining a new disease category with a small sample size. Thus, the study group comprised 163 patients: 109 patients (66.9%) in the inactive SLE group (group 1), 33 patients (20.2%) in the active SLE group (group 2), and 21 patients (12.9%) in the PE group (group 3) (Supplementary Figure S1). Among all patients classified as having active disease, 20 of 33 patients (60.6%) were diagnosed with active nephritis due to at least one renal variable in SLEPDAI. A total of 527 blood samples were obtained, with the following distribution: 12 patients (7.4%) had five samples collected, 51 patients (31.3%) had four samples collected, 67 patients (41.1%) had three samples collected, 29 patients (17.8%) had two samples collected, and 4 patients (2.4%) had one sample collected.

The clinical and demographic characteristics of the patients are shown in Table 1. The average duration (in years) between SLE diagnosis and pregnancy was shorter in the active disease group (5.2 years) than in the other two groups (P = 0.001). The frequency of associated antiphospholipid syndrome (APS), history of nephritis, and chronic hypertension were similar among the three groups.

Table 1.

Clinic and demographic variables and gestational outcomes of patients with inactive SLE, active SLE, and SLE and preeclampsia^*

	Inactive SLE (n = 109)	Active SLE (n = 33)	SLE and Preeclampsia (n = 21)	P value a
Clinical and demographic variables
Age mean ± SD, y	28.6 ± 5.6	27.2 ± 6.2	28.2 ± 6.5	0.48
Race, n (%)
Black person and Brazilian Mestizo	72 (66)	27 (81.8)	17 (80.9)
White person	37 (34)	6 (18.2)	4 (19.1)	0.12
Primigravida	42 (38.5)	13 (39.4)	7 (33.3)	0.88
Duration of SLE, mean ± SD, y	9.45 ± 6.2	5.2 ± 4.6	8.3 ± 6.0	0.001
Antiphospholipid syndrome, n (%)	4 (3.7)	2 (6.0)	3 (14.3)	0.14
SLICC/ACR‐DI ≥ 1, n (%)	18 (16.5)	7 (21.2)	6 (26.6)	0.40
History of nephritis	43 (39.4)	19 (57.6)	11 (52.4)	0.14
Classes III or IV	17 (39.5)	7 (36.8)	6 (54.5)	0.60
Chronic hypertension, n (%)	23 (21.1)	9 (27.3)	7 (33.3)	0.42
Obesity	32 (29.3)	10 (30.3)	7 (33.3)	0.93
Gestational results
Prematurity, n (%)	7 (6.4)	12 (36.4)	16 (76.2)	<0.0001
Intrauterine growth restriction, n (%)	11 (10)	8 (24.2)	13 (61.9)	<0.0001
Stillbirth / neonatal death, n (%)	1 (0.9)	2 (6.0)	3 (14.2)	0.05
Gestational age at delivery, b mean ± SD, wk	38.0 ± 2.2	36.9 ± 2.2	32.6 ± 4.6	<0.0001
Birth weight, b mean ± SD, g	2,964.1 ± 542.6	2,640.7 ± 587.9	1,579.1 ± 853.9	<0.0001
Small for gestational age newborn, n (%)	6 (5.5)	6 (18.2)	16 (76.2)	<0.0001
5th min Apgar score b mean ± SD	8.9 ± 0.8	8.9 ± 0.6	8.2 ± 1.2	0.002
Admission to NICU, b n (%)	3 (2.7)	8 (25.8)	14 (73.6)	<0.0001

^*Bolded values are statistically significant differences between the three groups (P < 0.05). ACR‐DI, American College of Rheumatology Damage Index; NICU, neonatal intensive care unit; SLE, systemic lupus erythematosus; SLICC, Systemic Lupus Erythematosus Collaborating Clinics.

^a The chi‐square test was used for categorical variables and analysis of variance (ANOVA) for continuous variables.

^b Stillbirths were excluded from this analysis.

The incidence of prematurity, intrauterine growth restriction, and small for gestational age newborn was very high in the PE group (76.2%, 61.9%, and 76.2%, respectively) and differed significantly from the other two groups (P < 0.0001 for all variables). These results were probably related to the higher occurrence of early PE cases (57.1% were born earlier than 34 weeks) and the need based on clinical indications to deliver the newborn sooner. The average gestational age at birth in this group was 32.6 weeks, and the mean birthweight was 1,579.1 g, resulting in a greater need for admission to the neonatal intensive care unit (73.6%). The analysis of these results also revealed statistically significant differences when compared with those of the other groups.

The medians and IQRs of serum sFlt1 levels, PlGF levels, and sFlt‐1:PlGF ratios estimated in the samples collected throughout pregnancy are shown in Tables 2, 3, and 4. In preliminary analysis, each sample was evaluated independently. All patients showed rising sFlt1 levels during pregnancy, but those who progressed to PE presented more pronounced elevations, with statistically significant differences (P = 0.01 to P < 0.001) in the samples obtained between 24 and 36 weeks, compared with the other two groups (Table 2). The medians PlGF levels in patients with and without active SLE were not significantly different. These levels presented an initial increase after 14 weeks, and a subsequent decrease after 30 weeks. Patients predisposed to developing PE had significantly lower medians values (P = 0.001 to <0.001) than those in the other two groups (Table 3). The median sFlt1:PlGF ratios in patients with and without SLE activity showed an initial drop after 14 weeks, followed by an increase after 30 weeks. The difference between the two groups was not statistically significant. The median sFlt1:PlGF ratios throughout pregnancy were significantly higher in patients with PE than in those with inactive or active SLE (P = 0.007 to < 0.001) (Table 4). A subanalysis of sFlt1:PlGF ratio including only patients with active nephritis compared to patients who developed PE provided similar results (Supplementary Table S1). The analysis using linear models with random effects (mixed‐effects models), showed that patients in the PE group had average sFlt1 levels and sFlt1:PlGF ratio values that were 2.96 times (95% confidence interval [CI] 2.30–3.80; P < 0.001) and 18.3 times (95% CI 12.2–27.4; P < 0.001) higher than those in the inactive SLE group, respectively (Table 5).

Table 2.

Comparison of median serum soluble Flt1 (sFlt1) levels (pg/mL) according to the outcome group (inactive SLE, active SLE or SLE and preeclampsia) during different periods of pregnancy^*

Sample period	Inactive SLE, median (IQR)	Active SLE, median (IQR)	SLE and preeclampsia, median (IQR)	P value (Kruskal‐Wallis)	P value (Wilcoxon comparisons 2 × 2)
14–18 wk	1,636.0 (1,010.0–2,268.0) (n = 65)	1,782.0 (1,428.0–2,817.0) (n = 18)	2,043.0 (1,849.0–3,177.5) (n = 11)	0.06	Inactive SLE × active SLE: 0.660; inactive SLE × SLE and preeclampsia: 0.083; active SLE × SLE and preeclampsia: p > 0.99 a
24–26 wk	1,541.0 (1,011.6–2,125.5) (n = 91)	2,018.1 (1,419.2–3,363.8) (n = 24)	6,909.0 (2,990.8–8,844.2) (n = 16)	<0.001	Inactive SLE × active SLE: 0.04; inactive SLE × SLE and preeclampsia: <0.001; active SLE × SLE and preeclampsia: 0.01
30–32 wk	1,766.0 (1,208.9–2,504.0) (n = 98)	1,749.0 (1,396.4–2,926.0) (n = 26)	6,875.0 (4,585.0, 13,221.0) (n = 13)	<0.001	Inactive SLE × active SLE: >0.9; inactive SLE × SLE and preeclampsia: <0.001; active SLE × SLE and preeclampsia: <0.001
34–36 wk	2,694.0 (1,851.0–3,867.5) (n = 90)	2,569.5 (2,060.8–3,393.2) (n = 24)	5,442.0 (4,759.0–8,406.0) (n = 9)	<0.001	Inactive SLE × active SLE: >0.9; inactive SLE × SLE and preeclampsia: 0.001; active SLE × SLE and preeclampsia: <0.001
38–40 wk	2,839.0 (2,022.0–4,209.5) (n = 35)	3,570.0 (3,543.0–3,790.0) (n = 5)	7,343.5 (6,946.2, 7,740.8) (n = 2)	0.06	Inactive SLE × active SLE: 0.549; inactive SLE × SLE and preeclampsia: 0.144; active SLE × SLE and preeclampsia: 0.286 a

^*Bolded values are statistically significant differences between the three groups (P < 0.05). IQR, interquartile range; SLE, systemic lupus erythematosus.

^a Some results in the Kruskal‐Wallis test had borderline P values in relation to the 5% significance level. Therefore, these comparisons 2 × 2 were maintained.

Table 3.

Comparison of median serum placenta growth factor (PlGF) levels (pg/mL) according to the outcome group (inactive SLE, active SLE or SLE and preeclampsia) during different periods of pregnancy^*

Sample period	Inactive SLE, median (IQR)	Active SLE, median (IQR)	SLE and Preeclampsia, median (IQR)	P value (Kruskal‐Wallis)	P value (Wilcoxon comparisons 2 × 2)
14–18 wk	174.0 (128.5–267.0) (n = 65)	162.6 (119.0–208.8) (n = 18)	63.0 (39.5–101.0) (n = 11)	<0.001	Inactive SLE × active SLE: >0.9; inactive SLE × SLE and preeclampsia: <0.001; active SLE × SLE and preeclampsia: 0.001
24–26 wk	551.2 (384.1–819.4) (n = 91)	389.0 (300.9–534.1) (n = 24)	73.2 (23.6–237.6) (n = 16)	<0.001	Inactive SLE × active SLE: 0.03; inactive SLE × SLE and preeclampsia: <0.001; active SLE × SLE and preeclampsia: <0.001
30–32 wk	745.5 (397.5–1,085.7) (n = 98)	433.9 (199.3–783.5) (n = 26)	80.0 (58.0–153.5) (n = 13)	<0.001	Inactive SLE × active SLE: 0.03; inactive SLE × SLE and preeclampsia: <0.001; active SLE × SLE and preeclampsia: <0.001
34–36 wk	489.8 (304.0–918.5) (n = 90)	285.8 (206.5–485.2) (n = 24)	87.0 (67.0–116.0) (n = 9)	<0.001	Inactive SLE × active SLE: 0.04; inactive SLE × SLE and preeclampsia: 0.001; active SLE × SLE and preeclampsia: <0.001
38–40 wk	304.9 (177.5–603.1) (n = 35)	284.0 (251.9–307.9) (n = 5)	87.8 (82.7–92.9) (n = 2)	0.14	Inactive SLE × active SLE: 0.55; inactive SLE × SLE and preeclampsia: 0.18; active SLE × SLE and preeclampsia: 0.29 a

^*Bolded values are statistically significant differences between the three groups (P < 0.05). IQR, interquartile range; SLE, systemic lupus erythematosus.

^a Some results in the Kruskal‐Wallis test had borderline P values in relation to the 5% significance level. Therefore, these comparisons 2 × 2 were maintained.

Table 4.

Comparison of median serum soluble Flt1:placenta growth factor (sFlt1:PlGF) ratio levels according to the outcome group (inactive SLE, active SLE or SLE and preeclampsia) during different periods of pregnancy^*

Sample period	Inactive SLE, median (IQR)	Active SLE, median (IQR)	SLE and preeclampsia, median (IQR)	P value (Kruskal‐Wallis)	P value (Wilcoxon comparisons 2 × 2)
14–18 wk	9.2 (5.1–11.6) (n = 65)	10.0 (7.3–14.1) (n = 18)	28.0 (22.2–81.5) (n = 11)	<0.001	Inactive SLE × active SLE: 0.81; inactive SLE × SLE and preeclampsia: <0.001; active SLE × SLE and preeclampsia: 0.007
24–26 wk	2.7 (1.7–3.6) (n = 91)	5.5 (3.2–7.2) (n = 24)	97.2 (13.4–481.9) (n = 16)	<0.001	Inactive SLE × active SLE: <0.001; inactive SLE × SLE and preeclampsia: <0.001; active SLE × SLE and preeclampsia: 0.001
30–32 wk	2.1 (1.6–4.5) (n = 98)	4.2 (2.4–12.1) (n = 26)	118.5 (33.7–174.0) (n = 13)	<0.001	Inactive SLE × active SLE: 0.01; inactive SLE × SLE and preeclampsia: <0.001; active SLE × SLE and preeclampsia: <0.001
34–36 wk	4.4 (2.6–12.7) (n = 90)	11.1 (5.1–16.4) (n = 24)	86.5 (47.8–125.5) (n = 9)	<0.001	Inactive SLE × active SLE: 0.11; inactive SLE × SLE and preeclampsia: 0.001; active SLE × SLE and preeclampsia: <0.001
38–40 wk	9.3 (3.9–24.2) (n = 35)	12.6 (10.9–14.0) (n = 5)	85.8 (76.3–95.3) (n = 2)	0.07	Inactive SLE × active SLE: >0.9; inactive SLE × SLE and preeclampsia: 0.036; active SLE × SLE and preeclampsia: 0.286 a

^*Bolded values are statistically significant differences between the three groups (P < 0.05). IQR, interquartile range; SLE, systemic lupus erythematosus.

^a Some results in the Kruskal‐Wallis test had borderline P values in relation to the 5% significance level. Therefore, these comparisons 2 × 2 were maintained.

Table 5.

Estimates of the parameters of the adjusted linear models with random effects for the logarithm of sFlt1, PlGF, and sFlt1:PlGF ratio according to the outcome group (inactive SLE, active SLE or SLE and preeclampsia)^*

Variable	sFlt1		PlGF		sFlt1:PlGF
	Exp (beta) (95% CI)	P value	Exp (beta) (95% CI)	P value	Exp (beta) (95% CI)	P value
Outcome group
Inactive SLE	–		–		–
Active SLE	1.21 (0.99–1.48)	0.07	0.72 (0.55–0.94)	0.02	1.68 (1.21–2.32)	0.002
SLE and preeclampsia	2.96 (2.30–3.80)	<0.001	0.16 (0.11–0.22)	<0.001	18.3 (12.2–27.4)	<0.001
Intercept variance (between groups)	0.22		0.36		0.47
Residual variance (within groups)	0.13		0.28		0.62
Intraclass correlation	0.63		0.56		0.43

^*CI, confidence interval; Exp, exponential of adjusted coefficients; PlGF, placenta growth factor; sFlt1, soluble Flt1; SLE, systemic lupus erythematosus.

The estimated values and CIs for sFlt1, PlGF, and sFlt1:PlGF ratio calculated using linear models with random effects, considering the interaction between each of the samples and the outcome groups, are presented in Figure 1A–C and Supplementary Tables S2, S3, and S4. The predicted mean concentrations of sFlt1 levels in patients with inactive and active SLE remained stable up to 30 weeks, whereas after this gestational age, there were mean increases of 205 pg/mL per week in group 1 and 193.5 pg/mL in group 2 until the end of pregnancy. However, the differences were not statistically significant between the groups. In contrast, patients who developed PE showed an increase in sFlt1 levels from the 24th week, with an average increase of 680 pg/mL until the end of pregnancy, whereas differences were statistically significant between this group and groups 1 and 2 (Figure 1A).

Figure 1.

Predicted values with respective 95% confidence intervals of (A) sFlt1, (B) PlGF, and (C) sFlt1/PlGF ratio using linear models with random effects, considering the interaction between each of the samples and the outcome group. PlGF, placenta growth factor; sFlt1, soluble Flt1; SLE, systemic lupus erythematosus.

Concerning the estimated mean levels of PlGF, patients in the two groups without PE had increasing estimated mean levels of PlGF for up to 30 weeks (from 194 pg/mL to 681 pg/mL in group 1 and from 185 pg/mL to 401 pg/mL in group 2), followed by a decline in these concentrations, with no statistically significant differences between them. In contrast, patients who later developed PE showed nearly stable average PlGF concentrations (70, 80, 68, 58, and 47 pg/mL) throughout pregnancy. These values were significantly lower than those observed in patients with inactive or active SLE. This significant difference was observed in the first samples analyzed (14–18 weeks) and persisted until delivery. The curves of the estimated mean values of the sFlt1:PlGF ratio in patients with inactive and active SLE showed an initial decline for up to 30 weeks and then raised. However, the curve of these values in patients who later developed PE showed a different pattern, with a steady increase from the beginning of the second trimester (14–18 weeks). The difference between the mean values of the estimated sFlt1:PlGF ratio in the PE group was statistically significant compared to that in the other two groups.

DISCUSSION

Few studies have evaluated whether SLE activity, especially active LN, could compromise the ability of angiogenic and antiangiogenic factors to differentiate the clinical manifestations of SLE from those of PE. Comparison of the three groups (inactive SLE with no PE, active SLE with no PE, and SLE complicated by PE) revealed that SLE activity did not interfere with the longitudinal behavior of angiogenic and antiangiogenic markers. Analysis of the trajectories of these angiogenic markers during pregnancy showed no differences between the first two groups; however, levels of angiogenic markers in both groups were significantly different from those in patients who developed PE.

Some studies have investigated sFlt1 and PlGF levels in normal pregnancies and pregnancies complicated by PE.^3–5,16 In a case‐control study, Levine et al ⁵ observed that in normal pregnancies, sFlt1 levels remained stable during the early and middle stages of pregnancy, followed by a steady increase from 33 to 36 weeks. This increase in sFlt1 levels was accompanied by a downward trend in serum PlGF levels. The authors speculated that, at the end of an uncomplicated pregnancy, the control of placental vascular growth is probably slowed down by the reversal of angiogenic to antiangiogenic state. However, in pregnancies predisposed to develop PE, this change occurs earlier and more abruptly, exaggerating the normal processes of managing placental growth and function. Several studies have shown that the serum concentration of sFlt1 is elevated in patients with PE, even before the clinical diagnosis of the disease, and this elevation can precede the onset of clinical manifestations by 6 to 11 weeks. ⁵ , ¹⁷ Another important characteristic related to the imbalance of these angiogenic markers in patients with PE is the association between sFlt1 serum levels and disease severity; that is, serum levels of sFlt1 are higher in early‐onset PE (<34 weeks) than in late‐onset PE (>34 weeks). ¹⁶ Unfortunately, none of these studies included patients with SLE.

Elucidating the role of angiogenic and antiangiogenic cytokine serum levels behavior between nonpregnant and pregnant patients with SLE is warranted because pregnant patients with SLE, mainly those with LN, are at an increased risk of PE. Moreover, its clinical manifestations (hypertension and proteinuria) are similar to those of active LN, and the differential diagnosis between these two conditions remains a major challenge. In a cross‐sectional study, our group previously analyzed the serum levels of VEGF, PlGF, and sFlt1 in nonpregnant patients with inactive and active SLE and compared the results with those of healthy patients in the control group. We found that patients with active SLE had higher serum levels of both angiogenic and antiangiogenic angiogenic markers than patients with inactive SLE or healthy patients. ⁶ To evaluate whether serum levels of VEGF, PlGF, and sFlt1 are useful in differentiating PE from active LN during pregnancy in patients with SLE, we conducted a second cross‐sectional study comparing the mean serum levels of these angiogenic markers in patients with inactive disease, active nephritis, and SLE with superimposed PE. Comparison of the three groups revealed that patients with PE had significantly lower serum PlGF levels, whereas sFlt1 and sFlt1:PlGF ratios were significantly higher. In addition, we observed an increase in the serum levels of VEGF in patients with active LN, which was unexpected in patients with PE. In this study, we concluded that these angiogenic markers may be useful in the differential diagnosis of PE and active LN. ¹⁰

Few studies have evaluated the longitudinal behavior of serum angiogenic factors in pregnant patients with SLE. ⁷ , ¹⁸ Leaños‐Miranda et al ¹⁸ investigated whether these angiogenic markers were associated with the risk of developing PE in patients with SLE. They conducted a case‐control study of 42 patients who developed PE and 75 patients with normal pregnancies. As observed in our study, lower levels of PlGF, higher levels of sFlt1 and soluble endoglin (sEng), and a higher sFlt1:PlGF ratio were found in patients prone to developing PE than in those with uncomplicated pregnancies. Another important observation was that in patients who developed early‐onset or late‐onset PE, abnormal cytokine fluctuations became significant from 12 weeks of pregnancy onward. ¹⁸ Mayer‐Pickel et al ⁷ longitudinally (12–36 weeks) measured the serum levels of PlGF, sFlt1, and sEng in patients with SLE and APS to identify, early on, the imbalance in the serum levels of these angiogenic factors in patients who later developed PE. The authors observed that in both patients with SLE and APS, the imbalance of these angiogenic markers occurred early in pregnancy and suggested their potential utility as early predicting factors for the development of PE in these patients. ⁷

Using data and samples from the predictors of pregnancy outcome: biomarkers in antiphospholipid antibody syndrome and systemic lupus erythematosus study, Kim et al ¹¹ selected 492 pregnant patients with SLE and/or APS to assess, early in pregnancy, the potential of angiogenic factors (PlGF, sFlt1, and sEng) to identify patients who were more likely to present adverse pregnancy outcomes (APOs). Circulating levels of sFlt1, PlGF, and sEng were measured monthly, and the patients were classified as having severe APO (PE < 34 weeks, fetal or neonatal death, indicated preterm delivery at <30 weeks) and as having moderate APO (PE ≥ 34 weeks, indicated preterm delivery at 30–36 weeks, growth restriction without PE). The authors observed that between 12 and 15 weeks, serum sFlt1 levels were the strongest predictor of severe APO, and at 16 to 19 weeks, the combination of sFlt1 and PlGF was even more predictive. In addition, they found that in patients who developed moderate APOs, antiangiogenic factors increased later in pregnancy, suggesting that the time of onset and duration of dysregulation of angiogenic factors are related to the severity of APOs.

In studies that longitudinally assessed the serum levels of these angiogenic markers in pregnant patients with SLE, patients with active disease were either excluded or analyzed together with patients with inactive disease, which may have impeded adequate interpretation of the fluctuations in these cytokine levels due to different clinical scenarios during pregnancy. Our study is the first to compare the longitudinal behavior of PlGF and sFlt1 in pregnant patients with SLE who developed PE with those in pregnant patients with active and inactive SLE.

Most studies that have analyzed the serum levels of angiogenic factors during different gestational periods ⁵ , ⁷ , ¹⁸ calculated the medians and IQRs of the variables. Comparative tests were performed to evaluate the statistical significance of the differences between groups. Nevertheless, it is important to note that repeated measurements violate the independence assumption as they are correlated (not independent) observations. In our case, these angiogenic markers were observed in the same individual because our investigation was focused on their trajectories. We considered the autocorrelation between repeated measures in the same individual, because a value obtained for an individual at a certain time point is influenced by previous measurements and influences subsequent ones. Therefore, these values were not independent of each other. ¹⁹

In their pivotal study, Levine et al ⁵ showed that the behavior of PlGF in pregnant patients without SLE who later developed PE followed a pattern similar to that of patients who had a normal course of pregnancy, that is, an increase in mean concentrations until the beginning of the third trimester and a subsequent decline, but with statistically significant lower levels. The PE “pattern” could be identified early in pregnancy, even in the beginning of the second trimester. ⁵ , ²⁰ In contrast, in the present study, the concentration of PlGF showed a nearly linear pattern with consistently low mean levels (70, 80, 68, 58, and 47 pg/mL). We speculate that this difference is probably due to a greater number of patients developing early and more severe PE. In a cohort of pregnant patients without SLE, Chappell et al ²¹ found that in patients at <35 weeks with suspected PE, a PlGF concentration <100 pg/mL predicted severe PE requiring delivery within 14 days. In another study, Parchem et al ²² found that PlGF levels <100 pg/mL could predict adverse maternal gestational outcomes. Moreover, in patients with severe or early‐onset PE, PlGF concentrations may be so low that they are undetectable using available PlGF assays. ²³ The results of this study reinforce the use of these angiogenic markers, which are commercially available, in predicting PE in patients with lupus, as well.

The main limitation of this study is that the patients were recruited from a single center, which can limit the generalization of the results, and the relatively low number of patients with PE. However, the differences in the analyzed cytokine levels between the groups were statistically significant. Nevertheless, this study has some positive aspects: (1) to the best of our knowledge, this is the largest cohort to analyze angiogenic and antiangiogenic factors in patients with SLE, which also included active disease as a variable, a factor to be considered in clinical practice; (2) all cases were reviewed by obstetricians and rheumatologists experienced in pregnancies with SLE, which ensures the appropriate group allocation; and (3) this is the first study to use a linear model with random effects to analyze these biomarkers longitudinally in a cohort of pregnant patients with SLE.

We have shown that throughout pregnancy in patients with SLE predisposed to develop PE, sFlt1 levels, and the sFlt1:PlGF ratio are elevated, whereas PlGF levels are reduced. The last two changes were detected before the clinical manifestation of the disease and at the beginning of the second trimester. We also demonstrated that SLE activity did not interfere with the behavior of these cytokine levels during pregnancy, presenting a pattern similar to that observed in pregnant patients with inactive SLE and significantly different from that found in patients prone to developing PE. Indeed, even patients with active SLE and superimposed PE, presented cytokines trajectories similar to those with PE and inactive SLE (data shown in Supplementary Table S5 and Supplementary Figure S2). These results support our proposal to measure serum levels of these angiogenic markers in pregnant patients with SLE to establish an early and accurate differential diagnosis between active disease and PE ¹⁰ and to predict adverse events in pregnancy. ¹¹

AUTHOR CONTRIBUTIONS

All authors contributed to at least one of the following manuscript preparation roles: conceptualization AND/OR methodology, software, investigation, formal analysis, data curation, visualization, and validation AND drafting or reviewing/editing the final draft. As corresponding author, Dr de Jesús confirms that all authors have provided the final approval of the version to be published, and takes responsibility for the affirmations regarding article submission (eg, not under consideration by another journal), the integrity of the data presented, and the statements regarding compliance with institutional review board/Declaration of Helsinki requirements.

ROLE OF THE STUDY SPONSOR

GSK had no role in the study design or in the collection, analysis, or interpretation of the data, the writing of the manuscript, or the decision to submit the manuscript for publication. Publication of this article was not contingent upon approval by GSK.

Supporting information

Disclosure form.

Appendix S1: Supplementary Information