Objective
Several conditions are associated with increased preeclampsia (PE) risk. Whether altered maternal angiogenic factor levels contribute to risk in these conditions is unknown. Our objective was to compare angiogenic biomarker patterns in high-risk pregnancies and low-risk controls.
Study Design
We conducted a planned secondary analysis of a 2-center observational study of angiogenic biomarkers in high-risk women. A total of 156 pregnant women with a PE risk factor and 59 low-risk controls were studied. Serial maternal serum samples were collected during 3 gestational windows: 23-27 weeks, 28-31 weeks, and 32-35 weeks. Soluble fms-like tyrosine kinase 1 (sFlt1), soluble endoglin (sEng), and placental growth factor (PlGF) were measured by enzyme-linked immunosorbent assay. Geometric mean angiogenic biomarker levels and angiogenic ratio (sFlt1 + sEng):PlGF were compared with low-risk controls for each risk group, at each gestational window.
Results
Gestational biomarker patterns differed in PE risk groups as compared with low-risk controls. Women with multiple gestations had markedly higher sFlt1 and sEng at all gestational windows. Women with prior PE had higher sFlt1 and angiogenic ratio, and lower PlGF, from 28 weeks onward. Women with chronic hypertension had significantly higher angiogenic ratio for all 3 gestational windows, but differences disappeared when women with PE were excluded. Obese and nulliparous women had significantly lower PlGF, but no differences in the angiogenic ratio.
Conclusion
High-risk groups have altered angiogenic biomarker patterns compared with controls, suggesting that altered production or metabolism of these factors may contribute to PE risk, particularly in women with multiple gestations and prior PE.
Women with chronic hypertension (cHTN), prepregnancy diabetes mellitus, obesity, multiple gestations (MGs), or preeclampsia (PE) in a prior pregnancy have a substantially increased risk of PE compared with women without such risk factors. The mechanisms by which these conditions increase PE risk are unknown. Dysregulated placental production of angiogenic factors, including soluble fms-like tyrosine kinase 1 (sFlt1), placental growth factor (PlGF), and soluble endoglin (sEng), contribute to endothelial dysfunction in PE by antagonizing endothelial-protective vascular endothelial growth factor, PlGF, and TGF-beta in the maternal circulation. Circulating levels of these angiogenic factors are altered weeks before the onset of PE in low-risk, nulliparous women. The angiogenic factor ratio has shown promise as a composite indicator of overall balance between circulating proangiogenic (PlGF) and antiangiogenic (sFlt1 and sEng) activity, and is more strongly predictive of PE in normal-risk women than any single biomarker. However, few studies have reported angiogenic factor levels in high-risk groups.
We hypothesized that gestational angiogenic biomarker profiles differ between low-risk controls and high-risk women. These differences may provide insights into the mechanism of risk predisposition in these groups. The goal of this study was to compare gestational patterns of sFlt1, PlGF, and sEng in women with cHTN, diabetes mellitus, obesity and nulliparity, MGs, and prior PE with low-risk women.
Materials and Methods
Study population
This was a planned secondary analysis of a 2-center observational cohort study of angiogenic biomarkers in high-risk women. The purpose of the primary study was to determine the use of angiogenic biomarkers for predication of PE in high-risk women. Women presenting to the University of Massachusetts Memorial Health Care or the George Washington University Medical Faculty Associates for prenatal care between September 2007 and June 2010 were considered for enrollment. Inclusion criteria were: pregnancy at or before 27 weeks and 6 days’ gestation, and eligibility into either (1) the low-risk control (LRC) cohort or (2) the high-risk cohort.
Inclusion in the high-risk cohort required the presence of at least one of the following: (1) nulliparous (no prior pregnancies beyond 20 weeks’ gestation) with prepregnancy body mass index (BMI) ≥30 kg/m 2 , (2) pregestational diabetes mellitus requiring oral hypoglycemic or insulin therapy before conception, (3) cHTN diagnosed or confirmed at screening by presence of blood pressure (BP) 140/90 mm Hg or greater on at least 2 occasions at least 4 hours apart before 20 weeks’ gestation and/or use of antihypertensive medications, (4) MGs confirmed by ultrasound evaluation and/or (5) previous PE reported by subject and/or medical record review, using diagnostic criteria outlined in following section. For the purposes of this analysis, where our goal was to describe the angiogenic biomarker patterns of specific risk groups, we performed a post hoc exclusion of women with more than 1 risk factor.
Inclusion in the low-risk control cohort required prepregnancy BMI less than 26 and absence of any risk factors described above. Prior pregnancy was not an exclusion criterion for the low-risk cohort.
Exclusion criteria for both cohorts included any 1 of the following: (1) age <20 or >40 years, (2) preexisting proteinuria (≥300 mg/24 hour from timed urine collection or protein:creatinine ratio ≥0.3), (3) prior diagnosis of systemic lupus erythematosus or antiphospholipid antibody syndrome, (4) significant concern about compliance or ability to complete study protocol, (5) use of antiretroviral medications, (6) history of organ transplantation, (7) known active illicit drug abuse or methadone maintenance, (5) expected delivery outside participating facilities, (6) inability to understand English, and/or (7) inability to provide informed consent. The institutional review boards of the University of Massachusetts Medical School and George Washington University approved the study, and all subjects provided informed consent.
Baseline demographic data and medical history were collected on enrollment by study personnel through personal interview and medical record review. Data collected included maternal age, race/ethnicity, tobacco and other substance use, medical problems, and obstetric history. Baseline data addressing absence or presence of risk factors included height, weight, number of fetuses by ultrasound, BP, and urine protein testing. Gestational age was calculated based on first trimester ultrasound or clinical dating that concurred with second trimester ultrasound.
Serum sampling and immunoassay
Serum specimens were collected at three prespecified gestational windows: 23-27 completed weeks, 28-31 weeks, and 32-35 weeks’ gestation. After phlebotomy, blood samples were immediately centrifuged, aliquoted, and frozen at −80°C until time of assay. Assays were performed less than 5 years after collection and each serum aliquot was thawed only once. Enzyme-linked immunosorbent assays (ELISA) for human sFlt1, PlGF, and sEng were performed in duplicate using commercial kits (R&D Systems, Minneapolis, MN) by an investigator blinded to risk group and pregnancy outcomes. Samples were repeated if there was greater than 10% variability between duplicates. Plates were repeated if the interassay variability was >15% based on an interassay standard. Interassay and intraassay variability were 4.9% and 2.5% for sFlt1, 8.3% and 1.8% for PlGF and 3.7% and 2.6% for sEng, respectively. Samples collected after the diagnosis of PE were not included in analyses.
Diagnosis of PE
PE was defined according to published guidelines as follows. In women without cHTN, PE was defined as the new onset of hypertension and proteinuria after 20 weeks’ gestation. Hypertension was either systolic BP ≥40 mmHg or diastolic BP ≥90 mmHg or greater on 2 occasions at least 4 hours apart. Proteinuria was excretion of ≥300 mg protein in a 24-hour urine collection, urine protein:creatinine ratio ≥0.30, or urine dipstick 1+ or greater on 2 occasions at least 4 hours apart, with no evidence of urinary tract infection. In women with cHTN, the diagnosis of PE required new onset proteinuria after 20 weeks’ gestation. Gestational hypertension was new onset hypertension without proteinuria after 20 weeks’ gestation. Although the diagnosis of PE required 2 abnormal BP readings, the onset of PE was defined as the time of the first elevated BP or urinary protein measurement leading to the diagnosis.
Statistical methods
Continuous variables were summarized using means and standard deviations, and pairwise comparisons of each high-risk subgroup with low-risk control subjects were made using Wilcoxon 2-sample rank sum tests. Categoric variables were summarized using frequencies, and pairwise comparisons of each high-risk subgroup with low-risk controls were made using Fisher exact tests. In longitudinal analyses, we estimated linear mixed models for each biomarker as a function of gestational window, low-risk control/high-risk subgroup, and their interaction. Each biomarker was log-transformed to handle right-skewness, and estimated means were backtransformed as geometric means and 95% confidence intervals for purposes of presentation. At each gestational window, pairwise comparisons of low risk controls with each high-risk subgroup were tested, adjusting for maternal age, current smoking, and race/ethnicity; P values were not adjusted for multiple comparisons because each pairwise comparison was of a priori interest. Analyses were performed both including and excluding subjects who subsequently developed PE. At each gestational window, and for each risk group, pairwise comparisons of the angiogenic ratio among subjects who did vs did not develop PE were tested using linear mixed modeling of log-transformed biomarkers, adjusting for maternal age, race/ethnicity, and current smoking. Statistical significance was set at P < .05 for all comparisons.
Results
Characteristics of the study subjects
A total of 258 women met inclusion criteria and contributed at least 1 serum specimen in the prespecified gestational windows. Of these, 43 subjects were excluded because they had more than one PE risk factor. Table 1 compares the clinical characteristics of the 156 high-risk subjects and 59 low-risk control subjects included in the analysis. Compared with low-risk control subjects, women in the high-risk groups differed with regard to both baseline characteristics and pregnancy outcomes. Specifically, women with cHTN were older, had more previous pregnancies/less nulliparity, higher prepregnancy BMI, earlier gestational age at delivery, lower birthweight, and were more likely to develop PE. Women with diabetes had a higher body mass index, and earlier gestational age at delivery. Women with prior PE had more previous pregnancies/less nulliparity, higher baseline BMI, earlier gestational age at delivery, and lower birthweight. Among women with prior PE, 35.7% classified their prior PE episode as “severe,” 18% of prior PE episodes were complicated by premature delivery (<37 weeks), and 6% by severe prematurity (<34 weeks). Obese and nulliparous women had fewer prior pregnancies/more nulliparity and higher baseline BMI. Women with MGs were older, had higher baseline BMI, earlier gestational age at delivery, lower mean birthweight, and were more likely to develop PE. One woman in the low-risk control group and 19 women in the high-risk groups developed PE.
| Characteristic | Low risk controls (LRC) (n = 59) | Hypertension (HTN) (n = 22) | Diabetes mellitus (DM) (n = 12) | Prior preeclampsia (Prior PE) (n = 42) | Obese and nulliparous (Ob&Nul) (n = 49) | Multiple gestatations (MG) (n = 31) |
|---|---|---|---|---|---|---|
| Maternal age, y, Mean (SD) | 30.7 (5.5) | 33.3 (4.9) a | 34.1 (2.6) | 30.1 (5.3) | 30.3 (4.3) | 34.6 (3.9) a |
| Gravity (number pregnancies), Mean (SD) | 2.4 (1.4) | 4.1 (2.9) a | 2.7 (1.4) | 3.2 (1.6) a | 1.7 (0.9) a | 2.3 (1.9) |
| Body mass index, kg/m 2 | 24.8 (2.2) | 30.3 (6.3) a | 30.4 (6.7) a | 29.0 (5.6) a | 38.0 (6.3) a | 28.2 (6.9) a |
| Race/ethnicity, % (n) | ||||||
| White | 61.0 (36) | 36.4 (8) | 66.7 (8) | 52.4 (22) | 65.3 (32) | 80.7 (25) |
| Hispanic | 10.2 (6) | 4.6 (1) | 16.7 (2) | 14.3 (6) | 8.2 (4) | 3.2 (1) |
| Black | 23.7 (14) | 54.6 (12) | 16.7 (2) | 33.3 (14) | 24.5 (12) | 12.9 (4) |
| Asian | 5.1 (3) | 4.6 (1) | 0.0 (0) | 0.0 (0) | 2.0 (1) | 3.2 (1) |
| Current smoker, % (n) | 8.5 (5) | 9.1 (2) | 0.0 (0) | 7.1 (3) | 6.1 (3) | 0.0 (0) |
| Gestational age at delivery, wks, Mean (SD) | 39.3 (1.7) | 37.9 (2.5) a | 37.2 (3.4) a | 37.6 (3.1) a | 39.7 (1.2) | 35.9 (2.9) a |
| Birthweight, g, Mean (SD) | 3316 (529) | 2990 (643) a | 3078 (1058) | 2988 (762) a | 3492 (472) | 2313 (487) a,b |
| Preeclampsia, % (n) | 1.7 (1) | 27.3 (6) a | 8.3 (1) | 11.9 (5) | 6.1 (3) | 12.9 (4) a |
| Onset <34 wks | 0.0 (0) | 13.6 (3) a | 8.3 (1) | 9.5 (4) a | 0.0 (0) | 6.5 (2) a |
| Onset 34-36.7 wks | 0.0 (0) | 0.0 (0) | 0.0 (0) | 0.0 (0) | 0.0 (0) | 6.5 (2) |
| Onset 37+ wks | 1.7 (1) | 13.6 (3) | 0.0 (0) | 2.4 (1) | 6.1 (3) | 0.0 (0) |
a P value for difference from healthy controls < .05, using Fisher exact test or Wilcoxon rank-sum test
Angiogenic factors and ratio in high-risk vs low-risk pregnancies
Figure 1 , A-D, compares geometric mean biomarker levels for the 5 high-risk groups as compared with low-risk controls for each biomarker (sFlt1, sEng, PlGF) and angiogenic ratio (sFlt1+sEng):PlGF by gestational age window. Figure 2 presents the same comparisons, excluding women who developed PE.
Multiple gestations
Women with MGs had higher sFlt1 ( Figure 1 , A) and sEng ( Figure 1 , B) levels in all gestational windows ( P < .0001) as compared with low-risk controls (LRC) and with the other high-risk groups. PlGF levels ( Figure 1 , C) in the MG group were significantly higher in the 23-27 week window ( P = .0011), and decreased through gestation; differences from the LRC group were not significant for subsequent windows. The angiogenic ratio was significantly higher in MG as compared with LRC for the 28-31 week ( P = .0004) and the 32-36 week ( P < .0001) windows ( Figure 1 , D). Exclusion of women who developed PE did not significantly affect these results ( Figure 2 ).
Prior PE
Women with prior PE (PE) had higher sFlt1 ( P < .05), lower PlGF ( P < .05), and higher angiogenic ratio ( P < .02) in the 28-31 and the 32-35 week windows as compared with LRC ( Figure 1 ). sEng tended to be higher in these windows as well, though the difference was of borderline significance in the 32-35 week window ( P = .023 at 28-31 weeks, P = .051 at 32-35 weeks). In the 23-27 week window, there were no significant differences from the LRC group for any biomarker. When women who developed PE were excluded ( Figure 2 ), overall patterns were similar, but differences from the LRC group were no longer statistically significant in the 28-31 week window for sFlt1 ( P = .121) and PlGF ( P = .108). The angiogenic ratio remained significantly higher for the latter 2 gestational windows ( P = .038 at 28-31 weeks, P = .012 at 32-35 weeks) after exclusion of women with PE.
Diabetes mellitus
sFlt1 and sEng were higher in women with DM as compared with LRC for the first 2 gestational windows ( P < .05), but were not significantly different from LRC for the 32-35 week window. PlGF tended to be lower, and the angiogenic ratio tended to be higher, as compared with LRC for all 3 windows; these differences did not reach statistical significance. Exclusion of women who developed PE attenuated the differences between DM and LRC groups with regard to sFlt1 and sEng, and differences were no longer statistically significant, with the exception of sEng in the 28-31 week window ( P = .033) ( Figure 2 ).
Chronic hypertension
Women with cHTN tended to have higher sFlt1 and lower PlGF as compared with LRC; differences reached statistical signficance for the 28-32 week window for sFlt1 ( P = .048), and the 23-28 week ( P = .005) and 32-36 week ( P = .046) windows for PlGF. There were no significant differences in sEng levels for any gestational age. The angiogenic ratio was significantly higher in cHTN compared with LRC for all 3 gestational windows ( P = .003, P = .024, and P = .015 for the 23-27, 28-31, and 32-35 week windows, respectively). Analyses excluding women who developed PE showed no significant differences between the cHTN and the LRC groups for any biomarker at any gestational window.
Obese and nulliparous
Unlike the other high risk groups, sFlt1 levels in women with obesity and nulliparity were not significantly different from low-risk controls ( Figure 1 ). sEng was significantly lower only in the 23-27 week window ( P = .0033). In contrast, PlGF was significantly lower than LRC for all 3 gestational windows ( P = .035, P = .008, and P = .014). There were no significant differences in the angiogenic ratio at any gestational age. Results were similar after exclusion of women who developed PE ( Figure 2 ).
Angiogenic ratio according to PE outcome
Table 2 compares the angiogenic ratio for women who did vs did not develop PE within each risk group. The angiogenic ratio was higher in women who developed PE vs those who did not among women with cHTN, DM, and prior PE, and MGs. Because of a small number of subjects with PE in each individual risk group, power was limited and statistical significance was not captured for each window/risk group, however, the ratio was >2-fold higher in women who developed PE for most comparisons. Obesity and nulliparity was the striking exception, with similar angiogenic ratio observed in women who did vs did not develop PE.
