Background
Placental lesions consistent with maternal vascular underperfusion (MVU) are thought to be pathogenically linked to preeclampsia, small-for-gestational-age newborns, fetal death, and spontaneous preterm labor and delivery; yet, these lesions cannot be diagnosed antenatally. We previously reported that patients with such conditions and lesions have an abnormal profile of the angiogenic placental growth factor (PlGF) and antiangiogenic factors (eg, soluble vascular endothelial growth factor receptor [sVEGFR]-1).
Objective
The objectives of this study were to: (1) examine the relationship between the maternal plasma PlGF/sVEGFR-1 concentration ratio (referred to herein as angiogenic index-1) and the burden of histologic placental features consistent with MVU; and (2) test the hypothesis that angiogenic index-1 can identify patients in the midtrimester who are destined to deliver before 34 weeks of gestation with multiple (ie, ≥3) histologic placental features consistent with MVU.
Study Design
A 2-stage case-cohort sampling strategy was used to select participants from among 4006 women with singleton gestations enrolled from 2006 through 2010 in a longitudinal study. Maternal plasma angiogenic index-1 ratios were determined using enzyme-linked immunosorbent assays. Placentas underwent histologic examination according to standardized protocols by experienced pediatric pathologists who were blinded to clinical diagnoses and pregnancy outcomes. The diagnosis of lesions consistent with MVU was made using criteria proposed by the Perinatal Section of the Society for Pediatric Pathology. Weighted analyses were performed to reflect the parent cohort; “n*” is used to reflect weighted frequencies.
Results
(1) Angiogenic index-1 (PlGF/sVEGFR-1) concentration ratios were determined in 7560 plasma samples collected from 1499 study participants; (2) the prevalence of lesions consistent with MVU was 21% (n* = 833.9/3904) and 27% (n* = 11.4/42.7) of women with ≥3 MVU lesions delivered before 34 weeks of gestation; (3) a low angiogenic index-1 (<2.5th quantile for gestational age) in maternal plasma samples obtained within 48 hours of delivery had a sensitivity of 73% (n* = 8.3/11.4; 95% confidence interval [CI], 47–98%), a specificity of 94% (n* = 3130.9/3316.2; 95% CI, 94–95%), a positive likelihood ratio of 12.2, and a negative likelihood ratio of 0.29 in the identification of patients who delivered placentas with ≥3 MVU lesions at <34 weeks; (4) prospectively, at 20-23 weeks of gestation, a maternal plasma concentration of angiogenic index-1 <2.5th quantile identified 70% (n* = 7.2/10.3; 95% CI, 42–98%) of patients who delivered placentas with ≥3 MVU lesions before 34 weeks (specificity, 97% [n* = 2831.3/2918; 95% CI, 96–98%]; positive likelihood ratio, 23; negative likelihood ratio, 0.31); and (5) among women without obstetrical complications who delivered at term, angiogenic index-1 was lower in women with than without placental lesions consistent with MVU ( P < .05).
Conclusion
Maternal plasma angiogenic index-1 (PlGF/sVEGFR-1) is the first biomarker for the burden of placental lesions consistent with MVU. We propose that an accumulation of these lesions in placentas delivered before 34 weeks is a histologic counterpart of an antiangiogenic profile.
Introduction
The uteroplacental circulation is established around the end of the first trimester. Physiologic transformation of the spiral arteries increases the size of these vessels, allowing blood to flow into the intervillous space, where oxygen and nutrients are transported to the fetus. Disruption of maternal vascular development is thought to result in reduced blood supply to the placenta. Histologic placental features consistent with maternal vascular underperfusion (MVU) are associated with preeclampsia, intrauterine growth restriction, fetal death, and delivery of small-for-gestational-age (SGA) newborns. These conditions contribute to a substantial fraction of perinatal morbidity and mortality, often mediated by indicated preterm delivery. Uteroplacental vasculopathy is also associated with spontaneous preterm labor (PTL) with intact membranes and preterm prelabor rupture of membranes (pPROM) ; thus, maternal vascular obstructive lesions, bleeding/vessel integrity, and lack of physiologic conversion of maternal spiral arteries might constitute or interact in pathways to spontaneous as well as indicated preterm delivery, possibly contributing to an even larger fraction of adverse pregnancy outcomes. Indeed, we propose that processes resulting in maternal vascular lesions contribute to many of the “great obstetrical syndromes.”
Our group has demonstrated that an imbalance in maternal plasma concentrations of the angiogenic placental growth factor (PlGF) and antiangiogenic factors (eg, soluble vascular endothelial growth factor receptor [sVEGFR]-1) is characteristic of a large fraction of women who have or will develop preeclampsia. We have also found significant differences in angiogenic and antiangiogenic factor distributions among women with uncomplicated pregnancies and those who are or will be affected by spontaneous PTL with intact membranes, fetal death, massive perivillous fibrin deposition, twin-to-twin transfusion syndrome, and delivery of SGA newborns. Moreover, we have reported that differences in angiogenic and antiangiogenic factor concentrations are greater when patients with these obstetrical syndromes have placental lesions consistent with MVU. Therefore, we hypothesize that the ratio of angiogenic to antiangiogenic factor concentrations in maternal plasma reflects the burden of lesions consistent with MVU.
The objectives of this study were to: (1) examine the relationship between the maternal plasma PlGF/sVEGFR-1 concentration ratio (referred to herein as angiogenic index-1) and the burden of histologic placental features consistent with MVU, irrespective of clinical diagnosis; and (2) test the hypothesis that angiogenic index-1 can identify patients in the midtrimester who are destined to deliver before 34 weeks of gestation with multiple (ie, ≥3) histologic placental features consistent with MVU.
Materials and Methods
Study design and participants
A 2-stage case-cohort sampling strategy was used to select participants into this prospective nested longitudinal case-cohort study from among 4006 women with singleton gestations. These women, who had been enrolled between 6-22 weeks of gestation at Hutzel Women’s Hospital, Detroit, MI, from 2006 through 2010, were followed up until delivery. Exclusion criteria were multiple gestations and any of the following at enrollment: active vaginal bleeding, obstetrical complications, serious medical illness (renal insufficiency, congestive heart disease, and chronic respiratory insufficiency), chronic hypertension requiring medication, asthma requiring systemic steroids, requirement of antiplatelet or nonsteroidal antiinflammatory drugs, active hepatitis, or fetal anomalies.
Figure 1 describes the selection of study participants. In the first sampling stage, 1000 women were randomly selected from among 2893 who had venipuncture samples collected in at least 3 of 7 predefined gestational age intervals (8-15.9, 16-19.9, 20-23.9, 24-27.9, 28-31.9, 32-36.9 and ≥37 weeks). In the second sampling stage, all remaining women who had any of the following diagnoses at delivery were selected from among the 3006 women who were not selected in the first stage of sampling: preeclampsia, PTL, fetal death, pPROM, and delivery of a newborn weighing <5th centile for gestational age. The most centrally located venipuncture sample within each of the 7 intervals defined by gestational age for each patient was used for analysis, and in cases of a tie, the first sample obtained was selected.
All patients provided written informed consent, and the use of clinical data and biological specimens for research purposes was approved by the Institutional Review Boards of Wayne State University and the Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Department of Health and Human Services.
Sample collection and immunoassays
Venipuncture was performed and blood was collected into tubes containing EDTA at enrollment and during subsequent examinations scheduled every 4 weeks until the 24th week of gestation, and biweekly thereafter until delivery. Samples were centrifuged at 4°C, pipetted, and stored at –70°C. Maternal plasma concentrations of PlGF and sVEGFR-1 were determined by immunoassays as described previously (R&D Systems, Minneapolis, MN). The interassay and intraassay coefficients of variation were 1.4% and 3.9% for sVEGFR-1 and 6.02% and 4.8% for PlGF, respectively. The sensitivities of the assays were 16.97 pg/mL for sVEGFR-1 and 9.52 pg/mL for PlGF. Laboratory personnel performing the assays were blinded to clinical information.
Histologic examination
Placentas were examined histologically according to standardized protocols by perinatal pathologists blinded to clinical diagnoses and obstetrical outcomes. Briefly, 3-9 sections of the placenta were examined, including at least 2 full-thickness sections of the placental disc, 2 umbilical cord sections, and 1 membrane roll from the extraplacental membranes. At least 1 full-thickness section was taken from the center of the placenta; others were taken randomly from the placental disc. Placental features consistent with MVU (referred to herein as MVU lesions) were diagnosed using criteria established by the Perinatal Section of the Society for Pediatric Pathology, as described in prior studies ( Supplemental Table 1 ).
Clinical definitions
Preeclampsia was defined as new-onset hypertension that developed after 20 weeks of gestation and with proteinuria. Hypertension was defined as systolic ≥140 and/or diastolic ≥90 mm Hg blood pressure, measured at 2 occasions, 4 hours to 1 week apart. Proteinuria was defined as a urine protein of ≥300 mg in a 24-hour urine collection, or 2 random urine specimens, obtained 4 hours to 1 week apart, showing ≥1+ by dipstick. Gestational hypertension was defined as blood pressure ≥140/90 mm Hg detected at >20 weeks of gestation without proteinuria, and without emergence of preeclampsia by 12 weeks’ postpartum. Delivery of SGA neonates was classified by birthweight <10th percentile for gestational age at delivery according to a US reference population and did not include co-occurring preeclampsia. The diagnosis of pPROM was determined with a sterile speculum examination with documentation of either vaginal pooling or positive nitrazine or ferning tests. Spontaneous PTL was defined as spontaneous onset of labor with intact membranes and delivery that occurred before the 37th week of gestation. Fetal death was defined as intrapartum or antepartum death of a fetus prior to delivery at ≥20 weeks of gestation, which was not a consequence of an induced termination of pregnancy.
Statistical analysis
The quantreg package under the R statistical environment was used to fit locally weighted linear quantile regression models to estimate reference biomarker quantiles (ie, centiles) as a function of gestational age at venipuncture, using information from women selected in the first stage of sampling who did not develop preeclampsia or deliver SGA newborns. A total of 100 models were fit to estimate quantiles for 100 equally spaced gestational ages occurring between the maximum and minimum observed values. Each model, for each specific gestational age estimate, used the entirety of data (ie, measurements across all gestational ages); yet, different weights were used for each measurement in each model. The weights for each measurement were determined using a Gaussian distribution, so that it decreased exponentially as a function of the distance between the gestational age of each measurement and that for which the quantiles were estimated. The spread of the Gaussian distribution was set such that the weights of measurements that were 3 weeks away from the gestational age for which centiles were estimated were about 50% of that for measurements at the targeted gestational age. The resulting locally estimated centiles were then smoothed using a 5th-degree polynomial function of gestational age, to allow interpolation of the centiles at any desired value of gestational age.
Weighting was used to reflect the parent cohort: 723 of 747 patients with conditions identified in the parent cohort and overselected into the case-cohort had venipuncture samples available, these patients received a weight of 1.03; 776 of 3259 patients without overselected conditions were included in the case-cohort study, these patients received a weight of 4.20. The sum of weighted frequencies is equal to the total sample size of the parent cohort (n = 4006); “n*” is used to denote weighted frequencies, whereas “n” is used to denote unweighted frequencies.
Locally weighted regression models were fit for descriptive purposes to show between-group differences in biomarker concentration ratio (PlGF/sVEGFR-1) distributions expressed as a function of gestational age at venipuncture. These models were not weighted by inverse sampling probability, since we wanted to show actual patient-specific biomarker profiles expressed as a function of gestational age, and thought it most appropriate that the accompanying group mean patterns directly characterize the observed patient-specific profiles (hence, confidence intervals [CI] are wider, and thus, more likely to overlap, indicating bias toward the null hypothesis—no association). Otherwise, analyses were performed with inverse probability weighting.
Generalized linear models with robust estimators of variance were fit to examine differences in biomarker ratios adjusting for potentially confounding factors. Proportions and 95% CIs or medians and interquartile ranges were calculated to describe categorical and arithmetic variables, respectively. Binomial and multinomial logistic regression models were fit to determine magnitudes of association. Positive likelihood ratio (LR+) was calculated as, sensitivity ÷ (1-specificity). Negative likelihood ratio (LR–) was calculated as (1-sensitivity) ÷ specificity. Statistical significance was defined using a 5% threshold for type I error (ie, P < .05 and 95% CI for magnitudes of association that do not include the null hypothesis [ie, an odds ratio [OR] of 1.0]). Statistical analyses were performed using SAS software, version 9.4. (SAS Institute, Cary, NC).
Materials and Methods
Study design and participants
A 2-stage case-cohort sampling strategy was used to select participants into this prospective nested longitudinal case-cohort study from among 4006 women with singleton gestations. These women, who had been enrolled between 6-22 weeks of gestation at Hutzel Women’s Hospital, Detroit, MI, from 2006 through 2010, were followed up until delivery. Exclusion criteria were multiple gestations and any of the following at enrollment: active vaginal bleeding, obstetrical complications, serious medical illness (renal insufficiency, congestive heart disease, and chronic respiratory insufficiency), chronic hypertension requiring medication, asthma requiring systemic steroids, requirement of antiplatelet or nonsteroidal antiinflammatory drugs, active hepatitis, or fetal anomalies.
Figure 1 describes the selection of study participants. In the first sampling stage, 1000 women were randomly selected from among 2893 who had venipuncture samples collected in at least 3 of 7 predefined gestational age intervals (8-15.9, 16-19.9, 20-23.9, 24-27.9, 28-31.9, 32-36.9 and ≥37 weeks). In the second sampling stage, all remaining women who had any of the following diagnoses at delivery were selected from among the 3006 women who were not selected in the first stage of sampling: preeclampsia, PTL, fetal death, pPROM, and delivery of a newborn weighing <5th centile for gestational age. The most centrally located venipuncture sample within each of the 7 intervals defined by gestational age for each patient was used for analysis, and in cases of a tie, the first sample obtained was selected.
All patients provided written informed consent, and the use of clinical data and biological specimens for research purposes was approved by the Institutional Review Boards of Wayne State University and the Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Department of Health and Human Services.
Sample collection and immunoassays
Venipuncture was performed and blood was collected into tubes containing EDTA at enrollment and during subsequent examinations scheduled every 4 weeks until the 24th week of gestation, and biweekly thereafter until delivery. Samples were centrifuged at 4°C, pipetted, and stored at –70°C. Maternal plasma concentrations of PlGF and sVEGFR-1 were determined by immunoassays as described previously (R&D Systems, Minneapolis, MN). The interassay and intraassay coefficients of variation were 1.4% and 3.9% for sVEGFR-1 and 6.02% and 4.8% for PlGF, respectively. The sensitivities of the assays were 16.97 pg/mL for sVEGFR-1 and 9.52 pg/mL for PlGF. Laboratory personnel performing the assays were blinded to clinical information.
Histologic examination
Placentas were examined histologically according to standardized protocols by perinatal pathologists blinded to clinical diagnoses and obstetrical outcomes. Briefly, 3-9 sections of the placenta were examined, including at least 2 full-thickness sections of the placental disc, 2 umbilical cord sections, and 1 membrane roll from the extraplacental membranes. At least 1 full-thickness section was taken from the center of the placenta; others were taken randomly from the placental disc. Placental features consistent with MVU (referred to herein as MVU lesions) were diagnosed using criteria established by the Perinatal Section of the Society for Pediatric Pathology, as described in prior studies ( Supplemental Table 1 ).
Clinical definitions
Preeclampsia was defined as new-onset hypertension that developed after 20 weeks of gestation and with proteinuria. Hypertension was defined as systolic ≥140 and/or diastolic ≥90 mm Hg blood pressure, measured at 2 occasions, 4 hours to 1 week apart. Proteinuria was defined as a urine protein of ≥300 mg in a 24-hour urine collection, or 2 random urine specimens, obtained 4 hours to 1 week apart, showing ≥1+ by dipstick. Gestational hypertension was defined as blood pressure ≥140/90 mm Hg detected at >20 weeks of gestation without proteinuria, and without emergence of preeclampsia by 12 weeks’ postpartum. Delivery of SGA neonates was classified by birthweight <10th percentile for gestational age at delivery according to a US reference population and did not include co-occurring preeclampsia. The diagnosis of pPROM was determined with a sterile speculum examination with documentation of either vaginal pooling or positive nitrazine or ferning tests. Spontaneous PTL was defined as spontaneous onset of labor with intact membranes and delivery that occurred before the 37th week of gestation. Fetal death was defined as intrapartum or antepartum death of a fetus prior to delivery at ≥20 weeks of gestation, which was not a consequence of an induced termination of pregnancy.
Statistical analysis
The quantreg package under the R statistical environment was used to fit locally weighted linear quantile regression models to estimate reference biomarker quantiles (ie, centiles) as a function of gestational age at venipuncture, using information from women selected in the first stage of sampling who did not develop preeclampsia or deliver SGA newborns. A total of 100 models were fit to estimate quantiles for 100 equally spaced gestational ages occurring between the maximum and minimum observed values. Each model, for each specific gestational age estimate, used the entirety of data (ie, measurements across all gestational ages); yet, different weights were used for each measurement in each model. The weights for each measurement were determined using a Gaussian distribution, so that it decreased exponentially as a function of the distance between the gestational age of each measurement and that for which the quantiles were estimated. The spread of the Gaussian distribution was set such that the weights of measurements that were 3 weeks away from the gestational age for which centiles were estimated were about 50% of that for measurements at the targeted gestational age. The resulting locally estimated centiles were then smoothed using a 5th-degree polynomial function of gestational age, to allow interpolation of the centiles at any desired value of gestational age.
Weighting was used to reflect the parent cohort: 723 of 747 patients with conditions identified in the parent cohort and overselected into the case-cohort had venipuncture samples available, these patients received a weight of 1.03; 776 of 3259 patients without overselected conditions were included in the case-cohort study, these patients received a weight of 4.20. The sum of weighted frequencies is equal to the total sample size of the parent cohort (n = 4006); “n*” is used to denote weighted frequencies, whereas “n” is used to denote unweighted frequencies.
Locally weighted regression models were fit for descriptive purposes to show between-group differences in biomarker concentration ratio (PlGF/sVEGFR-1) distributions expressed as a function of gestational age at venipuncture. These models were not weighted by inverse sampling probability, since we wanted to show actual patient-specific biomarker profiles expressed as a function of gestational age, and thought it most appropriate that the accompanying group mean patterns directly characterize the observed patient-specific profiles (hence, confidence intervals [CI] are wider, and thus, more likely to overlap, indicating bias toward the null hypothesis—no association). Otherwise, analyses were performed with inverse probability weighting.
Generalized linear models with robust estimators of variance were fit to examine differences in biomarker ratios adjusting for potentially confounding factors. Proportions and 95% CIs or medians and interquartile ranges were calculated to describe categorical and arithmetic variables, respectively. Binomial and multinomial logistic regression models were fit to determine magnitudes of association. Positive likelihood ratio (LR+) was calculated as, sensitivity ÷ (1-specificity). Negative likelihood ratio (LR–) was calculated as (1-sensitivity) ÷ specificity. Statistical significance was defined using a 5% threshold for type I error (ie, P < .05 and 95% CI for magnitudes of association that do not include the null hypothesis [ie, an odds ratio [OR] of 1.0]). Statistical analyses were performed using SAS software, version 9.4. (SAS Institute, Cary, NC).
Results
Descriptive characteristics
The maternal plasma angiogenic index-1 concentration ratio was determined in 7560 venipuncture samples collected across gestation from the 1499 women included in this study ( Figure 1 ). Descriptive characteristics of the case-cohort and the parent cohort study populations are presented in Table 1 : 92% of patients selected into the case-cohort were identified as African American, 39% were nulliparous, and their median gestational age at delivery was 39.3 weeks, closely reflecting the distributions in the parent cohort (91%, 37%, and 39.1, respectively). Subsequent analyses used information from the case-cohort study population.
| Descriptive characteristic | Stage 1, random sample n = 1000 | Stage 2, oversample n = 499 | Weighted case-cohort a n = 1499 | Parent cohort n = 4006 | ||||
|---|---|---|---|---|---|---|---|---|
| Age, y, median (IQR) | 23 | (20–27) | 23 | (20–28) | 23 | (20–27) | 23 | (20–27) |
| African American race | 93% | n = 927 | 93% | n = 463 | 92% | n* = 3700 | 91% | n = 3641 |
| Nulliparity | 38% | n = 384 | 39% | n = 195 | 39% | n* = 1545 | 37% | n = 1483 |
| Prepregnancy body mass index, median (IQR) | 26.6 | (22.5–32.5) | 26.8 | (22.7–32.5) | 26.8 | (22.6–32.6) | 26.8 | (22.6–32.4) |
| Any tobacco use during pregnancy | 21% | n = 207 | 25% | n = 124 | 20% | n* = 801 | 22% | n = 845 |
| Preterm labor with intact membranes | 7% | n = 65 | 28% | n = 137 | 5% | n = 209.7 | 5% | n = 209 |
| Preterm prelabor rupture of membranes | 3% | n = 34 | 16% | n = 78 | 3% | n* = 115.7 | 3% | n = 115 |
| Preeclampsia | 8% | n = 77 | 30% | n = 147 | 6% | n* = 231 | 6% | n = 224 |
| Small-for-gestational-age newborn | 15% | n = 149 | 37% | n = 185 | 14% | n* = 541 | 13% | n = 502 |
| Gestational hypertension | 9% | n = 94 | 5% | n = 26 | 10% | n* = 384 | 8% | n = 341 |
| Gestational age at delivery, median (IQR) weeks | 39.1 | (37.9–40.1) | 36.4 | (33.6–38.4) | 39.3 | (38–40.3) | 39.1 | (37.7–40.1) |
| Birthweight, median (IQR) grams | 3170 | (2792–3482) | 2370 | (1800–2745) | 3190 | (2830–3490) | 3150 | (2765–3475) |
The prevalence of having 1, 2, or ≥3 placental lesions consistent with MVU was 15.9% (n* = 620/3904), 4.3% (n* = 171.2/3904), and 1.1% (n* = 42.7/3904), respectively; Table 2 shows the specific lesions and clinical diagnoses of women in each of these group. Increased syncytial knots and increased intervillous fibrin were the lesions consistent with MVU most frequently diagnosed among women with ≥2 of such histologic placental features.
| Histologic placental feature/clinical diagnosis | No. of MVU features identified | |||
|---|---|---|---|---|
| 0 n* = 3070 | 1 n* = 620 | 2 n* = 171 | ≥3 n* = 43 | |
| % | % | % | % | |
| Villous changes | ||||
| Remote villous infarct, n* = 60.5 | – | 5.4 | 9.1 | 26.8 |
| Recent villous infarct, n* = 32.3 | – | 2.4 | 7.3 | 12.1 |
| Increased syncytial knots, n* = 356.2 | – | 30.5 | 76.3 | 85.3 |
| Villous agglutination, n* = 43.4 | – | 1.5 | 13.4 | 26.9 |
| Increased intervillous fibrin, n* = 260 | – | 22.4 | 58.0 | 44.0 |
| Distal villous hypoplasia, n* = 14.5 | – | 0.3 | 3.0 | 16.9 |
| Vascular lesions | ||||
| Persistent muscularization of basal plate arteries, n* = 223.7 | – | 29.0 | 13.4 | 48.9 |
| Mural hypertrophy of decidual arterioles, n* = 61.6 | – | 5.1 | 11.6 | 24.4 |
| Acute atherosis of basal plate arteries and/or decidual arterioles, n* = 45.98 | – | 3.4 | 7.9 | 26.8 |
| Clinical diagnosis at delivery | ||||
| Preeclampsia, n* = 231.4 | 4.0 | 8.0 | 16.9 | 48.4 |
| Small for gestational age, n* = 541.1 | 10.9 | 19.5 | 26.7 | 46.1 |
| Preterm labor with intact membranes, n* = 209.7 | 4.5 | 7.7 | 6.0 | 4.8 |
| Preterm prelabor rupture of membranes, n* = 115.7 | 2.6 | 3.5 | 4.2 | 0.0 |
| Fetal death, n* = 24.8 | 0.4 | 1.3 | 1.8 | 4.8 |
Burden of histologic placental features consistent with MVU and gestational age at delivery
We first tested the hypothesis that an accumulating burden (ie, increasing number) of MVU lesions in the placenta would be associated with increasing risk of early vs late preterm delivery. Table 3 shows the magnitudes of association (ORs) between the number of histologic placental features consistent with MVU and gestational age at delivery. The greater the number of histologic features consistent with maternal vascular obstruction, the stronger the magnitude of association (OR) with having delivered preterm, particularly before 34 weeks of gestation. Women whose placentas had ≥3 histologic features consistent with MVU were 2.8 (95% CI, 1.1–7), 13.4 (95% CI, 5.9–30), and 8.1 (95% CI, 2.4–28) times more likely than those without such features to have delivered preterm at 34.1-36.9, 28-34.0, or 20-27.9 weeks of gestation, respectively. Women with 2 placental features consistent with MVU were 1.8 (95% CI, 1.1–3), 4.9 (95% CI, 2.7–9), and 7.2 (95% CI, 3.8–14) more likely than those without such lesions to have delivered at 34.1-36.9, 28-34.0, or 20-27.9 weeks. Multivariable adjustment for maternal age, race, nulliparity, and prepregnancy body mass index did not appreciably alter the association between burden of MVU and gestational age at delivery ( Supplemental Table 2 ).
| No. of histologic features consistent with maternal vascular underperfusion, n | GA 20–27.9 weeks n* = 80.7 | GA 28–33.9 weeks n* = 133.4 | GA 34–36.9 weeks n* = 292.3 | |||
|---|---|---|---|---|---|---|
| OR | (95% CI) | OR | (95% CI) | OR | (95% CI) | |
| 0, n* = 3070 | 1 | reference | ||||
| 1, n* = 620 | 2.7 | (1.6–4.7) | 3.4 | (2.3–5) | 1.6 | (1.2–2.2) |
| 2, n* = 171.2 | 7.2 | (3.8–14) | 4.9 | (2.7–8.8) | 1.8 | (1.1–3.1) |
| ≥3, n* = 42.7 | 8.1 | (2.4–28) | 13.4 | (5.9–30) | 2.8 | (1.1–7) |
Angiogenic index-1 and MVU lesions in patients who deliver within 48 hours of venipuncture
To determine whether there was a relationship between angiogenic index-1 and the burden of placental lesions consistent with MVU, we first restricted the analysis to women who had a blood samples collected within 48 hours of delivery to maintain a meaningful temporal relationship between maternal plasma concentrations of angiogenic and antiangiogenic factors and histologic findings in the placenta. Of the 1499 women in this study, 1204 (80%) had samples obtained within 48 hours of delivery. The weighted proportions of these women who had 1, 2, or ≥3 lesions consistent with MVU were 16% (n* = 541.8/3327.6), 4% (n* = 145.2/3327.6), and 1.2% (n* = 38.6/3327.6), respectively.
The odds of having 0, 1, 2, or ≥3 placental lesions comparing women with and without angiogenic index-1 ratios <10th reference quantile for gestational age within 48 hours of delivery differed significantly before 34 weeks and after 34 weeks of gestation, with and without multivariable adjustment for potential confounders ( P < .0001 for both). Patients with angiogenic index-1 ratios <10th reference quantile for gestational age within 48 hours of delivery were 2 (95% CI, 1.3–4), 8 (95% CI, 5–15), 10 (95% CI, 4–24), and 17 (95% CI, 5–64) times more likely than those with higher ratios to deliver before 34 weeks with 0, 1, 2, or ≥3 placental lesions consistent with MVU, adjusting for maternal age, race, nulliparity, and prepregnancy body mass index. Figure 2 illustrates the weighted proportions of patients who delivered placentas with 0, 1, 2, or ≥3 MVU lesions before 34 weeks and at ≥34 weeks of gestation with angiogenic index-1 ratios <2.5th reference quantile for gestational age. The greater the number of MVU lesions, the higher the proportion of patients with low angiogenic index-1 ratios (Cochran-Armitage trend test; P < .0001 for both comparisons). The angiogenic index-1 ratio was <2.5th reference quantile for gestational age within 48 hours of delivery in 73% of patients with ≥3 lesions consistent with MVU who delivered before 34 weeks ([n* = 8.3/11.4; 95% CI, 47–98%]; specificity 94% [n* = 3130.9/3316.2; 95% CI, 94–95%]; LR+, 12.2; LR–, 0.29). In contrast to women who delivered before 34 weeks, a smaller fraction of those who delivered at ≥34 weeks with MVU lesions had angiogenic index-1 ratios <2.5th reference quantile for gestation within 48 hours of delivery ( Figure 2 ).
Plasma angiogenic index-1 concentration ratios and delivery of placentas with histologic features consistent with MVU before 34 weeks
Given the association between increasing burden of MVU and early preterm delivery (ie, before 34 weeks), and the association between low angiogenic index-1 ratios within 48 hours of delivery and increasing burden of MVU, we sought to determine whether low midpregnancy angiogenic index-1 ratios are associated with higher risks of delivering placentas at <34 weeks with increasing burden of MVU. Table 4 describes the relative odds of delivering placentas before 34 weeks of gestation with 0, 1, 2, or ≥3 histologic placental features consistent with MVU, comparing women with low (ie, <2.5th or <10th reference quantile) to those with higher plasma angiogenic index-1 ratios at 20-23 weeks of gestation. The magnitude of association (OR) with a low angiogenic index-1 strengthened with increasing numbers of histologic indicators of MVU in the placenta (ie, dose-response association). Women with biomarker ratios <10th reference quantile at 20-23 weeks of gestation were 6.6 (95% CI, 3.7–12), 7.4 (95% CI, 3–18), and 19 (95% CI, 5–73) times more likely than those with higher ratios to deliver placentas <34 weeks with 1, 2, or ≥3 placental features consistent with MVU, respectively. Multivariable adjustment for potential confounders did not meaningfully alter the estimated magnitudes of association ( Supplemental Table 3 ). When excluding women who developed preeclampsia or delivered SGA newborns, those with low compared to high biomarker ratios were still significantly more likely to deliver placentas <34 weeks with features consistent with MVU (OR, 4.4; 95% CI, 2.3–8.5), but not without placental features consistent with MVU (OR, 1.4; 95% CI, 0.7–2.9).
| Angiogenic index-1 ratio | Gestational age at delivery before 34 weeks with | |||||||
|---|---|---|---|---|---|---|---|---|
| 0 MVU n* = 77.6 | 1 MVU n* = 46.6 | 2 MVU n* = 19.6 | ≥3 MVU n* = 10.3 | |||||
| OR | (95% CI) | OR | (95% CI) | OR | (95% CI) | OR | (95% CI) | |
| <10th Quantile, n* = 348 | 1.4 | (0.7–2.7) | 6.6 | (3.7–12) | 7.4 | (3.0–18) | 19.3 | (5.1–73) |
| ≥10th Quantile, n* = 2580.3 | 1 | Reference | ||||||
| <2.5th Quantile, n* = 93.8 | 1.6 | (0.5–5.2) | 7.2 | (3.1–16) | 18 | (6.7–48) | 91 | (24–352) |
| ≥2.5th Quantile, n* = 2834.4 | 1 | Reference | ||||||
The specific lesions consistent with MVU most strongly associated with low angiogenic index-1 at 20-23 weeks of gestation were recent villous infarct, distal villous hypoplasia, and acute atherosis of the basal plate arteries and/or decidual arterioles ( Supplemental Table 4 ). These 3 lesions were also more common among those with ≥3 histologic features consistent with MVU who delivered before 34 weeks as compared to those who delivered after 34 weeks (recent villous infarct: 27%, n* = 3.1/11.4 vs 7%, n* = 2.1/31.3; distal villous hypoplasia: 55%, n* = 6.2/11.4 vs 3%, n* = 1/31.3; and acute atherosis of the basal plate arteries and/or decidual arterioles: 36%, n* = 4.1/11.6 vs 23%, n* = 7.3/31.3).
Women who delivered placentas with 1, 2, or ≥3 histologic features consistent with MVU before 34 weeks had plasma angiogenic index-1 concentration ratios around the expected median before 15 weeks of gestation ( Figure 3 , A). By contrast, at 20-23 weeks, 70% of patients with samples in this interval who delivered placentas with ≥3 features consistent with MVU before 34 weeks had biomarker ratios <2.5th reference quantile ( Figure 3 , B) ([n* = 7.2/10.3, 95% CI, 42–98%]; specificity, 97% [n = 2831.3/2918; 95% CI, 96–98%]; LR+, 23; LR–, 0.31). No patient who delivered placentas before 34 weeks with ≥3 histologic indicators of MVU had sVEGFR-1 concentrations <2.5th reference quantile, and only half had PlGF concentrations <2.5th reference quantile. The pattern of association between angiogenic index-1 at 24-28 weeks and subsequent delivery of placentas at <34 weeks with 2 MVU features was similar to the pattern observed at 20-23 weeks for patients whose placentas had ≥3 MVU lesions: 63% (n* = 10.3/19.6; 95% CI, 31–75%) of women who delivered placentas with 2 MVU lesions at <34 weeks of gestation had an angiogenic index-1 <10th reference quantile ( Figure 3 , C), and 53% (n* = 10.3/19.6; 95% CI, 31–74%) had ratios <2.5th reference quantile for gestational age. None of the patients who delivered placentas before 34 weeks with 2 MVU lesions had sVEGFR-1 concentrations <2.5th reference quantile, and 47% had PlGF concentrations <2.5th reference quantile.
