Summary
Background
Fetal growth restriction is a major determinant of perinatal morbidity and mortality. This condition has no gold standard definition, but a widely used proxy is delivery of a small for gestational age infant (<10th percentile) combined with an adverse pregnancy outcome. Effective screening for fetal growth restriction is an area of unmet clinical need. We aimed to determine the diagnostic effectiveness of a combination of ultrasonic fetal biometry and measurement of the ratio of soluble fms-like tyrosine kinase receptor 1 (sFLT1) to placental growth factor (PlGF) in predicting adverse pregnancy outcomes associated with delivery of a small for gestational age infant.
Methods
In this prospective cohort study, using serial antenatal blood sampling and blinded ultrasound scans, we investigated the association between the combination of an elevated sFLT1/PlGF ratio (>85th percentile) and ultrasonically suspected small for gestational age (<10th percentile) at both 28 and 36 weeks of gestational age. The outcome following the 28 week measurement was preterm delivery of a small for gestational age infant. The outcome following the 36 week measurement was subsequent delivery of a small for gestational infant associated with maternal pre-eclampsia or perinatal morbidity or mortality. All definitions of exposure and outcome were predefined before we did our data analysis.
Findings
Between Jan 14, 2008, and July 31, 2012, we recruited 4512 nulliparous women. 4098 women (91%) had a sFLT1/PlGF ratio measurement and estimated fetal weight at 28 or 36 weeks of gestational age, and outcome data available. 3981 women were analysed for 28 weeks of gestational age measurements and 3747 women were analysed for 36 weeks of gestational age measurements. At 28 weeks, 47 (1%) of 3981 women had the combination of ultrasonic small for gestational age and an elevated sFLT1/PlGF ratio. The positive likelihood ratio for preterm delivery of a small for gestational age infant associated with this combination was 41·1 (95% CI 23·0–73·6), the sensitivity was 38·5% (21·1–59·3), the specificity was 99·1% (98·7–99·3), and the positive predictive value was 21·3% (11·6–35·8). At 36 weeks, 102 (3%) of 3747 women had the combination of ultrasonic small for gestational age and an elevated sFLT1/PlGF ratio. The positive likelihood ratio for delivery of a small for gestational age infant associated with maternal pre-eclampsia or perinatal morbidity or mortality was 17·5 (95% CI 11·8–25·9), the sensitivity was 37·9% (26·1–51·4), the specificity was 97·8% (97·3–98·3), and the positive predictive value was 21·6% (14·5–30·8). The positive likelihood ratios at both gestational ages were higher than previously described definitions of suspected fetal growth restriction using purely ultrasonic assessment.
Interpretation
The combination of ultrasonically suspected small for gestational age plus an elevated sFLT1/PlGF ratio in unselected nulliparous women identified a relatively small proportion of women who have high absolute risks of clinically important adverse outcomes. Screening and intervention based on this approach could result in net benefit and this could be an appropriate subject for a randomised controlled trial.
Funding
NIHR Cambridge Comprehensive Biomedical Research Centre, Medical Research Council, and Stillbirth and neonatal death society (Sands).
Introduction
Fetal growth restriction is fetal growth that has failed to reach its genetically determined potential. However, this condition has no gold standard definition and is studied using proxies, principally small for gestational age, generally defined as babies who have birthweights below the tenth percentile for those of the same gestational age, plus either ultrasonic evidence of fetal growth restriction (slow growth or uteroplacental insufficiency) or associated complications, including preterm birth, pre-eclampsia, stillbirth, perinatal asphyxia, and neo-natal morbidity.1 Multiple aspects of obstetric care are influenced by suspected fetal growth restriction, including antenatal assessment of fetal wellbeing, maternal assessment for pre-eclampsia, the interpretation and response to fetal monitoring in labour, and the timing and method of delivery.2 However, screening for fetal growth restriction in the USA and UK is based on clinical grounds;3, 4 women are selected for ultrasonographic fetal biometry on the basis of previous risk factors, acquired complications of pregnancy, or if they appear small for dates on clinical examination. Multiple studies5, 6, 7, 8, 9 have reported that this approach has poor sensitivity for detecting small babies, and undiagnosed fetal growth restriction is a common finding in audits of perinatal deaths.10, 11 An alternative to this approach is universal ultrasound. However, a meta-analysis12 of randomised controlled trials, including 13 studies that recruited around 35 000 women in total, did not show any improvement in outcomes. Specifically, use of routine late pregnancy ultrasound was not associated with a reduced risk of perinatal death (stillbirth or neonatal death) or perinatal morbidity (eg, depressed Apgar score, neonatal unit admission, and need for resuscitation). Moreover, a study from France,13 where late pregnancy ultrasound scan is routine, showed high iatrogenic prematurity among false positives. Therefore, better screening for fetal growth restriction is an area of unmet clinical need.
Research in context.
Evidence before this study
We reviewed evidence about screening and intervention for fetal growth restriction included in the NICE Antenatal Care Guideline and the Cochrane library. Fetal growth restriction is a major cause of potentially preventable perinatal morbidity and mortality. However, there is no clear evidence that screening for this condition results in improved clinical outcomes. The NICE Antenatal Care Guideline did systematic reviews of the diagnostic and clinical effectiveness of screening for small babies. These reviews did not identify any large-scale study of the diagnostic effectiveness of screening for fetal growth restriction that included blinding of the test result and the NICE Antenatal Care Guideline identified development of better methods of screening as a priority area for future research. A meta-analysis of 13 randomised controlled trials, which recruited around 35 000 women in total, evaluated the effect of universal ultrasound on obstetric practice and did not show any improvement in pregnancy outcomes. Furthermore, there is evidence that use of universal ultrasound, one of the most obvious approaches to screening, could actually lead to harm through the adverse effects of iatrogenic prematurity on false positives. A crucial element for any screening study is a test that has a low false positive rate and a high positive likelihood ratio. Many previous studies have reported associations between placental biomarkers and the risk of fetal growth restriction; however, the positive predictive values for biomarkers on their own were generally low. We previously hypothesised that a combination of ultrasonic and biochemical screening might offer clinically useful prediction when screening low-risk women.
Added value of this study
The key finding of our study is that a combination of ultrasonic suspicion of a baby being small and a blood test indicating placental dysfunction (elevated soluble fms-like tyrosine kinase 1 to placental growth factor ratio) was strongly predictive of clinically important adverse pregnancy outcomes. For preterm birth of a small baby, the positive likelihood ratio was 41 and for term delivery of a small baby with either maternal pre-eclampsia or perinatal morbidity or mortality, the positive likelihood ratio was 17·5. Only 1–3% of the population screened positive at each gestational age. Both associations were far stronger than the previously described best performing methods using purely ultrasonic assessment and the false positive rate was much lower.
Implications of all the available evidence
Trials of screening and intervention based on a combination of ultrasonic suspicion of a baby being small and a blood test indicating placental dysfunction have a realistic chance of showing clinical effectiveness in reducing adverse pregnancy outcomes associated with fetal growth restriction. Such trials should be an area for future research.
Expression microarray studies identified upregulation of fms-like tyrosine kinase 1 (FLT1) as a key change in the placental transcriptome in pre-eclampsia and this was mirrored by elevated soluble FLT1 (sFLT1) protein in the mother's blood.14 sFLT1 binds placental growth factor (PlGF); hence, pre-eclampsia is associated with an elevated sFLT1/PlGF ratio.14 A large-scale, prospective study15 showed that an elevated sFLT1/PlGF ratio provides clinically useful assessment of risk in women with suspected pre-eclampsia, and this test is now recommended in the UK by the National Institute for Health and Care Excellence (NICE).16 A large body of evidence also implicates dysfunction of the placenta in many cases of fetal growth restriction, including those leading to stillbirth.17 However, whether the sFLT1/PlGF ratio might be useful in screening for fetal growth restriction is unclear. We studied a large prospective cohort of unselected nulliparous women with a singleton pregnancy to investigate the combination of ultrasonic fetal biometry and the sFLT1/PlGF ratio as a screening test to predict delivery of a small for gestational age infant combined with an adverse pregnancy outcome (preterm birth, perinatal morbidity or mortality, and maternal pre-eclampsia).
Methods
Study design and participants
The pregnancy outcome prediction (POP) study was a prospective cohort study of nulliparous women attending the Rosie Hospital, Cambridge (UK) for their dating ultrasound scan between Jan 14, 2008, and July 31, 2012. Women with a viable singleton pregnancy were eligible for inclusion. The study has been previously described in detail.6, 18, 19 Women were ineligible for the study if they had any previous births, liveborn or stillborn, at or after 24 weeks of gestational age. Women were recruited following their dating ultrasound scan (which was scheduled for approximate gestational age 12 weeks) and had blood taken on the day of recruitment and at three subsequent visits at the National Institute of Health Research (NIHR) Cambridge Clinical Research Facility (at approximately 20 weeks, 28 weeks, and 36 weeks of gestational age). At each of these visits an ultrasound scan was also done, and at the 20 weeks of gestational age visit a questionnaire was completed through an interview by a research midwife or research sonographer to retrieve demographic data and medical history.
Cambridgeshire 2 Research Ethics Committee gave ethical approval for the study (reference number 07/H0308/163) and all participants provided written informed consent. The reporting of this study conforms to the STARD 2015 guidelines for reporting diagnostic accuracy studies and we have included a completed checklist (appendix).
Procedures
Our analysis focused on the research ultrasound scan data and blood sampling done at 28 and 36 weeks of gestational age. The conduct, analysis, and reproducibility of the ultrasound data have previously been described in detail.6 Fetal biometry was done at 20, 28 and 36 weeks of gestational age. The measurements made were the fetal biparietal diameter, the head circumference, the abdominal circumference, the femur length and the amniotic fluid index. At each gestational age, we calculated an estimated fetal weight using the equations described by Hadlock and colleagues,20 and converted this into a percentile for gestational age.21 Additionally, Doppler flow velocimetry of the uterine and umbilical arteries was done. Women and clinicians were masked to the results of the research ultrasound scan unless there was one or more of the following: breech presentation (at 36 weeks of gestational age only), previously unrecognised congenital anomaly, previously unrecognised placenta praevia, or severe oligohydramnios, defined as an amniotic fluid index less than 5.
Serum samples were collected as previously reported,18 stored at −80°C, and sFLT1/PlGF were measured using Roche Elecsys assays on the electrochemiluminescence immunoassay platform Cobas e411 (Roche Diagnostics, Basel, Switzerland).22 Using this system, the intra-assay coefficient of variation for human serum samples is less than 2% for sFLT1/PlGF, and the inter-assay coefficients of variation are 2·3–4·3% for the sFLT1 assay and 2·7–4·1% for the PlGF assay.
Women were screened with ultrasound and blood testing at 28 and 36 weeks of gestational age. At 28 weeks, we defined screen positive as an estimated fetal weight below the tenth percentile and a sFLT1/PlGF ratio greater than 5·78. At 36 weeks, we defined screen positive as an estimated fetal weight below the tenth percentile and a sFLT1/PlGF ratio greater than 38. The 38 threshold had been previously described in multicentre development and validation studies.15 However, we did not use greater than 38 for the 28 weeks of gestational age measurement because it would represent an extreme elevation at that age (>99·5th percentile22). The two thresholds of sFLT1 to PlGF ratio used were the 85th percentile at each of the gestational ages.
We compared the associations with the above definitions of screen positivity with purely ultrasonic methods. Specifically, we analysed ultrasonic estimated fetal weight below the tenth percentile plus low ab-dominal circumference growth velocity as defined in a previous analysis of the cohort,6 and definitions of fetal growth restriction (both early onset and late onset) based on biometry, growth velocity, and Doppler flow velocimetry described following a Delphi procedure23 (see appendix for diagnostic criteria).
All classification of outcomes was done blind to the results of the research ultrasound scan and maternal serum concentrations of sFLT1/PlGF. The definitions of all exposures, outcomes, and statistical methods were agreed before data analysis and are documented in signed and dated analysis plans (appendix).
Outcomes
The primary outcome following the 28 weeks of gestational age assessment was subsequent preterm delivery (<37 weeks of gestational age) of a small for gestational age infant (birthweight below the tenth percentile). Preterm small for gestational age was defined using a customised reference standard using the Gestation-Related Optimal Weight version 6.7.8.1 bulk calculator (Perinatal Institute, Birmingham, UK). Our rationale for using this reference standard was that this is a widely used method for calculating birthweight percentile using a fetal growth standard, and use of a fetal growth standard might be more appropriate than a birthweight-based standard when assessing preterm births. The basis for this view is the observation that babies born preterm are more likely to have reduced growth velocity;24, 25 hence, it is problematic to derive a normal birthweight range for preterm deliveries on the basis of observed birthweights. A sensitivity analysis addressed the effect of using a non-customised, birthweight-based growth standard in the definition of preterm small for gestational age.
The primary outcome following the 36 weeks of gestational age assessment was subsequent delivery of a small for gestational age infant with complications (perinatal morbidity, non-anomalous perinatal death, or maternal pre-eclampsia). At this gestational age, a small for gestational age infant was defined as an infant with a birthweight below the tenth percentile for sex and gestational age using a population-based UK reference range.26
We defined perinatal morbidity as at least one of the following: a 5-min Apgar score less than 7, metabolic acidosis (cord pH <7·10 with a base deficit >10 mEq/L), or admission to the neonatal unit (ie, the intensive care unit, high dependency unit, or special care baby unit) at term for 48 h or more within 48 h of delivery. We also studied a subgroup of infants who had severe adverse perinatal outcome, defined as at least one of the following: non-anomalous stillbirth or term livebirth associated with neonatal death, hypoxic ischaemic encephalopathy, severe metabolic acidosis (a cord blood pH <7·0 and a base deficit of >12 mEq/L), or treatment with inotropes or mechanical ventilation. Pre-eclampsia was defined and classified on the basis of the 2013 American College of Obstetricians and Gynecologists criteria (appendix).27
Statistical analysis
The sample size for the present study was determined by recruitment to the POP study. Power calculations showed that with 4000 women in the analysis and a four-times increase in risk associated with an exposure affecting 5% of the population, we had 74% power to detect an association if the incidence of the condition was 0·65%, and 94% power if the incidence was 1·5%. We excluded women who had missing screen positive status or outcome from the analyses (appendix). We assessed screening performance using 2 × 2 tables and standard summary statistics when the exposures were treated categorically. We estimated 95% CIs for the positive and negative likelihood ratios using the method described by Simel and colleagues.28 For other screening statistics (sensitivity, specificity, and positive and negative predictive values), we calculated 95% CIs using a method (logit-transformed) to ensure the interval was always within the possible range (0–100%).29 When screening measurements were treated as continuous variables, we quantified discrimination by the area under the receiver operating characteristic (ROC) curve and we compared the areas under the curve using the De Long test. Additionally, we did time-to-event analysis. Delivery with the given outcome was the event and delivery without the outcome was treated as competing risk. This analysis was done using Stata package stcrreg, followed by stcurve, with option cif to generate cumulative incidence plots of the outcome from the time of measurement, using gestational age as the timescale. The stcrreg package uses a semiparametric partial likelihood based time-to-event regression model with competing risks, and the method is described in detail elsewhere.30 Analyses were done using Stata 14.2 and Stata 15.1.
Role of the funding source
The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. US and GCSS (corresponding author) had full access to all the data in the study. The corresponding author had final responsibility for the decision to submit for publication.
Results
Between Jan 14, 2008, and July 31, 2012, we identified 8028 eligible women, of whom we recruited 4512 to the study. A study profile has previously been published.6 The study group for the present analysis included 4098 women (91% of those recruited) who had a measurement of the sFLT1/PlGF ratio and estimated fetal weight at 28 or 36 weeks of gestational age and outcome data available. We included 3981 women in the analysis based on 28 weeks of gestational age measurements and 3747 women were included based on 36 weeks of gestational age measurements. In total, 84 (2%) women had one of the primary outcomes, and we tabulated the characteristics of the study group based on this (table 1).
Table 1.
Study cohort characteristics by primary outcome status
| Preterm small for gestational age (n=26) | Small for gestational age plus complication*(n=58) | All others (n=4014) | ||
|---|---|---|---|---|
| Maternal characteristics | ||||
| Age†, years | 30 (25–35) | 31 (27–35) | 30 (27–33) | |
| Age when stopped full-time education, years | 19 (18–22) | 21 (18–22) | 21 (18–23) | |
| Data missing | 1 (4%) | 1 (2%) | 120 (3%) | |
| Height, cm | 164 (160–167) | 162 (158–167) | 165 (161–169) | |
| Deprivation quartile‡ | ||||
| 1 (lowest) | 4 (15%) | 18 (31%) | 975 (24%) | |
| 2 | 5 (19%) | 8 (14%) | 962 (24%) | |
| 3 | 9 (35%) | 18 (31%) | 957 (24%) | |
| 4 (highest) | 7 (27%) | 11 (19%) | 956 (24%) | |
| Data missing | 1 (4%) | 3 (5%) | 164 (4%) | |
| Ethnicity | ||||
| Non-white | 1 (4%) | 1 (2%) | 225 (6%) | |
| White | 25 (96%) | 57 (98%) | 3721 (93%) | |
| Data missing | 0 | 0 | 68 (2%) | |
| Married | 17 (65%) | 38 (66%) | 2733 (68%) | |
| Smoker | 1 (4%) | 9 (16%) | 191 (5%) | |
| Any alcohol consumption | 1 (4%) | 4 (7%) | 181 (5%) | |
| Data missing | 0 | 0 | 1 (<1%) | |
| Body-mass index, kg/m2 | 27 (22–30) | 24 (22–27) | 24 (22–27) | |
| Data missing | 0 | 0 | 1 (<1%) | |
| Type 1 or type 2 diabetes | 0 | 0 | 14 (<1%) | |
| Chronic hypertension | 6 (23%) | 8 (14%) | 198 (5%) | |
| Renal disease | 0 | 2 (3%) | 38 (1%) | |
| Birth outcomes | ||||
| Birthweight, g | 1840 (1510–2130) | 2588 (2335–2845) | 3433 (3125–3745) | |
| Birthweight, centile (population-based) | 11·0 (3·0–15·1) | 3·4 (1·2–5·7) | 44·7 (25·2–67·0) | |
| Birthweight, centile (customised) | 3·9 (0·8–5·2) | 2·5 (0·7–6·5) | 45·9 (24·6–69·7) | |
| Gestational age, weeks | 34·2 (33·3–36·0) | 40·1 (39·0–41·1) | 40·3 (39·3–41·1) | |
| Induction of labour | 4 (15%) | 26 (45%) | 1280 (32%) | |
| Method of delivery | ||||
| Spontaneous vaginal | 4 (15%) | 19 (33%) | 1977 (49%) | |
| Operative vaginal | 3 (12%) | 15 (26%) | 948 (24%) | |
| Pre-labour caesarean | 18 (69%) | 12 (21%) | 378 (9%) | |
| Intrapartum caesarean | 1 (4%) | 12 (21%) | 701 (17%) | |
| Data missing | 0 | 0 | 10 (<1%) | |
| Gestational age at scan, weeks | ||||
| 12 week scan | 12·6 (12·0–13·1) | 12·7 (12·1–13·1) | 12·6 (12·1–13·1) | |
| 20 week scan | 20·3 (20·0–20·7) | 20·3 (20·1–20·6) | 20·3 (20·0–20·6) | |
| Data missing | 0 | 0 | 4 (<1%) | |
| 28 week scan | 28·1 (27·9–28·3) | 28·3 (28·0–28·6) | 28·3 (28·0–28·6) | |
| Data missing | 0 | 0 | 20 (<1%) | |
| 36 week scan | 36·0 (35·7–36·1) | 36·1 (36·0–36·4) | 36·1 (36·0–36·4) | |
| Data missing | 20 (77%) | 0 | 209 (5%) | |
| Exposures at 28 weeks of gestational age | ||||
| Estimated fetal weight centile | 13·2 (4·1–28·1) | 17·4 (10·1–29·2) | 39·6 (22·0–60·9) | |
| Estimated fetal weight below the tenth percentile | 12 (46%) | 13 (22%) | 303 (8%) | |
| Data missing | 0 | 1 (2%) | 25 (1%) | |
| sFLT1, pg/mL | 2408 (1407–4322) | 1982 (1349–2581) | 1489 (1078–2038) | |
| Data missing | 0 | 0 | 110 (3%) | |
| PlGF, pg/mL | 137 (51–222) | 315 (181–546) | 542 (374–789) | |
| Data missing | 0 | 0 | 110 (3%) | |
| sFLT1/PlGF ratio | 15·9 (5·5–70·3) | 5·6 (2·8–9·8) | 2·7 (1·8–4·2) | |
| sFLT1/PlGF ratio >5·78 | 19 (73%) | 28 (48%) | 548 (14%) | |
| Data missing | 0 | 0 | 110 (3%) | |
| Estimated fetal weight below the tenth percentile and sFLT1/PlGF ratio >5·78 | 10 (38%) | 4 (7%) | 33 (1%) | |
| Data missing | 0 | 1 (2%) | 116 (3%) | |
| Ultrasonic estimated fetal weight below the tenth percentile and lowest decile of abdominal circumference growth velocity | 3 (12%) | 4 (7%) | 83 (2%) | |
| Data missing | 0 | 1 (2%) | 44 (1%) | |
| Delphi procedure definition of fetal growth restriction | 12 (46%) | 8 (14%) | 99 (2%) | |
| Data missing | 1 (4%) | 2 (3%) | 128 (3%) | |
| Exposures 36 weeks of gestational age | ||||
| Estimated fetal weight centile | 0·9 (0·5–2·2) | 5·9 (2·4–11·7) | 37·3 (18·8–63·3) | |
| Estimated fetal weight below the tenth percentile | 6 (23%) | 39 (67%) | 492 (12%) | |
| Data missing | 20 (77%)§ | 0 | 221 (6%) | |
| sFLT1, pg/mL | 5896 (4250–6314) | 5284 (3356–9051) | 3013 (2209–4267) | |
| Data missing | 21 (81%) | 0 | 301 (7%) | |
| PlGF, pg/mL | 97 (77–102) | 139 (71–209) | 260 (155–447) | |
| Data missing | 21 (81%) | 0 | 301 (7%) | |
| sFLT1/PlGF ratio | 58·0 (55·3–65·0) | 40·4 (17·1–102·2) | 11·7 (5·3–24·5) | |
| sFLT1/PlGF ratio >38 | 4 (15%) | 31 (53%) | 531 (13%) | |
| Data missing | 21 (81%) | 0 | 301 (7%) | |
| Estimated fetal weight below the tenth percentile and sFLT1/PlGF ratio >38 | 4 (15%) | 22 (38%) | 76 (2%) | |
| Data missing | 21 (81%) | 0 | 330 (8%) | |
| Ultrasonic estimated fetal weight below the tenth percentile and lowest decile of abdominal circumference growth velocity | 4 (15%) | 18 (31%) | 142 (4%) | |
| Data missing | 20 (77%) | 0 | 236 (6%) | |
| Delphi procedure definition of fetal growth restriction | 6 (23%) | 35 (60%) | 385 (10%) | |
| Data missing | 20 (77%) | 1 (2%) | 280 (7%) | |
Data are expressed as median (IQR) or n (%). For fields where there is no category labelled as missing, data were 100% complete. All maternal characteristics, except age at recruitment, were defined by self-report in a questionnaire at 20 weeks, from examination of the clinical case record, or linkage to the hospital's electronic databases. For birthweight, a customised reference was used to define preterm small for gestational age (ie, the primary outcome for the analysis based on the 28 weeks of gestational age measurements) and a population-based reference was used to define small for gestational age plus complication at or near term (ie, the primary outcome for the analysis based on the 36 weeks of gestational age measurements). sFLT1=soluble fms-like tyrosine kinase 1. PlGF=placental growth factor.
Complication was defined as perinatal morbidity, mortality, or pre-eclampsia.
Age at recruitment.
Socioeconomic status was quantified using the Index of Multiple Deprivation (IMD) 2007, which is based on census data from the area of the mother's postcode.31
The high rate of missing values reflects the fact that 77% of these women delivered before their scheduled 36 weeks of gestational age research visit.
At 28 weeks of gestational age, 320 women (8%) had an ultrasonic estimated fetal weight below the tenth percentile, 593 women (15%) had an sFLT1/PlGF ratio greater than 5·78, and 47 women (1%) screened positive for both tests (table 2). The positive likelihood ratio for preterm delivery of a small for gestational age infant associated with this combination was 41·1 (95% CI 23·0–73·6), the positive predictive value was 21·3% (11·6–35·8), and the screen positive group included over a third of the women who had a preterm delivery of a small for gestational age infant (table 2). Delivery occurred 2 or more weeks after the 28 week gestational age measurements in around 95% of cases (figure 1). At 36 weeks of gestational age, 521 (14%) women had an ultrasonic estimated fetal weight below the tenth percentile, 563 (15%) women had an sFLT1/PlGF ratio greater than 38, and 102 women (3%) screened positive in both tests (table 3). The positive likelihood ratio for delivery of a small for gestational age infant associated with maternal pre-eclampsia, or perinatal morbidity or mortality associated with the combination, was 17·5 (95% CI 11·8–25·9; table 3). The positive predictive value was 21·6% (14·5–30·8), and the screen positive group included over a third of the women who had a delivery of a small for gestational age infant with complications. Delivery occurred 2 or more weeks after the 36 weeks of gestational age measurements in around 80% of cases (figure 1).
Table 2.
Ultrasonic and biochemical screening test diagnostic effectiveness at 28 weeks of gestational age for preterm delivery of a small for gestational age infant (n=3981)
| True positive/false positive | True negative/false negative | Positive likelihood ratio (95% CI) | Negative likelihood ratio (95% CI) | Sensitivity (95% CI) | Specificity (95% CI) | Positive predictive value (95% CI) | Negative predictive value (95% CI) | |
|---|---|---|---|---|---|---|---|---|
| Ultrasonic estimated fetal weight below the tenth percentile | 12/308 | 3647/14 | 5·9 (3·9–9·1) | 0·58 (0·41–0·83) | 46·2% (27·3–66·2) | 92·2% (91·3–93·0) | 3·8% (2·1–6·5) | 99·6% (99·4–99·8) |
| sFLT1/PlGF ratio >5·78 | 19/574 | 3381/7 | 5·0 (3·9–6·4) | 0·31 (0·17–0·59) | 73·1% (51·7–87·3) | 85·5% (84·4–86·6) | 3·2% (2·1–5·0) | 99·8% (99·6–99·9) |
| Ultrasonic estimated fetal weight below the tenth percentile and sFLT1/PlGF ratio >5·78 | 10/37 | 3918/16 | 41·1 (23·0–73·6) | 0·62 (0·46–0·84) | 38·5% (21·1–59·3) | 99·1% (98·7–99·3) | 21·3% (11·6–35·8) | 99·6% (99·3–99·8) |
| Ultrasonic estimated fetal weight below the tenth percentile and lowecst decile of abdominal circumference growth velocity* | 3/85 | 3852/23 | 5·3 (1·8–15·8) | 0·90 (0·79–1·04) | 11·5% (3·5–32·1) | 97·8% (97·3–98·3) | 3·4% (1·1–10·3) | 99·4% (99·1-99·6) |
| Delphi procedure definition of early fetal growth restriction* | 12/106 | 3747/13 | 17·4 (11·1–27·3) | 0·53 (0·37–0·78) | 48·0% (28·4–68·2) | 97·2% (96·7–97·7) | 10·2% (5·8–17·2) | 99·7% (99·4–99·8) |
Small for gestational age was defined as a customised birthweight below the tenth percentile (see Methods). sFLT1=soluble fms-like tyrosine kinase 1. PlGF=placental growth factor.
See appendix for definitions.
Figure 1.
Cumulative incidence graphs
(A) Preterm delivery of a small for gestational age infant (birthweight below the tenth customised percentile) following 28 weeks of gestational age measurement by screening status. Screen positivity was defined as estimated fetal weight below the tenth percentile for gestational age, combined with a sFLT1/PlGF ratio greater than 5·78. Number at risk 3934 screen negatives and 47 screen positives. (B) Delivery of a small for gestational age infant plus a complication following 36 weeks of gestational age measurement by screening status. Screen positivity was defined as an estimated fetal weight below the tenth percentile for gestational age, combined with sFLT1/PlGF ratio more than 38. Number at risk 3645 screen negatives and 102 screen positives. Small for gestational age was defined as birthweight below the tenth population-based percentile and a complication was defined as one or more of the following: non-anomalous perinatal death, any neonatal morbidity, or maternal pre-eclampsia. The number at risk increases before it decreases because the majority of women had their scan and blood sample taken after 28 weeks and zero days (A) or after 36 weeks and zero days (B). sFLT1=soluble fms-like tyrosine kinase 1. PlGF=placental growth factor.
Table 3.
Ultrasonic and biochemical screening test diagnostic effectiveness at 36 weeks of gestational age for subsequent delivery of a small for gestational age infant associated with either maternal pre-eclampsia or perinatal morbidity or mortality (n=3747)
| True positive/false positive | True negative/false negative | Positive likelihood ratio (95% CI) | Negative likelihood ratio (95% CI) | Sensitivity (95% CI) | Specificity (95% CI) | Positive predictive value (95% CI) | Negative predictive value (95% CI) | |
|---|---|---|---|---|---|---|---|---|
| Ultrasonic estimated fetal weight below the tenth percentile | 39/482 | 3207/19 | 5·1 (4·2–6·3) | 0·38 (0·26–0·54) | 67·2% (53·8–78·3) | 86·9% (85·8–88·0) | 7·5% (5·5–10·1) | 99·4% (99·1–99·6) |
| sFLT1/PlGF ratio >38 | 31/532 | 3157/27 | 3·7 (2·9–4·8) | 0·54 (0·41–0·72) | 53·4% (40·3–66·1) | 85·6% (84·4–86·7) | 5·5% (3·9–7·7) | 99·2% (98·8–99·4) |
| Ultrasonic estimated fetal weight below the tenth percentile and sFLT1/PlGF ratio >38 | 22/80 | 3609/36 | 17·5 (11·8–25·9) | 0·63 (0·52–0·78) | 37·9% (26·1–51·4) | 97·8% (97·3–98·3) | 21·6% (14·5–30·8) | 99·0% (98·6–99·3) |
| Ultrasonic estimated fetal weight below the tenth percentile and lowest decile of abdominal circumference growth velocity* | 18/143 | 3532/40 | 8·0 (5·3–12·1) | 0·72 (0·60–0·85) | 31·0% (20·2–44·4) | 96·1% (95·4–96·7) | 11·2% (7·1–17·1) | 98·9% (98·5–99·2) |
| Delphi procedure definition of late fetal growth restriction* | 35/377 | 3257/22 | 5·9 (4·7–7·4) | 0·43 (0·31–0·60) | 61·4% (47·9–73·4) | 89·6% (88·6–90·6) | 8·5% (6·2–11·6) | 99·3% (99·0–99·6) |
Small for gestational age was defined using a non-customised birthweight standard (see Methods). sFLT1=soluble fms-like tyrosine kinase 1. PlGF=placental growth factor.
See appendix for definitions.
We compared the predictive ability of a combination of ultrasonic small for gestational age and an elevated sFLT1/PlGF ratio with two other previously described indicators of fetal growth restriction—low abdominal circumference growth velocity6 and a composite measure generated by a Delphi procedure.23 At both 28 and 36 weeks of gestational age, the positive predictive value was around twice as high for combined ultrasonic and biochemical screening compared with the best performing purely ultrasonic method (Table 2, Table 3).
We did a prespecified subgroup analysis in which we analysed the relationship between the sFLT1/PlGF ratio and the risk of pre-eclampsia, with or without small for gestational age, using ROC curve analysis (figure 2). The association between the sFLT1/PlGF ratio at 28 weeks of gestational age and pre-eclampsia leading to preterm birth was very strong when the infant was small for gestational age, compared to when the infant was not small for gestational age (figure 2). Similarly, the sFLT1/PlGF ratio at 36 weeks of gestational age was very strongly associated with pre-eclampsia when combined with small for gestational age (figure 2).
Figure 2.
ROC curve analysis
(A) Relationship between the sFLT1/PlGF ratio at 28 weeks of gestational age and the risk of pre-eclampsia with delivery of a preterm small for gestational age infant (birthweight below the tenth customised percentile; n=12 cases). (B) Relationship between the sFLT1/PlGF ratio at 28 weeks of gestational age and the risk of pre-eclampsia with delivery of a preterm non-small for gestational age infant (birthweight at or above the tenth customised percentile; n=14 cases). (C) Relationship between the sFLT1/PlGF ratio at 36 weeks of gestational age and the risk of pre-eclampsia with delivery of a small for gestational age infant (birthweight below the tenth population-based percentile; n=21 cases). (D) Relationship between the sFLT1/PlGF ratio at 36 weeks of gestational age and the risk of pre-eclampsia with delivery of a non-small for gestational age infant (birthweight at or above the tenth population-based percentile; n=203 cases). ROC=receiver operating characteristic. sFLT1=soluble fms-like tyrosine kinase 1. PlGF=placental growth factor.
We did a series of other prespecified secondary and sensitivity analyses (Table 4, Table 5, Table 6). On its own, PlGF at 28 weeks of gestational age was equally predictive of preterm small for gestational age as the sFLT1/PlGF ratio, whereas sFLT1 was a considerably weaker predictor of preterm small for gestational age than the sFLT1/PlGF ratio (Table 4, Table 5). Similarly, at 36 weeks of gestational age PlGF was equally predictive of late fetal growth restriction as the sFLT1/PlGF ratio. However, sFLT1 was only slightly weaker or similar as a predictor compared to the sFLT1/PlGF ratio at 36 weeks (Table 5, Table 6). Adjustment of the PlGF/sFLT1 values for the exact gestational age or maternal weight at the measurement made little or no difference to the results (Table 4, Table 5, Table 6). We also found that use of a non-customised and birthweight-based standard to define preterm delivery of a small for gestational age infant resulted in a stronger association with the combination of estimated fetal weight below the tenth percentile and an elevated sFLT1/PlGF ratio; the positive likelihood ratio was 53·4 (29·6–96·4) and the sensitivity was 54% (appendix). The combination was similarly predictive of the primary outcome when the women who had their research scan result revealed were excluded (9 women at 28 weeks and 214 women at 36 weeks of gestational age; 5% in total); positive likelihood ratios were 41·0 (22·9–73·4) for preterm delivery of small for gestational age infant and 17·4 (11·5–26·4) for subsequent delivery of a small for gestational age infant with a complication. Other prespecified secondary analyses and additional analyses are presented in the appendix.
Table 4.
Biochemical screening diagnostic effectiveness for preterm delivery of a small for gestational age infant using absolute concentrations, and gestational age and maternal weight adjusted multiples of the median of sFLT1 and PlGF at 28 weeks of gestational age
| True positive/false positive | True negative/false negative | Positive likelihood ratio (95% CI) | Negative likelihood ratio (95% CI) | Sensitivity | Specificity | Positive predictive value | Negative predictive value | |
|---|---|---|---|---|---|---|---|---|
| Highest decile sFLT1/PlGF ratio | 19/377 | 3578/7 | 7·7 (6·0–9·9) | 0·30 (0·16–0·56) | 73·1% | 90·5% | 4·8% | 99·8% |
| Highest decile sFLT1 | 12/385 | 3570/14 | 4·7 (3·1–7·3) | 0·60 (0·42–0·85) | 46·2% | 90·3% | 3·0% | 99·6% |
| Lowest decile PlGF | 21/376 | 3579/5 | 8·5 (6·9–10·5) | 0·21 (0·10–0·47) | 80·8% | 90·5% | 5·3% | 99·9% |
| Highest decile sFLT1 (multiples of the median) | 13/385 | 3570/13 | 5·1 (3·5–7·6) | 0·55 (0·38–0·81) | 50·0% | 90·3% | 3·3% | 99·6% |
| Lowest decile PlGF (multiples of the median) | 20/377 | 3578/6 | 8·1 (6·4–10·2) | 0·26 (0·13–0·51) | 76·9% | 90·5% | 5·0% | 99·8% |
Multiples of the median were adjusted for the exact gestational age and maternal weight. The cutoff points of the highest decile were 6·90 for the sFLT1/PlGF ratio, 2788·5 for the sFLT1, and 1·83 for the sFLT1 multiples of the median. The cutoff points for the lowest decile of the PlGF were 252·7 and 0·48 for the PlGF multiples of the median. sFLT1=soluble fms-like tyrosine kinase 1. PlGF=placental growth factor.
Table 5.
Biochemical screening diagnostic effectiveness for subsequent delivery of a small for gestational age infant associated with either maternal pre-eclampsia, perinatal morbidity, or mortality using absolute concentrations, and gestational age and maternal weight adjusted multiples of the median of sFLT1 and PlGF at 36 weeks of gestational age
| True positive/false positive | True negative/false negative | Positive likelihood ratio (95% CI) | Negative likelihood ratio (95% CI) | Sensitivity | Specificity | Positive predictive value | Negative predictive value | |
|---|---|---|---|---|---|---|---|---|
| Highest decile sFLT1/PlGF ratio | 25/347 | 3342/33 | 4·6 (3·4–6·3) | 0·63 (0·50–0·79) | 43·1% | 90·6% | 6·7% | 99·0% |
| Highest decile sFLT1 | 24/349 | 3340/34 | 4·4 (3·2–6·0) | 0·65 (0·52–0·80) | 41·4% | 90·5% | 6·4% | 99·0% |
| Lowest decile PlGF | 25/348 | 3341/33 | 4·6 (3·3–6·2) | 0·63 (0·50–0·79) | 43·1% | 90·6% | 6·7% | 99·0% |
| Highest decile sFLT1 (multiples of the median) | 24/349 | 3340/34 | 4·4 (3·2–6·0) | 0·65 (0·52–0·80) | 41·4% | 90·5% | 6·4% | 99·0% |
| Lowest decile PlGF (multiples of the median) | 24/349 | 3340/34 | 4·4 (3·2–6·0) | 0·65 (0·52–0·80) | 41·4% | 90·5% | 6·4% | 99·0% |
Multiples of the median were adjusted for exact gestational age and maternal weight. The cutoff points of the highest decile were 49·8 for the sFLT1/PlGF ratio, 6266 for the sFLT1, and 2·06 for the sFLT1 multiples of the median. The cutoff points for the lowest decile of the PlGF were 104·4 and 0·42 for the PlGF multiples of the median. sFLT1=soluble fms-like tyrosine kinase 1. PlGF=placental growth factor.
Table 6.
ROC curve analysis of continuous biochemical exposures measured at 28 and 36 weeks of gestational age and respective primary outcomes
|
28 weeks of gestational age |
36 weeks of gestational age |
|||
|---|---|---|---|---|
| Area under ROC curve (95% CI) | p value from comparison with sFLT1/PlGF ratio* | Area under ROC curve (95% CI) | p value from comparison with sFLT1/PlGF ratio* | |
| sFLT1/PlGF ratio | 0·89 (0·82–0·97) | NA | 0·77 (0·70–0·84) | NA |
| sFLT1 | 0·74 (0·62–0·85) | 0·0012 | 0·73 (0·66–0·81) | 0·089 |
| PlGF | 0·89 (0·81–0·98) | 0·93 | 0·75 (0·69–0·82) | 0·42 |
| sFLT1 (multiples of the median) | 0·75 (0·64–0·86) | 0·0024 | 0·73 (0·65–0·80) | 0·032 |
| PlGF (multiples of the median) | 0·88 (0·79–0·97) | 0·56 | 0·76 (0·70–0·83) | 0·82 |
The area under the ROC curve for sFLT1/PlGF ratio to predict the Delphi procedure definition of ultrasonic diagnosis of fetal growth restriction was 0·55 (95% CI 0·50–0·61) at 28 weeks of gestational age and 0·57 (95% CI 0·54–0·60) at 36 weeks of gestational age. ROC=receiver operating characteristic. sFLT1=soluble fms-like tyrosine kinase 1. PlGF=placental growth factor. NA=not applicable. Multiples of the median were adjusted for gestational age and maternal weight.
p value is from the De Long test for the equality of the area under the ROC curve.
Discussion
In our study, we evaluated the combination of ultrasonic suspicion of small for gestational age plus an elevated sFLT1/PlGF ratio as a screening test for delivery of a small for gestational age infant combined with an adverse pregnancy outcome. We found that this combination was strongly predictive for preterm delivery of a small for gestational age infant when measured at 28 weeks of gestational age and for term delivery of a small for gestational age infant with maternal pre-eclampsia or perinatal morbidity or mortality when measured at 36 weeks of gestational age. Screening for fetal growth restriction is an area of unmet clinical need. Previous studies using purely ultrasonic methods of screening did not show that screening prevented adverse outcomes, such as perinatal morbidity and mortality.12 Detailed analysis of data from France,13 where universal ultrasound screening was implemented despite the absence of supportive trial data, indicated an excess of adverse outcomes (eg, caesarean delivery, need for neonatal resuscitation, and admission to a neonatal unit) among pregnancies incorrectly suspected to be affected by a small for gestational age infant (ie, false positives). Screening did not appear to result in an overall improvement in outcomes, as any benefit arising from identifying true positives was outweighed by harm caused to false positives. Using the combination of ultrasonic fetal biometry and the sFLT1/PlGF ratio, we found that the positive predictive value was more than 20% at 28 and 36 weeks of gestational age, and the false positive rate was 0·9% at 28 weeks and 2% at 36 weeks. Hence, this approach identified women with high absolute risks of adverse pregnancy outcome, with very low false positive rates in the healthy population.
Clinically effective screening also requires that the screening test is coupled with an intervention. At 28 weeks of gestational age, women who were identified as high risk by this screening test could be offered enhanced fetal assessment with the option of medically indicated delivery, as done for women with suspected fetal growth restriction identified using targeted ultrasound. High-quality evidence has been reported that assessment of high-risk pregnancies using serial umbilical artery Doppler reduces the risk of perinatal death.32 Moreover, identification of high-risk fetuses could allow optimisation of other aspects of care, including use of antenatal steroids and timing and method of delivery. At 36 weeks of gestational age, identifying late fetal growth restriction has the advantage that the primary intervention to prevent stillbirth (ie, medically indicated delivery, principally induction of labour) has reduced potential to cause harm through iatrogenic prematurity. We believe that a randomised controlled trial assessing clinical outcomes after screening and intervention using this approach is an appropriate area for future research.
An issue with some studies reporting novel diagnostic methods is that the definition of screen positive is based on observed patterns of association within the population. This approach can lead to overly optimistic estimates of screening performance, as there is the potential to set the thresholds defining screening status based on associations observed in a given study. Therefore, the findings of such studies require external validation. However, this was not the case in our study. The definition of screen positive was based on two previously described and widely accepted thresholds, namely an estimated fetal weight below the tenth percentile and an sFLT1/PlGF ratio greater than 38. Estimated fetal weight below the tenth percentile is a widely accepted definition for small for gestational age.1 The sFLT1/PlGF ratio greater than 38 threshold was obtained from previous multicentre development and validation studies.15 The only modification made in our study was to use a different threshold at 28 weeks of gestational age, as the greater than 38 level of the ratio represented a much more extreme elevation at 28 weeks than at 36 weeks. However, this and all other thresholds and definitions used in our study were predefined in an analysis plan that was finalised before any analysis of the associations with fetal growth restriction. Given that all analyses were prespecified, there is no capacity for data-driven analyses in our primary exposures and outcomes, hence there is no novel algorithm that requires external validation.
We did a series of secondary analyses. In contrast to the primary analysis, these analyses should be treated as hypothesis generating rather than hypothesis testing. One such analysis addressed use of lower thresholds of estimated fetal weight and sFLT1/PlGF ratio to define screen positives. We found that when the thresholds for screen positives were lowered to cause a similar false positive rate to the Delphi-derived ultrasonic definition of fetal growth restriction, the positive and negative likelihood ratios were similar for the two methods. Any false result in a diagnostic test has the potential to cause harm. However, the relative concerns about false positives and false negatives depend on the context. The Delphi-derived definition of fetal growth restriction was obtained from a process of surveying maternal fetal medicine specialists. Such professionals will generally be scanning in the context of a high-risk pregnancy and the result of a negative test might be to reduce the level of surveillance in a high-risk woman. In this context, a false negative result might be more concerning than a false positive. However, the context for our analysis is screening low-risk women. The result of a negative test would simply be that a woman would continue to receive the low-risk care that she would have received had she not had the test. In the context of screening low-risk women, false positive results might be of particular concern, as they will generate unnecessary anxiety and health-care costs, and could lead to short-term and long-term harm through iatrogenic prematurity. Hence, we believe that the screening performance achieved using a combination of estimated fetal weight and the sFLT1/PlGF ratio could be particularly suitable for screening low-risk populations. Another aspect of this approach is that the test could be applied sequentially; women could initially have the blood test and ultrasound might then only be done if the sFLT1/PlGF ratio is above the given threshold. This approach would lead to less pressure on imaging services. By contrast, screening using the Delphi-derived ultrasonic approach would require serial use of ultrasound in the whole population. Whether health-care systems have the capacity to deliver these imaging procedures at this scale is unclear. Moreover, the Delphi-derived ultrasonic approach also depends on uteroplacental Doppler, which is highly operator dependent, and implementation of these methods across the entire obstetric population might be problematic.
We found that the combination of small for gestational age and pre-eclampsia was exceptionally strongly associated with the sFLT1/PlGF ratio. At both 28 and 36 weeks of gestational age, the area under the ROC curve was 0·95, indicating that elevation of the ratio was almost invariably observed in this group. This observation is consistent with the hypothesis that there are distinct placental and maternal phenotypes of pre-eclampsia33 and suggests that the sFLT1/PlGF ratio might also help to define different subtypes of disease. The pathophysiological pathways leading to fetal growth restriction are complex and have recently been reviewed.34 The sFLT1/PlGF ratio clearly identifies one phenotype of fetal growth restriction with extremely high predictive accuracy. Other markers might reflect different causes of fetal growth restriction. For example, we have previously shown that maternal concentrations of delta-like homologue-1 (DLK1) were invariably lower in small for gestational age infants with abnormal umbilical artery blood flow compared with matched controls.35 Related animal studies showed that maternal DLK1 was derived from the fetus, unlike sFLT1, which is synthesised by the placenta. We speculate that panels of biochemical markers might allow identification of multiple different causes of fetal growth restriction and measurement of such panels, combined with ultrasonic estimated fetal weight, might result in future tests with higher sensitivity but similarly low false positive rates. Finally, secondary analyses showed that PlGF on its own was similarly predictive for these outcomes as the sFLT1/PlGF ratio, particularly at 28 weeks of gestational age, and further studies might address whether the combination of ultrasonic suspicion of small for gestational age plus a low PlGF might be a more cost-effective method of screening for fetal growth restriction.
Our study had several strengths. First, the study was sufficiently large to be powered for relatively uncommon adverse pregnancy outcomes. Second, the results of ultrasonographic fetal biometry were blinded. Few studies have evaluated prenatal ultrasonography with blinding, which is one of the key characteristics for a study of diagnostic effectiveness. Third, the biochemical analyses of the sFLT1/PlGF ratio also had a number of strengths. These analyses were also done blind to clinical outcome; were done using an automated, clinical grade, and widely available immunoassay platform; the ratio can readily be calculated without complex corrections for maternal characteristics; and the threshold used for screen positivity at 36 weeks of gestational age was previously generated for pre-eclampsia from a large-scale study and validated in another study.15, 22 Fourth, the primary exposures, primary outcomes, and analyses were defined in prespecified analysis plans. Finally, as the cohort was selected on the basis of nulliparity and singleton pregnancy alone, it included many low-risk women. Studies can overestimate the positive predictive value of a screening test by extrapolating the results of selected and high-risk cohorts to the general population.36 Therefore, given that our study used a cohort with a large proportion of low-risk women drawn from a largely healthy population, the combination of estimated fetal weight below the tenth percentile and sFLT1/PlGF ratio above the 85th percentile might be at least equally predictive in other populations, and could even be more strongly predictive of fetal growth restriction when investigated in other settings.
Our study also has limitations. We revealed the results of the research ultrasound scan in 5% of the study group (most cases were for breech presentation at 36 weeks of gestational age). However, the main findings were the same when the analysis was confined to women who did not have their research scan results revealed. Second, the study was confined to first and singleton pregnancies. Hence, the results cannot be extrapolated to multiparous women or multiple pregnancies and further studies will be required for these populations. However, the selection of nulliparous women was purposeful, as these women have higher rates of adverse outcomes and the most informative method of assessing risk of obstetric complications—namely, past obstetric history—is not available. Another issue that limits the external validity of the data for some populations is that the pregnancy outcome prediction study cohort was largely of white ethnicity. However, both elements of the screening test are likely to be applicable in other populations, as fetal growth is quite consistent across different ethnic groups37 and previous reports have indicated that the sFLT1/PlGF ratio is similarly associated with pre-eclampsia in black Africans38 and with placental dysfunction in a US population where more than 90% of women were African American.39 A further limitation is the relatively small number of events following each of the screening assessments, despite the large number of women in the cohort. The low event rate partly reflects the fact that the outcomes studied were clinically significant, hence uncommon. The small number of cases introduces some uncertainty in the exact screening properties of the test, and this is illustrated by the 95% CI. However, examination of the 95% CI shows that the lower limit of the positive predictive value at both gestational ages is more than 10. This observation indicates that screen positive women are at significant risk. Secondly, the lower limit of the specificity is more than 97% at both timepoints, indicating that the false positive rate is less than 3% even with the most pessimistic assessment of our results. A final relative weakness is that the data were generated using a single immunoassay platform, and the optimal cutoff sFLT1/PlGF ratio might differ when other assays are used.
In conclusion, our study showed that the combination of an ultrasonically suspected small for gestational age fetus and an elevated maternal sFLT1/PlGF ratio is strongly predictive of adverse pregnancy outcome associated with fetal growth restriction. The predictive performance of this approach improves upon previously described methods of screening using purely ultrasonic methods. We hypothesise that screening using a combination of ultrasound and measurement of placental biomarkers in the mother's blood could improve maternal and perinatal outcomes when coupled with an appropriate management plan for women with positive screening results. This novel method of screening could be tested against the current standard of care in a randomised controlled trial where the primary outcome was delivery of a small for gestational age infant associated with an adverse pregnancy outcome.
For the STARD 2015 guidelines see http://www.stard-statement.org/
For more on the Gestation-Related Optimal Weight (UK) bulk calculator see http://www.gestation.net/
Acknowledgments
Acknowledgments
We are extremely grateful to all the participants in the pregnancy outcome prediction study. We thank Leah Bibby, Samudra Ranawaka, Katrina Holmes, Josephine Gill, and Ryan Millar for technical assistance with the biochemical assays. We thank Alison Dacey, Amanda Rowley, Maxine Snaith, Sam Everett, Sonia Cumming, and Jan Nelder for performing the research ultrasound scans. This study was funded by the National Institute for Health Research (NIHR) Cambridge Comprehensive Biomedical Research Centre (Women's Health theme), and project grants from the Medical Research Council (UK; grant no G1100221) and the Stillbirth and neonatal death society (Sands). The study was also supported by Roche Diagnostics (provision of equipment and reagents for analysis of soluble fms-like tyrosine kinase 1 and placental growth factor), by GE Healthcare (donation of two Voluson i ultrasound systems for this study), and by the NIHR Cambridge Clinical Research Facility, where all research visits took place. US and GCSS had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Elements of the content of this paper were presented at the Society for Maternal-Fetal Medicine, Las Vegas (2017) and Dallas (2018).
Contributors
GCSS conceived and designed the study. FG, EC, and MH acquired the data. US, DSC-J, and GCSS analysed and interpreted that data. FG, US, and GCSS drafted the manuscript. All authors critically reviewed the manuscript for important intellectual content and approved the final version for publication.
Declaration of interests
FG, US, EC, DSC-J, and GCSS report grants from the National Institute for Health Research (NIHR) Cambridge Comprehensive Biomedical Research Centre, grants from the Medical Research Council, grants from the Stillbirth and neonatal death society (Sands), non-financial support from Roche Diagnostics Ltd, non-financial support from GE Healthcare, and non-financial support from NIHR Cambridge Clinical Research Facility, all during the conduct of the study. GCSS reports grants from the Scottish Government (Chief Scientist Office Division), grants and personal fees from GlaxoSmithKline, and personal fees from Roche Diagnostics, all outside the submitted work. Additionally, GCSS has a patent pending (PCT/EP2014/062602). MH reports being an employee of Roche Diagnostics and holding stock in F Hoffmann-La Roche.
Supplementary Material
References
- 1.Gaccioli F, Aye ILMH, Sovio U, Charnock-Jones DS, Smith GCS. Screening for fetal growth restriction using fetal biometry combined with maternal biomarkers. Am J Obstet Gynecol. 2018;218:S725–S737. doi: 10.1016/j.ajog.2017.12.002. [DOI] [PubMed] [Google Scholar]
- 2.Adelson P, Spurrett B, Trudinger B, Frommer M. A New South Wales population-based study of stillbirths weighing 2500 g or more. Aust NZ J Obstet Gynaecol. 1993;33:166–173. doi: 10.1111/j.1479-828x.1993.tb02385.x. [DOI] [PubMed] [Google Scholar]
- 3.National Institute for Health and Care Excellence Antenatal care for uncomplicated pregnancies. Clinical guideline (CG62) March, 2008. https://www.nice.org.uk/guidance/cg62 [PubMed]
- 4.American College of Obstetricians and Gynecologists ACOG practice bulletin no. 134: fetal growth restriction. Obstet Gynecol. 2013;121:1122–1133. doi: 10.1097/01.AOG.0000429658.85846.f9. [DOI] [PubMed] [Google Scholar]
- 5.Gardosi J. Intrauterine growth restriction: new standards for assessing adverse outcome. Best Pract Res Clin Obstet Gynaecol. 2009;23:741–749. doi: 10.1016/j.bpobgyn.2009.09.001. [DOI] [PubMed] [Google Scholar]
- 6.Sovio U, White IR, Dacey A, Pasupathy D, Smith GCS. Screening for fetal growth restriction with universal third trimester ultrasonography in nulliparous women in the Pregnancy Outcome Prediction (POP) study: a prospective cohort study. Lancet. 2015;386:2089–2097. doi: 10.1016/S0140-6736(15)00131-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Bais JM, Eskes M, Pel M, Bonsel GJ, Bleker OP. Effectiveness of detection of intrauterine growth retardation by abdominal palpation as screening test in a low risk population: an observational study. Eur J Obstet Gynecol Reprod Biol. 2004;116:164–169. doi: 10.1016/j.ejogrb.2004.01.037. [DOI] [PubMed] [Google Scholar]
- 8.Sparks TN, Cheng YW, McLaughlin B, Esakoff TF, Caughey AB. Fundal height: a useful screening tool for fetal growth? J Matern Fetal Neonatal Med. 2011;24:708–712. doi: 10.3109/14767058.2010.516285. [DOI] [PubMed] [Google Scholar]
- 9.Chauhan SP, Rouse DJ, Ananth CV. Screening for intrauterine growth restriction in uncomplicated pregnancies: time for action. Am J Perinatol. 2013;30:33–39. doi: 10.1055/s-0032-1321497. [DOI] [PubMed] [Google Scholar]
- 10.Saastad E, Vangen S, Frøen JF. Suboptimal care in stillbirths— a retrospective audit study. Acta Obstet Gynecol Scand. 2007;86:444–450. doi: 10.1080/00016340701207724. [DOI] [PubMed] [Google Scholar]
- 11.Stacey T, Thompson JM, Mitchell EA, Zuccollo JM, Ekeroma AJ, McCowan LM. Antenatal care, identification of suboptimal fetal growth and risk of late stillbirth: findings from the Auckland Stillbirth Study. Aust NZ J Obstet Gynaecol. 2012;52:242–247. doi: 10.1111/j.1479-828X.2011.01406.x. [DOI] [PubMed] [Google Scholar]
- 12.Bricker L, Medley N, Pratt JJ. Routine ultrasound in late pregnancy (after 24 weeks' gestation) Cochrane Database Syst Rev. 2015;6 doi: 10.1002/14651858.CD001451.pub4. CD001451. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Monier I, Blondel B, Ego A, Kaminiski M, Goffinet F, Zeitlin J. Poor effectiveness of antenatal detection of fetal growth restriction and consequences for obstetric management and neonatal outcomes: a French national study. BJOG. 2015;122:518–527. doi: 10.1111/1471-0528.13148. [DOI] [PubMed] [Google Scholar]
- 14.Hod T, Cerdeira AS, Karumanchi SA. Molecular mechanisms of preeclampsia. Cold Spring Harb Perspect Med. 2015;5:10. doi: 10.1101/cshperspect.a023473. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Zeisler H, Llurba E, Chantraine F. Predictive value of the sFlt-1:PlGF ratio in women with suspected preeclampsia. N Engl J Med. 2016;374:13–22. doi: 10.1056/NEJMoa1414838. [DOI] [PubMed] [Google Scholar]
- 16.National Institute for Health and Care Excellence PlGF-based testing to help diagnose suspected pre-eclampsia. Diagnostics guidance (DG23) May, 2016. https://www.nice.org.uk/guidance/dg23
- 17.Smith GC, Fretts RC. Stillbirth. Lancet. 2007;370:1715–1725. doi: 10.1016/S0140-6736(07)61723-1. [DOI] [PubMed] [Google Scholar]
- 18.Pasupathy D, Dacey A, Cook E, Charnock-Jones DS, White IR, Smith GC. Study protocol. A prospective cohort study of unselected primiparous women: the pregnancy outcome prediction study. BMC Pregnancy Childbirth. 2008;8:51. doi: 10.1186/1471-2393-8-51. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Gaccioli F, Lager S, Sovio U, Charnock-Jones DS, Smith GCS. The pregnancy outcome prediction (POP) study: investigating the relationship between serial prenatal ultrasonography, biomarkers, placental phenotype and adverse pregnancy outcomes. Placenta. 2017;59:S17–S25. [Google Scholar]
- 20.Hadlock FP, Harrist RB, Sharman RS, Deter RL, Park SK. Estimation of fetal weight with the use of head, body, and femur measurements—a prospective study. Am J Obstet Gynecol. 1985;151:333–337. doi: 10.1016/0002-9378(85)90298-4. [DOI] [PubMed] [Google Scholar]
- 21.Hadlock FP, Harrist RB, Martinez-Poyer J. In utero analysis of fetal growth: a sonographic weight standard. Radiology. 1991;181:129–133. doi: 10.1148/radiology.181.1.1887021. [DOI] [PubMed] [Google Scholar]
- 22.Sovio U, Gaccioli F, Cook E, Hund M, Charnock-Jones DS, Smith GC. Prediction of preeclampsia using the soluble fms-like tyrosine kinase 1 to placental growth factor ratio: a prospective cohort study of unselected nulliparous women. Hypertension. 2017;69:731–738. doi: 10.1161/HYPERTENSIONAHA.116.08620. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Gordijn SJ, Beune IM, Thilaganathan B. Consensus definition of fetal growth restriction: a Delphi procedure. Ultrasound Obstet Gynecol. 2016;48:333–339. doi: 10.1002/uog.15884. [DOI] [PubMed] [Google Scholar]
- 24.Gaillard R, Steegers EA, de Jongste JC, Hofman A, Jaddoe VW. Tracking of fetal growth characteristics during different trimesters and the risks of adverse birth outcomes. Int J Epidemiol. 2014;43:1140–1153. doi: 10.1093/ije/dyu036. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Partap U, Sovio U, Smith GC. Fetal growth and the risk of spontaneous preterm birth in a prospective cohort study of nulliparous women. Am J Epidemiol. 2016;184:110–119. doi: 10.1093/aje/kwv345. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Freeman JV, Cole TJ, Chinn S, Jones PR, White EM, Preece MA. Cross sectional stature and weight reference curves for the UK, 1990. Arch Dis Child. 1995;73:17–24. doi: 10.1136/adc.73.1.17. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.American College of Obstetricians and Gynecologists. Task Force on Hypertension in pregnancy Report of the American College of Obstetricians and Gynecologists' Task Force on Hypertension in Pregnancy. Obstet Gynecol. 2013;122:1122–1131. doi: 10.1097/01.AOG.0000437382.03963.88. [DOI] [PubMed] [Google Scholar]
- 28.Simel DL, Samsa GP, Matchar DB. Likelihood ratios with confidence: sample size estimation for diagnostic test studies. J Clin Epidemiol. 1991;44:763–770. doi: 10.1016/0895-4356(91)90128-v. [DOI] [PubMed] [Google Scholar]
- 29.Dean N, Pagano M. Evaluating confidence interval methods for binomial proportions in clustered surveys. J Surv Stat Methodol. 2015;3:484–503. [Google Scholar]
- 30.Fine JP, Gray RJ. A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc. 1999;94:496–509. [Google Scholar]
- 31.Noble M, McLennan D, Wilkinson K, Whitworth A, Barnes H, Dibben C. Department for Communities and Local Government; London: 2008. The English indices of deprivation 2007. [Google Scholar]
- 32.Alfirevic Z, Stampalija T, Dowswell T. Fetal and umbilical Doppler ultrasound in high-risk pregnancies. Cochrane Database Syst Rev. 2017;6 doi: 10.1002/14651858.CD007529.pub4. CD007529. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Ness RB, Roberts JM. Heterogeneous causes constituting the single syndrome of preeclampsia: a hypothesis and its implications. Am J Obstet Gynecol. 1996;175:1365–1370. doi: 10.1016/s0002-9378(96)70056-x. [DOI] [PubMed] [Google Scholar]
- 34.Kingdom JC, Audette MC, Hobson SR, Windrim RC, Morgen E. A placenta clinic approach to the diagnosis and management of fetal growth restriction. Am J Obstet Gynecol. 2018;218:S803–S817. doi: 10.1016/j.ajog.2017.11.575. [DOI] [PubMed] [Google Scholar]
- 35.Cleaton MA, Dent CL, Howard M. Fetus-derived DLK1 is required for maternal metabolic adaptations to pregnancy and is associated with fetal growth restriction. Nat Genet. 2016;48:1473–1480. doi: 10.1038/ng.3699. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Usher-Smith JA, Sharp SJ, Griffin SJ. The spectrum effect in tests for risk prediction, screening, and diagnosis. BMJ. 2016;353:i3139. doi: 10.1136/bmj.i3139. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Papageorghiou AT, Ohuma EO, Altman DG. International standards for fetal growth based on serial ultrasound measurements: the Fetal Growth Longitudinal Study of the INTERGROWTH-21st Project. Lancet. 2014;384:869–879. doi: 10.1016/S0140-6736(14)61490-2. [DOI] [PubMed] [Google Scholar]
- 38.Meeme A, Buga GA, Mammen M, Namugowa AV. Angiogenic imbalance as a contributor to the pathophysiology of preeclampsia among black African women. J Matern-Fetal Neonatal Med. 2017;30:1335–1341. doi: 10.1080/14767058.2016.1212832. [DOI] [PubMed] [Google Scholar]
- 39.Korzeniewski SJ, Romero R, Chaiworapongsa T. Maternal plasma angiogenic index-1 (placental growth factor/soluble vascular endothelial growth factor receptor-1) is a biomarker for the burden of placental lesions consistent with uteroplacental underperfusion: a longitudinal case-cohort study. Am J Obstet Gynecol. 2016;214:629.e1–629.e17. doi: 10.1016/j.ajog.2015.11.015. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
