Reducing stillbirths: screening and monitoring during pregnancy and labour

Background Screening and monitoring in pregnancy are strategies used by healthcare providers to identify high-risk pregnancies so that they can provide more targeted and appropriate treatment and follow-up care, and to monitor fetal well-being in both low- and high-risk pregnancies. The use of many of these techniques is controversial and their ability to detect fetal compromise often unknown. Theoretically, appropriate management of maternal and fetal risk factors and complications that are detected in pregnancy and labour could prevent a large proportion of the world's 3.2 million estimated annual stillbirths, as well as minimise maternal and neonatal morbidity and mortality. Methods The fourth in a series of papers assessing the evidence base for prevention of stillbirths, this paper reviews available published evidence for the impact of 14 screening and monitoring interventions in pregnancy on stillbirth, including identification and management of high-risk pregnancies, advanced monitoring techniques, and monitoring of labour. Using broad and specific strategies to search PubMed and the Cochrane Library, we identified 221 relevant reviews and studies testing screening and monitoring interventions during the antenatal and intrapartum periods and reporting stillbirth or perinatal mortality as an outcome. Results We found a dearth of rigorous evidence of direct impact of any of these screening procedures and interventions on stillbirth incidence. Observational studies testing some interventions, including fetal movement monitoring and Doppler monitoring, showed some evidence of impact on stillbirths in selected high-risk populations, but require larger rigourous trials to confirm impact. Other interventions, such as amniotic fluid assessment for oligohydramnios, appear predictive of stillbirth risk, but studies are lacking which assess the impact on perinatal mortality of subsequent intervention based on test findings. Few rigorous studies of cardiotocography have reported stillbirth outcomes, but steep declines in stillbirth rates have been observed in high-income settings such as the U.S., where cardiotocography is used in conjunction with Caesarean section for fetal distress. Conclusion There are numerous research gaps and large, adequately controlled trials are still needed for most of the interventions we considered. The impact of monitoring interventions on stillbirth relies on use of effective and timely intervention should problems be detected. Numerous studies indicated that positive tests were associated with increased perinatal mortality, but while some tests had good sensitivity in detecting distress, false-positive rates were high for most tests, and questions remain about optimal timing, frequency, and implications of testing. Few studies included assessments of impact of subsequent intervention needed before recommending particular monitoring strategies as a means to decrease stillbirth incidence. In high-income countries such as the US, observational evidence suggests that widespread use of cardiotocography with Caesarean section for fetal distress has led to significant declines in stillbirth rates. Efforts to increase availability of Caesarean section in low-/middle-income countries should be coupled with intrapartum monitoring technologies where resources and provider skills permit.


Introduction
Although most pregnancies progress normally, some are more complex because of antenatal or intrapartum conditions that place the mother, the developing fetus, or both at a higher risk for complications than pregnancies without these conditions. Pre-existing chronic conditions, as well as conditions that arise during pregnancy, can threaten the life and health of the fetus or the mother. Maternal hypertension, diabetes mellitus, renal disease, and autoimmune disorders, as well as placentation abnormalities and congenital anomalies, are examples of conditions that can place the pregnancy at high risk of fetal compromise. Fetal growth restriction arising from placental insufficiency is a significant cause of perinatal mortality (stillbirth or neonatal death) and morbidity (complications of prematurity) internationally [1]. Additionally, if not detected and addressed promptly, fetal hypoxia resulting from placental dysfunction or poor fetal tolerance of labour can cause stillbirth, neonatal death, or physical and developmental disabilities in the child [2].
Relatively non-invasive techniques exist to screen for a number of these conditions during the antenatal and intrapartum periods. These screening tools can also be used to monitor fetal well-being via assessment of fetal movement, heart rate, and/or growth; and feto-placental and/or uteroplacental circulatory dynamics, whether routinely at antenatal care (ANC) visits or via more complex screening tests in high-risk and post-term pregnancies [3]. Despite widespread clinical use of many of these techniques, the sensitivity and predictive value of these tests and methods are often too poor to reliably detect problems. Prompt detection of risk factors and complications is also critical, as measures of fetal distress or compromise associated with certain high-risk conditions may rapidly lead to fetal demise. Certain maternal or fetal problems may prompt the need for pharmacological intervention, early delivery, or surgical delivery (Caesarean section) rather than vaginal delivery. Optimising gestational age at delivery and judicious timing of corticosteroid administration are key challenges in responding to fetal compromise arising pre-term. The appropriate use of accurate screening and monitoring technologies can facilitate timely referral to facilities capable of providing operative delivery or other interventions for complications prior to or during labour. On the other hand, screening and monitoring techniques during pregnancy and the intrapartum period could inadvertently result in avoidable perinatal deaths, either because the technique itself is harmful or because it increases the risk of inappropriate or unnecessary use of drugs, induction of labour, early delivery, or Caesarean section.
Most studies of fetal screening and monitoring to date have been conducted in high-resource settings. Theoretically, evidence-based screening and monitoring techniques that are already in widespread use in high-income countries could be promoted to prevent stillbirth and other adverse pregnancy outcomes in low-/middleincome countries. We focus here on monitoring methods during pregnancy and the intrapartum periods, including identification and care of high-risk pregnancies and advanced monitoring techniques, with attention given where relevant to the feasibility and potential impact of implementing these techniques in low-resource settings where most stillbirths occur.

Methods
This is the fourth in a series of papers on the evidence for interventions that impact stillbirths. Details of the search strategy and review procedures for this paper are described in detail in Paper 1 of this series [4]. Each study was assigned a level of evidence (LOE) based on its design strength, size, and findings. The cumulative strength of the body of evidence for each intervention was then graded as A, B, C, or D using the SIGN grading system; impact estimates for each intervention were further cumulatively assessed as having no/negative, uncertain, some or clear evidence of benefit.
We reviewed 14 screening and monitoring interventions for evidence of impact which are included in this paper (Table 1). For most of these interventions, we first reviewed studies reporting how effectively a given screen-ing or monitoring test detected potential risk to the fetus (primarily observational studies), followed by studies that assessed the utility and/or impact of screening or monitoring interventions in preventing adverse outcomes, for which randomised controlled trials (RCTs) were most informative.

Pregnancy risk screening Background
Early identification of high-risk pregnancies can theoretically facilitate monitoring, referral and prompt initiation of therapy. Multiple screening and scoring systems have been developed to assess obstetric risk generally [5,6], as well as the risk of preterm labour, Caesarean delivery, and other maternal and fetal outcomes. Risk scores using these systems can range from simple additive scores to the products of more complex multivariable models that quantify risk factors according to their association with adverse outcomes [7]. An effective risk screening system, particularly if convenient to implement and relatively nondependent on diagnostic technologies, would be particularly useful in low-resource settings to help providers identify high-risk pregnancies and refer them for appropriate facility-based care, and to help facilities allocate scarce resources.

Literature-based evidence
Ten observational studies met our inclusion criteria; none tested interventions for pregnancies scored as high-risk (Table 2). Most risk scoring systems were originally devel-oped and tested in high-resource settings. At a national hospital in New Zealand, Pattison et al. [8] developed and tested an antepartum risk scoring system (N = 29,101 consecutive pregnancies) using prior obstetric history and current pregnancy risk factors, where a fetal risk score ≥ 3 denoted high risk. One-third of the total population (N = 10,859) was scored as high-risk, and 90% of those who had a perinatal death were identified using the scoring system. Women with an antepartum risk score of 7 or more (very high risk) had a perinatal mortality rate of 200/ 1000, whereas the low risk group of 18,242 (63%) had a perinatal mortality rate of 4.1/1000. The system clearly identified the population at risk of fetal or early neonatal loss, but could not effectively predict the need for intervention, as 60% of the low-risk group had a complicated pregnancy requiring intervention [LOE: 2-]. The same research group later used this dataset to develop a statistically derived antenatal risk scoring system using data on 27 antenatal variables from 20,985 pregnancies [9]. Tested on 3120 subsequent pregnancies, the scoring system had a positive predictive value of 0.73 in early pregnancy and 0.91 at onset of labour. Although only 16% of pregnancies were classified as high-risk at onset of labour, 87% of adverse outcomes occurred within this group. The positive predictive value of this system was higher than any previously reported statistically derived score, but requires that clinicians be able to sum logistic coefficients (basic statistical analysis), which requires more training than some other systems [LOE: 2-].
In the UK, an effort by Chard et al. [10] to calculate obstetric risk scores from individual risk factors (N = 2029 preg-

Identification and care of high-risk pregnancies
Pregnancy risk screening Fetal movement counting Routine ultrasound scanning Doppler velocimetry Pelvimetry Detection and management of maternal diabetes mellitus

Advanced monitoring in pregnancy
Antenatal fetal heart rate monitoring using cardiotocography Fetal biophysical profile test scoring Vibroacoustic stimulation Amniotic fluid volume assessment Home versus hospital-based bed rest and monitoring in high-risk pregnancy In-hospital fetal surveillance unit

Monitoring during the intrapartum period
Use of the partograph Cardiotocography with or without pulse oximetry  Abraham et al. 1991 [18] India. Health centre setting. Prospective cohort study. Health workers at 6 primary health centres used a home-based mothers card with pregnant, mostly illiterate women (N = 2446).
Assessed the association of perinatal mortality with risk factors recorded on a home-based mother's card to pregnant women on which risk factors and ANC attendance were documented.
Weighted risk factor method clearly superior to unweighted risk factor method in primiparae.
No difference in multiparae.
Assessed the utility of Edwards' scoring system adapted to a Korean setting in identifying high-risk pregnancy. Risk scoring included demographic, obstetric, medical, and miscellaneous factors.
560 infants (42.7%) were born to mothers with risk-scores greater than 7, and 753 infants (57.3%) were born to mothers with risk-scores less than 7.
Tested the predictive value of Coopland's obstetric risk in anticipating adverse outcome (perinatal death, birthweight < 2500 g, 5-min Apgar score < 7, or newborn transferred to a level 2 or level 3 nursery.
There was a clear relationship between risk score and probability of adverse outcome. Good sensitivity could be achieved only at the expense of a very high false-positive rate, however. Risk scoring no more effective than a policy that would refer all women with standard obstetric risk factors; majority of adverse outcomes occurred in women identified as lowrisk.
nant women) found that risk scores were useful only for identifying the small group of women at particularly high risk of adverse fetal outcomes. For most women, risk scores were uninformative [LOE: 2-].
In the USA, Morrison et al. [11] found that perinatal mortality was significantly higher in the high-versus the lowrisk groups identified with the application of a simplified risk scoring system, where high risk was a score of 3 or greater (69/1000 versus 7/1000, respectively, P < 0.0001). Seventy percent of perinatal deaths occurred in the highrisk group, which was 19% of the total group screened [LOE: 2-].
Other studies implemented risk scoring systems in more remote or low-resource settings in low-/middle-income countries. Attempting to predict intrapartum complications in rural Zimbabwe where most women receive care from nurse-midwives, Majoko et al. [12] employed antenatal risk assessment at the first antenatal visit based on medical and demographic measures and obstetric history (N = 5223 women at 12 health centres). All high-risk women (N = 2890) were encouraged to seek hospital delivery. Of the 924 (17.7%) women who experienced complications, 577 (62.4%) had had risk markers identified at booking; however, only 20% (577/2890) classified as high risk developed intrapartum complications. This risk screening system had a likelihood ratio of 1.16, indicating it was ineffective in identifying women at risk of pregnancy complications and generated too large a risk group for referral [LOE: 2-].
In India, Talsania et al. [13,14] reported on the application of an antenatal risk scoring system (N = 687 women), and observed that no stillbirths occurred among women with no identified risk factors, whereas among women in the highest risk group, 20% had stillbirths. Perinatal mortality was 84 In a rural primary care setting in the USA, Lefevre et al. [15] found a clear relationship between risk score and probability of adverse outcome, but cautioned that good sensitivity could be achieved only with a very high falsepositive rate, as the majority of adverse outcomes occurred in women identified as low-risk [LOE: 2-] In a remote area of Australia, Humphrey et al. [16] employed pregnancy risk scoring (N = 2875 women with singleton births), and found that during the study period, hospital and regional perinatal mortality rates fell by more than half. Women with low-risk scores had a statistically significantly lower incidence of preterm birth, leading the authors to conclude that risk scoring can be of benefit in allocating scarce resources [LOE: 2-].
Several other retrospective analyses attempted to associate perinatal mortality with the presence of specific risk factors. In Guadeloupe, West Indies, using data from the 1984-85 Guadeloupean Perinatal Audit, de Caunes et al. [17] observed that perinatal mortality was associated with a specific combination of risk factors representing maternal demographic, socioeconomic, obstetric history and risk characteristics measurable at the first antenatal visit, leading the authors to advocate for risk assessments specific to pregnancy outcomes within specific populations [LOE: 3-].
Using risk screening as a strategy to facilitate monitoring and referral, Abraham et al. [18] adapted the Home Based Mothers Card recommended by the World Health Organ- Talsania et al. 1994 [14] India (Ahmedabad). N = 687 indigent women enrolled during first trimester. Women scored as no, mild, moderate, or severe risk based on sociodemographic and obstetric data.
Assessed association of risk factors and risk scoring with perinatal mortality.

Conclusion
A number of studies reviewed were able to successfully identify women at high risk of obstetric complications. However, despite good sensitivity, risk scoring systems [7] have poor positive predictive value in anticipating adverse birth outcomes, particularly when used well before term or in populations significantly different from the population in which the system was developed [19]. This limitation of risk scoring systems limits the impact of their use.
No studies were found that effectively incorporated risk screening with appropriate interventions to demonstrate a possible impact on stillbirth or perinatal mortality rates compared to a control group. The evidence for risk screening at the community level yielded a Grade C assessment.

Fetal movement counting Background
Monitoring fetal movements using counting strategies is an indirect measure of central nervous system integrity and fetal responsiveness. Commonly employed in clinical practice, fetal movement counting is a simple and inexpensive means of monitoring fetal well-being [20]. The rationale for fetal movement counting is that decreased fetal movements signal decreased oxygenation, which often precedes fetal demise [1]. Kick charts or other recording strategies involve a pregnant woman in the second half of pregnancy monitoring fetal movements, documenting the frequency of movements she feels, and reporting these counts to her physician. Changes in these counts, particularly decreases, indicate possible fetal compromise, and thus alert care providers to the need for further diagnostic tests such as non-stress testing or the biophysical profile. Cessation of movement can indicate impending fetal death, while gradual diminishment of activity can indicate chronic fetal compromise [21]. Fetal movement monitoring may be used routinely, or only in high-risk pregnancies. There are many different counting methods, and fetal movement monitoring has a wide following among clinicians, who perceive the practice to serve as an early warning system for fetal compromise. A criticism of the practice is that it may cause undue worry for the pregnant woman [22].

Literature-based evidence
The literature search identified one Cochrane review comprised of three RCTs; and seven observational and intervention studies (Table 3).
Several trials in high-risk pregnancies documented an association between poor fetal movements and stillbirth/ perinatal mortality rates. A trial among high-risk women (N = 110) by Lema [23] found that fetal movement was a predictor of stillbirth rate, documenting differential rates [12/ 2-]. Two other observational studies found no difference in perinatal mortality between groups with good versus poor fetal movements measured using kick charts [25,26].
Other observational studies, RCTs, and reviews assessed impact on perinatal mortality of interventions using kick charts. A before-after study by Moore et al [27] introduced formal fetal movement monitoring into clinical practice, and stillbirth rates declined from 8.7 to 2.1/1000 over the course of the study (χ 2 = 6.8; P < 0.01). A Cochrane review by Mangesi et al [22] included 3 trials that tested strategies of routine kick counting, but varied study designs precluded outcome pooling (Additional file 1). Unfortunately, no included trials compared fetal movement counting with no fetal movement counting, and all studies showed nonsignificant impact on perinatal outcomes, including stillbirth incidence. [LOE: 1++].
The largest RCT testing kick charts [28] assessed the impact of the use of kick charts on unexplained stillbirth in normally-formed singleton pregnancies (N = 68654), and found no difference in rates of fetal death between intervention and control groups ( [29][LOE: 1+]. However, of the 17 women in the study randomised to kick charts who alerted their provider about decreased fetal movement and subsequently delivered a stillborn baby, none had an emergency delivery, as follow-up testing using cardiotocography resulted in false negatives for all 17 women [30].

Conclusion
The existence of a Cochrane review of multiple RCTs [22] yields a Grade B evidence rating. In keeping with the NICE guidelines based largely on 1 study [28], evidence from these studies indicates a lack of impact of fetal movement monitoring on stillbirth or perinatal mortality. Despite indirect evidence that formal movement monitoring using a counting method is more effective than mothers' subjective assessments of fetal movement in identifying babies at risk of intrauterine death, false negatives on subsequent fetal assessment tests and clinical error may be responsible for the lack of impact on perinatal mortality, as suggested by Del Mar et al [30]. Monitoring does appear to be of some value in high-risk pregnancies [23,24], particularly those in which there is suspicion of placental insufficiency. Routine fetal movement monitoring is currently recommended only for high-risk pregnancies, particularly those in which there is clinical suspicion of restricted fetal growth or placental dysfunction revealed through ultrasonographic or Doppler studies. Further studies are warranted to determine which methods of fetal movement counting prove most effective in identifying complications (sensitivity and specificity) early enough for interventions to prevent stillbirth, as well as acceptability to and feasibility for women. Universal fetal movement monitoring for all pregnancies is unsupported by scientific evidence.

Routine use of ultrasound scanning Background
Ultrasound scans during pregnancy are widely used, even in many resource-poor settings, but availability and quality of ultrasound machines vary, and ultrasound operators in some settings may lack the ability to accurately interpret imaging. Diagnostic ultrasound examination may be employed to date pregnancies, identify multiple pregnancies, document placental location, identify fetal anomalies (particularly when the fetus is suspected to be at high risk of malformation), identify fetal growth restriction or abnormal amniotic fluid volume, or to investigate clinical complications (e.g., bleeding). Some clinicians have postulated that routine use of ultrasound in all pregnancies could identify problems in asymptomatic pregnancies, whether early or late in gestation [31].

Literature-based evidence
We identified two Cochrane reviews and five other intervention/observational studies; we also conducted an independent meta-analysis incorporating 9 RCTs (Table  4).
One observational study of routine ultrasound use suggested that ultrasound may help to identify some highrisk pregnancies. In Egypt, Mahran et al [32] reported that routine ultrasound was superior to fundal palpation in identifying fetal growth restriction (89.7% versus 34.7% of growth-restricted infants identified accurately with each method, respectively). No observational studies reported any statistically significant impact of routine ultrasound scanning on subsequent stillbirth rates. A Cochrane review on the impact of ultrasound during pregnancy is in progress [33].
Another Cochrane review by Neilson [31] reviewed adequately controlled trials of routine ultrasound imaging in early pregnancy (N = 9) (Additional file 2). The study found that routine ultrasound examination was associated with earlier detection of multiple pregnancies and reduced rates of induction of labour for post-term pregnancy, but ultrasound had no impact on PMR [Odds ratio (OR) = 0. 86 [36] where placental grading information from ultrasound screening was either given to a clinician (intervention) or withheld (controls), the antepartum SBR (excluding lethal malformations) was 0/1014 versus 9/1011 among intervention versus controls, respectively (P < 0.05) [LOE: 2+].

New meta-analysis
We also conducted an independent, new meta-analysis for the purposes of this review, as we identified 9 RCTs (N = 35,049 women) reporting an impact on perinatal mortality rate of ultrasound in early pregnancy versus no or selective use of ultrasound in early pregnancy (before 24 weeks) ( Figure 1). We found no significant difference between the 2 groups when the results were pooled (OR = 0.89, 95% CI: 0.70-1.14).

Conclusion
There is no clear evidence that ultrasound examination during pregnancy is harmful [31], and its assumed benefits include (1) better gestational age assessment; (2) earlier detection of multiple pregnancies; (3) determination of placental location to rule out placenta praevia; (4) detection of clinically unsuspected fetal malformation when termination of pregnancy is more feasible, and (5) monitoring of fetal growth for pregnancies at increased risk of fetal growth restriction or macrosomia. The reduced incidence of induction of labour for apparent post-term pregnancy in the routinely scanned groups presumably results from better gestational age dating, and twin pregnancies being detected earlier. Neither of these effects has been shown to improve fetal outcome, but much larger numbers of participants would be required to accurately measure this outcome.
Based on the results of 2 Cochrane reviews, our meta-analysis, and other RCTs (overall Grade B evidence), there is no evidence that routine ultrasonography has any impact on perinatal mortality compared to the selective use of ultrasonography based on clinician judgement. It may be that routine ultrasound cannot reliably detect complications, or that high rates of false positives expose higher numbers of babies to iatrogenic intervention (particularly the risk of iatrogenic preterm birth in the event of inaccurate gestational age dating). The routine use of early ultrasonography in pregnancy cannot be recommended to prevent stillbirth, as there is no evidence of its benefit in preventing stillbirth or perinatal mortality. There is a need for decision analysis studies subsequent to diagnostic ultrasound, like the study of indicated Caesarean section for fetal macrosomia diagnosed by ultrasound by Rouse et al [37], but which report perinatal mortality outcomes. Clinics and hospitals, particularly those in resource-constrained settings, must assess whether the potential benefits are worth the cost of routine ultrasound screening for all pregnant women.

Doppler velocimetry Background
In many high-income countries, Doppler ultrasound studies are used as a non-invasive means to assess the suf-ficiency of uterine and umbilical cord blood flow. These velocimetry studies can improve management of pregnancies by aiding identification of fetuses at highest risk of adverse outcomes associated with pre-eclampsia, fetal growth restriction, and congenital malformations. Management of pre-term pregnancies with signs of fetal growth restriction and pre-eclampsia is complex, especially before 32 weeks gestation. The risk of prolonged hypoxia and acidaemia leading to stillbirth or neonatal death if the pregnancy is allowed to progress must be balanced against the risks of neonatal morbidity and mortality associated with prematurity if early delivery is chosen. Doppler ultrasonographic evaluation may aid determinations of the degree to which the fetus may be or become compromised.
Early in normal pregnancy, trophoblasts invade the maternal uterine spiral arteries and reduce resistance to uterine blood flow. Impeded flow measured by uterine artery Doppler suggests a failure of this trophoblastic invasion, which is associated with subsequent preeclampsia, fetal growth restriction, and stillbirth [38]. Uterine artery Doppler studies could therefore be helpful in identifying women likely to develop pre-eclampsia or to have a growth-restricted fetus [39].
Observational Doppler studies of the umbilical artery, first conducted in the 1970s, consistently showed a correlation between extremely abnormal waveforms and adverse outcomes, including fetal growth restriction and stillbirth [40][41][42]. In growth-restricted fetuses, results of umbilical and fetal Doppler waveform analyses suggest progressive severity of fetal compromise [38,43]. Initially, Forest plot of results of meta-analysis of perinatal mortality rates in women examined by routine vs. selective ultrasound in early pregnancy Figure 1 Forest plot of results of meta-analysis of perinatal mortality rates in women examined by routine vs. selective ultrasound in early pregnancy. umbilical artery velocity waveforms show increased resistance; subsequent deterioration is indicated by absent or even reversed end diastolic flow in the umbilical artery. Later, fetal middle cerebral artery flow shows decreased resistance, indicating brain sparing, and eventually, abnormal venous Doppler results (ductus venosus waveforms and umbilical vein pulsatility) suggest fetal cardiac dysfunction. Consequent central nervous system damage then manifests as non-reactive results to fetal tests of wellbeing, but there is wide variability in the timeline of fetal progression to severe compromise [38,43].

Study or Subgroup
Despite these well-established markers of fetal compromise in Doppler testing, it is not clear whether abnormal results of different modalities of Doppler ultrasound lead to improved perinatal outcomes and prevention of stillbirths, nor are the most appropriate indications and tim-ing of testing known. Additionally, if a fetus is not seriously compromised, Doppler ultrasound may potentially cause iatrogenic harm in suggesting the need for inappropriate early delivery.

Literature-based evidence
The literature search identified three systematic reviews, one Cochrane protocol, and 10 other intervention/observational studies (Table 5 and Table 6).

Uterine artery Doppler waveform analysis
Several studies assessed whether uterine artery Doppler velocimetry in unselected populations could identify high-risk pregnancies, particularly those at risk of stillbirth. In a systematic review, Papageorghiou et al [44] reviewed 15 studies of routine Doppler assessments in pregnancy in unselected populations, and found that  increased impedance to uterine artery flow was associated with increased risk of pre-eclampsia, fetal growth restriction and perinatal death, and that Doppler diagnosis of impeded artery flow appropriately identified ~40% of women who subsequently developed pre-eclampsia (6fold increased risk with positive Doppler) and ~20% of fetal growth restriction cases An intervention RCT attempting to use Doppler in interventions to reduce rates of pre-eclampsia in France and Belgium by the Essai Régional Aspirine Mère-Enfant (ERASME) Collaborative Group [46] randomised women to either uterine Doppler (N = 1253) or placebo (N = 617). Women with abnormal waveforms in the intervention group were given low-dose aspirin to prevent preeclampsia; but despite the aspirin prescription, rates of   [51].
Umbilical artery Doppler alone may not be any more effective than routine ultrasonography in some diagnostic assessments of fetal growth. An RCT of Doppler in twin pregnancies in New Zealand, Australia and Southeast Asia where all pregnancies were studied ultrasonographically to monitor fetal growth showed no statistically significant difference in perinatal mortality between groups on which umbilical artery Doppler was performed versus those with no Doppler [52].

Conclusion
Doppler ultrasound is a relatively new technique that has been applied to the study of fetal, placental and uterine circulatory dynamics. Despite its novelty, it has been evaluated by more RCTs than has any other biophysical test of fetal growth or well-being. In low-risk or unselected populations, there is little evidence that any form of routine Doppler velocimetry contributes to reductions in stillbirth rates (overall Grade C evidence). This lack of impact may be complex: Doppler ultrasound may not identify a sufficient proportion of flow abnormalities to measurably impact stillbirth incidence; Doppler-detected abnormalities may not be subsequently monitored appropriately with other tests of fetal well-being and serial Doppler testing; intervention based on abnormal Doppler may not work; or high rates of false-positives may unnecessarily expose the fetus to the risk of preterm birth, particularly if gestational age dating is inaccurate. Additionally, most existing studies are underpowered to detect small impacts on perinatal or maternal outcomes.
Uterine artery Doppler waveform analysis accurately identifies compromised fetuses at risk of stillbirth, especially in cases of placental underperfusion associated with preeclampsia and/or growth restriction, but no studies have shown any ability of subsequent intervention to prevent stillbirth. More studies are needed into the optimal timing of monitoring and intervention in cases of abnormal uterine artery waveforms.
Of all Doppler diagnostic studies of the fetus, umbilical artery and ductus venosus Doppler velocimetry are most predictive of fetal compromise associated with fetal growth restriction, and there is some evidence that timely and appropriate intervention for abnormal umbilical artery or ductus venosus waveforms can prevent stillbirths [53]. More evidence should soon be available: in addition to the Cochrane review of umbilical and fetal Doppler velocimetry, results are forthcoming of the multi-centre Trial of Umbilical and Fetal Flow in Europe (TRUFFLE) group, an RCT of timing of delivery in early pre-term fetal growth restriction based on early and late fetal Doppler venous changes versus cardiotocography. Such studies may shed light on the most appropriate and effective methods of fetal surveillance and optimal uses of Doppler velocimetry, including multi-vessel analysis [54].
Further studies are needed to assess whether such decision analytical models based on Doppler and other fetal surveillance findings for fetal growth restriction, pre-eclampsia, or congenital abnormalities could have any impact on stillbirths. Further research might also investigate other applications of Doppler velocimetry, including identify-ing women who should be given other screening tests, the comparative efficacy of Doppler versus other fetal surveillance methods, and contributing to the study of the pathophysiology of impaired placentation, uteroplacental and fetoplacental haemodynamics, and pre-eclampsia.

Pelvimetry
Background Pelvimetry, or pelvic measurement in pregnant women with the intention of predicting likely cephalopelvic disproproportion of cephalic presentations (and thus their subsequent need for Caesarean section), can be performed by clinical manual examination, or with imaging techniques including conventional x-ray, computerised tomography scanning, or magnetic resonance imaging. Successful detection and management of cephalopelvic disproportion is thought to reduce the risk of obstructed labour and intrapartum stillbirth.

Literature-based evidence
We identified 1 Cochrane review and 1 other observational study of pelvimetry reporting perinatal outcomes ( Table 7). The observational study by Fine et al [55] retrospectively analysed studies of x-ray pelvimetry (N = 100 trials) in cephalic presentations, comparing 3 prognostic indicators for vaginal delivery: the Thoms method, the modified Ball technique, and manual assessment. Neither pelvimetric method was significantly more accurate than manual assessment of the pelvis in predicting obstetric outcome, nor was any one method superior to the other.
When implemented as part of a strategy of pregnancy management, pelvimetry appears to have no impact on stillbirth. A Cochrane review by Pattinson et al. [56] assessed RCTs implementing x-ray pelvimetry in cephalic

Conclusion
There is little support for the use of pelvimetry to predict the need for Caesarean section in women with fetuses with cephalic presentations, as the dynamic and individual nature of maternal tissue changes during labour and fetal head moulding in the birth canal make antenatal pelvimetry a poor predictor of cephalopelvic disproportion. The practice may result in inadvertent harm to the mother by significantly increasing her risk of having a Caesarean section, without increasing the benefit to fetus or neonate, as pelvimetry shows no impact on stillbirth incidence. However, deficiencies in the existing studies included in Pattinson's meta-analysis should be noted: Parsons [57] attributed the increased perinatal mortality and birth asphyxia reported by Crichton [58] to lack of availability of electronic fetal monitoring, not to cephalopelvic disproportion. Even more problematic, the 2 deaths documented by Richards [59] occurred in utero prior to labour, which imply that cephalopelvic disproportion was not implicated. Treatment allocation strategies in all trials were of poor quality.
Given the deficiencies in existing studies, it remains plausible that other forms of imaging could effectively diagnose true cephalopelvic disproportion and avert stillbirth via Caesarean section in these cases. We classified the overall evidence as Grade C and found no evidence in support of using pelvimetry for preventing stillbirths.

Detection and management of maternal diabetes mellitus Background
In pregnant women, pre-existing diabetes mellitus can cause severe complications for both mother and child during pregnancy and delivery, including congenital malformations, hypertension, pre-eclampsia, macrosomia and intrauterine fetal death [60][61][62][63]. Macrosomia (large size for gestational age) increases the risk to the fetus of birth trauma, including shoulder dystocia, bone fractures and brachial plexus injury, in addition to obstructed labour. Good metabolic control in the mother from prior to conception through the postpartum period reduces the risk of these complications [64][65][66]. Because pre-gestational diabetes is a known risk factor for stillbirth, women with this condition are usually offered intensive surveillance and management during pregnancy, which may include glycaemic control efforts through diet, exercise, and/or insulin therapy with glucose monitoring, frequent fetal surveillance using tests of fetal well-being, and/or induction at or before term. Despite widespread practice of this protocol, a recent UK study demonstrated a 4-fold higher rate of stillbirth among women with pre-gestational diabetes compared to non-diabetic women, with 83% of stillbirths unrelated to congenital malformations [60].
There is also substantial confusion surrounding optimal screening for and management of glucose intolerance and gestational diabetes. Gestational diabetes and impaired glucose tolerance are relatively common and, like pre-gestational diabetes, have been linked with adverse perinatal outcomes including stillbirth and shoulder dystocia as a consequence of fetal macrosomia [67,63,68]. However, how to best screen for gestational diabetes is controversial. The American Diabetes Association, the American College of Obstetricians and Gynecologists, and the World Health Organization recommend universal screening for gestational diabetes based on the conclusion that selective screening is inadequate [69][70][71][72], while the US Preventive Services Task Force concluded that there is insufficient evidence of benefit to justify universal screening for gestational diabetes. As with management of pregestational diabetes, intensive management of gestational diabetes includes glucose monitoring, dietary regulation, and tests of fetal well-being. The need for insulin therapy is usually less for gestational diabetes than with pre-gestational diabetes.

Literature-based evidence
Our literature search identified 4 reviews and 12 other intervention and observational studies reporting perinatal outcomes after management of women with any form of diabetes mellitus (Table 8).

Intensive management
Tuffnell et al. [73] undertook a systematic review of RCTs (N = 3) of strategies for intensive management of women with gestational diabetes and/or impaired glucose tolerance in pregnancy (N = 223), including obstetric monitoring, dietary regulation, and insulin therapy in some cases. A number of studies compared different strategies of management of diabetic pregnant women, including glycaemic monitoring and control, diet and exercise regimens, and insulin treatments. In a prospective population-based study of intensive management of women with diabetes To evaluate the impact on perinatal mortality of a focused preconceptional and early pregnancy programme for Type 1 diabetes including strict glucose control and antepartum fetal surveillance from 32 weeks gestation until delivery.
PMR: 3% vs. 2% vs. 0% before, during, and after the programme, respectively.  In the US, McElvy et al. [78] conducted an evaluation of a programme that worked to achieve strict glycaemic control before and during early pregnancy in women with pre-gestational diabetes (N = 306), with intensive antepartum surveillance from 32 weeks gestation until delivery. Perinatal mortality declined steadily from 3% before the programme to a rate comparable to general population levels after completion of the programme [LOE: 1-]. In Soweto, South Africa, Huddle et al [79] reported that a programme involving gestational diabetic women given special services during pregnancy observed lower rates of perinatal mortality among women participating in the service compared to non-users of the service (3.7% versus 15.6%, respectively), but no statistical data was furnished [LOE : 3].
A descriptive study by Landon et al. [80] of non-stress testing (with delivery for non-reactive tests) in women with pre-gestational diabetes found no differences in glycaemic parameters in women delivered for suspected fetal compromise compared to a non-intervention group, suggest-ing the possibility that glycaemic control was unrelated to fetal compromise [LOE : 3].
Several studies assessed the impact on perinatal outcomes of mode of insulin treatment. Mukhopadhyay et al. [81] conducted a meta-analysis of RCTs (6 trials) comparing the impact of continuous subcutaneous insulin infusion versus multiple-dose insulin on glyacemic control and pregnancy outcome in diabetic women. Neither glycaemic control nor pregnancy outcomes were different between treatment groups [LOE: 1++] (Additional file 9). In Israel, Hod et al. [82] conducted a small RCT (N = 322 women) comparing the impact on fetal and perinatal outcomes of insulin aspart, a fast-acting insulin analogue, versus human insulin therapy in women with pre-gestational diabetes. A slight trend toward lower perinatal mortality was observed in the insulin aspart group compared to the human insulin group (14/1000 versus 22/1000, respectively), but no statistical significance data was reported [LOE: 1-]. A study of insulin therapy using either pump infusion or split-dose therapy found no difference in rates of perinatal mortality in the treated groups compared to non-diabetic controls [83].

Conclusion
While data from RCTs is scarce, the limited evidence, largely from clinical management of diabetic patients (Grade C evidence), suggests that pregnant women who do not have vascular complications and with good glycaemic control do not as a group have an increased risk of stillbirth above the general population [85]. In settings with poor access to care and high prevalence of gestational diabetes, perinatal mortality has been shown to be reduced with diabetes screening and treatment [79].
Although retrospective studies reveal declines in the incidence of stillbirth among diabetic women in low-prevalence populations where diabetes care is available [86], few prospective trials have been able to show any impact on stillbirths or perinatal mortality, largely owing to the high sample sizes required for such studies [63]. Adequate glycaemic control and monitoring during pregnancy are a reasonable means for reducing stillbirths that has been shown to reduce many complications that may be related to stillbirth, including congenital anomalies and macrosomia. Where feasible, particularly in high-prevalence settings, large studies with rigorous designs may help confirm this recommendation.

Advanced monitoring in pregnancy
Antenatal fetal heart rate monitoring using cardiotocography Background Antenatal fetal surveillance studies are widely used to assess fetal well-being and to identify the compromised fetus. Fetal heart rate patterns are a principal component of the majority of tests of fetal well-being, and cardiotocography, where it is available, is the method frequently used to electronically record the fetal heart rate. In highincome countries, beginning in the third trimester, cardiotocographic monitoring (electronic fetal monitoring) has largely replaced intermittent auscultation (periodic listening to the fetal heart rate using a stethoscope or handheld Doppler device) for monitoring fetal heart rate. During the antenatal period, external cardiotocography is used, which employs a Doppler ultrasound transducer; the mother's uterine contractions are also monitored using a pressure transducer, with both transducers strapped to the mother's abdomen while recording is in process.
Antenatal fetal heart monitoring includes both non-stress tests (NST) and contraction stress tests (CST). Fetal heart rate patterns are classified as either reassuring (reactive), non-reassuring (non-reactive) or abnormal, considering heart rate baseline, variability, and decelerations. In highincome countries, particularly the US, the NST is widely accepted as a primary fetal surveillance tool in high-risk pregnancies. It may be used in combination with ultrasound testing to observe fetal movement and amniotic fluid indices. When an NST is non-reactive, its poor predictive value (< 50%) warrants confirmatory testing with CST or a complete BPP [87]. During a CST, uterine contractions are induced (usually with intravenous oxytocin, but contractions can be generated using nipple stimulation) and subsequent fetal heart accelerations and decelerations are monitored to detect possible cases of uteroplacental insufficiency. It is important to note that antenatal fetal heart monitoring is gestational-age dependent; at 28 weeks' gestation, only about 60% of normally formed fetuses will have reactive non-stress testing due to immaturity of the fetal autonomic nervous system rather than placental insufficiency.
Variation between practitioners in interpretation of cardiotocographic tracings can potentially lead to inappropriate intervention, or false reassurance without appropriate intervention. Tracings are notoriously difficult to interpret, as fetal sleep patterns, labour progress, external stimuli, and opiate administration to the mother can alter fetal heart rate changes in the absence of fetal hypoxia or distress [2]. Additionally, as indications of fetal distress often indicate early delivery, the gestational age threshold at which antepartum testing is initiated will depend on an institution's ability to care for very preterm neonates, which, in general, is earlier in high-income settings than in lower-resource settings [21].

Literature-based evidence
We identified 1 Cochrane review and 4 observational studies testing the impact of antepartum cardiotocography (including both NST and CST) on perinatal outcomes (

Conclusion
There are relatively few RCTs of antepartum fetal heart rate surveillance (Grade C evidence), though observational studies clearly indicate a correlation between non-reassuring cardiotocographic traces and adverse perinatal outcomes, including stillbirth. Although there is a Cochrane review on antepartum cardiotocography [91] showing a trend toward increased risk of perinatal mortality, the study was underpowered to detect such an impact, and the causes of the deaths observed appeared unlikely to have been preventable. While RCTs are lacking to evaluate an impact of NSTs in reducing stillbirth, apparent reductions in stillbirth rates have followed the incorporation of the NST into protocols for management of high-risk pregnancy in the United States [21]. As with fetal movement monitoring and umbilical Doppler velocimetry, the rate of false-positives with a non-reactive NST is high enough that additional testing, whether CST or BPP, is advised to guard against unnecessary intervention.
When used experimentally as a primary fetal surveillance tool, CST use was associated with lower rates of antepartum stillbirth [92], suggesting that it could be useful for primary fetal surveillance testing in pregnancies at highest risk of stillbirth, including those of women with hypertension, diabetes, and intrauterine growth restriction. As it typically uses intravenous oxytocin to stimulate uterine contractions, the CST is more technologically intensive than the NST and can induce labour; thus, it is contraindicated for patients with placenta previa, placental abruption, previous classical Caesarean section, and premature rupture of membranes [87]. The CST should be conducted cautiously in cases of multiple pregnancy, incompetent cervix, and polyhydramnios.

Fetal biophysical profile (BPP) test scoring for assessing high-risk pregnancy Background
The BPP was conceptually derived from the Apgar score that is used to rate the condition of the newborn [93]. Using NST monitoring with cardiotocography and fetal ultrasound, the BPP collects 5 indicators of fetal wellbeing: fetal heart rate reactivity, breathing movements, gross body movements, muscular tone and qualitative amniotic fluid volume. Indications of fetal compromise often suggest the need for early delivery by induction of labour or Caesarean section.

Literature-based evidence
We identified one Cochrane review and four other observational studies of BPP reporting perinatal mortality outcomes (

Conclusion
Compared with conventional fetal monitoring, which is based primarily on cardiotocography/NST, BPP appears to offer no improvement in pregnancy outcomes (Grade C evidence). Although rates of false negative test results of the BPP are low, rates of false positives are high. The total number of subjects included in the meta-analysis remains small (N = 2974), and some of the studies carry a high risk of allocation bias. Although there is no consensus on opti-mal protocol for fetal monitoring, studies demonstrate that it is reasonable to implement the modified BPP (which consists of an NST and amniotic fluid volume measurement only), with full BPP or other tests such as the CST reserved for abnormal test results [98]. This passive and rapid approach to monitoring has been shown to be more cost-effective without diminishing its predictive value [99]. For reasons of cost and simplicity, the modified BPP may also prove to be more implementable than the full BPP in low-/middle-income country contexts with ultrasound and cardiotocographic capacity [100,101].

Vibroacoustic stimulation Background
Adjunctive vibroacoustic stimulation with the NST is used less often than CST or BPP after non-reassuring NST results. Vibroacoustic stimulation provokes fetal response using a sound-emitting device placed on the maternal abdomen near the fetal head, along with simultaneous cardiotocography to document fetal movements and heart traces. Theoretically, the resultant startle reflex in the fetus and subsequent fetal heart rate acceleration or transient tachycardia following vibroacoustic stimulation provide reassurance of fetal well-being and minimise unnecessary intervention for nonreactive tests [102].

Literature-based evidence
Our literature search identified one Cochrane review on vibroacoustic stimulation and four other observational/ interventional studies (Table 11). In a Cochrane review of vibroacoustic stimulation in conjunction with tests of fetal well-being (9 RCTs, N = 4838 women), Tan et al. [102] found that fetal vibroacoustic stimulation reduced the incidence of non-reactive antenatal cardiotocography tests (

New meta-analysis
We conducted a meta-analysis using available RCTs comparing outcomes (intrauterine or perinatal deaths) of pregnancies monitored by vibroacoustic stimulation compared to mock or no stimulation (4 trials, N = 1935 women in the vibroacoustic group, N = 2052 women in the mock or no stimulation group) (Figures 2 and 3

Conclusion
The benefits of using fetal vibroacoustic stimulation adjunctively with NSTs must be weighed with respect to the effect on the predictive reliability of the tests and the safety of the procedure. Vibroacoustic stimulation appears to offer maternal, fetal, and clinical benefits by decreasing the incidence of non-reactive cardiotocography and reducing the testing time. Although there are several RCTs and a Cochrane review evaluating vibroacoustic stimulation during pregnancy (Grade B evidence), the non-significant impact on perinatal mortality does not support including this intervention presently. Further RCTs are needed to determine optimal intensity, frequency, duration and position of vibroacoustic stimulation, as well as efficacy, predictive reliability, safety (in terms of fetal hearing impairment and neurological development) and perinatal outcome.

Amniotic fluid volume assessment Background
Amniotic fluid protects and supports the fetus during pregnancy. Both low amniotic fluid (oligohydramnios) and high amniotic fluid (polyhydramnios) are abnormal and potentially place the fetus at risk of adverse outcomes. Oligohydramnios can be associated with maternal conditions including pre-eclampsia, or with placental membrane rupture, fetal growth restriction, post-term pregnancy, fetal kidney problems, or fetal or placental abnormalities [105,106]. Polyhydramnios is frequently associated with maternal diabetes mellitus, maternal cardiac problems, twin-twin transfusion syndrome, or fetal or placental malformations, but in 50-60% of cases, it is idiopathic [107][108][109][110]. Until ultrasound became available, the invasive nature of amniotic fluid testing limited its clinical usefulness [111]. Using ultrasound, multiple methods for measuring amniotic fluid have been developed and are used to identify at-risk pregnancies. The amniotic fluid index (AFI) is a measure of the amount of amniotic fluid which is calculated by summing centimeters of depth of 4 different pockets of fluid; alternatively, the single deepest vertical pocket or maximum pool depth may be used. At term, many clinicians will induce labour or perform Caesarean section after diagnosis of decreased amniotic fluid volume to prevent an adverse pregnancy outcome. However, using assessments of amniotic fluid volume to predict fetal complications is controversial, and the utility of amniotic fluid assessment is different when used prior to versus after the onset of labour. Here, we examine the potential for assessment of amniotic fluid volume, or interventions to achieve normal amniotic fluid volume, to effectively detect high-risk pregnancy or fetal distress and subsequent adverse perinatal outcomes. We also examined the impact of interventions to achieve normal amniotic fluid volume on perinatal outcomes

Literature-based evidence
Our literature search identified two Cochrane reviews, one other review, and 15 other interventional or observational trials of amniotic fluid assessment to identify polyhydramnios or oligohydramnios that reported perinatal outcomes (Table 12 and Table 13).

Polyhydramnios
Several observational studies explored the association of polyhydramnios with perinatal mortality, with inconsistent findings. Several observational studies [112] observed that polyhydroamnios was an independent risk factor for both perinatal mortality and intrapartum morbidity  logistic regression analysis of SGA and appropriate-weight fetuses with and without polyhydramnios determined that the combination of SGA and polyhydramnios was an independent risk factor for perinatal mortality compared to normal weight-for-age infants with normal AFI values (OR = 20.55; CI: 12.6-33.4) [114]. The association may not apply to certain subsets of high-risk pregnancies, as a similar study among women with gestational diabetes who had AFI performed antenatally found no elevated risk of perinatal morbidity or mortality among pregnancies complicated by polyhydramnios compared to gestational diabetics without polyhydramnios [LOE: 2-]. No intervention studies subsequent to diagnosis of polyhydramnios were identified.

Oligohydramnios
Observational studies of oligohydramnios cases demonstrated a consistent elevated risk of poor perinatal outcomes. Anandakumar et al. [115] studied how AFI used with non-stress cardiotocography and fetal acoustic stimulation for non-reactive NSTs might predict adverse pregnancy outcome in high-risk pregnancies (N = 565 women). Of the 4 perinatal deaths in the group with low AFI, 3 had had a reactive NST within 7 days of fetal death [LOE: 2-]. In Italy, Locatelli et al. [116] reported rates of oligohydramnios twice as high in pregnancies with poor perinatal outcome than pregnancies with no adverse outcome [LOE: 2-]. However, oligohydramnios identified during the intrapartum period did not appear associated with adverse perinatal outcome in high-risk pregnancies: Baron et al [117] observed that women with low AFI were    One study found that in addition to the absolute volume of AFI, the distribution of amniotic fluid by intrauterine quadrants is predictive of fetal outcome. Myles et al. [118] observed higher rates of meconium staining, poor 1minute Apgar scores, fetal heart decelerations, and fetal acidaemia in pregnancies where fluid volume was higher in the upper quadrants than in pregnancies where fluid volume was higher in lower quadrants [LOE: 2+].

Meta-analysis (Forest plot, Fixed effects model) of impact of vibroacoustic stimulation versus mock or no stimulation on intra-uterine or perinatal deaths
A number of observational studies, RCTs and a systematic review compared AFI measurement to other methods of measuring amniotic fluid in ability to predict poor perinatal outcome. In an observational study in Egypt, Youssef et al. [119] (N = 174 women) found that the AFI was more sensitive than the single largest vertical pocket measurement in predicting perinatal mortality and multiple meas-ures of perinatal morbidity. Morris et al. [120] documented similar findings, but the sensitivity of AFI < 5 cm for major adverse fetal outcome was only 28 percent. A Cochrane review [121] compared the evidence for the predictive value of 2 methods of amniotic fluid assessment (AFI versus single deepest vertical pocket) on adverse pregnancy outcomes (Additional file 15). Compiling the evidence from available RCTs (4 trials, N = 3125 women), the authors found that AFI versus single deepest vertical pocket led to significantly more cases of oligohydramnios being diagnosed (RR [Random] = 2.33, 95% CI: 1.67-3.24), more frequent labour induction (RR [Fixed] = 2.10, 95% CI: 1.60-2.76) and higher rates of Caesarean delivery for fetal distress (RR [Fixed] = 1.45, 95% CI: 1.07-1.97). They recommended that the single deepest vertical pocket measurement be used because AFI resulted in increased diagnoses of oligohydramnios and rates of labour induction with no improvement in perinatal outcomes compared to single deepest vertical pocket, suggesting better predictive value of single deepest vertical Assessed the association of idiopathic polyhydramnios on perinatal outcomes.
PMR: 2-fold to 5-fold increase in risk in polyhydramnios vs. normal AFI groups, respectively.
Compared the effect in the group with increased amniotic fluid volume (exposed) vs. normal amniotic fluid volume (unexposed) by sonographic assessment. pocket. However, others have expressed concern that inter-observer reliability is lower with single deepest vertical pocket measurement (kappa = 0.33) compared to AFI (kappa = 0.72) [122]. An RCT by Alfirevic et al [123] randomised women to fetal monitoring by either AFI with computerised cardiotocography, or maximum pool depth determined by computerised cardiotocography, but found no statistically significant difference in perinatal outcome in the 2 groups. Another RCT by Oral et al [124] (N = 101 pregnancies) compared fetal monitoring (with cardiotocography) by either AFI or maximal vertical pocket. Measurement of maximal amniotic fluid vertical pocket proved slightly better than AFI in identifying postterm pregnancies with abnormal perinatal outcomes.
Only one intervention RCT compared AFI to no amniotic fluid assessment. Comparing AFI on admission (intervention) to no AFI (controls), Chauhan et al [125] found that women in the intervention group with a diagnosis of low AFI (measured as ≤ 5 cm or ≤ 5 th percentile) were no more likely to have Caesarean section for fetal distress, neonatal acidosis, or Apgar score < 7 at 5 minutes than untested controls or women with normal AFI results; no perinatal mortality statistics were reported [LOE: 1+].
One intervention strategy to improve perinatal outcomes is to augment amniotic fluid volume in cases of oligohydroamnios. Hofmeyr and Gülmezoglu [126] undertook a systematic review of RCTs (2 trials, N = 78 women) assessing the impact of maternal hydration status (requesting that women drink 2 litres of water prior to a repeat ultrasound) on AFI (Additional file 16). In women with and without oligohydramnios, drinking water was associated with an increase in amniotic volume (WMD for women with oligohydramnios = 2.01, 95% CI: 1.43-2.60; WMD for women with normal AFI = 4.5, 95% CI: 2.92-6.08). Intravenous hypotonic hydration in women with oligohydramnios was associated with an increase in amniotic fluid volume (WMD = 2.3, 95% CI: 1.36-3.24), but isotonic intravenous hydration had no measurable effect.

Conclusion
Polyhydramnios is a clear risk factor for perinatal mortality, whether associated with congenital malformations, placental insufficiency, or of idiopathic origin. However, no studies reported the impact of interventions subsequent to a diagnosis of polyhydramnios after amniotic fluid assessment, so the impact on stillbirth of amniotic fluid screening for polyhydramnios remains unclear.
Although robust RCTs are limited, very low AFI values (oligohydramnios) are frequently associated with poor pregnancy outcomes, and in these cases a reactive NST loses its usual reassuring value (Grade C evidence). Where feasible, amniotic fluid volume estimation may be helpful for identifying severe oligohydramnios, but further research is needed to document subsequent intervention and perinatal mortality outcomes to determine the costbenefit ratio of utilizing amniotic fluid assessment procedures. Amniotic fluid assessment is complicated by high variability of sequential measurements and use of different measurement methods, which can compromise the accuracy of the test. No particular method of amniotic fluid volume assessment appears superior to another [123,124], though using maximum vertical pool depth rather than AFI appears to limit unnecessary inductions of labour and Caesarean section by reducing diagnoses of oligohydramnios. There is a need for further research to test the impact of interventions to prevent or treat oligohydroamnios, particularly in the antepartum period with intact membranes, on perinatal outcomes.

Home versus hospital bed rest and monitoring for high-risk pregnancies Background
Women with high-risk pregnancies, especially multiple pregnancies and pregnancies complicated by hypertensive disorders, are frequently admitted to hospital for bed rest and monitoring. However, in some instances, more limited monitoring in facilities and home-based bed rest or reduced physical activity may be as effective as hospitalbased monitoring. Home-based activity modification, accompanied by outpatient surveillance and hospital admissions only for complications, would offer cost savings over hospital admission, reduced burden on hospital resources and personnel, and reduced disruption to the life of the mother posed by lengthy hospital stays.

Literature-based evidence
The literature search identified two Cochrane reviews and two other interventional studies comparing home-based versus hospital-based methods of care for high-risk pregnancies (Table 14). The first Cochrane review [127] assessed the effectiveness, in management of multiple gestations, of hospital-based bed rest versus home bed rest (reduced physical activity) with admission only for complications (6 RCTs, N>600 women, N>1400 infants) (Additional file 17). Hospital bed rest had no advantage compared to domiciliary bed rest in reducing perinatal mortality (OR = 0.89, 95% CI: 0.43-1.85), but birth weights were slightly higher in the hospitalised group. In the subset of trials of uncomplicated twin [128] and triplet [129]  Another Cochrane review [130] evaluated the impact of different degrees of bed rest compared with routine activity, as well as the effects of hospital-based versus home- PMR: Maximal amniotic fluid vertical pocket appeared to be slightly better than AFI for identifying the post-term pregnancy at risk for abnormal perinatal outcome.
To evaluate the role of the AFI, used along with NST and fetal acoustic stimulation test, when required, in prediction of adverse pregnancy outcome.
Compared rates of adverse fetal and neonatal outcomes in women diagnosed with oligohydramnios via AFI (cases) vs. women with normal AFI (controls). To compare predictive ability single ultrasound scan to detect a single deepest pool of AFI<2 cm (exposed) vs. AFI<5 cm (unexposed) in anticipating subsequent adverse pregnancy outcome.
PMR: 0 in both groups. An AFI <5 cm but not a single deepest pool <2 cm was significantly associated with birth asphyxia or meconium aspiration. Sensitivity of AFI < 5 cm for major adverse outcome: 28  To determine whether gestational diabetes (GDM) complicated with hydramnios is associated with higher rates of perinatal morbidity and mortality than those with normal amniotic fluid (AFI).
No significant differences in rates of antepartum fetal death.
Observational study. Fetuses (N = 174) within one wk of delivery.
The AFI was more sensitive in predicting mortality (87.5%) and the following measures of perinatal morbidity: low 5-minute Apgar score (88.8%), fetal distress during labour (86.6%), meconium-stained amniotic fluid (63.6%), and the presence of fetal growth restriction (79.4%).  In The Netherlands, an RCT [131] compared domiciliary care to hospital-based care in high-risk pregnant women (N = 150). There was 1 perinatal death in each group (1/ 77 versus 1/74 in home-versus hospital-based groups, respectively), and the only stillbirth occurred in the domicilary care group (1/77 versus 0/74 in home-versus hospital-based groups, respectively); the sample was too small to assess statistical significance [LOE: 1-]. A large RCT from China [132] (N = 13,006 women) compared a protocol of self-monitoring at home to a control group managed at hospitals. Both perinatal mortality and fetal deaths were lower among the self-monitoring group (peri-natal mortality: 6.30% versus 10.92% in self-monitoring versus control groups, respectively, P < 0.05; fetal deaths: 3.23% versus 6.34%, respectively, P < 0.05).

Conclusion
The limited evidence available (Grade D evidence) suggests that home-based care for certain subsets of women with uncomplicated high-risk pregnancy has no apparent disadvantage compared with hospital-based bed rest and monitoring in terms of impact on stillbirths and perinatal mortality. The number of studies on this subject is small, however, and limited to the conditions of multiple gestation and hypertensive disorders of pregnancy, exclusively in high-resource settings. As one RCT above indicated, bed rest in hospital for non-proteinuric hypertension may be superior to home-based care [133], but insufficient evidence exists to recommend hospitalization in this case, as the study was underpowered. The Cochrane review found that in multiple pregnancy, bed rest (in hospital) appears to confer little advantage over no modification of physical activity; bed rest had no impact on rates of perinatal mortality, stillbirth, or pre-term birth, but may improve fetal growth [127]. The potential cost savings and increased convenience to women of home-based bed rest and monitoring indicates that further large studies of the efficacy of bed rest as an intervention as well as monitoring at home versus in hospital are needed, particularly studies that include economic analyses of costs to mothers and hospitals.

In-hospital fetal surveillance units Background
A fetal surveillance unit provides a wide range of maternal and fetal diagnostic tests, often on an outpatient basis, of particular benefit in identifying and monitoring high-risk pregnancies. Antenatal fetal surveillance protocols, which may include multiple surveillance methods at a specified frequency, have the potential to impact perinatal mortality, neonatal morbidity, birth weight in cases of fetal growth problems, rates of prematurity, and length of hospital admission. Assessment of fetal condition can usually be performed on an outpatient basis, with admission reserved for delivery, which is usually less disruptive to the pregnant woman's family [134]. Specially designated antenatal surveillance units that offer predominantly outpatient services also offer operating cost savings compared with standard hospital admission.

Literature-based evidence
Our literature search identified 1 Cochrane review, 1 Cochrane protocol and 2 other intervention studies (Table 15). A Cochrane review [135] evaluated the use of antenatal outpatient day care units as an alternative to inpatient care for women with complicated pregnancy (1 RCT, N = 54 women) [136] (Additional file 19). Day care unit assessment for non-proteinuric hypertension reduced subsequent inpatient stay (difference in mean stay: 4.0 days; 95% CI: 2.1-5.9 days). The rate of induction of labour was much higher in the hospital-based inpatient care group compared to the day care unit group (OR = 4.9, 95% CI: 1.6-13.8). There were no perinatal deaths in either group [LOE: 1+]. The Cochrane protocol, for a review currently in progress, proposes to assess the impact of different specified regimens of fetal surveillance for impaired fetal growth on maternal and perinatal outcomes and length of hospital admission [137]. The 2 intervention studies adopted different approaches. Soothill et al [134] conducted a before-after intervention study comparing the number and length of antenatal admissions in the 5 months after the opening of a perinatal care unit providing largely outpatient-based services to records from the 6 months prior to the opening. They found no significant change in stillbirth rate (6/1294 versus 8/1372 after versus before, respectively; rate difference (RD) = 0.0012, 95% CI: -0.0043-0.0067) [LOE: 2+], suggesting that fetal surveillance on an outpatient basis was equally effective as hospital admission in managing highrisk pregnancy. Menzies et al. [138] performed a similar study of all women admitted with pre-eclampsia to a tertiary-level perinatal unit before and after introducing standardised assessment and surveillance, and reported that perinatal outcomes did not change [LOE: 2-].

Conclusion
If patients could be consistently monitored on an outpatient basis, obstetricians would be more willing to admit patients only for delivery, and to perform more fetal assessments and maternal tests as outpatient services. Outpatient fetal surveillance offers benefits in terms of cost and convenience, and could improve health facility organisation, and improved record-keeping would streamline obstetric and perinatal audit, data quality for research, and teaching opportunities [135]. However, the evidence base for the benefits of outpatient surveillance of high-risk pregnancies is relatively limited, as the single Cochrane review on the subject included 1 small RCT that reported no perinatal deaths (Grade C evidence). The reduced rates of induction of labour are of interest and require further trials reporting labour induction rates with sufficient power to detect differences in subsequent perinatal outcomes. At this time, even in high-resource settings, in-hospital fetal surveillance units cannot be recommended as an intervention to prevent stillbirths.

Monitoring in labour
Use of the partograph Background A partograph, alternatively called a partogram, is a simple pre-printed paper form on which midwives and obstetricians record labour observations. The tool provides a continuous pictorial overview of the progress of labour, while monitoring maternal and fetal well-being. The partograph distinguishes between the latent and active phases of labour. The active labour section has 2 straight lines called the alert and action lines. The alert line reflects a modification of the mean rate of cervical dilatation of the slowest 10% of primigravid women in the active phase of labour (1 cm per hour). Slower progress than this crosses the alert line on the partograph, which may prompt initiation of the process of transfer to a facility with emergency obstetric capacity in preparation for intervention for prolonged labour. Depending on the partograph version, the 'action line' is 2 to 4 hours to the right of the alert line. Labour crossing this line suggests primary inefficient uterine activity and prompts immediate appropriate management of slow progress of labour, usually via amniotomy, oxytocin infusion, or both.
Some evidence suggests that midwives and physicians find the partograph practical in terms of ease of use, time resourcefulness, continuity of care and educational assistance [139], which may contribute to positive maternal and fetal outcomes. Partographs are also inexpensive and relatively simple to use, making their use attractive in lowresource settings where other intrapartum monitoring technologies are unavailable or prohibitively expensive. In higher-resource settings, the partograph can be implemented alongside other fetal surveillance tests such as car-diotocography to provide more information for decisionmaking. However, some practitioners view the partograph as inappropriately restrictive and formulaic, prompting intervention prematurely [139], factors which could impact both clinical and maternal psychological outcomes. Various versions of the partograph are marked by different slope and position of the action line, which is likely to impact labour augmentation interventions, Caesarean section rates, and maternal satisfaction.

Literature-based evidence
We identified 1 Cochrane review including 5 RCTs and quasi-RCTs (N = 6963 women), and 3 intervention/observational studies (Table 16). The Cochrane review [140] compared the impact of use of the partograph in comparison with no partograph, as well as different versions of the partograph, for monitoring the progress of spontaneous labour at term (Additional file 20). None of the included studies reported intrapartum stillbirth rates, though several reported neonatal outcomes. The randomised controlled trials (RCTs) in the review that compared partograph use to no partograph use (2 trials; N = 1590 women) [141,142] showed no impact of partograph use on Caesarean section rate (risk ratio (RR) = 0.64, 95% confidence interval (CI): 0.24-1.70), instrumental vaginal delivery (RR = 1.00, 95% CI: 0.85-1.17) or Apgar score < 7 at 5 minutes (RR = 0.77, 95% CI: 0.29-2.06).
The Cochrane review [140] also compared the impact of different versions of the partograph with different action line placement on labour outcomes. Two RCTs, both conducted in high-resource settings, compared different partograph versions with the action line either 2 hours or 4 hours after the alert line (N = 3601 women) [143,144], and found no difference in Caesarean section rates between labours monitored with the 2 different partographs (RR = 1.06, 95% CI: 0.85-1.32 [NS]). Women in the 2-hour action line group were significantly more likely to receive oxytocin augmentation (RR = 1.14, 95% CI: 1.05-1.22). One RCT, also in a high-resource setting, compared a 2-hour versus a 3-hour action line (N = 617 women) [143]. There was no difference in Caesarean section (RR = 0.78, 95% CI: 0.51-1.18). Maternal satisfaction with care was higher in the 2-hour action line group than the 3-hour action line group (RR = 0.49, 95% CI: 0.27-0.90). There was no difference in neonatal outcomes. One RCT in a high-resource setting compared the impact of partographs with a 3-hour versus a 4-hour action line (N = 613) [143]. The Caesarean section rate was significantly lower in the 4-hour action line group (RR = 1.70, 95% CI: 1.07-2.70) but there was no difference in neonatal outcomes. One RCT in a low-resource setting compared a partograph with no action line (alert line only) to a partograph with an action line (N = 694 women) [145]. The Caesarean section rate was lower in the group with no action line (RR = 0.68, 95% CI: 0.50-0.93), but there was no difference in neonatal outcomes.
The review also pooled the results from 3 RCTs to compare shorter (2-hour action line or alert line only) versus longer time-to-intervention (4-hour action line or alert plus action lines) and found no differences between the groups for Caesarean section rate, Apgar score or instrumental delivery, but early intervention in the low-resource setting reduced the Caesarean section rate.
A large multicentre study [146] compared the impact of partograph use in multiple hospitals in Southeast Asia (N = 35,484 women). The reported stillbirth rate was 0.3% in the group for which the partograph was used, versus 0.5% in the control group; this small difference was not assessed for statistical significance because partograph use was introduced in stages. A sub-study using the same WHO dataset of breech birth management using the partograph [147] found a non-significant decrease in intrapartum stillbirth after the introduction of the partograph compared with before (1.1% versus 1.9% after versus before, respectively).
In Indonesia, Fahdhy et al [148] compared the use of midwives trained to employ the partograph versus standard midwifery care without the partograph. Seventy-one of 304 labours plotted on the partograph progressed beyond the alert line; the study reported significant decreases in obstructed labour, oxytocin use, and Apgar score less than 7 at 1 minute, but there were only non-significant reductions in fetal death and early neonatal death [adjusted OR for fetal death = 0.62, 95% CI: 0.17-2.19 (NS); adjusted OR for neonatal death = 0.70, 95% CI: 0.16-3.11(NS)].

Conclusion
Overall, there were no significant differences in maternal or perinatal outcomes with the use of partograph versus no partograph, and no evidence that any particular version of the partograph is better than another in preventing perinatal mortality. Partographs may be comparatively more effective in low-resource settings, as the studies from Africa and Mexico in the Lavender review [140], as well as data from Southeast Asia [146] that showed reduced Caesarean section rates with use of the partograph and early intervention for slow progress of labour. The data from Southeast Asia and Indonesia also showed trends toward improved birth outcomes [146]. Our overall assessment of the grade of evidence of studies of partograph is Grade C. Given the limitations of the studies included and the potential impact of organisational issues, e.g. guidelines on partograph use, a large cluster-RCT in low-resource settings is recommended to compare partograph versus no partograph use, specifically including stillbirths as a reported outcome.

Intrapartum cardiotocography with and without pulse oximetry Background
Methods to assess fetal heart rate and levels of oxygenation of fetal blood are monitoring strategies intended to identify early signs of fetal compromise as a result of oxygen shortage, or fetal hypoxia [2]. Severe and prolonged hypoxia is associated with stillbirth, early neonatal death, and long-term physical or mental disability, including cerebral palsy, if the baby survives. Fetal heart rate patterns are classified as either reassuring, nonreassuring or abnormal, considering heart rate baseline, variability, and decelerations. While nonreassuring tests may become reassuring with simple change in maternal position, abnormal fetal heart rate alterations or low blood oxygen levels are a frequent indication for Caesarean or instrumental delivery [149].
In many hospitals in high-income countries including the US and Canada, cardiotocography is widely used to monitor fetal heart rate in labour, though intermittent auscultation (periodic listening to the fetal heart rate using a stethoscope or handheld Doppler device) is occasionally performed. In addition to external cardiotocography, which can be employed continuously or intermittently during labour, internal cardiotocography can be performed by attaching a sensor to the fetal presenting part, usually its head. This can be done only if the membranes are ruptured, as electrode must be embedded in the baby's scalp. Intermittent auscultation via stethoscope or handheld Doppler is more common than electronic fetal monitoring methods in low-resource settings where fetal heart rate monitoring is available.
Currently, cardiotocography is recommended in high-risk pregnancy and labours induced or augmented with oxytocin [150]. Because of the poor ability of cardiotocography alone to detect true fetal distress, the practice of pulse oximetry is thought to provide additional helpful information to corroborate cardiotocographic traces. Pulse oximetry is intended as a follow-up procedure in the presence of a nonassuring cardiotocographic test, and is intended to improve the accuracy of the assessment of fetal wellbeing in the intrapartum period [151]. In this procedure, a sensor is attached via a clip to the fetus, generally on the scalp, cheek, temple, or back; oxygenation values exceeding 30% are considered reassuring even when a cardiographic trace is nonreassuring [152]. While non-reassuring tests may become reassuring with simple change in maternal position, abnormal fetal heart rate alterations are a frequent indication for Caesarean or instrumental delivery [149].
There is some danger that improper interpretation of cardiotocographic tracings can lead to inappropriate intervention, or false reassurance that delays necessary intervention. There is a need for sensitive and specific methods of using cardiotocography, potentially in conjunction with pulse oximetry, to improve detection of fetal compromise due to hypoxia. Improved detection of hypoxia, primarily by expediting delivery, could improve outcomes and prevent stillbirth in these instances.

Literature-based evidence
The literature review identified two Cochrane reviews on intrapartum cardiotocography, one Cochrane review examining pulse oximetry in conjunction with cardiotocography, and two other observational/interventional studies ( A third Cochrane review [149] assessed RCTs that comparied maternal and fetal outcomes after fetal pulse oximetry was used in labour (with or without concurrent use of cardiotocography or auscultation), compared with cardiotocography alone (Additional file 23). Adjunctive fetal pulse oximetry with cardiotocography was associated with significantly decreased rates of Caesarean section for nonreassuring fetal status compared to cardiotocography alone in groups without fetal blood sampling prior to study entry (RR = 0.68, 95% CI: 0.47-0.99), and in groups with fetal blood sampling prior to study entry (RR = 0.03, 95% CI: 0.00-0.44). Based on 2 trials reporting outcomes, there was no impact of adding fetal pulse oximetry to car-

Conclusion
Although there are several studies of fetal pulse oximetry and intrapartum cardiotocography, few reported stillbirth or perinatal mortality as outcomes. The available results show no statistically significant impact on stillbirths or perinatal mortality, whether cardiotocography is used alone or in conjunction with fetal pulse oximetry in labour. Continuous cardiotocography appears associated with increased rates of operative delivery and lower rates of neonatal seizures compared to no or intermittent cardiotocography, but has no demonstrated impact on rates of perinatal mortality acidosis, dystocia, or long-term physical or developmental outcomes. There is currently no evidence of benefit from randomised or quasi-randomised studies for intrapartum cardiotocography alone or in conjunction with electrocardiogram or fetal pulse oximetry in preventing stillbirth.
One drawback of intrapartum cardiotocography is that its poor predictive value of true fetal distress and hypoxia is not sufficiently enhanced by any adjunctive technologies at this time. Pulse oximetry appears safe, though longterm developmental studies have not been performed, but it does not have any clear impact on stillbirths or perinatal mortality [149]. Adjunctive electrocardiogram possibly reduces neonatal encephalopathy, academia, and instrumental delivery, but these findings need additional research to verify. Minimally invasive technologies and tests are needed in addition to intrapartum cardiotocography to more accurately identify intrapartum hypoxia and dystocia.
There are robust studies and Cochrane reviews on fetal pulse oximetry and intrapartum cardiotocography (Grade B evidence). The non-significant impact of cardiotocography with or without fetal pulse oximetry on stillbirth/perinatal mortality in these studies suggests no evidence of benefit. In high-income countries, documented stillbirth rates, particularly stillbirths associated with intrauterine asphyxia, in high-income countries have followed increasing prevalence of cardiotocographic fetal monitoring with Caesarean section for fetal distress [155][156][157], though this association may be confounded. Electronic fetal monitoring (with access to operative delivery) may an important tool in the arsenal of strategies to prevent stillbirth, though this is unproven. More precise tests for fetal distress and indications for Caesarean section are needed.

Summary
Although a reasonable number of Cochrane reviews and RCTs were available which assessed the impact of one or more implementation strategies for many screening and monitoring interventions, none of the interventions we reviewed demonstrated convincing evidence of impact on stillbirths or perinatal mortality. A number of studies that did report statistically significant impact on stillbirth or perinatal mortality, or which suggested large magnitude differences between intervention and control/comparison groups, lacked sufficient rigor to justify recommending the routine use of the interventions evaluated.
The evidence for all interventions reviewed in this paper is summarised in Table 18. Of the range of monitoring interventions evaluated, fetal movement counting and Doppler monitoring were promising for further evaluation in high-risk pregnancies in low-resource settings. Low amniotic fluid measurements were strongly predictive of stillbirth, but interventions to restore adequate amniotic fluid volume or to deliver the baby based on identification of oligohydramnios have not been systematically tested to conclude whether amniotic fluid assessment is a useful diagnostic tool that leads to actions which prevent stillbirth.

Research gaps
Despite the existence of many tools, devices, and techniques for monitoring pregnancy for complications, there is a dearth of rigorous evidence that any screening and monitoring intervention has a direct impact on stillbirth rates in unselected and low-risk populations. This stems from several common shortcomings in the design and interpretation of studies investigating the use of screening and monitoring techniques. First, the small size and insufficient rigor of many screening and monitoring studies to date renders them underpowered to detect significant differences in perinatal mortality subsequent to their implementation. Many studies assessing the impact of the interventions covered in this review did not report stillbirths or perinatal mortality at all. For virtually all the interventions included in this review (with the exception of x-ray pelvimetry), there is a need for RCTs sufficiently powered to detect statistically significant impact on stillbirth or perinatal mortality, if such differences exist. Second, for screening and monitoring interventions to reduce stillbirth incidence, they must effectively identify women at higher risk of stillbirth in time for an appropriate and effective intervention to be provided. Most studies we reviewed limited their analysis to the first step of successful screening or monitoring: detection of women at increased risk of stillbirth or perinatal death. A number of studies showed that positive test results corresponded fairly well with increased risk of adverse pregnancy outcome. However, a trial of an otherwise effective screening method will show no impact on stillbirths if the subsequent intervention fails to prevent stillbirth, either because intervention is ineffective or provided too late. In most of the studies we reviewed, few women who screened positive were treated according to an established protocol. This flaw precludes assessment of whether the lack of impact of any given screening technique is attributable to a failed screening method or a failed intervention. This fact highlights the need for time-to-decision and time-to-intervention studies, as well as analyses that consider which interventions are employed, and protocols for employing them, subsequent to adverse findings of screening and monitoring studies. The quality, appropriateness, and timeliness of the intervention care provided must also be considered along with the effectiveness of screening or monitoring in detecting true complications.
In addition to these tests of diagnostic accuracy and intervention effectiveness, other types of studies are also needed, including innovative pilot studies, safety studies, and effectiveness trials, to bolster the weak evidence base for screening and monitoring interventions during labour (Table 19). In high-income countries, there is a particular need for better understanding of placental pathophysiology, as up to half of unexplained stillbirths show signs of growth restriction, much of which is attributed to placental insufficiency [158]. Placental dysfunction is often observed in other major known causes of stillbirth including pre-eclampsia and abruption, suggesting the need for accurate placental function or placental biomarker screening tests, especially those appropriate for use in unselected populations. One such test in early pregnancy which measures pregnancy associated plasma protein-A has been associated with a 40-50-fold increase in risk of stillbirth due to placental dysfunction in a selected population, but it is unclear what interventions could prevent stillbirth among women with such positive test results [159]. Appropriate management of pregnancies complicated by placental dysfunction and/or fetal growth restriction requires further research, particularly when detected too early in pregnancy to consider early delivery.
Additional questions surround which screening and monitoring interventions could be most effective in low-/middle-income country settings where 98% or more of the global burden of stillbirths occurs [160]. Screening and monitoring are often technology-dependent, as illustrated by the cardiotocographic machine required for the BPP, CST, and NST; and the use of video ultrasound machines for amniotic fluid assessment. This makes some of these interventions impractical or unaffordable in low-resource or remote settings at the present time. However, certain technologies, such as handheld Doppler ultrasound and video ultrasound scanning machines, are increasingly available in low-resource settings; there is a need for studies to define the most cost-effective ways of using these and other available technologies to identify high-risk pregnancies. Additionally, other techniques, such as fetal movement monitoring and the partograph, require little investment other than training of midwives or mothers, and may be appropriate for use in high-risk pregnancies, if these pregnancies can be identified and if the use of these techniques result in improved outcomes, which has not yet been conclusively demonstrated. Additionally, there is some potentially promising evidence that fetal surveillance typically provided on an inpatient basis, often with bed rest, for high-risk pregnancies can be offered on an outpatient basis with bed rest or reduced activity at home without deleterious impact on pregnancy outcomes; this evidence requires confirmatory effectiveness trials in low-resource settings before programmatic adoption of this approach could be recommended.
Practicality, logistical feasibility, and cost-effectiveness are chief concerns in many low-resource areas. Where ANC services are available and widely utilised, the prospects for identifying and monitoring high-risk pregnancy improve. Along with interventions that improve detection of risk factors and complications, improved management and referral of women with complications are critical to manage obstetric risk in low-/middle-income countries. Unfortunately, in areas where ANC attendance is poor, few of these screening and monitoring techniques are likely to be implementable.

Implications for programmes and clinical practice
A few monitoring approaches showed promise for use in high-risk pregnancies, including fetal movement monitoring and umbilical Doppler velocimetry. Before recommending these interventions for widespread use, further large RCTs of sufficient rigour to detect differences in stillbirth and perinatal outcomes are needed to determine the utility and effectiveness of these monitoring approaches and interventions used after positive tests.
This review identified several interventions -for example, amniotic fluid measurement for signs of oligohydramnios, especially as pregnancy approaches or exceeds term; routine ultrasound scanning, and intrapartum cardiotocography -that are in widespread use in high-resource settings, but for which rigorous evidence is lacking. For these interventions, their continued use is reasonable even though the evidence base for impact on stillbirths is lacking. Intrapartum cardiotocography in particular has been shown in RCTs to elevate the risk of operative delivery. Prevailing wisdom credits cardiotocography and available Caesarean section with the diminution of stillbirth rates in high-income countries in recent decades; however, the role of cardiotocography in reducing stillbirth should be confirmed with more rigorous evidence. Efforts to achieve glycaemic control in cases of maternal diabetes mellitus are also encouraged, despite the lack of strong evidence showing an impact of such interventions on stillbirth, on the grounds that women with good glycaemic control experience fewer complications and negative outcomes of pregnancy.
Particularly in pregnancies deemed high-risk based on screening test results, clinical progression of pregnancy, or reproductive history, multiple antepartum testing modalities can minimise deficiencies in sensitivity and predictive value when used in combination to assess fetal wellbeing and abnormalities. The BPP reflects one standardised strategy by which tests of fetal well-being have been packaged to permit a more comprehensive assessment of fetal well-being. In cases where one test raises the possibility that a pregnancy is high-risk, multiple tests of fetal well-being should be considered if time allows, along with the patient's overall clinical history, to determine the appropriateness of intervention. Multiple tests are recommended in many protocols for identifying and making clinical decisions about management of pregnancies with suspected fetal growth restriction [161]. In these cases the collective results of a battery of tests including all the com- ponents of the BPP and Doppler velocimetry, as well as ultrasound biometry, can be informative in decision-making to maximise gestational age and inform the strategic timing of corticosteroid administration and early delivery via induction or Caesarean section. However, conflicting test results -a recent study showed only 44.5% concordance between BPP and Doppler assessments of compromise in growth-restricted fetuses -can complicate interpretation [162].
Recommendations for clinical decision-making protocols and introduction of interventions in low-/middle-income countries are more difficult. Resource constraints preclude the implementation of many of the interventions reviewed in this paper. The interventions requiring the lowest investment in resources and training -fetal movement monitoring and the partograph -show some evi-dence of benefit in high-risk pregnancies, but more studies are needed. Where safe Caesarean section is available, ensuring that intermittent or continuous intrapartum fetal heart rate monitoring is performed may improve targeting of Caesarean section while improving birth outcomes.
Screening and monitoring interventions cannot be effective without responsive, quality care and careful clinical management, particularly for high-risk pregnancies. Health systems quality improvement activities, including boosting provider skill in interpreting test results and determining appropriate intervention, timely response including swift referral, and evaluative strategies to document outcomes after intervention are essential complementary activities to ensure that an effective screening or

Pilot/cohort studies of interventions
• Alternative imaging or diagnostic technologies (alternatives to X-ray pelvimetry) to predict cephalopelvic disproportion in the antepartum period • Pathophysiology of impaired placentation and identification of clinical markers of poor placentation/perfusion to develop tests of stillbirth risk • Development of optimal methodologies for vibroacoustic stimulation (frequency, placement, amplitude, etc), and studies of efficacy and predictive reliability • Safety of vibroacoustic stimulation (auditory function, cognitive development) and pulse oximetry (cognitive development) • Interventions to prevent and treat oligohydramnios, particularly in cases of intact membranes • Non-interventional sensitivity, specificity, and predictive value testing of untested screening techniques in unselected populations (low-and high-risk pregnant women) • Development of predictive variables for stillbirth at term • Low-tech strategies, such as the partograph, for identifying high-risk pregnancies in low-resource settings • New adjunctive techniques to improve the positive predictive value of fetal distress and hypoxia of cardiotocography

Well-designed RCTs of interventions powered to detect stillbirth rates
• Community-based pregnancy risk screening schemes • Formal fetal movement monitoring in high-risk pregnancies ‫ؠ‬ Comparisons of different methods ‫ؠ‬ Impact of timing from monitoring-to-intervention on perinatal mortality • Optimal combinations of tests to screen for fetal growth restriction • Optimal management of fetal growth restriction and timing of delivery • Ultrasound assessment of placental appearance (lesions and calcifications) in high-risk pregnancy • Ability of uterine artery Doppler ultrasound in combination with other testing for pre-eclampsia prediction and subsequent development of prevention measures for women at highest risk • Optimising glycaemic control in managing diabetes mellitus in pregnancy • Assessment of stillbirth risk in instances of gestational diabetes and impaired glucose tolerance (little data compared to pre-existing diabetes mellitus) • Usefulness of BPP in identifying fetal compromise • Vibroacoustic stimulation studies in labour • Impact of in-hospital fetal surveillance units on stillbirth outcomes • Partograph versus no partograph use

Effectiveness and cost-effectiveness trials in large populations/at scale
• Cost-benefit analyses of routine ultrasound for gestational age dating and multiple pregnancy detection in resource-poor settings • Cost-effectiveness studies of fetal surveillance units in hospitals • Safety of out-of-hospital bed rest and outpatient fetal surveillance in high-risk pregnancies in resource-poor settings (including economic analyses) monitoring technique translates to measurable improvements in perinatal mortality outcomes.

Conclusion
Screening pregnancies to identify risk factors, complications, or indications of fetal distress and providing appropriate surveillance for identified risk factors theoretically encourages appropriate use of interventions, including induction of labour, Caesarean section, and pharmacological treatment, to prevent fetal loss and adverse maternal or neonatal outcomes. Screening and monitoring are futile, however, in the absence of effective interventions to act promptly and appropriately to manage identified risk factors and complications. The weak evidence in this review for impact of screening and monitoring interventions on stillbirth incidence is due at least in part to limitations in study designs that result in uninterpretable data. Screening tests may be ineffective, inappropriate, or occur too late to prevent adverse outcomes, but more problematically, most studies did not report the impact of intervention (particularly standardised intervention) following positive screening test results. Effective interventions exist to prevent stillbirth associated with many maternal infections or conditions (e.g., syphilis, malaria, pre-eclampsia) and placental dysfunction (e.g., oligohydramnios, placental insufficiency, abruption). Because screening and monitoring techniques during pregnancy also pose the risk of inappropriate or unnecessary use of drugs, induction of labour, iatrogenic preterm birth, or Caesarean section, it is practically, ethically, and economically important to validate the sensitivity, specificity, and positive predictive value of available screening and monitoring techniques to protect women and their babies from iatrogenic harm; and to ensure the wise use of scarce medical resources in low-and middle-income settings.