Reducing stillbirths: interventions during labour

Instrumental vaginal deliveries, which make up a subset of operative deliveries, are procedures involving traction applied to the fetal head, for indications including maternal exhaustion or other compromise (e.g., maternal heart disease), fetal distress or heart rate abnormalities (often associated with prolonged second stage of labour), and fetal malposition [6]. Rates of instrumental deliveries range from 5–20% in high-income countries [6]. Before an instrumental delivery should be attempted, the fetal head must be engaged and its position known, the cervix fully dilated, and the membranes ruptured. Traction is generally applied to the fetal head with either forceps or a vacuum extraction device (also called a ventouse). The use of these devices poses the potential for injury to the mother or the baby. Forceps have long been associated with birth trauma, particularly when used for rotational procedures, but low overall morbidity rates are reported for well-trained forceps practitioners [7]. While the association with birth injury is fairly well established, whether forceps use or vacuum extraction is preferable for preventing stillbirth or perinatal mortality is unclear.

Our systematic review included 2 Cochrane reviews and 5 other observational and intervention studies (Table 2). No studies were found that compared stillbirth outcomes with versus without instrumental vaginal delivery. One Cochrane review by Johanson et al. [7] compared the impact of vacuum extraction to forceps delivery, and included 7 trials reporting perinatal mortality outcomes (Additional file 1). Vacuum extraction compared to forceps delivery was associated with significantly less maternal trauma (OR = 0.41, 95% CI: 0.33–0.50) and less general and regional anaesthesia. Additionally, the risk of Caesarean section showed a trend towards being lower among the vacuum extractor group than the forceps group (OR = 0.56, 95% CI: 0.31–1.02 [NS]). Vacuum extraction was more likely to fail when used for assisted vaginal delivery than forceps (OR = 1.69, 95% CI: 1.31–2.19). Perinatal mortality rates were not statistically significantly different between the two instrumental methods (7 trials, N = 1800, OR = 0.80, 95% CI: 0.18–3.52 vacuum extraction vs. forceps, respectively). The vacuum extractor was associated with an increase in neonatal cephalhaemotomas and retinal haemorrhages, but serious neonatal injury was uncommon with either instrument and Apgar scores at 1 and 5 min were comparable [LOE: 1+]. The other Cochrane review, also by Johanson et al. [8], compared the impact of use of soft versus rigid vacuum extractor cups; only one trial (N = 72) in Malaysia reported perinatal death as an outcome but found no significant difference between soft and hard cups (OR = 1.26, 95% CI: 0.08–20.85 [NS]) [LOE: 1+](Additional file 2).

An RCT by Weerasekera et al. [9] not included in the above reviews compared the outcomes associated with vacuum and forceps deliveries (N = 442 women) in the second stage of labour. There were no significant differences between the two methods in the incidence of third-degree perineal tears, postpartum haemorrhage or ruptured uterus, but cervical tears were slightly higher in the forceps group. Cephalhaematoma incidence was higher among the vacuum extraction group, but there were no significant differences between the groups in babies needing resuscitation at birth, admission to neonatal intensive care unit, stillbirth or neonatal death rates [LOE: 2+]. Another RCT from Pakistan reported similar success rates with both instruments [10].

Multiple observational studies have evaluated the complications and outcomes related to vacuum deliveries either alone or in comparison with alternative methods of instrumental deliveries [11, 12]. In a prospective study of 167 vacuum extractions (6.3% of total deliveries) at the national hospital in Kenya, Gachiri et al. [13] documented perinatal morbidity and mortality rates of 16.2% and 4.8% respectively [LOE: 3]. Lurie et al [14] conducted a decision-to-delivery time analysis and found that it was faster to undertake forceps deliveries compared to vacuum extractions. While some studies suggested higher rates of vaginal tears with forceps [15], in general no differences in complication rates were found in other studies [16], and no studies found differences in stillbirth incidence.

We identified 7 trials comparing vacuum extraction versus forceps-assisted deliveries that reported stillbirth incidence (N = 632 vacuum, N = 611 forceps). Our meta-analysis found no evidence of differential impact of either method on stillbirths (OR = 0.60, 95% CI: 0.07–5.00) (Figure 1).

Although vacuum extraction was associated with a trend toward lower Caesarean section rates and fewer significant maternal injuries and less anaesthetic requirement than forceps delivery, there was no difference in rates of intrapartum stillbirth or perinatal mortality. Vacuum extraction tended to be associated with increased, albeit low, risk of neonatal cephalohaematoma and retinal haemorrhage. The lower rate of Caesarean section despite higher failure rate among vacuum extractions may be due to superiority of the vacuum for managing certain fetal malpositions (such as deflexed occipital posterior position, for example), or more likely, because following a failed vacuum extraction, delivery is usually by forceps, while failed forceps is usually followed by Caesarean section [7, 17]. The reduced maternal morbidity and limited, largely short-term risk of neonatal complications associated with vacuum extraction suggest that although there is no evidence that either method is superior to the other in preventing stillbirths, vacuum extraction may be preferable in areas where vacuum extractors are available and practitioners are suitably trained to perform vacuum extraction [17]. In other areas where forceps deliveries are the norm and health practitioners do not have training in vacuum extraction, the significant investment needed for the purchase of vacuum extractors and quality training to capacitate practitioners to perform the procedure may be prohibitive for certain low-resource settings. Relatively inexpensive manual vacuum extractors are available that may require less training to use than forceps and may be a useful alternative to forceps to facilitate rapid delivery in the presence of signs of fetal distress in the second stage of labour; however, these require further evaluation.

The vast majority of the world's one million intrapartum stillbirths each year occur in low-/middle-income countries marked by high rates of unmet obstetric need [18, 19], suggesting that many of these deaths could be prevented with improved obstetric care [20, 21]. The availability of comprehensive essential obstetric care (EOC), supported by emergency transport, skilled providers, and aftercare, is regarded as critical for effective maternal health services in obstetric emergencies [22]. Essential obstetric care (EOC) refers to elements of obstetric care needed for the management of normal and complicated pregnancy, delivery and the postpartum period. Basic EOC can be performed in primary health care facilities, and includes administration of anti-biotics, oxytocics, anti-convulsants including magnesium sulfate, manual removal of the placenta, treatment for incomplete miscarriage, post-abortion care, and instrumental vaginal delivery with forceps or vacuum extractor. Comprehensive EOC includes all basic EOC functions plus Caesarean section, anaesthesia, and blood transfusion, and generally requires a secondary or higher-level health facility. The subset of comprehensive EOC interventions used to respond to unexpected intrapartum complications such as haemorrhage and obstructed labour is referred to as emergency obstetric care (EmOC) [23], and includes some elements of basic EOC such as manual removal of the placenta and medical treatment in labour, as well as all anaesthesia, blood transfusion, and Caesarean section.

Non-availability of EmOC, especially Caesarean section, in resource-poor settings has been implicated as a risk factor for intrapartum stillbirths, particularly those associated with prolonged labour and its associated fetal asphyxia, infection, and birth trauma [24‐28]. Reductions in intrapartum stillbirth rates observed in the UK have been attributed to more liberal use of Caesarean section [20], though this is controversial. There is evidence to suggest that the optimal Caesarean section rate to reduce the number and proportion of intrapartum stillbirths lies between 5 and 10 percent [3]. Recent data from global estimates of Caesarean section availability at population level indicate that many countries fall far short of this range, while others exceed this, to the possible detriment of maternal health outcomes [29]. Some studies have shown that it can be comparatively safe to deliver in rural hospitals even if they lack Caesarean section capability [30].

Quality of EmOC is also critical. These issues have been documented in high-income countries, particularly among rural populations [26, 31, 32], but the data associating quality of EmOC with stillbirth or perinatal mortality rates has not been compiled.

Because the components of EmOC are by definition life-saving interventions, RCTs of EmOC versus no EmOC would be unethical, so the evidence base for improving access to and quality of EmOC, especially Caesarean section, consists largely of observational studies.

The literature on EmOC considered two primary issues: availability and optimal rates of Caesarean section (and associations with intrapartum stillbirth rates); and the impact of quality of obstetric care on perinatal outcomes. We identified 40 studies reporting stillbirth or perinatal mortality outcomes that reviewed or implemented interventions to provide EmOC (Table 3).

In low-/middle-income countries, prolonged and/or obstructed labour complicated by fetal asphyxia, fetal or placental infection, and birth trauma is a major cause of stillbirth, often arising from a small maternal pelvis due to childhood malnutrition, which subsequently leads to cephalopelvic disproportion during delivery [33]. Timely delivery, often by Caesarean section or instrumental vaginal delivery, can reduce associated intrapartum stillbirth, and is largely credited for the relatively low rates of intrapartum stillbirth in high-income countries [3].

Using regression analysis of data from WHO and other sources to examine the association between stillbirth rates and obstetric care, Goldenberg et al. [3] and McClure et al. [21] observed that Caesarean section availability in low-/middle-income countries was associated with reductions in intrapartum stillbirths (Figures 2, 3 and 4). They [3] reported that intrapartum stillbirths dropped by 1.61 per 1000 births for every one percentage point increase in Caesarean section from 0 to 8 percent. Thereafter, they observed a small, non-significant increase in intrapartum stillbirths for each percent increase in Caesarean section. Intrapartum stillbirth rates correlated more closely with the measures of obstetric care in the regression analysis than did antepartum stillbirth rates, corroborating the theory that obstetric care availability and quality improvements will impact intrapartum stillbirth incidence more than antepartum stillbirth incidence. There was no relationship between Caesarean section and intrapartum stillbirths in high-income countries, which all had rates of Caesarean section of 15 percent or more. The WHO recommends Caesarean section rates of 10 to 15 percent, although they also note that 'countries with some of the lowest perinatal mortality rates in the world have a Caesarean rate of less than 10 percent,' and that perinatal mortality declines are steep until Caesarean section rates reach approximately 8 percent of deliveries, after which point the relationship becomes less clear [34]

Using data from the early 1980s in Australia, Alessandri et al. [35] performed a case-control study of intrapartum stillbirths, and could not identify any specific antenatal risk factors that predicted these demises. However, they did suggest that because the odds of emergency Caesarean section were significantly higher among stillbirth cases than live born controls, prompt Caesarean section might reduce intrapartum stillbirths, the risk of which was strongly associated with placental abruption or other placental problems, fetal distress, umbilical cord prolapse, and vaginal breech delivery. Other studies have confirmed that in many high-income country settings, liberal policies of Caesarean section are associated with positive effects on fetal survival, even when performed well before term [36‐39].

While Caesarean section can be a life-saving intervention for both mother and child, its liberal use exposes some proportion of mothers and babies who do not need the procedure to unnecessary risks of morbidity. Additionally, in some low-/middle-income countries, particularly in low-resource areas, there is some evidence that the practice of Caesarean section may be associated with an increased risk of perinatal mortality [40, 41], suggesting that Caesarean section may be performed too late or by inadequately skilled practitioners in these settings. Additionally, the elevated risk of uterine rupture in subsequent pregnancies after a Caesarean section should be a consideration in providing the procedure in remote areas. In addition to the risk of uterine rupture, the data are conflicting about whether vaginal labour after previous Caesarean is associated with increased risk of perinatal mortality [42‐46]. Because of the known risk of uterine rupture, particularly among women whose prior Caesarean incision was classical rather than lower-segment, it is recommended that women who have had prior classical Caesarean section have immediate access to EmOC in subsequent pregnancies to reduce the risk of maternal mortality and stillbirth.

A number of studies associated suboptimal care, particularly inadequate, inappropriate, or delayed care of complications such as obvious fetal distress, placental abruption, breech presentation, twin pregnancy, or eclampsia, with increased perinatal mortality [47, 48]. Several studies comparing hospitals or assessing health systems found that quality of care provided in facilities was directly associated with perinatal mortality [49‐51, 25]. Recent improvements in term stillbirth rates, especially intrapartum stillbirths, suggest that more rapid performance of Caesarean section in cases of placental abruption has resulted in a lower incidence of abruption-associated stillbirth [52]. In Virginia, USA, Cruikshank and Linyear [53] observed that intrapartum stillbirths associated with abruption were characterised by a fairly long delay between the appearance of fetal compromise and delivery, suggesting that timely Caesarean section could have prevented these deaths. In Finland, Korhonen and Kariniemi [54] found that having a surgical team in-hospital to perform Caesarean sections versus having the team on-call resulted in lower perinatal mortality (P = 0.05) and prevented all cases of fetal death and hypoxic ischemic encephalopathy (versus 3/41 in controls and 1/41 in controls, respectively).

Jehan et al. [55] recently reviewed stillbirths in rural Sindh Province, Pakistan, and found that a large proportion of these stillbirths occurred in facility settings with ostensibly skilled care providers. Because the rate of fetal mortality in labour seems to be reduced with adequate care, Kiely et al [56] have proposed using fetal deaths in labour as an epidemiologic measure of the quality of obstetric care. They found that delivery in a level 1 (primary care) hospital was associated with a 60% increase in intrapartum stillbirth compared to a level 3 facility (equipped for perinatal intensive care). Similarly, a retrospective analysis [57] found that rural residence was associated with elevated stillbirth rates (OR = 1.20; 95% CI: 1.09–1.32, P < 0.001) and neonatal death in hospital (adjusted OR = 1.26; 95% CI: 1.07–1.48) compared to urban residence, which granted women better access to higher-level emergency obstetric care.

There is some limited evidence that Caesarean section capability is not the only component of EmOC with the potential to reduce perinatal mortality. Advances in labour management and perinatal care have also contributed to significant declines in perinatal mortality. Steyn et al [58] observed that over a 20-year period in South Africa, the Caesarean section rate remained constant, but perinatal mortality decreased by 50 percent. In a remote area of the US in a high-risk Native American obstetric population, Leeman et al [30] documented a Caesarean section rate one-third of the national average without any adverse impact on perinatal mortality rate (PMR), which was the same as the national average. Even in remote areas without Caesarean section capability, intrapartum stillbirths and perinatal deaths can be significantly reduced with improved quality of obstetric and perinatal care alone. Grzybowski [59] undertook a prospective study to determine whether a small, isolated hospital without Caesarean section capability and which handled fewer than 50 births annually could provide safe obstetric and perinatal care. Over the 5-year study period, there were 6 perinatal deaths, for a perinatal mortality rate of 20.8 per 1000 (95% CI: 4.4–37.2 per 1000); however, the wide confidence intervals indicate the study was underpowered to measure perinatal mortality, and the perinatal mortality rate was still more than double that of many high-income countries, including the US [60].

While varied in design and lacking the rigor of RCTs, evidence from available, largely observational studies, taken together, indicates that availability of facilities capable of providing EmOC with trained care providers who are able to undertake safe and timely Caesarean section for appropriate indications is critical for reducing the risk of intrapartum stillbirths. It is not possible, however, based on the available data to ascertain the relative contribution of various components of EmOC to the mortality reductions observed. The need for Caesarean section capacity in rural settings is apparent from the data. The risk of uterine rupture and placental invasion of the uterine scar, and the higher rate of stillbirth in subsequent pregnancies after Caesarean section [61] suggests that liberal Caesarean section policies are unjustifiable, especially where access to emergency obstetric care is limited or home birth is common. Medically unnecessary Caesarean sections would place a large group of women at risk of uterine rupture if they do not or cannot access EmOC in subsequent pregnancies. A Caesarean section rate of 10–15% as recommended by WHO appears appropriate, particularly in resource-constrained settings; it is paramount that the procedure is performed for indications of fetal distress or other standard indications, rather than for the convenience of the provider or on maternal request. Additionally, several studies reported significant improvement in perinatal mortality rates in facilities without Caesarean section capability, and others demonstrated that Caesarean section provided too late or in too remote a setting increased perinatal mortality, suggesting that quality of obstetric care, rather than mere availability of Caesarean section, is key in preventing stillbirths and perinatal deaths. Despite the obvious lack of RCTs, the evidence in support of increased emergency Caesarean section availability to reduce stillbirths is strong (Grade B); EmOC is needed in all low-/middle-income country health systems.

Induction of labour, a common intervention in obstetric practice, is indicated when it is determined that the fetus or mother will more likely have a healthy outcome than if birth is delayed. While many studies have compared different regimens and administration techniques of drugs to ripen the cervix and induce labour, few have evaluated the outcomes associated with induction versus expectant management of spontaneous labour for different indications [62]. Induction is frequently practised at term or post-term, or in cases in which the fetus is suspected to be macrosomic and therefore likely to require complicated operative delivery [63]. Additional common indications for induction of labour include pre-eclampsia, cases of premature rupture of membranes (PROM) where labour does not quickly become spontaneously established thereafter, and twin pregnancy. The process of induction of labour should only be considered when vaginal delivery is felt to be the appropriate route of delivery. Many different agents and methods can be used to induce and augment labour, including drugs such as oxytocin or prostaglandins, and physical interventions including sweeping the membranes or early amniotomy. Little is known about the impact of induction for different indications on stillbirth and perinatal outcomes.

Our literature search identified 5 Cochrane reviews and 3 other interventional/observational studies (Table 4).

We conducted a meta-analysis of trials (3 RCTs, N = 1770 women) of planned induction of labour using prostaglandins versus expectant management in women with PROM. No significant decrease in perinatal mortality was found (OR = 0.50, 95% CI: 0.05–5.53) in the planned induction versus expectantly managed groups, respectively (Figure 5).

The evidence from this review indicates that induction of labour does not impact the likelihood of delivery by Caesarean section. Two trials, however, showed a significant reduction in the risk of Caesarean section from a policy of routine induction of labour among low-risk women with post-term pregnancy [72, 73], though performance bias may have skewed the results in the Hannah trial.

Induction of labour for suspected fetal macrosomia in non-diabetic women has not been shown to alter the risk of maternal or neonatal morbidity, but the power of the included studies to show a difference in rare events is limited. Large trials are ongoing to address this question. Induction of labour appears to be an appropriate intervention in post-term pregnancy at 41 completed weeks of gestation or later. The Cochrane review by Gülmezoglu et al. [64] comparing elective induction of labour with expectant management showed a statistically significant reduction in perinatal mortality and meconium aspiration syndrome. Although increasingly more commonplace, elective inductions are not advised before 39 weeks gestation given the potential, albeit low, for complications associated with prematurity. Even in cases of post-term pregnancy, 500 inductions may be required to prevent one perinatal death. The reduction in stillbirth incidence noted in the Gülmezoglu review was not statistically significant, likely because the number of stillbirths was too small.

Larger RCTs are needed to address the impact of induction of labour on stillbirth and perinatal mortality for the following indications: suspected macrosomia in non-diabetic mothers, multiple pregnancy, and mild pre-eclampsia.

There are a variety of drugs, including oxytocin and a number of prostaglandins and prostaglandin analogues, and many different administration methods including intracervical, vaginal, oral, and intravenous (IV) routes, for the induction of labour. Drugs to induce labour can have adverse side effects, fail to induce labour, or cause dysfunctional labour or hyperstimulation of the uterus leading to fetal distress and Caesarean section. Fetal and possibly maternal death is possible if Caesarean section is not available or delayed when these drugs are used. The state of cervical ripening and favourability for induction should be assessed before a regimen is selected, as oxytocin induction, in particular, often fails unless the cervix is ripe. In women with unfavourable cervical ripening, different prostaglandin drugs, including prostaglandin F2-alpha, prostaglandin E2 (dinoprostone), and prostaglandin E1 (misoprostol) promote cervical ripening and initiation of labour [74]. Misoprostol is the only prostaglandin analogue that is effective in inducing labour without gastrointestinal side effects when given as an oral preparation. It is inexpensive and stable at room temperature, making it an easily administered intervention appealing for use in low-resource settings, though the risk of uterine hyperstimulation at high doses may produce increased risk of maternal or perinatal death [75]. Insertion of a Foley catheter has also been shown to be as effective as prostaglandin E2 for stimulating preinduction cervical ripening, potentially providing an effective, safe, non-pharmacological mechanical method of preinduction cervical ripening [76‐78].

Given the use of induction for certain indications including post-term pregnancy, pre-eclampsia, and PROM in which the fetus is at higher risk of perinatal death, as well as the potential for drugs used to induce labour to cause fetal distress, an assessment of the evidence of impact of specific drugs available for induction of labour on perinatal outcomes is warranted.

Our literature search identified 8 Cochrane reviews and 11 other studies (Table 5) assessing the impact of different drugs for cervical ripening and/or induction of labour. The trials identified are grouped below by the drugs being compared and the route of administration.

Both vaginal and oral misoprostol are more effective than placebo and comparable to intravenous oxytocin or vaginal prostaglandin E2 in inducing labour at term [75, 79]. Misoprostol at any dosage carries higher risk of meconium staining, but this was not associated with any adverse perinatal outcomes. Compared with vaginal prostaglandins, oral misprostol appears to reduce rates of Caesarean section [79] and vaginal misoprostol is associated with shorter labour, fewer side effects, and lower incidence of retained placenta, with no difference in perinatal mortality. However, there remain questions about the safety of both vaginal and oral misoprostol because of a relatively high rate of uterine hyperstimulation and the lack of appropriate dose ranging studies. While no increase in adverse fetal outcomes was reported in any of the misoprostol studies, the number of adverse fetal outcomes was too few to draw meaningful conclusions. There is no evidence to support the selective use of either oral or vaginal misoprostol on either perinatal mortality or stillbirths.

There was no statistically significant impact of any prostaglandin preparation on perinatal outcomes; it should be noted that most studies reviewed were small and the numbers of adverse outcomes too small to accurately measure differences in stillbirth or perinatal mortality rates between study arms. Vaginal prostaglandin E2, especially as vaginal tablets, appears to be an effective labour induction agent, improving the likelihood of vaginal delivery within 24 hours and reducing the need for augmentation with oxytocin, without increasing the risk of Caesarean section [90]. Vaginal prostaglandin E2 may reduce perinatal mortality, but the non-significant findings in the Cochrane review by Kelly et al indicate the need for further large RCTs. The intracervical mode of administration of prostaglandins is less used and less well evaluated than other modes of administration [62], and appears to be less effective in inducing labour than vaginal prostaglandin application, with no benefit for perinatal mortality. The data for perinatal mortality associated with extra-amniotic prostaglandins are extremely limited and require further trials [92].

Approximately 3–4% of term singleton pregnancies are complicated by breech presentation at term, and in most high-income countries, the majority of these births are handled via either planned or emergency Caesarean section. Whether planned caesarean section for breech presentation at term results in better perinatal outcomes is a contentious issue. An alternative to breech Caesarean section or breech vaginal delivery is external cephalic version, which has been shown to reduce numbers of breech births and breech Caesarean sections without impacting perinatal mortality [95, 96]. Until recently the major evidence for Caesarean section for breech presentation has been derived from patient case studies and register studies lacking sufficient rigor on which to make policy or clinical management recommendations. Here, we review the evidence for impact of planned Caesarean section for breech presentation on perinatal mortality and stillbirth rates.

Our literature search identified 1 Cochrane review and 3 other intervention/observational studies assessing the impact of planned Caesarean section for breech presentation on perinatal mortality outcomes (Table 6). The Cochrane review by Hofmeyr et al. [97] (Additional file 16) evaluated RCTs (3 trials, N = 2396 women), including the large Term Breech Trial [98], comparing planned Caesarean section with planned vaginal birth for singleton breech presentation at term. Perinatal or neonatal death (excluding fatal anomalies) or serious neonatal morbidity was reduced with planned Caesarean section compared to planned vaginal delivery [3/1166 (0.26%) vs. 14/1222 (1.15%), respectively; RR = 0.29, 95% CI: 0.10–0.86; 3 trials; N = 2388 women]. Compared with planned vaginal breech delivery, planned Caesarean section was associated with an increased risk of avoidable short-term maternal morbidity, including postpartum infection, haemorrhage, and anaemia (RR = 1.29, 95% CI: 1.03–1.61) [LOE: 1+].

In Switzerland, a cohort study by Irion et al. [99] of planned vaginal versus planned Caesarean section for breech singleton presentation at term (N = 705 women) found a non-significant increased risk of neonatal mortality among the group delivering vaginally, but all deaths were attributable to major malformations (1% vs. 0.3% in planned vaginal vs. Caesarean section groups, RR = 3.33, 95% CI: 0.37–29.60 [NS]) [LOE: 2-].

In The Netherlands, Molkenboer et al. [100] conducted a 2-centre retrospective matched cohort study including all singleton breech deliveries from 1998–2000 of at least 37 and less than 42 weeks' gestation, excluding antepartum stillbirths, but no perinatal deaths occurred in either group [LOE: 2-].

The above evidence for planned caesarean section delivery for singleton breech presentation at term indicates that this intervention is associated with a statistically significant reduction in perinatal and neonatal mortality and morbidity, though it is associated with increased short-term maternal morbidity associated with recovery from a Caesarean section. The largest study in the Cochrane review, the Term Breech Trial, demonstrates strong evidence of benefit of planned caesarean section in reducing perinatal mortality, though it should be noted that no point estimates for impact on stillbirths were reported. Additionally, a number of methodological and analytical criticisms have been levied at the largest study in the Cochrane review (the Term Breech Trial, whose principal investigator co-authored the Cochrane review) including a lack of adherence to inclusion criteria, great variability in level of care between facilities and labours (including insufficient monitoring in some cases), and a lack of clinicians with expertise in vaginal breech delivery, Glezerman M: Five years to the term breech trial: the rise and fall of a randomised controlled trial. Am J Obstet Gynecol 2006, 194:20-25.

Although planned Caesarean for breech presentation appears to be a promising intervention, in light of these criticisms and the downstream implications of recommending a policy of Caesarean section in low-resource settings, our recommendation for Caesarean section in term breech presentation carries several caveats. Given the risks associated with Caesarean section, a planned Caesarean section should only be considered after external cephalic version has been unsuccessful and sufficient time has been allowed for spontaneous version and cephalic birth to take place [95]. Caesarean section poses short-term risks of infection and haemorrhage, as well as long-term risks including uterine rupture (particularly where a classical incision is performed), placental invasion of the uterine scar, stillbirth, and maternal death, for women who cannot or do not access facilities with comprehensive essential obstetric care in subsequent pregnancies. For these reasons, in some low-resource settings, the complications associated with Caesarean section may pose greater risks to both mothers' and babies' lives than vaginal delivery for breech presentation. Vaginal birth for breech presentations is advised in these circumstances, as 97% of breech-position singleton infants will be born without complications [97]. Additionally, adoption of a policy of breech delivery by Caesarean section will ultimately lead to the disappearance of the specialised clinical skills required to perform vaginal breech delivery, meaning that women with breech fetuses who do deliver vaginally may face greater risk.

Magnesium sulphate has both anti-convulsant and tocolytic applications. Anti-convulsants, including magnesium sulphate, diazepam, and phenytoin, are a key strategy for preventing or stopping eclamptic seizures in pregnant women with pre-eclampsia and eclampsia, respectively, which account for 50,000 maternal deaths per year worldwide (10% of all direct maternal deaths) [101]. Magnesium sulphate is generally regarded as the first choice drug, and superior to diazepam or phenytoin, for controlling eclamptic fits and preventing associated maternal and fetal deaths [102‐104]. However, the use of magnesium sulphate presents potential hazards including severe maternal adverse effects such as respiratory and cardiac arrest, and risk of fetal neurological depression (e.g., decreased respiratory effort), suggesting that while magnesium sulphate may prevent some fetal and neonatal deaths due to eclampsia, it may present an increased risk of harm in some pregnancies.

Magnesium sulphate also has tocolytic properties: it relaxes smooth muscle and inhibits uterine contractile activity. Tocolytic drugs can theoretically prevent or delay labour and pre-term birth for women at high risk of pre-term delivery [105]. Ideally, a tocolytic agent would delay labour long enough to administer corticosteroids to hasten fetal lung maturation prior to delivery. Magnesium sulphate is widely used as a tocolytic agent for preventing pre-term birth in the United States [106]; below, we explore the evidence for the impact of its use on stillbirth.

Our literature search identified 4 Cochrane reviews on the role of magnesium sulphate for management of pre-eclampsia and eclampsia and 3 Cochrane reviews and 1 other study on its role in threatened preterm labour (Table 7). The role of antenatal magnesium supplementation in prevention of pre-eclampsia in magnesium-deficient populations has been discussed in a previous paper in this series [107].

Intravenous or intramuscular magnesium sulphate significantly decreases the risk of eclampsia. IV or intramuscular magnesium sulphate appears to be substantially more effective than diazepam and phenytoin for treatment of eclampsia, and is therefore the treatment of choice. No trials reported any statistically significant impact of magnesium sulphate on stillbirth incidence or perinatal mortality.

Among women deemed to be at risk of preterm birth, magnesium sulphate is ineffective at delaying or preventing labour, and while its use may be associated with side effects in the infant such as neurological depression, there is evidence that magnesium sulphate may also prevent moderate or severe cerebral palsy in live-born infants [112]. There is not enough evidence to show any difference between magnesium maintenance therapy and either placebo or no treatment, or alternative therapies (other tocolytics, for example) in preventing preterm birth after an episode of threatened preterm labour. Thus, despite good quality evidence and impact on maternal pre-eclampsia (Grade A), evidence for impact of magnesium therapy on stillbirth prevention is insufficient.

Long-term fetal hypoxia, which results from diminished oxygen and/or nutrient flow to the fetus from the mother, is often implicated in cases of impaired fetal growth. Impaired fetal growth may result from a number of factors, including fetal characteristics (e.g., congenital abnormalities), placental factors (e.g., small placenta, poor placentation), and maternal conditions (e.g., drug use, malnutrition, renal or vascular problems) [113]. Severe deprivation of oxygen and/or nutrients can result in hypoxia or nutrient deprivation so severe as to cause stillbirth. Several observational uncontrolled trials have evaluated the physiological basis for maternal hyperoxygenation (40–55% humidified oxygen by face mask at 8 litres per minute, 24 hours per day) as a way to improve oxygen flow to the fetus and thereby alleviate hypoxia and stimulate nutrient transfer [114‐118].

Our literature search identified 1 Cochrane review and 1 other intervention study (Table 8) that evaluated the impact of maternal hyperoxygenation on perinatal mortality. Say et al. [119] reviewed RCTs (3 RCTs, N = 94 women) comparing maternal oxygen therapy with no oxygen therapy in suspected impaired fetal growth, and found statistically significantly lower PMR in the oxygenation group compared to the non-oxygen group in all three included trials (RR = 0.50, 95% CI: 0.32–0.81). However, the reviewers cautioned that higher gestational age in the oxygenation groups might have accounted for the difference in mortality rates [LOE: 1++](Additional file 24).

A quasi-RCT by Battaglia et al. [120] attempted to compare the impact of maternal oxygenation versus no oxygen in a group of pregnant women (N = 38) prescribed bed rest, but no perinatal deaths occurred in either group [LOE: 2-].

The evidence for maternal hyperoxygenation in the Cochrane suggests that in pregnant women treated with long-term oxygen therapy, perinatal mortality is significantly reduced [119]. Methodological deficiencies in the trials included in the Cochrane review, including the greater average gestational age of fetuses in the oxygen intervention groups and the small size of the included trials, suggest that this promising finding must be confirmed with rigorous, large multicentre RCTs reporting stillbirths and early neonatal mortality separately. Interestingly, while there appeared to be an impact on perinatal mortality, maternal hyperoxygenation did not appear to improve fetal growth. We classify maternal hyperoxygenation as having some evidence of benefit in reducing perinatal mortality given the promising findings of the Cochrane review on this subject, but suggest that this intervention not be included in programs until confirmatory results become available, as some prior studies have suggested that maternal hyperoxygenation may inadvertently reduce uterine blood flow, and that long-term oxygen therapy may be associated with maternal pulmonary dysfunction [121‐123].

Amnioinfusion is a procedure whereby fluid (either saline or Ringer's lactate) is added to the uterine cavity transcervically via catheter, when the membranes have ruptured, or transabdominally using a needle if amniotic membranes are still intact. The procedure is often performed in an effort to prevent or relieve umbilical cord compression during labour associated with low amniotic fluid volume (oligohydramnios); the infusion of fluid also dilutes meconium detected in the amniotic fluid. Dilution of meconium is thought to minimize the risk of meconium aspiration, as thick meconium in cases of oligohydramnios is associated with significant perinatal mortality and morbidity [124]. Amnioinfusion may impact fetal survival via one or both of these mechanisms and the two can be difficult to differentiate. Amnioinfusion has also been used in an effort to prevent ascending infection in cases of premature rupture of membranes (PROM)[125], or to facilitate external cephalic version at term, though the impact of amnioinfusion for external cephalic version on perinatal outcomes has not been examined [126]. The evidence for whether amnioinfusion for this variety of indications has any impact on perinatal mortality outcomes has not been systematically compiled.

The literature search identified 2 Cochrane reviews and 11 other interventional/observational studies, presented below by indication (Table 9).

Our literature search identified ten randomised and quasi-randomised trials reporting an impact of amnioinfusion on stillbirths (9 trials, N = 1681 women) and perinatal mortality (10 trials, N = 3656 women). Pooled analysis of the impact of amnioinfusion on stillbirth incidence revealed a non-significant reduction in risk associated with amnioinfusion (RR [fixed] = 0.68, 95% CI: 0.19–2.41 [NS]; RR [random] = 0.69, 95% CI: 0.19–2.43 [NS]) (Figures 8 and 9). Comparing the impact of amnioinfusion versus controls without amnioinfusion on perinatal deaths yielded a trend toward reduced mortality (RR = 0.51, 95% CI: 0.25–1.04, RR [random] = 0.52, 95% CI: 0.25–1.09 [NS]) (Figures 10 and 11).

In low-resource settings, paediatric facilities for the management of meconium aspiration syndrome are scarce and interventions to prevent meconium aspiration are needed. Amnioinfusion, as one preventive option, is comparatively more feasible than management of meconium aspiration syndrome in settings with limited intrapartum facilities. Pooled data from the Hofmeyr meta-analysis support the use of amnioinfusion for meconium stained amniotic fluid to reduce the incidence of meconium aspiration syndrome, with a trend toward reduced perinatal mortality. However, the only large RCT on this subject by Fraser et al [129], performed after the Cochrane review, found no statistically significant impact on either meconium aspiration syndrome or perinatal mortality, suggesting the possibility of small study bias (an overrepresentation of published small trials favouring a treatment effect) in the Cochrane review [135].

For other antepartum and intrapartum indications for amnioinfusion, including PROM, oligohydramnios with intact membranes, alleviation of umbilical cord compression, and prevention of infection after membrane rupture, the available evidence is too limited to formulate recommendations on the use of amnioinfusion. Some small studies show promising differentials in mortality and other outcomes between intervention and control groups, but larger, more rigorous studies are needed to determine the impact of amnioinfusion on these outcomes.

Further studies of amnioinfusion are needed to confirm the obstetric indications and administration techniques by which amnioinfusion might reduce perinatal mortality. When performing amnioinfusion in low-resource settings, providers should remain vigilant about the risk of infection if aseptic conditions are not maintained.