Clinical Review State of the Art Review

Preterm birth and the role of neuroprotection

BMJ 2015; 350 doi: (Published 20 January 2015) Cite this as: BMJ 2015;350:g6661
  1. Eugene Chang, associate professor
  1. 1Department of Obstetrics and Gynecology, Medical University of South Carolina, Charleston, SC 29492, USA
  1. Correspondence to: E Chang changey{at}


Preterm birth remains a common complication of pregnancy and causes substantial neonatal morbidity and mortality. As improvements in the care of preterm neonates have outpaced efforts to prevent preterm birth, the numbers of survivors with neurologic sequelae that affect quality of life have increased. The main strategies to reduce the impact of neurologic complications of prematurity include prevention of preterm birth and protection of the developing fetal brain through antenatal administration of drugs. These strategies rely on a basic understanding of the intertwined pathophysiology of spontaneous preterm labor and perinatal brain injury, which will be reviewed here. The review will outline current methods for the prevention of prematurity and neuroprotection. The use of magnesium sulfate as a neuroprotective compound will be discussed, including concerns over its association with increased pediatric mortality and abnormalities in bone density.


Preterm birth (delivery at <37 weeks’ gestation) is a leading cause of neonatal morbidity and mortality, and despite substantial efforts its incidence has little changed. Although only slight headway has been made in reducing preterm birth, care of preterm neonates has continually improved, so that survival of those born at 23 weeks’ gestation is now common. Because improvements in the survival of preterm neonates have outpaced the prevention of preterm birth, the impact of long term complications on these survivors is increasingly recognized. The neurologic consequences of extreme prematurity range from mild behavioural and cognitive defects to severe disability. Perinatal neuroprotection aims to reduce these outcomes. This review will cover the neurologic sequelae related to prematurity, the pathophysiology of preterm birth and perinatal brain injury, current methods to prevent brain injury with a focus on the rationale for the use of magnesium sulfate, and emerging treatments for neuroprotection.


In 2010, about 14.9 million babies were born preterm worldwide—around 11.1% of all births.1 Complications from preterm birth are responsible for about 35% of neonatal deaths worldwide and are the second most common cause of death in children under 5 years of age.2 Preterm neonates have a median risk of 27.9% (interquartile range 18.6-46.4) for having at least one long term complication and 8.1% (3.7-10.2) for having multiple impairments.3 The most common sequelae noted in preterm neonates are learning difficulties, cognitive problems, developmental delay, cerebral palsy, and visual impairment.3

Sources and selection criteria

In September 2014, I conducted a Medline search using PubMed and Ovid with no date restrictions using the keywords “fetal”, “neuroprotection”, “preterm birth”, “prematurity”, “cerebral palsy”, and “perinatal brain injury”. For each section, keywords relating to neurodevelopmental consequences of prematurity, the pathophysiology of preterm birth and perinatal brain injury, and individual interventions for neuroprotection (such as progesterone, corticosteroids, magnesium sulfate, melatonin) were also searched. Priority was given to prospective human trials when examining neuroprotective agents or those that were being actively studied. When describing the incidence of neurodevelopmental sequelae, global studies were given precedence. With respect to pathophysiology, human studies were given priority, although animal studies were also deemed necessary for this review. I obtained additional articles by searching through reference lists of review articles identified in the above searches. I also searched using the keywords “N-acetylcysteine”, “erythropoietin”, “stem cells”, “melatonin”, and “neuroprotection”. Animal studies, reviews, experimental studies, and clinical trials were all considered.

Preterm birth


The World Health Organization defines preterm birth as delivery occurring at less than 37 weeks’ gestation. The global rate of preterm birth was 11.1% (range 9.1-13.4%) in 2010.1 Nationally, the rate ranges from 5% to 18%, with rates typically lowest in high income countries and highest in low income countries.1 However, some high income countries, such as the United States, also have high rates of preterm birth (12%).1

Thus far, the only effective interventions for preterm birth are its prevention through the use of progestins and cerclage in appropriately selected patients, smoking cessation, limits to the number of embryos transferred in assisted reproduction, and reduction of elective preterm birth in uncomplicated pregnancies.4 The widespread adoption of these interventions, however, would only slightly reduce the overall preterm birth rate.4 Therefore, a continued focus on prenatal and postnatal interventions to prevent complications related to prematurity is crucial.

Survival of preterm neonates

Although most preterm births occur after 32 weeks’ gestation (84%; 12.5 million in 2010, deliveries that occur earlier are responsible for most of the clinical and financial burden.1 Neonatal care has improved such that, in higher income countries, 90% of babies born before 28 weeks’ gestation now survive.1 Although encouraging, these babies are at risk of severe developmental disability and cerebral palsy. A prospective cohort study compared survival and neurodevelopmental outcome in babies born before 27 weeks in 1995 and 2006 in England.5 Although survival improved, there were no significant differences in rates of severe disability and cerebral palsy, underscoring the need for continued improvement in the care of the preterm neonate.

Neurodevelopmental consequences of prematurity

Neurodevelopmental consequences of prematurity include cerebral palsy, severe intellectual disability, sensorineural hearing loss or blindness, epilepsy, and more subtle behavioral and cognitive deficits. In 2007, the Institute of Medicine estimated that preterm birth costs $16.9bn (£10.7bn; €13.6bn) annually in the US, including about $1.7bn for lost household and labor market productivity and special education costs for preterm infants.6 The institute estimated that 42-47% of cases of cerebral palsy could be attributed to preterm birth. Similarly, preterm birth was an important factor in children with hearing (23% of cases), visual (37%), and cognitive (27%) impairments.

These disorders have a huge impact on families. In 2013, the lifetime cost for a child born with cerebral palsy was estimated at $1.1m in the US.7 In addition, a study that assessed the impact of preterm birth on families found that in the first year 11% had financial difficulties, 26% observed a decrease in social activities, and 28% reported that having a preterm child was emotionally challenging or difficult to accept.8 Clearly, the prevention of preterm birth and its sequelae would have a substantial impact on both society and individuals.

Cerebral palsy

Definition and risk factors

Cerebral palsy is defined as “a group of disorders of the development of movement and posture, causing activity limitation that are attributed to non-progressive disturbances that occurred in the developing fetal or infant brain” (fig 1).9 Usually, multiple factors (antenatal, intrapartum, and postnatal) act synergistically to lead to the development


Fig 1 Classification of cerebral palsy10

Preterm birth, perinatal infection, and chronic uteroplacental insufficiency that leads to impaired growth are other important factors in the development of cerebral palsy. Spastic diplegia is the most common type of cerebral palsy found in neonates born preterm. It is unclear whether the type and severity vary with the cause of preterm birth.


Cerebral palsy is usually seen in the setting of diffuse white matter injury or cystic periventricular leukomalacia (or both). It is also seen with intraparenchymal hemorrhage and intraventricular hemorrhage.11 12 Many studies have also found evidence of injury in the corticospinal tract, which carries fibers from the motor cortex to the spinal cord, in cerebral palsy.13 14 15 16 The posterior thalamic radiations that connect the thalamus to the posterior parietal and occipital cortices also show evidence of injury.17 18 Finally, neuronal loss is seen in the subplate, basal ganglia, and cerebellum.12 Thus, in the preterm neonate, injury in multiple areas within the brain lead to the clinical findings of cerebral palsy.


The prevalence of cerebral palsy is between 1.5 and 2.5 per 1000 live births and has remained stable. A systematic review found a worldwide prevalence of cerebral palsy of 2.11 per 1000 in 2013 and that prevalence is inversely related to gestational age and birth weight at delivery.19 Prevalence was highest in infants between 1000 g and 1499 g at birth (59.8/1000 live births) and lowest in those over 2500 g (1.33/1000 live births). Similarly, prevalence was higher in those born before 28 weeks’ gestation (111.8/1000) than in those born after 36 weeks (1.35/1000). The incidence of cerebral palsy is also significantly increased with impaired growth, probably as a result of hypoxia from chronic uteroplacental insufficiency. Infants under the third centile for weight have an increased risk of cerebral palsy (odds ratio 11.75, 95% confidence interval 6.22 to 12.08).20 Gestational age is a stronger predictor of cerebral palsy than impaired growth.20

Other neurodevelopmental consequences of prematurity

Motor dysfunction

Although many survivors of preterm birth show neuromotor abnormalities on examination, most do not have cerebral palsy.21 The prevalence of impairments in fine motor skills is 40-60% in infants born at less than 32 weeks’ gestation.22 One study reported poorer performance on a variety of motor tasks, overflow movements during motor tasks, sensorimotor difficulties, and visual spatial problems in 241 6 year olds born at less than 26 weeks’ gestation when compared with term controls.23 Developmental coordination disorder, a milder motor disorder than cerebral palsy that can interfere with daily activities, occurs in 18.3% of children born at less than 32 weeks’ gestation.24 Other motor abnormalities related to prematurity include mild gross motor delay, persistent neuromotor abnormalities such as asymmetries in movement patterns and tight heel cords, and functional impairments related to motor planning problems or sensorimotor integration.21

Sensorineural hearing loss and blindness

Visual and hearing impairment are also inversely related to gestational age and birth weight and their prevalence is increased in the setting of intraventricular hemorrhage or periventricular leukomalacia (or both).25 26 27 28 In one study of preschool children, 6% of those born at less than 28 weeks’ gestation had moderate to severe visual impairment and 4% had moderate to severe hearing impairment. The incidence dropped to 0.5% for moderate to severe visual changes and hearing loss at 28-32 weeks’ gestation.25 Another study of 1384 children aged 4-6 years who were under 1250 g at birth found that those with periventricular leukomalacia had the highest risk of visual impairment.26 A subsequent review estimated that worldwide about 3% of survivors born at less than 32 weeks’ gestation have visual impairment.27 Bilateral isolated hearing loss was seen in 2.2% of children born at less than 28 weeks’ gestation when they reached 2-3 years of age, and its incidence increased in the presence of intraventricular hemorrhage (2.34, 1.17 to 4.67).28

Cognitive and academic outcomes

Cognitive and academic impairments may be the most prevalent neurodevelopmental sequelae of prematurity.21 29 30 One study noted that at 2 years of age 54% of infants born at less than 27 weeks’ gestation had a Griffith mental development quotient more than 2 deviations below the mean, and that only 40% had normal cognitive abilities.29 In addition, more children had developmental delay than neurologic disabilities, including cerebral palsy. A second study found that cognitive impairment was the most common disability in children born at 30-34 weeks’ gestation.31 Finally, birth at less than 27 weeks’ gestation has been associated with an increase in the risk of autism spectrum disorders (adjusted hazard ratio 2.7, 1.5 to 5.0).32

The investigation of fetal neuroprotection is limited because cerebral palsy tends to be the sole clinical outcome studied and because considerable overlap exists between all of the neurodevelopmental outcomes. It is easy to see how visual and auditory deficits could contribute to academic difficulties, but even motor disorders can be associated with learning deficits. In a study of Australian 8-9 year olds born at less than 28 weeks’ gestation or with a birth weight below 1000 g, developmental coordination disorder was more common than in term controls and was associated with cognitive and academic deficits.33 In addition, the study described above in 6 year olds born at less than 26 weeks’ gestation found that impairment of motor, visuospatial, and sensorimotor functions independently contributed to poor classroom performance.23

When considering fetal neuroprotection it is therefore important to realize that its benefits may be underestimated because most neurodevelopmental sequelae of prematurity and more isolated or subtle outcomes have not been analyzed in clinical studies.34 35 36 37

Pathophysiology of preterm birth and perinatal brain injury

Spontaneous preterm birth is strongly associated with ascending intra-amniotic infection, which leads to maternal inflammation; this was reported as early as 1950.38 Positive amniotic fluid cultures are noted in 20-30% of women with preterm labor.39 Gestational age and intra-amniotic infection are inversely associated, with intra-amniotic infection increasingly seen as gestational age decreases. More than 85% of neonates born at less than 28 weeks’ gestation have histologic chorioamnionitis.40 In addition, maternal inflammation as defined by increased levels of interleukin 6 in the amniotic fluid is associated with adverse perinatal outcomes.41 The fetus can also develop an inflammatory response that leads to neurologic injury in this setting.

The fetal inflammatory response syndrome is characterized by raised fetal plasma interleukin 6.42 One study found that fetuses with fetal inflammatory response syndrome have higher neonatal morbidity than controls (77.8% v 29.7%; P<0.001),42 and that raised umbilical cord interleukin 6 was associated with periventricular leukomalacia (odds ratio 6.2, 2.0 to 19.1).43

Periventricular leukomalacia, the most common form of white matter injury, has two components—focal and diffuse. In the focal component, necrosis, with loss of all cellular elements in the deep periventricular white matter, leads to cystic disease. In the diffuse portion, loss of developing oligodendrocytes, astrogliosis, and microgliosis lead to diffuse white matter injury.44 Although the incidence of cystic periventricular leukomalacia has decreased, non-cystic periventricular leukomalacia is now seen in most infants born at under 1500 g.45 Figure 2 summarizes the pathogenesis of periventricular leukomalacia in relation to cerebral ischemia, systemic infection or inflammation, and vulnerability of oligodendroglial progenitor cells. Ischemia or inflammation can lead to microglial activation, excitotoxicity, and oxidative stress. These are the areas of potential intervention for neuroprotection, with excitotoxicity, inflammation, and oxidative stress currently being the most clinically relevant targets.


Fig 2 Summary of the pathogenesis of periventricular leukomalacia

Oligodendrocytes and oligodendroglial progenitor cells, which form white matter, are the main cell types injured by premature birth.46 47 Oligodendroglial progenitor cells are highly sensitive to injury between 24 and 32 weeks’ gestation, and this results in white matter injury.46 Inflammation leads to a reduction in peroxisomal proliferation in these cells and prevents their maturation.48 The consequence is that the oligodendrocyte lineage may be irreparably reduced or damaged when exposed to inflammation antenatally, making in utero interventions crucially important in the prevention of neurodevelopmental sequelae.49

Experimental models have shown that ischemia and inflammation cause cell death mainly through oxidative stress. Oligodendroglial progenitors, unlike mature oligodendrocytes, are sensitive to oxidative stress and vulnerable to attack by reactive oxygen species and reactive nitrogen species.47 50 Therefore, antioxidants may help prevent damage to these cells. In addition, cerebral ischemia can result in excitotoxicity and lead to glutamate toxicity, which can cause cell death.44

Cell death from excess glutamate is receptor and non-receptor mediated.44 The key receptors are the α-amino-3-hydroxy-5-methyl-4-isoazolepropionic acid/kainate type (AMPA/KA) and N-methyl-D-aspartic acid (NMDA) receptors.44 The presence of these receptors on oligodendroglial progenitor cells and their involvement in perinatal brain injury makes them a possible target for neuroprotection.51 Finally, with non-receptor mediated cell death, excess glutamate leads to blockade of cystine transport, depletion of intracellular glutathione, injury from oxidative stress, and cell death.44 52

Perinatal neurologic injury is caused not only by white matter injury but also by neuronal injury. Abnormalities of the cerebral cortex, thalamus, basal ganglia, and white matter neurons are also seen.53 Widespread neuronal and axonal injury often accompanies white matter injury, and the term “encephalopathy of prematurity” has been coined to reflect the wide array of abnormalities in the preterm neonatal brain.53 Neuronal-axonal disturbances may underlie neurologic sequelae such as adverse effects on cognition, attention, and behavior and are therefore also targets for neuroprotection.

Prevention of preterm birth: progesterone

Because preterm birth and neurodevelopmental outcomes are so strongly linked, strategies to prevent early delivery are paramount. Prevention of prematurity is beyond the scope of this review, however, progesterone for prevention of preterm birth warrants discussion. In a study of women with a history of spontaneous preterm birth, women were randomized to receive 17 α-hydroxyprogesterone caproate (17OHPC) or placebo from 16-20 weeks’ gestation until 36 weeks.54 Women taking 17OHPC had a significant reduction in preterm birth (relative risk 0.66, 0.54 to 0.81). Importantly, there was a significant reduction in birth at less than 32 weeks (0.58, 0.37 to 0.91). This, and the results of a recent systematic review,55 have led to the widespread use of 17OHPC for the prevention of preterm birth. Interestingly, progesterone may also have benefits with respect to neuroprotection.

Neuroprotective effects of progesterone

A recent review assessed the evidence for a role for progesterone as a neuroprotectant.56 Progesterone, and one of its derivatives in particular, allopregnanolone, is important for brain growth, neuronal and glial cell survival, and the repair of these cells after injury.57 Decreased allopregnanolone can lead to increased susceptibility to hypoxia induced excitotoxicity.57 Although no studies have examined progesterone for the prevention of cerebral palsy, allopregnanolone was shown to have potential neuroprotective benefits in an animal model of term birth asphyxia (hypoxia).58 In addition, in an animal model of white matter injury, progesterone reduced inflammation and improved myelination.59 Because the death of oligodendroglial progenitor cells (partly as a result of inflammation) leads to white matter injury and cerebral palsy, progesterone could be useful for perinatal neuroprotection in this setting through a reduction in inflammation. There is compelling evidence for progesterone as a neuroprotectant in other types of neurologic injury.56 58 59 Whether its use for prevention of preterm birth has independently had an impact on cerebral palsy is an interesting question for future research.

Neuroprotection with corticosteroids

Like progesterone, corticosteroids were first used for an alternative purpose antenatally. In 1972 it was reported that respiratory distress was reduced in preterm neonates exposed to prenatal corticosteroids.60 Over time, additional benefits of corticosteroids were recognized. In 1995, the National Institutes of Health (NIH) and the American College of Obstetricians and Gynecology (ACOG) convened and released a consensus statement recommending corticosteroids for the prevention of respiratory distress syndrome, intraventricular hemorrhage, and neonatal deaths.61

Several studies have shown a decrease in ultrasound detected intraventricular hemorrhage with steroids.62 A review of the effects of corticosteroids, which included 13 studies and 2872 infants born at less than 36 weeks’ gestation, found a significant decrease in intracranial hemorrhage (relative risk 0.54, 0.43 to 0.69).63 In addition, several studies have shown a reduction in periventricular leukomalacia with corticosteroids.62 The reductions in both these conditions might lead to an improvement in neurodevelopmental outcomes. Pooled data from four randomized trials of antenatal corticosteroids in infants born at 36 weeks’ gestation or less showed a significant reduction in cerebral palsy (odds ratio 0.59, 0.35 to 0.97).62 One study also found higher IQs in 14 year olds exposed to antenatal corticosteroids.64 These studies lend strong support to the notion that corticosteroids are clinically effective neuroprotective agents in preterm neonates.

Definition of viability and changing use of steroids

The NIH-ACOG guidelines initially recommended giving steroids between 24 and 34 weeks’ gestation to women in whom preterm delivery was anticipated. Over time, as the threshold for survival has become lower, steroids have been considered at earlier gestational ages. One retrospective cohort study showed that steroids given to women at 23 weeks’ gestation significantly reduced death in neonates (odds ratio 0.18, 0.06 to 0.54).65 In a more recent prospective cohort study of infants delivered at 22-25 weeks’ gestation, death or neurodevelopmental impairment was significantly lower with the use of antenatal steroids (0.58, 0.42 to 0.80).66 On the basis of these data, the use of steroids for improvement in survival and for neuroprotection is reasonable at 23 weeks’ gestation alongside counseling of women about the potential outcomes of resuscitation.

Neuroprotection with magnesium sulfate

Currently, the only clinically available agents for prenatal neuroprotection are corticosteroids and magnesium sulfate. Although antenatal corticosteroids are clearly warranted in those at risk of preterm birth, the use of magnesium sulfate is less clear. A case-control study published in 1995 first reported that magnesium sulfate might prevent cerebral palsy.67 Cases were very low birth weight (<1500 g) infants with a mean gestational age of 28.9 weeks and moderate-severe cerebral palsy who survived to 3 years, and controls were randomly selected very low birth weight (mean gestational age 28.4 weeks) infants. Both groups were graded according to their prenatal exposure to magnesium sulfate. Children with cerebral palsy were exposed to magnesium sulfate less often than the controls (odds ratio 0.14, 0.05 to 0.51).67 This subsequently led to three large, randomized placebo controlled trials,34 35 36 37 although little attention was given to understanding the pharmacokinetics and mechanism of action of magnesium sulfate as a neuroprotectant.

Concerns over the use of magnesium sulfate

Concerns about the safety of magnesium sulfate, especially in the setting of preterm labor, have been raised several times. One trial in particular is the source for most of this concern.68 The MagNET trial was designed to determine whether antenatal magnesium sulfate decreased the rate of cerebral palsy in preterm neonates. The study had four arms. In the tocolytic arms, women in preterm labor at less than 34 weeks’ gestation were randomized to receive magnesium sulfate (4 g bolus, then 2-3 g/h) or an alternative tocolytic (non-blinded); in the other two arms magnesium sulfate was studied solely as a neuroprotective agent. This part of the study was double blinded—women who were not eligible for tocolysis were randomized to receive a 4 g bolus of magnesium sulfate or saline.

An interim safety report found 10 deaths among those randomized to magnesium sulfate and one death in those randomized to saline.68 Overall, there were 150 neonates: 75 in the magnesium sulfate arm (65 singletons, 10 twin pairs) and 75 in the control arm (69 singletons, six twin pairs). The authors reported a significant difference in the risk of mortality between those receiving magnesium sulfate and controls (risk difference 10.7%, 2.9 to 18.5%; two sided Fisher’s exact test, P=0.02) and concluded that the use of magnesium sulfate in preterm gestations, primarily as a tocolytic, might be associated with increased pediatric mortality.68

These results raised concern about the role of magnesium sulfate in pediatric mortality. However, the details are important. In the magnesium sulfate group one infant who died was part of a twin set and had congenital anomalies, which seemed to be the cause of death rather than magnesium sulfate exposure. Two of the other deaths occurred in a twin pregnancy complicated by twin-to-twin transfusion syndrome in which one twin was a stillborn. Again, it is possible but unlikely that magnesium sulfate was the cause of death in these cases. Another death occurred at 260 days of age in a baby who had bronchopulmonary dysplasia with necrotizing pneumonia, hypertensive vasculopathy, and an atrial septal defect. Finally, three of the deaths were the result of sudden infant death, which is a common cause of death in premature children.

The authors of the MagNET trial also presented information from other contemporary trials examining magnesium sulfate.68 They identified five of 10 studies of magnesium sulfate that had data on pediatric mortality. None of those studies individually showed a significant association between exposure to magnesium sulfate and unexpected death. One of the studies included information on fetal, neonatal, and post-neonatal mortality.69 The authors included this study but not the others in a meta-analysis because it had complete information on total pediatric mortality and showed a significant increase in the risk of mortality with exposure to magnesium sulfate.

Again review of the deaths in that study is warranted. The study randomized 156 women at 24-34 weeks’ gestation to receive intravenous magnesium sulfate or no tocolysis.69 It found no difference in the duration of gestation, birth weight, neonatal morbidity, or perinatal mortality. Eight babies in the magnesium sulfate group and two in the control group died. A subsequent study described the causes of death.70 There were three malformation related deaths (all in the magnesium sulfate group), one related to abruption, two caused by extreme prematurity, and four related to well known complications of prematurity. Reconsideration of these data and those from the MagNET trial call into question the association between magnesium sulfate and pediatric mortality. Further study is warranted.

Although the association between magnesium sulfate and total pediatric mortality is unclear, reports have described skeletal abnormalities in fetuses exposed to magnesium sulfate in utero. This has prompted the Food and Drug Administration to change the classification of magnesium sulfate from pregnancy category A to D.71 Category A drugs are those in which adequate and well controlled studies have not demonstrated a risk to the fetus. Category D drugs are ones with evidence of human fetal risk, although they can be used when there is potential benefit.

On the basis of the available data, the FDA concluded that continuous administration of magnesium sulfate for longer than five to seven days should be avoided.71 The shortest duration and lowest dose that could result in harm to the fetus is unknown. Given the concerns over skeletal abnormalities and the potential association between magnesium sulfate and pediatric mortality, it seems reasonable to use the minimum dose needed for neuroprotection.

Unfortunately, little information is available on the therapeutic dosage of magnesium sulfate for neuroprotection or the acute dose that may result in fetal toxicity or adverse outcome. A study in mice reported that exposure to high doses of magnesium sulfate resulted in apoptotic cell death in the developing neonatal brain.72 Ideally, therefore, a phase I or II trial should have been performed to answer some of these questions. However, because this would be difficult in pregnant women, and because magnesium sulfate has long been used in obstetrics and is thought by many to be safe, investigators may have been justified in seeing further studies as unnecessary.

Magnesium: mechanism of action

A few studies have investigated how magnesium sulfate acts as a fetal neuroprotectant. Most of this work concerns its action as a non-competitive antagonist of the NMDA receptor.73 Although mature oligodendrocytes lack NMDA receptors, their presence on oligodendroglial progenitor cells accounts for the vulnerability of these progenitor cells to glutamate excitotoxicity.51 74 75 76 Magnesium sulfate may therefore partly work by preventing glutamate receptor mediated excitotoxicity, which leads to oligodendroglial progenitor cell death. Magnesium sulfate may also prevent neuronal cell death. A mouse model for preterm birth and fetal brain injury showed that magnesium sulfate ameliorated the morphologic changes in neurons in culture when exposed to intrauterine lipopolysaccharide.73 Although the mechanisms behind the neuroprotective effects of magnesium sulfate are still unclear and the optimum dose is unknown, several studies have assessed whether it is beneficial as a neuroprotectant.

Evidence supporting the use of magnesium

Table 1 summarizes the three major trials of magnesium sulfate.34 35 36 37 The first is a multicenter randomized trial performed in Australia and New Zealand (ACTOMgSO4) and reported in 2003.34 In total, 1062 women at less than 30 weeks’ gestation were randomized to receive a 4 g load of magnesium sulfate (n=535) or placebo (n=527) followed by a maintenance infusion (1 g/h) for up to 24 hours. The primary outcomes were total pediatric mortality, cerebral palsy, and the combination of death and cerebral palsy at 2 years of age. The study was powered to determine whether there was a 50% reduction in cerebral palsy.

Table 1

Findings of the three major trials of magnesium sulfate*

View this table:

No significant differences were found in the primary outcomes. Pediatric mortality occurred in 13.8% of the magnesium sulfate group versus 17.1% of the placebo group (relative risk 0.83, 0.64 to 1.09); cerebral palsy occurred in 6.8% versus 8.3% (0.83, 0.54 to 1.27); and the combined outcome occurred in 19.8% versus 24% (0.83, 0.66 to 1.03). Substantial motor dysfunction, a secondary outcome, was significantly decreased with magnesium sulfate (3.4% v 6.6%; 0.51, 0.29 to 0.91). Although the results of this study were negative, it was underpowered to detect more modest reductions in cerebral palsy.

The findings of the second randomized controlled trial (PREMAG) were published in 2006 and 2008.35 36 Magnesium sulfate (single 4 g bolus) was compared with placebo and the gestational age cut off was less than 33 weeks. The primary outcome in the 2006 study was neonatal mortality before discharge, severe white matter injury, and the combination of these outcomes. The study was powered to detect a 50% reduction in the primary outcome.

Two hundred and eight women were randomized to magnesium sulfate and 278 to placebo35; 688 infants (352 receiving magnesium sulfate and 336 receiving placebo) were studied. In this trial, 92.3% of neonates received the full dose of magnesium sulfate at a mean of one hour and 38 minutes before delivery. There were no significant differences in the primary outcomes. Similar to the first study, there were trends towards improvement in severe white matter injury with magnesium sulfate. Neonatal death occurred in 9.4% versus 10.4% (odds ratio 0.79, 0.44 to 1.44), severe white matter injury in 10.0 versus 11.7% (odds ratio 0.78, 0.47 to 1.31), and the combined outcome in 16.5% versus 17.8% (0.86, 0.55 to 1.34).

In 2008 data from 688 infants were reported.36 There were no significant differences in long term clinical outcomes: gross motor dysfunction (odds ratio 0.65, 0.41 to 1.02) or combined death and cerebral palsy (0.65, 0.42 to 1.03). Death and gross motor dysfunction (0.62, CI 0.41 to 0.93) as well as death, cerebral palsy, and cognitive dysfunction combined (0.68, 0.47 to 1.00) were decreased with magnesium sulfate. There was no significant difference in cerebral palsy alone (0.63, 0.35 to 1.15).

The results of a multicenter trial in the US were published in 2008.37 Women at risk of imminent delivery between 24 and 31 weeks’ gestation were randomised to receive either magnesium sulfate or placebo. Patients receiving magnesium sulfate were given a 6 g loading dose followed by a maintenance dose (2 g/h) for up to 12 hours. If the woman had not delivered or delivery was no longer felt to be imminent after 12 hours the infusion was stopped. Magnesium sulfate (or placebo) was restarted if delivery was again thought to be imminent. If more than six hours had elapsed since exposure to magnesium sulfate, another loading dose was given. The primary outcome was stillbirth or death before 1 year of age or moderate to severe cerebral palsy at 2 years of age or more. The study was powered to detect a 30% reduction in the primary outcome.

There were 2241 participants studied. No significant difference was seen in the primary outcome. The rate of death or moderate to severe cerebral palsy was 11.3% versus 11.7% (relative risk 0.97, 0.77 to 1.23). There was a non-significant increase in deaths with magnesium sulfate (9.5% v 8.5%; 1.12, 0.85 to 1.47). Contrary to the previous studies, the rate of moderate to severe cerebral palsy was significantly lower in patients treated with magnesium sulfate (1.9% v 3.5%; 0.55, 0.32 to 0.95). The investigators concluded that magnesium sulfate might be beneficial through reducing the likelihood of moderate to severe cerebral palsy.37

Several meta-analyses followed the publication of these three trials.77 78 79 All had similar findings, although the methods that they used were slightly different. The first meta-analysis found significant reductions in cerebral palsy, moderate to severe cerebral palsy, and substantial gross motor dysfunction but no significant difference in total pediatric mortality.77 The second found a reduction in cerebral palsy, moderate to severe cerebral palsy, and death or moderate to severe cerebral palsy if magnesium sulfate was given before 34 weeks.78 Finally, the Cochrane Database review found a reduction in cerebral palsy and substantial gross motor dysfunction without an increase in the risk of pediatric mortality.79

Protocol development

Although the above meta-analyses are compelling with regard to magnesium sulfate, controversy remains. Because the major randomized trials all used different dosing regimens of magnesium sulfate, the ideal dosing and timing remain unclear. In addition, different gestational ages at randomization leave clinicians confused about which patients should be candidates for neuroprotection. Finally, because of the lingering problems of safety, therapeutic level, and uncertain mechanism of action, many clinicians remain skeptical about this treatment.

To clarify the use of magnesium sulfate for neuroprotection, several societies have separately published practice guidelines (summarized in table 2).80 81 82 83 There are similarities in dosing but not with respect to gestational age at treatment. ACOG gave no specific dosing guidelines but stated that the evidence suggests a benefit with magnesium sulfate as a neuroprotectant, and it urged clinicians to develop specific guidelines in accordance with one of the larger trials.

Table 2

Published guidelines on the use of magnesium sulfate for neuroprotection

View this table:

Given ACOG’s recommendations and the uncertainty surrounding magnesium sulfate, it is important to consider the criteria for this treatment. In terms of dosing, it seems advantageous to use the lowest “therapeutic” dose possible. A single 4 g bolus, as studied in the PREMAG trial, seems the most logical and easy to implement dose.

Because a bolus dose without continuous infusion makes treatment logistically easier, it seems reasonable to extend treatment to a later gestational age. Oligodendroglial progenitor cells are thought to be most vulnerable to injury at 24-32 weeks’ gestation,44 46 so it seems logical to stop treating patients at 32 weeks. From a clinical perspective, however, subgroup analysis by the Society of Obstetricians and Gynaecologists of Canada demonstrated a significant benefit of treatment up to 34 weeks.80 Despite this, the society reached a consensus to use a cut-off gestational age of 32 weeks to strike a balance between appropriate use of magnesium sulfate and its overuse at later gestational ages.80 Finally, the Antenatal Magnesium Sulfate for Neuroprotection Guideline Development Panel in Australia concluded that because of limited resources, treatment should be considered only up to 30 weeks—the time at which magnesium sulfate has its greatest effect.83

Taking all of the above into account, a reasonable strategy would be a single 4 g bolus of magnesium sulfate over 30 minutes in patients up to 34 weeks’ gestation in whom delivery is felt to be imminent. Delivery is defined as being imminent if the woman presents in labor, with or without ruptured membranes, and the cervix is dilated by more than 4 cm.80 In planned preterm delivery, magnesium sulfate should be started in the active phase of labor or before a cesarean section at least two hours before delivery to allow for a similar exposure to fetuses as in the PREMAG trial.35

Emerging treatments

The use of magnesium sulfate is controversial and highlights the difficulties in studying neuroprotective agents for prenatal use. Although pharmacokinetic and mechanistic studies are ideal, they are time consuming, difficult, and expensive to conduct. However, several promising treatments have emerged, including N-acetylcysteine, erythropoietin, melatonin, and stem cells for neuroprotection. These are currently being investigated.


N-acetylcysteine has antioxidant and anti-inflammatory properties that might be useful in the prevention of preterm birth and perinatal brain injury.84 85 86 It has been shown to have neuroprotective benefits in an animal model of preterm perinatal brain injury.84 In the US, a clinical trial assessing whether N-acetylcysteine can prevent adverse neonatal outcomes in infection in women with preterm labor or preterm premature rupture of membranes is due to be completed in April 2015.


Erythropoietin, the primary cytokine in red cell maturation, is also a promising neuroprotective agent.87 This cytokine has several actions that may help prevent preterm brain injury. It decreases cell death, acts as an anti-inflammatory agent, increases neurogenesis, and protects developing oligodendrocytes.86 87 88 In vitro and in vivo studies in neonatal animal models suggest that erythropoietin has a neuroprotective benefit.87 A phase III trial is currently being conducted and should be completed in December 2018.


Melatonin is endogenously synthesized from the neurotransmitter serotonin and has multiple functions.89 It is a highly effective antioxidant and free radical scavenger, and it reduces the production of proinflammatory cytokines.90 91 It is an attractive candidate for neuroprotection because of its ability to cross physiologic barriers to reach subcellular compartments.92 93 It has been shown to be neuroprotective in animal models of neonatal hemorrhagic brain injury and periventricular leukomalacia.94 95 A recent pharmacokinetics study of melatonin administered to preterm neonates (<31 weeks) was recently published and provides information to guide therapeutic trials.96 Such trials are ongoing.

Umbilical cord blood stem cells

Umbilical cord blood stem cells are a promising treatment. Umbilical cord blood has a diverse population of progenitor and stem cells that may be useful for neuroprotection.97 98 Two specific cell populations, endothelial progenitor cells and mesenchymal stem cells, hold the most promise.97 Endothelial progenitor cells are mobilized in response to acute hypoxia. They maintain vascular integrity and homeostasis and mediate the response to vascular injury.97 Mesenchymal stem cells are multipotent cells that augment host repair and tissue recovery. They support re-myelination and the inhibition of apoptosis and inflammation and thus are attractive neuroprotectants.97 An animal study found that intranasal delivery of mesenchymal stem cells reduced white matter injury and motor deficits in neonatal ischemia, pointing to the potential of this intervention.99 Several ongoing trials are exploring the use of umbilical cord blood for the treatment of established ischemic injury and cerebral palsy but none exists for prevention.

Summary of emerging treatments

These candidate neuroprotective compounds require substantially more study before they can be used clinically. It is hoped that short term markers that predict future development of both mild and severe neurologic sequelae can be identified. This would greatly help in future studies and would enable more efficient translational research in the area of neuroprotection. It would also help identify compounds and determine appropriate sample sizes for future randomized trials.


The prevention of preterm birth remains a worldwide challenge. Given its relatively stable rate worldwide and the increased survival of preterm neonates, attention needs to be devoted to preventing the sequelae of prematurity. To be successful, the role of infection and inflammation in preterm birth and preterm brain injury needs to be recognized. In addition, the role of excitotoxicity and neuronal injury needs to be understood, especially when considering the potential of magnesium sulfate and other agents as neuroprotectants.

To prevent the neurodevelopmental complications of preterm birth, a two pronged approach is needed—firstly, the prevention of preterm birth with clinically proven interventions such as 17OHPC, and, secondly, the use of dedicated neuroprotective drugs (fig 3). Although several promising neuroprotective drugs have been identified, further study is needed. Clinicians need to be aware of the drugs that are currently available for neuroprotection. The judicious use of steroids in women at risk of preterm birth can reduce neurodevelopmental sequelae; these drugs should be administered according to published guidelines. Finally, magnesium sulfate should also be used for neuroprotection and local protocols developed with respect to its dosing and indications.


Fig 3 Use our interactive graphic to explore the neurodevelopmental consequences of prematurity and the different neuroprotective agents used. See

Future research questions

  • Because inflammation and infection are strongly associated with prenatal brain injury, neurodevelopmental outcomes in the offspring of women who have non-obstetric infections during pregnancy, such as pyelonephritis and influenza, should be studied. Would the administration of neuroprotective drugs in these women be of benefit?

  • What are the mechanism of action and pharmacokinetics of magnesium sulfate? A better understanding of these processes may ultimately improve dosing protocols

  • The impact of progesterone on neurodevelopmental outcomes needs further study. In particular, do micronized progesterone and 17-α hydroxyprogesterone caproate have different effects on neurodevelopment?

  • Newer drugs for neuroprotection are being studied; do drug combinations have synergistic benefits in terms of neuroprotection?


Cite this as: BMJ 2015;350:g6661


  • Contributors: EC is the sole contributor and will act as guarantor.

  • Competing interests: I have read and understood BMJ policy on declaration of interests and declare the following interests: none.

  • Provenance and peer review: Commissioned; externally peer reviewed.


View Abstract

Log in

Log in through your institution


* For online subscription