• Mortality;
  • neuroprotection;
  • newborn;
  • periventricular;
  • leucomalacia;
  • white matter


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Funding
  9. References
  10. Appendix

Objective  To evaluate whether magnesium sulphate (MgSO4) given to women at risk of very-preterm birth would be neuroprotective in preterm newborns and would prevent neonatal mortality and severe white-matter injury (WMI).

Design  A randomised study.

Setting  Eighteen French tertiary hospitals.

Population  Women with fetuses of gestational age < 33 weeks whose birth was planned or expected within 24 hours were enrolled from July 1997 to July 2003 with follow up of infants until hospital discharge.

Methods  Five hundred and seventy-three mothers were randomly assigned to receive a single 40-ml infusion of 0.1 g/ml of MgSO4 (4 g) solution or isotonic 0.9% saline (placebo) over 30 minutes. This study is registered as an International Standard Randomised Controlled Trial, number 00120588.

Main outcome measures  The primary endpoints were rates of severe WMI or total mortality before hospital discharge, and their combined outcome. Analyses were based on intention to treat.

Results  After 6 years of enrolment, the trial was stopped. Data from 688 infants were analysed. Comparing infants who received MgSO4 or placebo, respectively, total mortality (9.4 versus 10.4%; OR: 0.79, 95% CI 0.44–1.44), severe WMI (10.0 versus 11.7%; OR: 0.78, 95% CI 0.47–1.31) and their combined outcomes (16.5 versus 17.9%; OR: 0.86, 95% CI 0.55–1.34) were less frequent for the former, but these differences were not statistically significant. No major maternal adverse effects were observed in the MgSO4 group.

Conclusion  Although our results are inconclusive, improvements of neonatal outcome obtained with MgSO4 are of potential clinical significance. More research is needed to assess the protective effect of MgSO4 alone or in combination with other neuroprotective molecules.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Funding
  9. References
  10. Appendix

Cerebral palsy and neurodevelopmental disorders remain serious concerns among infants born very preterm. White-matter injury (WMI) is the main developmental stage-specific brain disease incriminated as the origin of these neurodisabilities.1,2 Some WMI features, such as focal periventricular cysts, can be identified by routine cranial ultrasound scan (CUS) during the neonatal period.3–5

In several human observational studies, prenatal administration of magnesium sulphate (MgSO4) for tocolysis or pre-eclampsia was reported to be associated with lower neonatal mortality and lower risk of cerebral palsy in very-low-birthweight children,6–9 even though these beneficial effects were not observed in other similar studies.10–12 Three randomised trials were undertaken in recent years, one in Australia and New Zealand, one in France, and one in the United States, to assess the effectiveness of MgSO4 in preventing neonatal mortality, perinatal cerebral injuries and/or cerebral palsy. The results of the first trial, performed by the Australasian Collaborative Trial of the MgSO4 Collaborative Group, are now available and showed a significantly lower rate of substantial gross motor dysfunction at 2 years in children born very preterm (<30 weeks of gestation) whose mothers had been given prenatal MgSO4 infusions rather than placebo saline infusions.13

Herein, we describe the results of the French PREMAG multicentre randomised trial that included women with no pregnancy-associated vascular disease and at risk of preterm delivery before 33 weeks of gestational age. Its goal was to assess the effectiveness of a single MgSO4 infusion in preventing mortality and/or WMI in newborns.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Funding
  9. References
  10. Appendix


Between July 1997 and July 2003, women pregnant with singleton, twin, or triplet very-preterm fetuses younger than 33 weeks of gestational age were eligible if birth was expected or planned within 24 hours. No lower limit was set for gestational age at enrolment, except those established at individual participating centres concerning viability. Gestational age was the best estimate of completed weeks of gestation based on early ultrasound and menstrual history. Furthermore, the women could not have received betamimetics, aminoglycosides, or steroids for at least 1 hour, and their signed written informed consent was required.

Women were not eligible if the fetus had severe malformations or chromosomal abnormalities or if they met at least one of the following criteria: hypotension, cardiac rhythm abnormalities, hydroelectrolyte abnormalities, renal insufficiency, ingestion during the last 24 hours of calcium channel blockers, digitalins or indomethacin, persistent signs of cardiovascular toxicity or tachycardia >1 hour after cessation of tocolytic intake, myasthenia, or indication for emergency caesarean section. Moreover, women with a pregnancy-associated vascular disease (i.e. pre-eclampsia, growth restriction, haemolysis, elevated liver-function test results, low-platelet syndrome, retroplacental haematoma) were not eligible, to restrict our study to newborns at higher risk of developing sequelae. Indeed, lower frequencies of periventricular leucomalacia and cerebral palsy were reported for infants whose preterm births resulted from their mothers’ pregnancy-associated vascular disease.14–16


The Research and Ethics Committee of the coordinating centre’s region, i.e. the Upper-Normandy region in France, approved the study protocol. Eighteen tertiary hospitals with a neonatal intensive care unit (NICU) accepted to participate in the study. Randomisation was stratified by centre, singleton/multiple pregnancy and gestational age (<27, 27–29, 30–32 weeks).

Randomisation numbers were generated by computer using variable block size from 2 to 16 depending on expected recruitment. Central telephone randomisation was managed by staff of the NICU at the Antoine-Béclère Hospital, Clamart, 24 hours a day, 7 days a week. Once treatment was assigned, the next corresponding unmasked treatment pack available from each participating centre’s supplies was used. Treatment packs were prepared by the coordinating centre’s pharmacy and sent ready for use to each participating centre’s pharmacy. The time of treatment assignment by phone was the point of randomisation, regardless of whether the infusion was subsequently started or completed.


Women received a single 40-ml infusion of either a 0.1 g/ml MgSO4 solution (i.e. 4 g or 16 mmol of MgSO4) or isotonic 0.9% saline over 30 minutes. The two solutions looked identical so that the women were unaware of whether they received a MgSO4 or placebo solution. Treatment assignment was single blind rather than double-blind for the following two reasons. First, the anaesthetists involved in the PREMAG trial planning required to know the treatment assigned to be able to take immediate action against the possible adverse effects of MgSO4. Second, double blinding was deemed impossible to maintain for anaesthetists or obstetricians because MgSO4 infusion is usually responsible for characteristic flushes.

At each participating centre, women and infants were cared for according to standard clinical practices. Maternal blood electrolyte and creatinine levels were determined before randomisation. Women’s pulse rate, blood pressure (Dinamap®; GE Healthcare, Little Chalfont, UK), respiratory rate, tendon reflexes, and any maternal adverse effects were recorded throughout the infusion. The infusion was stopped at the attending anaesthetist’s discretion. Fetal heart rate was monitored throughout labour. Mothers and their children were followed up until hospital discharge.


The primary outcomes were overall neonatal mortality before hospital discharge, detection of neonatal CUS abnormalities evocative of severe WMI, as defined below, and the combination of severe WMI, and/or neonatal mortality.

The secondary neonatal outcomes were the following: all CUS-diagnosed WMI, whether severe or moderate, as defined below, nonparenchymal haemorrhage, periventricular cavitary lesions and their extension (unilateral/bilateral, ≥1 location among frontal, parietal, and/or occipital lobes), clinical adverse effects of the infusion, and neonatal complications. The secondary maternal outcomes were cardiovascular and respiratory adverse effects of the infusion, postpartum haemorrhage, and/or maternal death.

For all surviving infants, CUS was conducted by a senior neonatologist or radiologist in each centre separately and in a blind manner relative to treatment allocation. They were planned according to the following schedule: one within the first week after birth, one between days 15 and 21 of life, and one 6 weeks after birth. An additional CUS was scheduled before discharge from the NICU for the most premature infants. At each participating centre, the total number of CUS performed and the numbers and types of abnormalities found were recorded by the infant’s neonatologist during hospitalisation. To minimise variations among centres in the detection of CUS abnormalities, two general variables were defined based on previous suggestions.4,17 Severe WMI was considered present when at least one of the three following parenchymal abnormalities was detected: cystic periventricular leucomalacia, periventricular parenchymal haemorrhagic involvement defined as a large unilateral parenchymal hyperdensity, or a large single unilateral porencephalic cyst caused by ischaemic–haemorrhagic infarction. WMI was considered present when severe WMI, as defined above, was present or at least one of the two following abnormalities was detected: echodensity persisting for >14 days without cyst formation or isolated ventricular dilatation with no associated intraventricular haemorrhage (IVH).18 Nonparenchymal haemorrhage was defined as germinal matrix layer haemorrhage or IVH with or without ventricular dilatation.

Statistical analysis

All statistical analyses were undertaken on an intention-to-treat basis. All comparisons between the MgSO4 and placebo groups took into account the correlations of outcomes of twins or triplets born to the same mother through a generalised estimating equation approach within logistic regression.19 In addition, comparisons of primary outcomes and all CUS secondary outcomes were adjusted for stratification variables, gestational age, singleton/multiple pregnancy, and birthweight, which was found to be predictive of the primary outcomes. The linear regression residuals (defined as the difference between observed birthweight and that predicted by the linear regression line) of birthweight versus gestational age were used to eliminate colinearities between the two variables. No appreciable change was obtained with further adjustment for short-term perinatal outcomes, namely the Apgar score, which was found to be predictive of the primary outcomes and prolonged (i.e. ≥18 hours) prelabour rupture of membranes (PPROM) or maternal–fetal infection (yes, no) that developed more frequently in the MgSO4 group. Odds ratios (OR) and their corresponding 95% confidence intervals were estimated. Stata version 8.2 software (StataCorp., College Station, TX, USA) was used, and the two-sided significance level was 0.05.

Sample size

The target sample size of 1106 newborns was calculated to detect a 50% reduction of the risk of severe WMI from 8 to 4%, with 80% power at the two-sided 0.05 level. This 50% reduction appeared achievable in view of much larger reductions of the risk of periventricular leucomalacia or cerebral palsy with magnesium sulphate reported in several observational studies.6,8,9 Expecting that 20 and 1% of mothers would give birth to twins and triplets, respectively, meant that 906 mothers had to be included.

Interim and safety analyses

No interim analysis of the primary outcomes was either planned or conducted. The trial was overseen by a steering committee that was kept informed of mother and infant complications. In the absence of major maternal adverse effects or immediate neonatal deaths imputable to MgSO4, the steering committee never held a formal meeting. In light of the report by Mittendorf et al.20 of increased mortality of infants born to women who received MgSO4 at tocolytic doses in the MAGNET randomised trial, the Upper-Normandy Research and Ethics Committee was consulted, and it authorised the continuation of the PREMAG Trial in January 1998.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Funding
  9. References
  10. Appendix

A total of 573 women were enrolled in the trial (Figure 1). Two of the 18 centres that accepted to participate in the study included no women and three included fewer than five women (two, three, or four women). Their data were not retained in the trial analysis according to a predefined rule, which left 564 women in 13 centres for analysis, of which 286 women were assigned to receive MgSO4 and 278 the placebo. At the time of randomisation, two women in the MgSO4 group and five in the placebo group, all with twin pregnancies, had one dead fetus. A total of 688 fetuses were living, 352 from women assigned to receive MgSO4 and 336 from placebo recipients. Outcome data were obtained until hospital discharge for all 564 women and 688 infants. After 6 years, the motivation of many investigators appeared to have dwindled severely so that the enrolment rate had decreased dramatically. Therefore, the recruitment of women was stopped before reaching the projected sample size.


Figure 1. Trial profile.

Download figure to PowerPoint

Baseline maternal and pregnancy characteristics

Most characteristics were similar for MgSO4 and placebo groups (Table 1). The median gestational age at trial entry was 30 weeks; gestational age was <27 weeks for 72 (12.8%) pregnancies, between 27 and 29 weeks for 195 (34.6%) and between 30 and 32 weeks for 297 (52.7%). One hundred and twenty-two (21.6%) women had multiple pregnancies (9 triplets and 113 twins). The main reasons for preterm birth (Table 1) were preterm labour, PPROM, antepartum haemorrhage and chorioamnionitis. Most women received tocolytics and antibiotics, and almost all received corticosteroids. The only notable difference between the two groups was a slightly higher proportion of women with PPROM in the MgSO4 group (53.9 versus 46.6%).

Table 1.  Baseline maternal and pregnancy characteristics and reasons for preterm birth in the PREMAG trial
CharacteristicsMgSO4 (n= 286)Placebo (n= 278)
  • Values are n (%) unless stated otherwise.

  • *

    Defined as the presence of at least two criteria among maternal pyrexia greater than 38°C, fetal tachycardia, meconium in amniotic fluid, bacteria in amniotic fluid, C-reactive protein level > 40 mg/l, or neutrophil count > 20 g/l within the last 48 hours.

  • **

    Uterine malformation, polyhydramnios, cervical incompetency, alloimmunisation, abdominal trauma, diabetes, pyelonephritis, or cholestasis.

Maternal age, mean ± SD (years)29.3 ± 5.329.5 ± 5.1
Gestational age at entry, median [range] (weeks)30 [24 weeks to 32 weeks 6 days]30 [23 weeks 4 days to 32 weeks 6 days]
Singleton pregnancy222 (77.6)220 (79.1)
Reasons for preterm birth
Preterm labour236 (84.0)242 (88.3)
PPROM187 (53.9)156 (46.6)
Chorioamnionitis*27 (9.5)34 (12.6)
Antepartum haemorrhage54 (19.0)54 (20.0)
Other**33 (9.8)43 (13.3)
Tocolysis190 (67.6)192 (70.8)
Antibiotics219 (77.1)207 (75.3)
Corticosteroids270 (95.1)261 (94.6)

Treatment and maternal adverse effects

Most women in both groups received the full dose of MgSO4 or placebo (92.3 or 89.2%, respectively). Among the 52 women who did not receive the full dose (22 in the MgSO4 group and 30 in the placebo group), 41 received no infusion before delivery (20 and 21, respectively). No women received MgSO4 for clinical reasons after enrolment.

In the MgSO4 group, no major maternal adverse effects (death, cardiac arrest, or prolonged mechanical ventilation) were observed. Two women suffered severe postpartum haemorrhages (one with placenta praevia, the other with a twin pregnancy). Six women (2.1%) experienced moderate adverse effects consisting of hypotension (n= 3), tendon reflex abolition (n= 2), or curarisation (n= 1). Mild adverse effects were observed in 41 women (14.3%) including flushes (n= 23), nausea or vomiting (n= 9), or headache (n= 4). In the placebo group, one woman with placenta accreta died following a major postpartum haemorrhage. No other women experienced any other major adverse events. One woman experienced tendon reflex abolition, and three women had mild adverse effects consisting of nausea or vomiting (n= 2), or headache (n= 1).

Labour and delivery outcomes, neonatal morbidity

Labour and delivery outcomes were similar for the two groups (Table 2), including times from infusion initiation to birth. Caesarean sections were performed on a large proportion (37.7%) of the women either prior to or at labour onset (28.4%) or during labour (9.2%). The only treatment group difference for neonatal morbidity was the significantly higher rate of maternal–fetal infections in the MgSO4 group (P= 0.01) (Table 3).

Table 2.  Labour and delivery outcomes
ParameterMgSO4PlaceboP value*
  • ROM, rupture of membranes.

  • Values are n (%) unless stated otherwise.

  • *

    Student’s t test or Pearson’s chi-square test as appropriate.

  • **

    Singleton pregnancies only.

Maternal outcomes, n286278 
Interval from infusion to delivery, median [range]1 hour 38 minutes [5 minutes to 25 hours 5 minutes]1 hour 30 minutes [8 minutes to 61 hours 30 minutes]0.21
Caesarean section116 (40.6)96 (34.7)0.15
No anaesthesia46 (16.2)58 (21.7)0.10
 Epidural165 (70.8)146 (71.2) 
 Spinal28 (12.0)34 (16.6) 
 General47 (20.2)29 (14.1) 
Fetal outcomes, n352336 
Interval from ROM to delivery, median [range] (hour)24 [0–1656]12 [0–2352]0.92
Abnormal fetal cardiotocogram95 (27.9)94 (29.9)0.76
Presentation 0.71
 Cephalic238 (69.2)229 (69.2) 
 Breech87 (25.3)88 (26.6) 
 Other19 (5.5)14 (4.2) 
Placenta weight,** median [interquartile range] (g)370 [300–450]350 [280–440]0.29
Table 3.  Characteristics of birth and neonatal morbidity
CharacteristicsMgSO4 (n= 352)Placebo (n= 336)P value
  • BP, blood pressure; HC, head circumference.

  • Values are n (%) unless stated otherwise.

  • *

    Defined as the presence of at least two criteria among clinical signs of infection, leucocyte count > 30 g/l or neutrophil count < 5 g/l, presence of bacteria in gastric or blood samples, or C-reactive protein level > 20 mg/l.

At birth
Gestational age, median [range]30 weeks 1 day [24 weeks 1 day to 32 weeks 6 days]30 weeks 1 day [23 weeks 4 days to 32 weeks 6 days]0.87
Male, n (%)193 (55.4)197 (58.9)0.46
Birthweight, median [interquartile range] (g)1350 [1080–1670]1415 [1120–1680]0.45
Length, median [interquartile range] (cm)40 [38–42]40 [38–42]0.50
HC, median [interquartile range] (cm)28 [26–29]28 [26–29]0.77
Apgar score < 745 (12.8)31 (9.2)0.11
Tracheal intubation and/or epinephrine115 (32.9)101 (30.1)0.66
In neonatal period
Respiratory distress syndrome145 (42.0)123 (37.8)0.43
Endotracheal ventilation191 (55.4)175 (53.8)0.89
Noninvasive ventilation235 (69.3)217 (67.4)0.57
Apnoea or bradycardia101 (29.8)76 (23.7)0.15
BP at admission in NICU, median [interquartile range] (mmHg)35 [31–40]36 [31–41]0.91
Hypotension41 (12.1)30 (9.3)0.31
Maternal–fetal infection*74 (21.2)46 (14.1)0.01
Necrotising enterocolitis9 (2.6)6 (1.9)0.50
Seizures7 (2.1)9 (2.9)0.50
Oxygen dependency at 36 weeks28 (8.4)31 (10.0)0.44

Primary outcomes

The rates of total mortality before hospital discharge, severe WMI, and the combination of severe WMI and/or death were all lower for the MgSO4 group, but no differences were statistically significant (Table 4). Among the fetuses alive at randomisation, 68 (9.9%) died. Five (0.7%) deaths occurred in utero and 63 (9.2%) during the postnatal period before discharge. The rates of respiratory and/or neurological postnatal death were similar for the two groups.

Table 4.  Primary outcomes
OutcomeMgSO4 (n= 352)Placebo (n= 336)P valueAdjusted OR [95% CI]
  1. Values are n (%) unless stated otherwise.

Total mortality33 (9.4)35 (10.4)0.450.79 [0.44–1.44]
In utero2 (0.6)3 (0.9) 
Postnatal31 (8.8)32 (9.5) 
 Neurological12 (3.4)15 (4.5) 
Severe WMI34 (10.0)38 (11.7)0.350.78 [0.47–1.31]
Severe WMI and/or death58 (16.5)60 (17.9)0.490.86 [0.55–1.34]

Among the 688 fetuses alive at randomisation, 23 (3.3%) infants never had CUS because they died too early. Among the 665 (96.7%) infants with available data, the severe WMI rate was lower in the MgSO4 group than in the placebo group (Table 5). The combined outcome was also lower in the MgSO4 group. A sensitivity analysis that included participants who had received the full treatment dose gave similar results for these primary outcomes.

Table 5.  Secondary cranial ultrasound outcomes
OutcomeMgSO4 (n= 341)Placebo (n= 324)Adjusted OR [95% CI]
  • SEH, subependymal haemorrhage.

  • Values are n (%) unless stated otherwise.

  • *

    In frontal, parietal, and/or occipital lobes.

All WMI54 (15.8)59 (18.2)0.79 [0.51–1.22]
Cysts27 (7.9)28 (8.6)0.94 [0.53–1.68]
Diagnosed at >day 82024 
Time to diagnosis >day 8, median [interquartile range] (days)29 [18–40]18 [18–40] 
Diameter > 3 mm1510 
Bilateral site2018 
More than one site*139 
Intraparenchymal haemorrhage8 (2.4)11 (3.4)0.42 [0.14–1.21]
Porencephalic cyst1 (0.3)3 (0.9)0.29 [0.03–2.82]
Nonparenchymal haemorrhage63 (18.5)71 (21.9)0.75 [0.50–1.11]
SEH45 (13.2)51 (15.7)0.77 [0.49–1.21]
Isolated IVH24 (7.0)23 (7.1)0.87 [0.47–1.62]
IVH with dilation11 (3.2)13 (4.0)0.76 [0.31–1.84]

WMI and nonparenchymal haemorrhages

Among the 665 infants with available data, no significant differences were observed for secondary CUS outcomes. The rate of all WMI, whether severe or not, was lower in the MgSO4 group than in the placebo group (Table 5) as was the rate of nonparenchymal haemorrhages.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Funding
  9. References
  10. Appendix

In our PREMAG trial assessing MgSO4 as a potential neuroprotective molecule before very-preterm birth, each primary outcome (i.e. neonatal mortality before discharge, severe neonatal WMI, and combined severe WMI and/or mortality) and several secondary neonatal outcomes (i.e. all WMI and nonparenchymal haemorrhages) tended to be lower for the MgSO4 than for the placebo group. Although none of these differences was statistically significant, the average reductions in these outcomes (21, 22, 14, 21 and 25% on the OR scale, respectively) are of potential clinical significance.

This study has two limitations associated with statistical power and use of severe WMI as a primary outcome. Because of difficulties in maintaining a high enrolment rate, recruitment was stopped after 6 years when only 564 mothers had been included, yielding 688 infants for analysis (62% of the target sample size). Ignoring correlations among infants of multiple pregnancies, this sample size achieved 80% power for detecting a 52% decrease in the severe WMI rate with MgSO4 (5.6 versus 11.7%), a 54% decrease in neonatal mortality before discharge (4.8 versus 10.4%), or a 40% decrease in the combined outcome rate (10.4 versus 17.9%). For differences actually observed during the trial, power was much lower. A higher recruitment would have been obtained if mothers with pregnancy-associated vascular disease had been included. Indeed, in the French multicentre EPIPAGE study, pregnancy-associated vascular disease was the main aetiology in 31% of very-preterm single births in the year 1997.16 However, few additional cases of WMI would have been expected because of the low rate of WMI among these infants.14–16 If MgSO4 had a stronger (respectively weaker) neuroprotective effect on preterm infants in case of pregnancy-associated vascular disease, our study would have underestimated (respectively overestimated) this effect, but there seem to be no data in the literature suggesting this potential differential effect.

The use of CUS-diagnosed severe WMI as a primary outcome was motivated by its rapid availability during the postnatal hospital stay and its association with subsequent neurological disorders.1,2 However, poor visualisation of diffuse WMI forms and problems of inter-observer variability have been reported with CUS.4,17,21 But since the radiologists or senior neonatologist interpreting CUS in each centre was not aware of treatment allocation, there is no reason to think that potential errors in WMI diagnosis occurred differentially between the two treatment groups. To check the validity of our results based on decentralised CUS reading, central reading by a single CUS expert (D.E.) blinded to the initial CUS diagnosis was performed for about half of the newborns (n= 347), and substantial agreement was demonstrated between initial decentralised reading and the central reading, with a kappa value of 0.65 (95% CI 0.53–0.78). Moreover, substituting the expert’s reading for the initial decentralised reading when they disagreed yielded lower severe WMI rates, with 59 (8.9%) cases of severe WMI overall: 31 (9.6%) in the placebo group and 28 (8.2%) in the MgSO4 group, but the adjusted OR changed only marginally from 0.78 to 0.79 (95% CI 0.45–1.37) and remained nonsignificantly different from 1 (P= 0.40). Use of magnetic resonance imaging may have yielded better sensitivity than CUS in detecting subtle forms of WMI, but it was not routinely available in all the centres in 1997 in France.22

The lack of a significant MgSO4 effect in our PREMAG trial, despite its reported anti-inflammatory and antiexcitotoxic properties,23–27 may be explained by the higher frequencies of PPROM and/or maternal–fetal infection observed in our MgSO4 group that may have counterbalanced the neuroprotective effect of magnesium. Indeed, inflammatory processes, such as PPROM, chorioamnionitis, or maternal–fetal infection, have been reported to be associated with higher risks of periventricular leucomalacia.16,28 The higher PPROM rate observed in our MgSO4 group may result from the moderate tocolytic effect of MgSO4,29 which may have enhanced in turn the risk of infection. However, the 4-g MgSO4 dose used in the PREMAG trial failed to stop preterm labour and was unable to significantly prolong the interval from treatment onset to birth. Moreover, the enhancement of proinflammatory cytokines associated with magnesium deficiency, reported for animal models, is at odds with a potential effect of MgSO4 favouring PPROM and maternal–fetal infection.30,31 Finally, the higher rates of PPROM and maternal–fetal infection observed in our MgSO4 group may be a spurious finding. Further adjustment for PPROM and maternal–fetal infection resulted in little change in comparisons between the MgSO4 and placebo groups (data not shown). When additional interaction terms between treatment group and PPROM or maternal–fetal infection were introduced, no significant improvements in the prediction of the primary outcomes were obtained, but more pronounced effects of MgSO4 were intimated. For instance, the primary outcomes were lower by an average of 56% (OR: 0.44; 95% CI 0.12–1.58) for neonatal mortality, 31% (OR: 0.69; 95% CI 0.23–2.08) for severe WMI and 35% (OR: 0.65; 95% CI 0.25–1.72) for the combined outcome measure in the case of maternal–fetal infection.

The following other reasons might have contributed to the lack of significant results. First, the timing of the MgSO4 infusion may have been inappropriate with respect to the initiation of the deleterious cascade of events leading to WMI. Second, the MgSO4 dose may have been too low because no maintenance infusion was used. However, transplacental passage of MgSO4 was documented by the significantly different blood magnesium concentrations in placebo- and MgSO4-treated newborns (data not shown). Third, MgSO4 may not be able to counteract all pathways leading to WMI. Indeed, WMI seems to be of multifactorial in origin, involving both preconceptional maternal factors, such as genetic polymorphisms, nutritional deficiency, or exposure to toxic agents, and perinatal factors, such as inflammatory processes, ischaemia–hypoxia, maternal and/or placental hormonal and growth factor deficiencies due to premature delivery, postnatal hypoxia or severe hypocapnia, nutritional deficiency, or adverse effects of postnatal treatment of neonates.3,5 The respective contributions of these various factors is unknown.

The absence of more frequent major maternal, fetal or infant adverse effects with MgSO4 in the PREMAG, as in the Australasian ACTOMgSO4 trial, is in sharp contrast with the alarming report on the MAGNET randomised trial showing apparently higher infant mortality rates for very-preterm newborns that had received prenatal MgSO4 treatment.20 This difference may be a consequence of the higher (tocolytic) dose of MgSO4 used in the MAGNET trial and to some methodological shortcomings of that study.32–35

The PREMAG trial shares many similarities with the results of the ACTOMgSO4 trial published in 2003.13 However, several notable differences merit being mentioned: in particular, the ACTOMgSO4 trial used different primary outcomes (infant mortality over 2 years, cerebral palsy rates at 2 years, and the two outcomes combined) and had larger sample (1062 women randomised and 1255 infants analysed). Hence, it is noteworthy that results of that trial and ours are qualitatively and quantitatively similar, notably showing a nonsignificant trend in favour of magnesium on all three primary outcomes, with 17% lower relative risks for all primary outcomes in the ACTOMgSO4 trial and 14–22% lower ORs in the PREMAG trial. When completed, the continuing follow up of the PREMAG trial will enable further comparisons with the ACTOMgSO4 trial regarding cerebral palsy and paediatric mortality at 2 years.

In conclusion, our findings suggest a neuroprotective effect of MgSO4 given before very-preterm birth but do not provide strong enough evidence for recommending widespread MgSO4 use in clinical practice. Stronger evidence might be gained by jointly analysing the data from ACTOMgSO4 and PREMAG trials and the randomised trial currently in progress in the USA. Another worthwhile undertaking would be to devise preventive approaches combining MgSO4 with other candidate molecules so as to block more pathways that could potentially lead to WMI and secondary brain-development disorders.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Funding
  9. References
  10. Appendix

We are grateful to Ms Nathalie Turbet-Deloff and Ms Véronique Chambaretaud for their administrative support. The authors thank Ms Janet Jacobson for editorial assistance. We are indebted to the women and their children who participated in this study.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Funding
  9. References
  10. Appendix

This research was funded by a 3-year grant, ‘Programme Hospitalier de Recherche Clinique’, from the French Department of Health obtained in 1997 and a grant from Rouen University Hospital obtained in 1997.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Funding
  9. References
  10. Appendix
  • 1
    Marlow N. Neurocognitive outcome after very preterm birth. Arch Dis Child Fetal Neonatal Ed 2004;89:F2248.
  • 2
    Livinec F, Ancel PY, Marret S, Arnaud C, Fresson J, Pierrat V, et al. Prenatal risk factors for cerebral palsy in very preterm singletons and twins. Obstet Gynecol 2005;105:13417.
  • 3
    Marret S, Marpeau L. Grand prematurity, risk of neuropsychic handicaps and neuroprotection. J Gynecol Obstet Biol Reprod (Paris) 2000;29:37384.
  • 4
    Kuban KC, Allred EN, Dammann O, Pagano A, Leviton A, Share J, et al. Topography of cerebral white-matter disease of prematurity studied prospectively in 1607 very-low-birthweight infants. J Child Neurol 2001;16:4018.
  • 5
    Larroque B, Marret S, Ancel PY, Arnaud C, Marpeau L, Supernant K, et al. White matter damage and intraventricular hemorrhage in very preterm infants: the EPIPAGE study. J Pediatr 2003;143:47783.
  • 6
    Nelson KB, Grether JK. Can magnesium sulphate reduce the risk of cerebral palsy in very low birth weight infants? Pediatrics 1995;95:2639.
  • 7
    Hauth JC, Goldenberg RL, Nelson KG, DuBard MB, Peralta MA, Gaudier FL. Reduction of cerebral palsy with maternal MgSO4 treatment in newborns weighing 500–1000 g. Am J Obstet Gynecol 1995;172:419.
  • 8
    Schendel DE, Berg CJ, Yeargin-Allsopp M, Boyle CA, Decoufle P. Prenatal magnesium sulphate exposure and the risk of cerebral palsy or mental retardation among very-low-birth-weight children aged 3 to 5 years. JAMA 1996;276:180510.
  • 9
    FineSmith RB, Roche K, Yellin PB, Walsh KK, Shen C, Zeglis M, et al. Effect of magnesium sulphate on the development of cystic periventricular leukomalacia in preterm infants. Am J Perinatol 1997;14:3037.
  • 10
    Rouse DJ, Hauth JC, Nelson KG, Goldenberg RL. The feasibility of a randomised clinical perinatal trial: maternal magnesium sulphate for the prevention of cerebral palsy. Am J Obstet Gynecol 1996;175:7015.
  • 11
    Canterino JC, Verma UL, Visintainer PF, Figueroa P, Klein SA, Tejani NA. Maternal magnesium sulphate and the development of neonatal periventricular leucomalacia and intraventricular hemorrhage. Obstet Gynecol 1999;93:396402.
  • 12
    O’Shea TM, Klinepeter KI, Dillard RG. Prenatal events and the risk of cerebral palsy in very low birth weight infants. Am J Epidemiol 1998;147:3629.
  • 13
    Crowther CA, Hiller JE, Doyle LW, Haslam RR; for the Australasian Collaborative Trial of Magnesium Sulphate Collaborative Group (ACTOMgSO4). Effect of magnesium sulphate given for neuroprotection before preterm birth: a randomised controlled trial. JAMA 2003;290:266976.
  • 14
    Zupan V, Gonzales P, Lacaze-Masmonteil T, Boithias C, D’Allest AM, Dehan M, et al. Periventricular leukomalacia: risk factors revisited. Dev Med Child Neurol 1996;38:10617.
  • 15
    Murphy DJ, Sellers S, MacKenzie IZ, Yudkin PL, Johnson AM. Case-control study of antenatal and intrapartum risk factors for cerebral palsy in very preterm singleton babies. Lancet 1995;346:144954.
  • 16
    Ancel PY, Marret S, Larroque B, Arnaud C, Zupan-Simunek V, Voyer M, et al. Are small for gestational age and maternal hypertension risk factors for severe intraventricular hemorrhage and cystic periventricular leucomalacia compared with other causes of very preterm birth? Results of the EPIPAGE cohort study. Am J Obstet Gynecol 2005;193:17884.
  • 17
    Reynolds PR, Dale RC, Cowan FM. Neonatal cranial ultrasound interpretation: a clinical audit. Arch Dis Child Fetal Neonatal Ed 2001;84:F925.
  • 18
    Ment LR, Vohr B, Allan W, Westerveld M, Katz KH, Schneider KC, et al. The etiology and outcome of cerebral ventriculomegaly at term in very low birth weight preterm infants. Pediatrics 1999;104:2438.
  • 19
    Hosmer DW, Lemeshow S. Applied Logistic Regression, 2nd edn. New York, NY: Wiley & Sons, 2000.
  • 20
    Mittendorf R, Pryde P, Khoshnood B, Kwang-Sun L. If tocolytic magnesium sulphate is associated with excess total pediatric mortality, what is its impact? Obstet Gynecol 1998;92:30811.
  • 21
    Courtney SE, Durand DJ, Asselin JM, Hudak ML, Aschner JC, Shoemaker CT, et al. High-frequency oscillatory ventilation versus conventional mechanical ventilation for very-low-birth-weight infants. N Engl J Med 2002;347:64752.
  • 22
    Inder TE, Scott J, Mogridge NB, Spencer C, Volpe JJ. Defining the nature of the cerebral abnormalities in the premature infant: a quantitative magnetic resonance imaging study. J Pediatr 2003;143:1719.
  • 23
    Mildvan A. Role of magnesium and other divalent cations in ATP-utilizing enzymes. Magnesium 1987;6:2833.
  • 24
    McIntosh TK, Vink R, Yamakami I, Faden AI. Magnesium protects against neurological deficit after brain injury. Brain Res 1989;482:25260.
  • 25
    McDonald JW, Johnston MV. Magnesium reduces N-methyl-D-aspartate-mediated brain injury in perinatal rats. Neurosci Lett 1990;109:2348.
  • 26
    Thordstein M, Bagenholm R, Thiringer K, Kjellmer I. Scavengers of free oxygen radicals in combination with magnesium ameliorate perinatal hypoxic–ischemic brain damage in the rat. Pediatr Res 1993;34:238.
  • 27
    Marret S, Gressens P, Gadisseux JF, Evrard P. Prevention by magnesium of excitotoxic neuronal death in the developing brain: an animal model for clinical intervention studies. Dev Med Child Neurol 1995;37:47384.
  • 28
    Wu YW, Colford JM Jr. Chorioamnionitis as a risk factor for cerebral palsy: a meta-analysis. JAMA 2000;284:141724.
  • 29
    Atkinson MW, Guinn D, Owen J, Hauth JC. Does magnesium sulphate affect the length of labor induction in women with pregnancy-associated hypertension? Am J Obstet Gynecol 1995;173:121922.
  • 30
    Shogi T, Miyamato A, Ishiguro S, Nishio A. Enhanced release of IL-1beta and TNF-alpha following endotoxin challenge from rat alveolar macrophages cultured in low-Mg(2+) medium. Magnes Res 2003;16:11119.
  • 31
    Maier JA, Malpuech-Brugere C, Zimowska W, Rayssiguier Y, Mazur A. Low magnesium promotes endothelial cell dysfunction: implications for atherosclerosis, inflammation and thrombosis. Biochem Biophys Acta 2004;1689:612.
  • 32
    Benichou J, Zupan V, Fernandez H, Marpeau L, Marret S. Tocolytic magnesium sulphate and paediatric mortality. Lancet 1998;351:2901.
  • 33
    Crowther C, Hiller J, Doyle L, Lumley J, Carlin J. Tocolytic magnesium sulphate and paediatric mortality. Lancet 1998;351:291.
  • 34
    Leveno KJ. Tocolytic magnesium sulphate and paediatric mortality. Lancet 1998;351:2912.
  • 35
    Grether J, Hirtz D, McNellis D, Nelson K, Rouse DJ. Tocolytic magnesium sulphate and paediatric mortality. Lancet 1998;351:292.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Funding
  9. References
  10. Appendix
PREMAG trial group
Steering committee

G Moriette (Port-Royal University Hospital, Paris) (chair), J Barrat (retired obstetrician), P Truffert (Lille University Hospital), E Jacz-Aigrain (Robert-Debré University Hospital, Paris) and J Seebacher (Pitié-Salepêtrière University Hospital, Paris).

Investigators (hospital, number of patients included)

C Savagnier, P Deschamps, C Monrigal, S Le Bouedec and P Gillard (Angers University Hospital, 10); H Fernandez, V Zupan, D Benhamou and J de Laveaucoupet (Antoine-Béclère University Hospital, Clamart, 90); A Burguet, JP Schaal, S Haïfi, A Menget, R Maillet and C Clément (Besançon University Hospital, 5); JN Leng, J Herovitz, D Dallay, C Billeaud and R Torreli (Bordeaux University Hospital, 16); B Guillois, M Dreyfus, A Mayaud, A Girard, G Kobilinsky and G Muller (Caen University Hospital, 25); B Delaporte, H Bruel, F Kilani, G Schweitzer, R Walch, L Vercoustre and JY Col (Le Havre General Hospital, 42); V Lacroze, U Simeoni, V Millet, C Gire, C d’Ercole, D Cohen, C Gamène, L Boubli, C Chau, JM Garnier, D Ortega, N Girard, R Bernard and P Devred (Marseille University Hospitals, 16); G Cambonnie, P Boulot, C Doré and JC Picaud (Montpellier University Hospital, 59); M Hoffet, H Daude, R Mangin and D Amram (Nimes University Hospital, 15); S Couderc, P Rozenberg, R Lenclen, A Paupe, C Fisher, P Narcy, B Guyot, P Sinda and N Tabary (Poissy General Hospital, 9); C Follet-Bouhamed, G Magnin, D Oriot, R Sarfati and J Demendion (Poitiers University Hospital, 64); D Pinquier, B Rachet and A Gravier (Rouen University Hospital Rouen, 163); D Astruc, E Boudier, F Schon and U Siméoni (Strasbourg University Hospital, 50).

Coordinating centre’s pharmacy and preparation of treatment packs

N Donnadieu (Rouen University Hospital).