Maternal adverse effects of different antenatal magnesium sulphate regimens for improving maternal and infant outcomes: a systematic review

Background Antenatal magnesium sulphate, widely used in obstetrics to improve maternal and infant outcomes, may be associated with adverse effects for the mother sufficient for treatment cessation. This systematic review aimed to quantify maternal adverse effects attributed to treatment, assess how adverse effects vary according to different regimens, and explore women’s experiences with this treatment. Methods Bibliographic databases were searched from their inceptions to July 2012 for studies of any design that reported on maternal adverse effects associated with antenatal magnesium sulphate given to improve maternal or infant outcomes. Primary outcomes were life-threatening adverse effects of treatment (death, cardiac arrest, respiratory arrest). For randomised controlled trials, data were meta-analysed, and risk ratios (RR) pooled using fixed-effects or random-effects models. For non-randomised studies, data were tabulated by design, and presented as RR, odds ratios or percentages, and summarised narratively. Results A total of 143 publications were included (21 randomised trials, 15 non-randomised comparative studies, 32 case series and 75 reports of individual cases), of mixed methodological quality. Compared with placebo or no treatment, magnesium sulphate was not associated with an increased risk of maternal death, cardiac arrest or respiratory arrest. Magnesium sulphate significantly increased the risk of 'any adverse effects’ overall (RR 4.62, 95% CI 2.42-8.83; 4 trials, 13,322 women), and treatment cessation due to adverse effects (RR 2.77; 95% CI 2.32-3.30; 5 trials, 13,666 women). Few subgroup differences were observed (between indications for use and treatment regimens). In one trial, a lower dose regimen (2 g/3 hours) compared with a higher dose regimen (5 g/4 hours) significantly reduced treatment cessation (RR 0.05; 95% CI 0.01-0.39, 126 women). Adverse effect estimates from studies of other designs largely supported data from randomised trials. Case reports supported an association between iatrogenic overdose of magnesium sulphate and life-threatening consequences. Conclusions Appropriate administration of antenatal magnesium sulphate was not shown to be associated with serious maternal adverse effects, though an increase in 'minor’ adverse effects and treatment cessation was shown. Larger trials are needed to determine optimal regimens, achieving maximal effectiveness with minimal adverse effects, for each antenatal indication for use. Vigilance in the use of magnesium sulphate is essential for women’s safety.

Death; given calcium gluconate; absent tendon reflexes; respiratory depression; caesarean Sequence generation: unclear risk of bias (not detailed). Allocation concealment: high risk of bias (by sealed opaque envelopes (cards marked A or B), cards not envelopes numbered consecutively; envelopes distributed in batches of 20, with equal A and B). Blinding: low risk of bias (personnel and women blinded). Incomplete outcome data/losses: unclear risk of bias (123 envelopes/data sheets could not be found; 14 additional women excluded; 83% of women included in analyses). Selective reporting: unclear risk of bias (outcomes not clearly pre-specified). Cox 1990 [3] USA, 1 hospital [1987][1988][1989] 156 women. Inclusions: PTL; 24-34 weeks GA; intact fetal membranes. Exclusions: ROM or maternal/fetal complications requiring delivery.
Cessation of therapy due AE; respiratory depression; systemic hypotension; generalised muscle weakness; caesarean Sequence generation: unclear (random number table). Allocation concealment: low risk of bias (consecutive numbered, sealed, envelopes). Blinding: high risk of bias (not detailed; Mg concentration measurement for intervention group suggests no blinding). Incomplete outcome data/losses: low risk of bias (outcomes reported for all neonates/mothers). Selective reporting: unclear risk of bias (outcomes not clearly pre-specified). Crowther 2003 [4] AUSTRALIA AND NEW ZEALAND, 16 hospitals 1996ZEALAND, 16 hospitals -2000 1062 women. Inclusions: < 30 weeks GA, singleton or higher order pregnancy, delivery expected < 24 hours. Exclusions: 2 nd stage of labour, received MgSO4 in this pregnancy, contraindications to MgSO4.
Intervention group (n=535): 4 g IV LD over 20 mins; 1 g/hour MD until birth or for 24 hours, which ever occurred first. Control group (n=527): equal volume of placebo solution (isotonic sodium chloride solution 9%).
Death; cardiac arrest; respiratory arrest; cessation of infusion due to AE; respiratory depression; hypotension; tachycardia; any AE; warmth over body; nausea; sleepiness; mouth dryness; dizziness; sweating; blurred vision; arm discomfort; caesarean; PPH; ICU admission Sequence generation: low risk of bias (central computer-generated randomisation sequence). Allocation concealment: low risk of bias (central telephone randomisation). Blinding: low risk of bias (women, caregivers and outcome assessors blinded). Incomplete outcome data/losses: low risk of bias (hospitalisation outcomes given for all women/infants). Selective reporting: low risk of bias (outcome pre-specified; no indication of selective reporting). Intervention group (n=5055): 4 g IV LD over 10-15 mins; then 1 g/hour IV or 10 g IM (5 g each buttock) with the initial LD, followed by 5 g IM MD every 4 hours; for 24 hours. Control group (n=5055): a placebo by identical regimen.
Death; cardiac arrest; respiratory arrest; cessation of therapy due to AE; given calcium gluconate; respiratory depression; absent/reduced reflexes; hypotension, palpitations or tachycardia; any AE; flushing; nausea and or vomiting; muscle weakness; drowsiness/confusion; headache; thirst; dizziness; Sequence generation: low risk of bias (central computer-generated randomisation sequence). Allocation concealment: low risk of bias (central pharmacy randomisation or consecutively numbered, sealed treatment packs). Blinding: low risk of bias (women, caregivers and outcome assessors blinded). Incomplete outcome data/losses: low risk of bias (5 women were excluded (2 in each group -no data; 1 in intervention group -wrong trial); follow up data available for 99.7% of women randomised before delivery and 98.6% of babies). Selective reporting: low risk of bias (outcomes prespecified; no indication of selective reporting). Intervention group (n=109): 6 g IV LD over 20 mins; 2 g/hour MD, until 12 hours postpartum or a total of 12 hours for those randomised after delivery. Control group (n=113): identical saline placebo.
Toxicity (absent tendon reflexes/respiratory depression); pulmonary oedema; caesarean Sequence generation: low risk of bias (central computer-generated simple randomisation sequence). Allocation concealment: low risk of bias (consecutively numbered, sealed opaque envelopes). Blinding: low risk of bias (investigators, women and outcome assessors blinded; if women developed severe PE after randomisation, group assignment revealed). Incomplete outcome data/losses: low risk of bias (no losses reported). Selective reporting: unclear risk of bias (outcomes not clearly pre-specified Intervention group (n=67): 6 g IV LD over 15-20 mins; 2 g/hour MD infusion continued until 12 hours postpartum. Control group (n=68): saline placebo solution by an identical regimen. Cessation of therapy due to AE; feeling warm and flushed; lethargy; slurred speech; caesarean; PPH Sequence generation: low risk of bias (computer generated table of random numbers). Allocation concealment: low risk of bias (sequentially numbered, sealed, opaque envelopes). Blinding: low risk of bias (women and caregivers blinded). Incomplete outcome data/losses: low risk of bias (no losses reported). Selective reporting: low risk of bias (outcomes pre-specified; no indication of selective reporting). Intervention group (n=72): 4 g IV LD over 15-20 mins; 2 g MD every 3 hours by IM or slow IV. 2 g IM/slow IV for recurrent convulsions. Control group (n=54): (Pritchard's regimen) 4 g IV LD slowly, with 10 g IM; 5 g IM every 4 hours in alternate buttocks. Both groups: for 24 hours after last convulsion or delivery, whichever was later.
Death due to toxicity; stopped dosage due to toxicity; skipped doses due to toxicity; respiratory depression; absent knee jerk; gluteal abscess; pulmonary oedema; PPH; acute respiratory failure Sequence generation and allocation concealment: unclear risk of bias (not detailed). Blinding: high risk of bias (not detailed -unlikely for personnel and women in view of the intervention; unclear for outcome assessors). Incomplete outcome data/losses: low risk of bias (outcomes complete). Selective reporting: unclear risk of bias (outcomes not clearly pre-specified). Unbalanced randomisation *Intervention group: 37 (51%) women with imminent eclampsia, 35 (49%) with eclampsia. Control group: 16 (30%) women with imminent eclampsia, 38 (70%) women with eclampsia. Intervention group (n=25): (Low-dose Dhaka regimen) 4 g IV and 6 g IM LD; 2.5 g IM every 4 hours for 24 hours. Control group (n=25): (High-dose regimen) 4 g IV and 8 g IM LD; 4 g IM every 4 hours for 24 hours. Additional 2 g given IV for recurrent seizures. Requirement for calcium gluconate; deferred doses due to AE; absent knee jerk reflex Sequence generation: low risk of bias (patients randomised using a 'Tippet table'). Allocation concealment: unclear risk of bias (not detailed). Blinding: high risk of bias (not detailed -unlikely for personnel and women in view of the intervention; unclear for outcome assessors). Incomplete outcome data/losses: low risk of bias (no losses to follow up detailed). Selective reporting: unclear risk of bias (outcomes not clearly pre-specified). Intervention group (n=50): 6 g IV LD; 2 g/hour MD -increased by 1 g/hour hourly until successful tocolysis or failure (max 4 g/hour). Control group (n=50): 4 g IV over 20 mins; 2 g/hour MD.

Lower dose versus higher dose magnesium sulphate IV maintenance
Cessation of therapy due to AE (severe hypotension, pulmonary oedema, respiratory depression); no AE; flushing; nausea/vomiting; headache; caesarean Sequence generation: low risk of bias (computer generated random number allocation). Allocation concealment: unclear risk of bias (consecutively numbers opaque envelopes -unclear if sealed). Blinding: high risk of bias (not detailed -unlikely for personnel and women in view of the intervention; unclear for outcome assessors). Incomplete outcome data/losses: low risk of bias (outcomes assessed for all women during hospitalisation). Selective reporting: unclear risk of bias (not all outcomes reported were pre-specified).
Intervention group (n=82): 4 g IV LD over 20 mins; 5 g/hour MD. Control group (n=78): 4 g IV LD over 20 mins; 2 g/hour MD. If after 1 hour the patient continued to have contractions, further cervical dilatation or effacement, MD was increased by 1 g/hour (max 6 g/hour). Cessation of therapy due to AE; no AE; flushing; headache; pulmonary oedema; caesarean Sequence generation: low risk of bias (computer generated random number allocation). Allocation concealment: unclear risk of bias (consecutively numbered, opaque envelopes -not detailed if sealed). Blinding: high risk of bias (not detailed -unlikely for personnel and women in view of the intervention; unclear for outcome assessors). Incomplete outcome data/losses: unclear risk of bias (12 women who were delivered due to failed tocolysis were excluded (8 intervention (10.3%); 4 control group (4.9%)); were more likely to develop pulmonary oedema). Selective reporting: unclear risk of bias (outcomes not pre-specified). Intervention group (n=72): 4 g IV slow bolus LD; 0.75 g/hour IV MD. Control group (n=72): (Pritchard's regimen) 4 g IV LD slowly and 10 g IM (5 g in each buttock); 5 g IM every 4 hours in alternate buttocks. Additional 2 g IV given for convulsions.

Magnesium sulphate IV maintenance versus IM maintenance
Death; MgSO4 toxicity; Caesarean; PPH Sequence generation: low risk of bias (computer generated randomisation). Allocation concealment: low risk of bias (sealed, sequentially, numbered, brown envelopes). Blinding: high risk of bias (women/caregivers not blinded due to the nature of the intervention). Incomplete outcome data/losses: unclear risk of bias (5/72 women in intervention, and 2/72 women in control group were excluded from analysis due to defective record-keeping). Selective reporting: low risk of bias (outcomes pre-specified; no evidence of selective reporting Intervention group (n=8): 6 g IV LD over 15 mins; 2 g/hour IV MD for 24 hours or until delivery. Control group (n=9): 4 g IV LD over 15 mins, and 5 g IM each buttock; 5 g in alternate buttocks every 4 hours for 24 hours or until delivery.
Cessation of therapy due to AE; clinical signs of toxicity (oliguria, decreased patellar reflexes); caesarean Sequence generation: unclear risk of bias (not detailed). Allocation concealment: unclear risk of bias (not detailed). Blinding: high risk of bias (not detailed -unlikely for personnel and women in view of the intervention; unclear for outcome assessors). Incomplete outcome data/losses: low risk of bias (outcomes complete (3 intervention and 2 control group missing 19 hour serum albumin and calcium)). Selective reporting: unclear risk of bias (outcomes not clearly pre-specified Sequence generation: unclear risk of bias ('patients randomised based on randomly permuted blocks'). Allocation concealment: unclear risk of bias (investigators given set of sealed envelopes identified by patient's number containing the allocation of the individual patient number to one of two groups). Blinding: high risk of bias (not detailed -unlikely for personnel and women in view of the intervention; unclear for outcome assessors). Incomplete outcome data/losses: low risk of bias (patient questionnaire assessment (20/23 RTU) and (21/23) 89% response). Selective reporting: low risk of bias (outcomes pre-specified -assessment of 'general tolerability' by investigators unclear). Use of patient questionnaire may have influenced the high frequencies of AE. Intervention group (n=101): 2 g/hour for 12 hours after delivery. Control group (n=99): 2 g/hour for 24 hours after delivery. All women received a 4 g IV LD followed by 2 g/hour maintenance prior to delivery.

MgSO4 toxicity; intolerance
Sequence generation: low risk of bias (computer-generated random number table in blocks of 10). Allocation concealment: low risk of bias (consecutively numbered, sealed, opaque envelopes). Blinding: high risk of bias (not detailed -unlikely for personnel and women in view of the intervention; unclear for outcome assessors). Incomplete outcome data/losses: low risk of bias (4 women (2%) lost from the control group). Selective reporting: unclear risk of bias (outcomes not clearly pre-specified; AE not clearly reported Padhar regimens. Group 1 (n=37): 2 g IV and 4 g IM LD; 4 g IM every 4 hours. Group 2 (n=58): 2 g IV and 8 g IM LD; 4 g IM every 4 hours. Both groups: for 24 hours after delivery or last convulsion which ever was later.
Respiratory depression (abnormal RR); absent knee jerks, and MD omitted Sequence generation and allocation concealment: high risk of bias (not randomised Group 1: women who came directly to the hospital; Group 2: women who had received diazepam/Phenergan at a referring hospital). Blinding: high risk of bias (not blinded). Losses to follow up: low risk of bias (no losses or exclusions; outcome data complete). Selective reporting: low risk of bias (outcomes pre-specified; no evidence of selective reporting Group 1 (n=97): (Intravenous push) 10 g IM LD; 2 g slow intravenous 'push' every 1-2 hours (mean: 30 g). Group 2 (n=47): 10 g IM LD; continuous MD 1 g/hour (mean: 46 g) Death; heat and flushing; respiratory effects Sequence generation and allocation concealment: high risk of bias (not randomised). Blinding: high risk of bias (not detailed and unlikely in view of the intervention). Losses to follow up: low risk of bias (no losses detailed). Selective reporting: unclear risk of bias (outcomes/AE not clearly reported). Groups reported to be 'clinically comparable' but this was unclear. Ales 1987 [26] USA 1981 RC 393 women with hypertension complicating pregnancy, with a focus on 178 women whose hypertension was first documented in labour. MgSO4 (n= 64) v no MgSO4 (n=114).

Not detailed
Caesarean; caesarean due to failure to progress Selection: low risk of bias (representative cohort, selected from all women who delivered during Jan-December 1981; non-exposed women selected from this same cohort; exposure ascertained from medical records). Comparability: unclear risk of bias (study controls for potential confounders in logistic regression analysis. Authors acknowledge the degree of cervical dilatation may have played a role in the failure of labour -this could not be accurately assessed; biases/practices of individual physicians represent another confounder). Outcome: unclear risk of bias (outcomes assessed by reference to medical records, no detail of whether this was independent blind assessment). Assaley 1998 [27] USA BAS 18 women with PE. MgSO4 (n=15) v no MgSO4 (n=3). 4 g IV LD; MD infusion of 1-2 g/hour, until 12-24 hours postpartum (81.3 g mean total, 52.5-145 g range).
Prolonged bleeding time Selection: unclear risk of bias (unclear). Comparability: unclear risk of bias (no baseline comparison of groups, or any control for potential confounders for bleeding time). Outcome: low risk of bias (objective outcomes unlikely to be affected by lack of blinding). Kynczl-Leisure 1996 [28] USA BAS 12 women with PE. MgSO4 (n=9) v no MgSO4 (n=3).
'standard convulsion prophylaxis treatment with magnesium sulfate' (15-40 hours duration) Prolonged bleeding time Selection: unclear risk of bias (unclear). Comparability: unclear risk of bias (no detail or control for potential confounding factors -treated women had lower bleeding time at baseline).Outcome: high risk of bias (whilst objective outcomes unlikely to be affected by lack of blinding, bleeding time measured at different times for 2 groups; 1 woman from exposed group was excluded as post-partum measurement was taken 5 hours after discontinuation -no change in bleeding time was noted for this patient).

Magee 2005 [29]
USA 1997-2001 RC 377 women who received IV MgSO4 who had a hypertensive disorder of pregnancy but no PTL (obtained from ICD-9 computerised database). Group 1 (n=162): received nifedipine as antihypertensive agent v Group 2 (n=32): received another antihypertensive v Group 3 (n=183): received no antihypertensive. 4 g IV bolus LD, followed by infusion of 2 g/hour Neuromuscular weakness (absent DTR, weakness, respiratory depression, neuromuscular blockage, calcium gluconate given); maternal hypotension; nausea/vomiting; drowsiness/confusion; dizziness; flushing; thirst, respiratory problems; dyspnoea; pulmonary oedema; maternal tachycardia; infusion rate decreased due AE; infusion rate stopped early due to AE Selection: low risk of bias (all women (cohort) drawn from same time period, and exposure ascertained from medical records). Comparability: high risk of bias (study does not control for potential confounding factors identified on baseline comparisons (neurological disorders, severe hypertension)). Outcome: low risk of bias (outcomes (AE) assessed/abstracted independently by 2 reviewers, with complete outcomes for all subjects (however limited by what was in case records)). Policy: 4 g IV LD over 20 mins; 2 g/hour MD which could be incrementally increased to achieve uterine quiescence for up to 4 g/hour. Visual disturbances; ileus; osteopenia; chest tightness; pulmonary oedema; vulvar oedema; hypocalcaemia; ≥ 1 AE; discontinuation of therapy due to AE Selection: low risk of bias (women from both groups drawn from same population/period; exposure obtained from secure medical records). Comparability: unclear risk of bias (study does not control for baseline imbalances -women in Group 1 more likely to have multiple gestations, be of an earlier GA). Outcomes: unclear risk of bias (2 authors collected data on a standard sheet to ensure uniformity from maternity record -i.e. not independent blind assessment of outcomes; a bias in reporting/checking for maternal AE on prolonged treatment is also possible) Phlebitis; signs/symptoms of toxicity; calcium gluconate used; errors associated with failure to reset pump after LD; errors associated with a change to the order Selection: low risk of bias (consecutive women in study period included (list generated from BC Perinatal Database Registry of obstetric patients)). Comparability: high risk of bias (no control for group imbalances -baseline characteristics, and additional aspects of the treatment received). Outcome: unclear risk of bias (Unclear who assessed outcomes in charts and whether this could be blinded; inconsistent documentation in the charts noted