Populations

Nine of the 25 included studies involved adults exclusively (Nannini 2000; Abreu‐Gonzalez 2002; Bessmertny 2002; Hughes 2003; Kokturk 2005; Gaur 2008; Gallegos‐Solórzano 2010; Goodacre 2013; Hossein 2016); and four included adults and children (Mangat 1998; Aggarwal 2006; Neki 2006; Sarhan 2016). Eight studies enrolled children (Meral 1996; Mahajan 2004; Ashtekar 2008; Khashabi 2008; Powell 2013; Mohammedzadeh 2014; Alansari 2015; Turker 2017); and in the remaining four studies the age of participants was not stated (Dadhich 2005; Drobina 2006; Ahmed 2013; Badawy 2014).

The severity of disease varied between studies (Table 1). Fourteen studies enrolled patients based on specific lung function criteria (Meral 1996; Mangat 1998; Nannini 2000; Bessmertny 2002; Hughes 2003; Mahajan 2004; Dadhich 2005; Neki 2006; Gaur 2008; Gallegos‐Solórzano 2010; Goodacre 2013; Powell 2013; Alansari 2015; Hossein 2016), while the remaining studies enrolled patients previously diagnosed with asthma using accepted clinical standards, or did not specify how asthma was diagnosed. Based on the baseline demographic data, 15 studies were considered to enrol severe acute exacerbations of asthma (FEV1 or PEF < 50% predicted at baseline or symptom criteria defined by BTS/SIGN guideline) (Meral 1996; Mangat 1998; Nannini 2000; Bessmertny 2002; Hughes 2003; Mahajan 2004; Dadhich 2005; Kokturk 2005; Aggarwal 2006; Neki 2006; Ashtekar 2008; Gaur 2008; Gallegos‐Solórzano 2010; Goodacre 2013; Powell 2013).

Sixteen studies recruited participants from emergency departments; two from outpatient or emergency departments (Badawy 2014; Sarhan 2016); and one in a children's assessment unit after general practitioner referral (Ashtekar 2008). Department of presentation was unclear in the remaining six studies (see Table 2).

Badawy 2014 recruited exclusively pregnant women. Due to concerns about baseline imbalance in this study, and the narrow population recruited, we did not include this study in our meta‐analyses and instead present the results narratively. The study has been included in another Cochrane Review that addresses asthma treatment options in pregnant women (Bain 2014).

Participants were excluded for a number of reasons including pre‐existing lung conditions and features of infection on examination. There was great variation in pharmaceutical exclusion due to drugs taken before recruitment (see Table 2).

Interventions

All studies used nebulised MgSO₄ in the intervention group but the comparison and placebo nebulised solutions varied (Table 3). Three studies compared MgSO₄ with β₂‐agonist directly with no placebo (Meral 1996; Mangat 1998; Neki 2006). Twelve studies compared β₂‐agonist with MgSO₄ to β₂‐agonist with placebo (normal saline) (Nannini 2000; Abreu‐Gonzalez 2002; Bessmertny 2002; Hughes 2003; Mahajan 2004; Kokturk 2005; Aggarwal 2006; Khashabi 2008; Ahmed 2013; Badawy 2014; Mohammedzadeh 2014; Turker 2017) . Five studies compared β₂‐agonist and ipratropium with MgSO₄ to β₂‐agonist and ipratropium with placebo (Ashtekar 2008; Gaur 2008; Gallegos‐Solórzano 2010; Powell 2013; Hossein 2016), and Drobina 2006 compared β₂‐agonist and ipratropium with MgSO₄ to β₂‐agonist and ipratropium only (i.e. no placebo). Alansari 2015 compared β₂‐agonist with MgSO₄ to β₂‐agonist with placebo (normal saline) after both groups had received one hour of therapy with combined β₂‐agonist and ipratropium, and thus is included in comparison one. Two studies had three groups and investigated MgSO₄ versus β₂‐agonist versus MgSO₄ plus β₂‐agonist (Dadhich 2005 and Sarhan 2016) and thus appear in comparisons 2 and 3. Goodacre 2013 studied one group with nebulised MgSO₄, β₂‐agonist, ipratropium and IV placebo, one with IV MgSO₄, β₂‐agonist, ipratropium and nebulised placebo, and a third group with β₂‐agonist, ipratropium and both nebulised and IV placebo. The comparison involving IV MgSO₄ has been covered in other reviews (Kew 2014; Griffiths 2016).

Most studies used 0.9% normal saline as placebo; Aggarwal 2006 used distilled water (as well as normal saline) for placebo and Abreu‐Gonzalez 2002 used ‘physiological serum’ as placebo.

We identified the following comparisons which have been used throughout the review to lend structure.

• MgSO₄ and β₂‐agonist and ipratropium versus placebo (saline) and β₂‐agonist and ipratropium (seven studies: Drobina 2006; Ashtekar 2008; Gaur 2008; Gallegos‐Solórzano 2010; Goodacre 2013; Powell 2013; Hossein 2016).

• MgSO₄ with β₂‐agonist versus placebo (saline) and β₂‐agonist (15 studies: Nannini 2000; Abreu‐Gonzalez 2002; Bessmertny 2002; Hughes 2003; Mahajan 2004; Dadhich 2005; Kokturk 2005; Aggarwal 2006; Khashabi 2008; Ahmed 2013; Badawy 2014; Mohammedzadeh 2014; Alansari 2015; Sarhan 2016; Turker 2017).

• MgSO₄ versus β₂‐agonist alone (five studies: Meral 1996; Mangat 1998; Dadhich 2005; Neki 2006; Sarhan 2016).

Dose, formulation and dose frequency of MgSO₄ differed, meaning that the overall dose of MgSO₄ given differed between studies (Table 3). Not all studies reported the concentration of MgSO₄ nebulised, but when the information was available most included studies used MgSO₄ of similar concentration and osmolality. However, dose per nebulisation and the number of nebulisations performed varied.

Ten studies nebulised three doses of MgSO₄ at 20 minutes intervals (Bessmertny 2002; Hughes 2003; Aggarwal 2006; Ashtekar 2008; Gallegos‐Solórzano 2010; Goodacre 2013; Powell 2013; Mohammedzadeh 2014; Alansari 2015; Turker 2017). Three studies nebulised four doses at 20 minute intervals (Mangat 1998; Neki 2006; Sarhan 2016). Kokturk 2005 nebulised hourly up to four hours after the initial treatment of three doses in one hour. Five studies nebulised only one treatment (Meral 1996; Nannini 2000; Abreu‐Gonzalez 2002; Mahajan 2004; Ahmed 2013). Khashabi 2008 gave two doses of treatment but the timing was unclear. Three studies were unclear how frequently the doses were given but probably only one dose was given (Dadhich 2005; Drobina 2006; Gaur 2008). One study gave up to three doses at 20 minute intervals (Badawy 2014). Hossein 2016 gave treatments every 20 to 60 minutes but the total number of doses given was unclear.

All control or placebo interventions were similar in appearance to the treatment drug. The most frequent placebo was saline. One study collected data on participants' ability to distinguish between the treatment and control, and noted no ability to discern (Hughes 2003). Even when not expressly stated, it can reasonably be assumed that the control (placebo) would be similar in appearance to the treatment drug (especially if given in a β₂‐agonist vehicle).

Co‐interventions

Co‐interventions used added complexity and heterogeneity to the review (Table 2). In 11 studies, systemic corticosteroids were administered to all participants, although the timing (before/after nebulised treatment) varied (Mangat 1998; Hughes 2003; Mahajan 2004; Kokturk 2005; Neki 2006; Ashtekar 2008; Gaur 2008; Badawy 2014; Alansari 2015; Sarhan 2016; Turker 2017). In one study, systemic corticosteroids were administered if there was no improvement after the three doses of study treatment (Bessmertny 2002). Overall, 15 studies routinely administered corticosteroids, but in different doses, routes and frequency. In three studies, corticosteroids were administered according to local standard/conventional treatment, or at the clinicians’ discretion (Aggarwal 2006; Goodacre 2013; Powell 2013). Meral 1996 gave no further medication as a co‐intervention. Six studies made no comments on co‐interventions (Nannini 2000; Abreu‐Gonzalez 2002; Dadhich 2005; Khashabi 2008; Ahmed 2013; Mohammedzadeh 2014).

Outcomes

A summary of the outcomes relevant to this review reported in the included studies is given in Table 4.

Ongoing trials and unpublished data

We have identified three ongoing studies relevant to this review (Motamed 2015; Saucedo 2015; Schuh 2016a).

Pulmonary function

Four studies including 1279 participants reported on FEV1 or PEF (Gaur 2008; Gallegos‐Solórzano 2010; Goodacre 2013; Hossein 2016).

Gallegos‐Solórzano 2010 and Gaur 2008, both adult studies, reported a greater %FEV1 in the MgSO₄ and β₂‐agonist and ipratropium group at 90 minutes (MD: 8.57, 95% CI 1.99 to 15.15; participants = 60) and 120 minutes (MD: 2.60, 95% CI 0.25 to 4.95; participants = 60) respectively with the pooled fixed‐effect result favouring the MgSO₄ intervention (MD 3.28, 95% CI 1.06 to 5.49; participants = 120; studies = 2; I² = 64%; Analysis 1.1).

Despite being similar‐sized studies, Gaur 2008 contributed nearly 90% of the weight to the pooled analysis due to reporting much smaller standard deviations. Sensitivity analysis with a random‐effects model results in reduced weighting for that study, and increased the size of the CI such that the lower confidence interval included no difference (MD 4.76, 95% CI −0.86 to 10.39; participants = 120; studies = 2).

Hossein 2016 reported a significantly greater per cent predicted PEF in the MgSO₄ and β₂‐agonist and ipratropium group at 20 minutes (MD 6.90, 95% CI 1.63 to 12.17) but an important between‐group difference was not found when 60 minute data were combined with 60 minute data from the large Goodacre 2013 trial (MD 0.05, 95% CI −2.33 to 2.42; participants = 636; studies = 2; I² = 67%). A random‐effects model substantially increases imprecision, and the effect estimate remains inconclusive. The Goodacre results taken individually did not demonstrate an important between‐group difference (MD −0.30%, 95% CI −2.71% to 2.11%).

Alansari 2015, Ashtekar 2008 and Powell 2013 did not report this outcome. Drobina 2006 reported that "peak flow measurements improved over time in both groups (p < 0.001). The addition of aerosolized magnesium sulfate did not result in a statistically significant increase in either the maximum or the average peak flow over time (p = 0.279 and p = 0.399, respectively)." As this research is only available in abstract form, it is unclear how many participants were in each group and no data were reported to include in the meta‐analysis.

Clinical severity scores

Powell 2013 reported a lower (therefore better) Yung asthma severity score (ASS) in children receiving MgSO₄ and β₂‐agonist and ipratropium compared to placebo (saline) and β₂‐agonist and ipratropium at 60 minutes (MD −0.23, 95% CI −0.48 to 0.02; participants = 472). The minimal important difference on this nine point scale is not known, but in Powell 2013 it was regarded as 0.5 by the trial steering group. Goodacre 2013 reported no significant difference in the change in dyspnoea visual analogue scale between the groups at 60 (MD 3.10, 95% CI −0.53 to 6.73; participants = 658) or 120 minutes (MD 3.12, 95% CI −1.35 to 7.59; participants = 619). The minimal important difference on the 100 mm scale used in Goodacre 2013 is thought to be 22 mm. When the '60 minute' data from each study are combined using an SMD analysis there is no between‐group difference (SMD 0.01, 95% CI −0.11 to 0.12; participants = 1130; studies = 2) but a high level of heterogeneity (I² = 83%). As the studies were of a similar size and weight in this analysis, a random‐effects model has little impact on the effect estimates, although the confidence interval is widened.

Ashtekar 2008 reported that there was no significant difference between the median area under the curve of ASS of the MgSO₄ compared with the placebo‐treated group (1530 versus 1355).

Of note: Powell 2013 performed subgroup analysis for asthma severity score, investigating whether participants with a more severe asthma exacerbation or an exacerbation of shorter duration derived more benefit. Although not powered to detect a difference, the study results support the hypothesis that children with more severe exacerbations and children with shorter duration of symptoms prior to presentation benefit more from inhaled MgSO₄ as measured using the Yung asthma severity score.

Admission to hospital

Gallegos‐Solórzano 2010 reported admissions to the emergency department and the general ward; Goodacre 2013 reported admissions to hospital, HDU and ICU; and Powell 2013 reported admissions to PICU/HDU or intubation.

Pooled results for adults and children for admissions from the emergency department at initial presentation suggests that admissions are decreased in those receiving MgSO₄, β₂‐agonist and ipratropium compared to placebo (saline), β₂‐agonist and ipratropium (RR 0.95, 95% CI 0.91 to 1.00; participants = 1308; studies = 4; I² = 52%) but the upper confidence interval reaches no difference (Analysis 1.4). The overall risk of admission was 82% on placebo which translates into a 78% risk of admission (95% CI 75% to 82%) with nebulised magnesium (Figure 3, summary of findings Table for the main comparison). The RR is the same if a random‐effects model is used, but the result is less precise (RR 0.95, 95% CI 0.87 to 1.05).

Figure 3

---

---

In the control group 82 people out of 100 had hospital admission , compared to 78 (95% CI 75 to 82) out of 100 for the active treatment group.

Gallegos‐Solórzano 2010, an adult study, also reported admission to the emergency department (RR 0.38, 95% CI 0.16 to 0.94; participants = 60).

Goodacre 2013 and Gallegos‐Solórzano 2010, both adult studies, reported on readmission after initial attendance (up to 7 days in Goodacre 2013; time point unclear in Gallegos‐Solórzano 2010). There was no significant difference in admission rates between the two groups, but the result is imprecise (RR 1.80, 95% CI 0.84 to 3.87; participants = 750; studies = 2; I² = 37%, Analysis 1.6).

Goodacre 2013 did not detect a between‐group difference for either HDU admission (RR 1.19, 95% CI 0.66 to 2.13; participants = 690) or ICU (RR 1.94, 95% CI 0.66 to 5.73; participants = 690) but events were infrequent and both results inconclusive (Analysis 1.5). Similarly Powell 2013, a study in children, reported HDU or ICU admissions/intubations and did not detect a between‐group difference (RR 1.48, 95% CI 0.79 to 2.79; participants = 505).

Duration of symptoms

Not reported.

Vital signs

Hossein 2016 and Goodacre 2013 reported on vital signs at various time points. For this update, we chose to present the 60‐minute time point in the meta‐analysis. There are concerns that administration of MgSO₄ may lead to an unwanted drop in blood pressure. Goodacre 2013 reported a significantly higher (therefore better in this context) diastolic blood pressure in participants receiving MgSO₄ and β₂‐agonist and ipratropium compared to placebo and β₂‐agonist and ipratropium at 60 minutes (MD 2.40, 95% CI 0.29 to 4.51; participants = 674), but this difference is unlikely to be clinically meaningful (Analysis 1.10). There was no significant difference in respiratory rate, heart rate or systolic and diastolic blood pressure at any other time point. Similarly there was no significant difference in the change in vital signs between groups at 60 or 120 minutes (Goodacre 2013; data not shown).

Drobina 2006 stated that vital signs were measured in the conference abstract but data were not reported.

Adverse events

Four studies including 2067 participants reported adverse events (Goodacre 2013; Powell 2013; Alansari 2015; Hossein 2016). During admission, more serious adverse events were reported in the placebo group compared to the MgSO₄ group, but only one of the two studies reporting this outcome contributed events and the confidence interval includes no difference (RD −0.03, 95% CI −0.06 to −0.00; participants = 557; studies = 2; I² = 0%; Analysis 1.11). There was no significant difference in adverse events between groups for any adverse event (RD 0.01, 95% CI −0.03 to 0.05; participants = 1197; studies = 2; I² = 0%; Analysis 1.12).

Goodacre 2013 reported a significantly higher 'all adverse event' rate within 30 days of primary attendance for those participants receiving MgSO₄ and β₂‐agonist and ipratropium compared to placebo (saline) and β₂‐agonist and ipratropium (RD 0.06, 95% CI 0.01 to 0.11; participants = 690; Analysis 1.14) but there was no difference in serious adverse events (Analysis 1.13). Hossein 2016 reported no serious adverse events in either group in the first 60 minutes of treatment.

Both Goodacre 2013 and Powell 2013 reported hypotension and flushing but did not detect a significant between‐group difference for either outcome (Analysis 1.15; Analysis 1.16).

Ashtekar 2008 reported that one child had a transiently low blood pressure and another had tingling of the fingers; both received nebulised MgSO₄. Drobina 2006 reported that there were no significant side effects noted in either treatment group, but did not report data. Gaur 2008 did not report adverse effects.

Gallegos‐Solórzano 2010 reported that the most common adverse reaction associated with MgSO₄ was a dry and bitter mouth, but no other side effect was associated with treatment. Electrocardiography (ECG) was abnormal in some participants (43% versus 36%): most commonly, sinus tachycardia (40% versus 36%). One person in the MgSO₄ group developed supraventricular extrasystole that did not require additional management. One from each group reported dizziness.

Pulmonary function

Eleven studies involving 589 participants reported at least one measure of lung function (Nannini 2000; Abreu‐Gonzalez 2002; Bessmertny 2002; Hughes 2003; Mahajan 2004; Kokturk 2005; Aggarwal 2006; Ahmed 2013; Badawy 2014; Mohammedzadeh 2014; Sarhan 2016).

Five studies reported per cent predicted FEV1; two at 120 minutes (Badawy 2014; Sarhan 2016), two at 60 minutes (Bessmertny 2002; Hughes 2003) and one at 20 minutes (Mahajan 2004). Due to our concerns about potential baseline imbalance and the population recruited in Badawy 2014 (all pregnant women) we chose not to include this study in the analysis.

Pulmonary function based on per cent predicted FEV1 was improved in those who received MgSO₄ and a β₂‐agonist compared to β₂‐agonist and placebo, but the confidence interval includes no difference (MD 3.34, 95% CI −1.58 to 8.26; participants = 208; studies = 4), with moderate between‐study heterogeneity identified (I² = 43%). We did not detect a difference between adults and children, but only one study involving children contributed to the analysis. We were only able to present subgrouping based on severity at presentation with three studies as we were unable to characterise severity in Sarhan 2016. Results suggest the combination of MgSO₄ and salbutamol may be more beneficial in people with a more severe exacerbation (FEV1 < 50% predicted) but the test for subgroup difference was negative (I² = 51.8%, P = 0.15). Badawy 2014 reports significantly higher post‐treatment per cent predicted FEV1 in pregnant women receiving MgSO₄ in addition to salbutamol compared to salbutamol alone; end‐point scores were 56.31% (SD 8.25) in the intervention group compared to 32.68% (SD 7.15) (P < 0.001, 30 participants), but baseline FEV1 in each group was not measured.

Six studies reported peak expiratory flow rate (PEF): one at 20 minutes (Nannini 2000), two at 60 minutes (Aggarwal 2006; Mohammedzadeh 2014), two at 120 minutes (Badawy 2014; Sarhan 2016), one at discharge (Kokturk 2005). Where studies reported at more than one time point we have extracted and presented the closest time point to 60 minutes. We excluded Badawy 2014 due to the reasons given above and Kokturk 2005 due to the incompatible time point. We have not pooled adults and children in this analysis. In adults, there was small improvement in PEF compared to the control group, but the confidence interval includes no difference (MD 11.91 L/min, 95% CI −4.12 to 27.95, participants 155, studies 3, I² = 13%, Analysis 2.3). A similar small and imprecise effect was seen in Mohammedzadeh 2014, the study involving children: MD 11.90 L/min, 95% CI −6.86 to 30.66, participants 80. Mohammedzadeh 2014 also reports a change from baseline "adjusted % PEF", which favours the combination of MgSO₄ and salbutamol compared to salbutamol alone (MD 6.70, 95% CI 3.80 to 9.60; data not displayed). Kokturk 2005 reported "both groups displayed comparable improvement in PEF (%) and clinical scores over 120 min" and displayed results graphically. Discharge mean (SD) per cent predicted PEF was similar in both groups (71.18 (11.55) and 70.50 (12.34)).

One study (Ahmed 2013; an abstract) reported "the percentage increase in peak flow" at 10 and 20 minutes, but the lack of clarity about whether this was a per cent change in the absolute values or a change in the per cent predicted meant we could not include data in the analyses. The study reported a greater improvement in the intervention group compared to the control group at both 10 and 20 minutes (20 (SD 4) vs 13 (3) at 10 minutes; and 35 (7) vs 24 (6) at 20 minutes).

Abreu‐Gonzalez 2002 (n = 24), also a conference abstract, selectively reports improvement in PEF at 30 minutes for people in the β₂‐agonist and MgSO₄ group compared to the placebo plus β₂‐agonist group and improvement in FEV1 at 45 minutes for those in the β₂‐agonist alone with placebo group compared to the β₂‐agonist and MgSO₄ group.

Clinical severity score

Four studies involving 575 participants reported clinical severity scores with enough detail for data extraction (Mohammedzadeh 2014; Alansari 2015; Sarhan 2016; Turker 2017), but we did not perform meta‐analysis. Pulmonary index score was significantly lower (therefore better) for those receiving MgSO₄ and β₂‐agonist compared to β₂‐agonist and placebo at 90 minutes in Mohammedzadeh 2014 (MD −0.90, 95% CI −1.43 to −0.37, participants = 80; ) but not at other time points. Sarhan 2016 reported no significant difference in change in Fischl index at two hours between the two groups (MD −0.10, 95% CI −1.22 to 1.02; participants = 20; studies = 1). Alansari 2015 reported PRAM asthma severity score at seven time points (from four to 48 hours). PRAM was significantly lower (therefore better) for those receiving placebo (saline) and β₂‐agonist and ipratropium compared to MgSO₄ and β₂‐agonist and ipratropium at the earliest time point (4 hours; MD −0.4, 95% CI −0.7 to −0.01; P = 0.05; participants = 365;) but not at later time points. Turker 2017 measured the modified pulmonary index score at 20, 40 and 120 minutes post baseline but it was only possible to extract data at 120 minutes; no between‐group difference was identified (MD 0.38, 95% CI −0.25 to 1.01; participants = 100; studies = 1).

Kokturk 2005 reported that both groups displayed comparable improvement in clinical scores over 120 minutes. Khashabi 2008 reported a non‐significant difference in respiratory distress scores, but it is not clear which score was used and the number of participants in each group is not clear. The author was contacted for further information before the 2012 update but no response was received.

Admission to hospital

Six studies involving 375 participants reported admissions for MgSO₄ and β₂‐agonist compared to β₂‐agonist and placebo (Nannini 2000; Hughes 2003; Mahajan 2004; Kokturk 2005; Aggarwal 2006; Turker 2017). Although there was a reduction in the risk of admission for people receiving the MgSO₄ intervention, the confidence interval includes no difference (risk ratio 0.78, 95% CI 0.52 to 1.15; participants = 375; studies = 6; I² = 0%). It should be noted that this analysis is dominated by evidence from adult studies. As we did not detect any statistical heterogeneity in the analysis we did not perform our prespecified subgroup analyses. One study in children reported that one child from each group required admission/re‐admission in the two weeks after discharge (Alansari 2015).

Vital signs

Three studies including 190 participants reported heart rate at 120 minutes (Aggarwal 2006; Badawy 2014; Sarhan 2016). As previously, we have not included Badawy 2014 in the analysis and combining the remaining two studies resulted in such substantial heterogeneity (I² = 82%) that we did not perform meta‐analysis (Analysis 2.5). Participants receiving MgSO₄ and β₂‐agonist had a slightly lower (therefore better) heart rate at 120 minutes compared to those receiving β₂‐agonist and placebo in Aggarwal 2006 (MD −2.70, 95% CI −6.15 to 0.75; participants = 100), but the opposite effect was seen in Sarhan 2016 (MD 22.60, 95% CI 1.61 to 43.59; participants = 20). Badawy 2014 reported a MD of −25.46 (95% CI −28.38 to −22.54) favouring the MgSO₄ group.

There was no significant difference between the two groups in systolic or diastolic blood pressure or respiratory rate in the studies reporting these outcomes.

Adverse events

Serious adverse events were reported by five studies including 243 participants (Nannini 2000; Bessmertny 2002; Hughes 2003; Mahajan 2004; Sarhan 2016). No serious events occurred in any of the studies (Analysis 2.9).

Three studies reported on mild‐moderate adverse events (Nannini 2000; Bessmertny 2002; Aggarwal 2006); and two reported all adverse events (Ahmed 2013; Alansari 2015). There was no significant difference in those experiencing one or more events between the two groups (RD −0.01, 95% CI −0.05 to 0.03; participants = 694; studies = 5; I² = 0%; Analysis 2.10); and no difference detected between adults and children.

Turker 2017 reports narratively that "no side effect caused by magnesium was observed in any of the patients in the study"; this data has not been included in the forest plots.

Aggarwal 2006 reported that tremor was the same in both groups. Khashabi 2008 reported no side effects but it is not clear how many participants were in each group, so their data were not included in this analysis. Kokturk 2005 reported that two participants in the MgSO₄ group and four in the placebo group required additional therapy. Two participants developed transient hypotension after receiving MgSO₄ and β₂‐agonist. No one needed nebulisation to be withheld. One participant in the placebo group suffered palpitations after the second salbutamol nebulisation. No other side effects were reported. Alansari 2015 reported that no participants experienced hypotension; one participant in the placebo group had excessive cough after the first nebulisation and was withdrawn by his parents; one participant in the MgSO₄ group experienced chest tightness and facial rash after the third nebulisation (both resolved after 30 minutes); and one placebo group participant required paediatric ICU admission for refractory status asthmaticus.

Pulmonary function

Four studies involving 133 participants reported PEF, but due to the range of time points and ways of reporting the outcome we have not presented the results on a forest plot. Mangat 1998 found no significant difference in % predicted PEF for MgSO₄ alone compared with β₂‐agonist alone at 60 minutes (MD 4.20, 95% CI −12.29 to 20.69; participants = 33). Neki 2006 shows a significant advantage for β₂‐agonist alone but the time point reported is unclear (MD −50 L/min, 95% CI −67.83 to −32.17; participants = 4). Meral 1996 reported the mean improvement per group in PEFR (%) at five minutes, favouring the β₂‐agonist group (MD −36.77, 95% CI −61.26 to −12.28; participants = 40). Sarhan 2016 reported change in per cent predicted PEF and found no significant difference between groups (MD −1.70, 95% CI −8.45 to 5.05; participants = 20).

Clinical severity score

The Fischl index (a composite of vital signs, PEF and clinical features ranging from 0 to 7 with scores > 4 indicating acute severe asthma) was reported in three studies (Mangat 1998; Neki 2006; Sarhan 2016); and the Davies, Leffert, Drabous score (a composite measure of retraction, nasal flaring, cyanosis and wheeze) in one study (Meral 1996). We pooled the results for studies reporting the Fischl Index, which suggests no significant between‐group difference (MD −0.13, 95% CI −0.62 to 0.36; participants = 93; studies = 3; I² = 0%).

Meral 1996 reported the maximum clinical severity score in the first hour (MD −3.20; 95% CI −17.62 to 11.22; participants = 40).

Admission to hospital

There was no significant difference in risk of admission between those receiving MgSO₄ alone compared with β₂‐agonist alone (RR 0.53, 95% CI 0.05 to 5.31; participants = 33, studies = 1; Analysis 3.2); however, the wide confidence interval indicates, due to there being few events in a small trial, that equivalence cannot be claimed. With a single trial contributing data (Mangat 1998), no additional analyses were possible. Three studies did not appear to measure or report this outcome (Meral 1996; Neki 2006; Sarhan 2016).

Vital signs

Sarhan 2016 reported a significantly lower (therefore better) mean heart rate in people receiving β₂‐agonist alone compared to MgSO₄ alone (MD 21.20, 95% CI 0.17 to 42.23; participants = 20). Respiratory rate was reported as significantly lower (therefore better) in those receiving MgSO₄ alone compared to β₂‐agonist alone (MD −2.40, 95% CI −3.91 to −0.89; participants = 60; studies = 2; I² = 0%)(Neki 2006; Sarhan 2016). There was no significant difference in systolic or diastolic blood pressure between those receiving MgSO₄ alone compared with β₂‐agonist alone in the one study which reported this (Analysis 3.5; Analysis 3.6).Two studies did not appear to measure or report these outcomes (Meral 1996; Mangat 1998).

Adverse events

Two studies reported that there were no serious adverse events in either arm (participants = 53; studies = 2) (Mangat 1998; Sarhan 2016). One study reported mild to moderate adverse events and no difference was detected between groups (RD −0.17, 95% CI −0.41 to 0.06; participants = 33; Analysis 3.8). Meral 1996 also reported that there were no adverse events in either group. One study did not report adverse events (Neki 2006).