Inhibition of Delayed Cerebral Ischemia by Magnesium Is Insufficient for Subarachnoid Hemorrhage Patients: A Network Meta-Analysis

Objective After subarachnoid hemorrhage, magnesium could reduce the incidence of delayed cerebral ischemia; however, it is still controversial. This study updated the results of recently published magnesium-related studies and conducted an exploratory analysis of the impact of application strategies and intervention factors on the results. Methods Public databases were searched from the date of their inception to May 10, 2021. Randomized controlled trials on magnesium agent-related regimens for subarachnoid hemorrhage patients were included. Results In total, 28 articles were included in the meta-analysis. For delayed cerebral ischemia, magnesium-related interventions significantly reduced the risk of delayed cerebral ischemia compared with nonmagnesium interventions (odds ratios: 0.40; 95% confidence interval: 0.28–0.56; p < 0.01). For cerebral vasospasm, a random effects model showed that magnesium significantly reduced the risk of cerebral vasospasm (odds ratios: 0.46; 95% confidence interval: 0.33–0.63; p < 0.01). In the subgroup analysis, intracranial magnesium (odds ratios: 6.67; 95% confidence interval: 1.14–38.83; p=0.03) and magnesium plus hydrogen (odds ratios: 10; 95% confidence interval: 1.59–62.73; p=0.01) produced significant results in improving the good recovery rate compared to the control. In the network meta-analysis, magnesium plus nimodipine and simvastatin even showed an effective trend in death/persistent vegetative status improvement. Conclusion This study supports the beneficial effect of magnesium in reducing the risk of delayed cerebral ischemia. Based on a single randomized controlled trial, immediate intracranial magnesium therapy with intravenous hydrogen after subarachnoid hemorrhage can increase the good recovery rate. Therefore, more high-quality studies are needed to confirm this finding.


Introduction
Subarachnoid hemorrhage (SAH) accounts for approximately 5% of all types of stroke incidence, however, this prevalence is higher among the young and middle-aged population and has an extremely poor prognosis [1,2]. SAH occurs in approximately 9 per 100,000 people every year, and half of SAH patients are younger than 55 years of age and have an extremely poor prognosis [3]. One-third of people die within three months after hemorrhage, and one-fifth of people with SAH need to rely on others for daily activities [4]. e current evidence-based treatment for SAH is neurosurgical clipping or endovascular coiling and administration of nimodipine [5]. However, patients still have a higher incidence of cerebral vasospasm (CVS) and delayed cerebral ischemia (DCI) [6,7]. It is currently believed that DCI is the main cause of death and neurological deficits in SAH patients [8].
Magnesium is a low-cost neuroprotective agent that has been successfully applied in eclampsia treatment. Eclampsia has the same pathophysiological characteristics as DCI after SAH [9]. A recent observational study supports that magnesium influences hemorrhage severity in patients with SAH, potentially through a hemostatic mechanism [10]. e effect of magnesium on SAH is still controversial. In an individual patient data meta-analysis of magnesium for SAH, it was believed that magnesium intervention in an earlier time window did not bring more beneficial DCI results [11]. Two meta-analyses reported that magnesium can reduce the risk of DCI [12,13], and one meta-analysis indicated that magnesium can significantly reduce the incidence of CVS [14]. However, magnesium application did not show benefits with respect to neurological recovery results and mortality. e incongruence between phase 2 and phase 3 clinical studies of magnesium for SAH was reviewed recently. However, it neglected the impact of combination drugs and infusion routes on the therapeutic effect of magnesium [15]. One of the major concerns arising from the magnesium for aneurysmal subarachnoid hemorrhage (MASH)-2 trial is that magnesium does not cross the blood-brain barrier well. e intracisternal (but not intravenous) magnesium infusion strategy reinspired enthusiasm for its clinical application [16]. In addition, the use of concomitant drugs in Japan, China (fasudil), Europe, and North America (nimodipine) will also affect magnesium treatment for SAH [16]. Another review indicated that the immediate intracisternal infusion of magnesium with intravenous hydrogen may be effective for treating early brain injury after SAH [17]. However, it is qualitative and did not analyze the impact of concomitant drugs. Finally, the review still believes that even if magnesium is not routinely used, it is still reasonable to maintain magnesium levels in the normal range because hypomagnesemia is associated with DCI and poor prognosis on SAH [17]. e current study updated the results of recently published magnesium-related randomized controlled trials (RCTs) and further evaluated the effects of magnesium application regimens on DCI, CVS, the modified Rankin score (mRS), the Glasgow outcome scale (GOS) scores, and mortality through a network meta-analysis. It also tried to analyze the impact of important factors among the studies on the above outcomes by meta-regression.

Search Strategy.
We searched all RCTs on the magnesium-related treatment of SAH published up to 24 May 2021. e searched public databases included PubMed, Embase, the Cochrane Library, Scopus, EBSCO, and the Chinese databases of China National Knowledge Infrastructure (CNKI), Wanfang, Chongqing VIP, and SinoMed. e keywords included subarachnoid hemorrhage, magnesium, and random * . To avoid omission, manual searches of references in related reviews were also performed.

Inclusion and Exclusion
Criteria. Two authors checked the literature according to the established inclusion and exclusion criteria. If there was a dispute, it was discussed with the third author to make a concordant decision. e inclusion criteria were as follows: (1) study researched SAH patients, (2) RCT design, (3) the intervention group used magnesium-related treatment, and the control group did not use magnesium in treatment or a different magnesium-related treatment from intervention group, (4) the study reported one of the following outcomes: frequency of DCI, CVS, good recovery(GR) patients according to mRS or GOS/ the Glasgow outcome scale extended (GOSE) assessment, death, or persistent vegetative status (PVS). e exclusion criteria were as follows: (1) the study included SAH patients who were younger than 18 years old, (2) the study researched magnesium intervention that included other stroke patients and did not report SAH patients' results separately, (3) post hoc research, (4) protocol, (5) non-RCTs, and (6) the study did not report the outcomes of interest. Although the search had no language restrictions, the included studies needed to at least have English abstracts. In addition, reviews, comments, and conference abstracts were also excluded.

Data Extraction.
e extracted information included the first author's name, publication year, research location, sample size, magnesium intervention time window, neurosurgical treatment, magnesium intervention, injection dosage and route, magnesium treatment duration, control treatment, and follow-up. e outcomes included the frequency of DCI, CVS, GR according to mRS or GOS/GOSE assessment, and death or PVS. e Cochrane bias risk assessment tool was used to evaluate the methodological quality of the included RCTs [18].

Statistical Methods.
Odds ratios (ORs) and 95% confidence intervals (CIs) were pooled for dichotomous outcomes, and the prediction interval (PI) was also reported. We evaluated heterogeneity by the chi 2 test and calculated I 2 . Significant heterogeneity was defined as an I 2 greater than 50%. e Mantel-Haenszel and Peto methods were used for the fixed effects model, and the Mantel-Haenszel method was used for the random effects model [19]. Funnel plots, Begg's test, and Egger's test were performed to identify potential publication bias. If the results had potential publication bias, the trim-and-fill method was used for correction. e pooled results were further analyzed by a subgroup analysis based on treatment strategies, and frequentist random effect network meta-analysis was also used to rank the effect of strategies based on mixed multiple treatment comparisons [20]. e methods for assessing the extent of the heterogeneity and inconsistency based on generalized Cochran's Q statistic were used for network meta-analysis. e P score was calculated to rank the intervention strategies, and k-means cluster analysis of multiple outcomes was performed [21]. Other important factors that potentially affect outcomes, including publication time, time window, magnesium dose, intervention duration, and follow-up time, were analyzed by metaregression. e software used for analysis included the "meta," "netmeta," and "pheatmap" packages in R language (version 4.0.5) and RevMan (version 5.3).

Results
After searching public databases, a total of 611 English publications and 597 Chinese publications were obtained. After removing duplicate publications, 216 English and 195 Chinese publications remained. After screening the titles 2 Evidence-Based Complementary and Alternative Medicine and abstracts, 70 English publications and 25 Chinese publications remained. After full-text screening, 67 publications were excluded for the following reasons: 13 publications were reviews, 12 publications were post hoc studies, 3 publications did not include SAH patients or report SAH patients separately, one publication was a protocol, 8 publications did not report the desired outcomes, 2 publications were not RCTs, 11 publications were conference abstracts, 16 publications were non-English articles without English abstracts, and one publication included SAH patients younger than 18 years. Finally, 28 papers were included in this meta-analysis   (Figure 1). e publication time ranged from 2002 to 2021. One study clearly excluded patients who needed neurosurgical surgery within 72 hours [24], eight studies did not mention surgery [23, 28, 37-40, 42, 45], and others performed neurosurgery for patients based on actual conditions. Two of the magnesium applied routes are intracranial [22,27]. e follow-up period ranged from 2 weeks to 1 year (Table 1).
Since all included studies were RCTs, the level of overall evidence was acceptable ( Figure 2). However, with the exception of several large-scale phase III clinical studies, the sample sizes in the remaining studies were relatively small. e assessment of DCI, CVS, and GR still suffers from subjective bias, which may cause the results to be more positive. In addition, several studies, including those with small sample sizes, may also impact the robustness of the results.
For DCI, the fixed effects model showed that magnesium-related interventions significantly reduced the risk of DCI compared with nonmagnesium interventions (OR: 0.40; 95% CI: 0.28-0.56; p < 0.01). In the subgroup analysis, the fixed effect model showed that magnesium plus nimodipine can significantly reduce the risk of DCI compared to nimodipine (OR: 0.41; 95% CI: 0.25-0.65; p < 0.01), and magnesium alone can also reduce the risk of DCI (OR: 0.23; 95% CI: 0.11-0.50; p < 0.01) compared to conventional treatment without clear combination drugs, such as nimodipine ( Figure 3). ere was no significant result in other subgroups. Publication bias analysis showed that there was potential bias (Egger's test, p � 0.0165) (Supplementary Figure 1(a)). After correction, the results were still considered stable (OR: 0.49; 95% CI: 0.31-0.77; p < 0.01).
For death or PVS assessment, the fixed effect model did not support that magnesium can significantly reduce the risk of death or PVS in SAH patients (OR: 0.72; 95% CI: 0.72-1.09; p � 0.27). In the subgroup analysis, only magnesium plus simvastatin and nimodipine had a tendency to reduce the risk, however, the difference was not significant (OR: 0.20; 95% CI: 0.04-1.02; p � 0.05) ( Figure 5). Potential publication bias was revealed by Egger's test (p � 0.029) (Supplementary Figure 1(c)), and the negative results were not changed after correction (OR: 0.97; 95% CI: 0.79-1.19; p � 0.79).
For the GR results based on the mRS evaluation, the random effects model (OR: 1.26; 95% CI: 0.90-1.77; p � 0.17) did not show a significant effect of magnesium application in improving GR. In the subgroup analysis, magnesium (OR: 6.67; 95% CI: 1.14-38.83; p � 0.03) and magnesium plus hydrogen (OR: 10; 95% CI: 1.59-62.73; p � 0.01) produced significant results compared to the control. However, these positive results were based on one study [22]. In this study, magnesium was used intracranially, and hydrogen was intravenously used in the magnesium plus hydrogen group. Because of the small number of studies, no publication bias analysis was performed. Based on the GOS/ GOSE assessment, a fixed effects model showed that magnesium did not significantly increase the frequency of GR persons (OR: 1.13; 95% CI: 0.87-1.46; p � 0.34). e subgroup analysis also did not show the advantages of magnesium application ( Figure 6).
Metaregression analysis was performed to compare magnesium plus nimodipine and nimodipine alone. However, the factors were not found to have a significant impact on the effect size in all analyzed results. e multivariate analysis was not performed further (Supplementary Table 4).

Discussion
In this study, we analyzed the effects of the magnesium application strategy on reducing the risk of DCI, CVS, PVS, and death, as well as on GR and GOSE for SAH patients by conventional meta-analysis with subgroup analysis. Furthermore, network meta-analysis was performed to compare the effects of different magnesium application strategies.
is work explained the reasons for the controversy about the effect of magnesium on SAH from the perspectives of different magnesium application strategies.
is study provides evidence for improving the magnesium application strategy in the treatment of SAH in the clinic.
is study supported that magnesium can significantly reduce the DCI risk. At the same time, magnesium can also reduce the CVS risk, however, this positive result may be because of potential publication bias. In the subgroup analysis, intracranial magnesium and magnesium plus hydrogen produced significant results in improving the GR rate compared to the control. In the network meta-analysis, magnesium plus nimodipine and simvastatin showed an effective trend in death/PVS outcome. In the comparisons of magnesium plus nimodipine and nimodipine alone, the metaregression analysis did not identify significant factors related to the outcome.
In the exploratory analysis, the advanced results of magnesium plus cinepazide are based on a Chinese study. Cinepazide maleate, a calcium antagonist, also has the ability to inhibit platelet aggregation and inflammatory factor formation. In clinical studies, there is still a lack of welldesigned studies on cinepazide for SAH. For ischemic stroke, an RCT showed that cinepazide maleate can improve the neurological function recovery and the activities of daily living in ischemic stroke patients who are better than those in the placebo group [50]. However, the therapeutic effectiveness of magnesium plus cinepazide in SAH still needs to be confirmed by more authoritative clinical studies.
Simvastatin application on the basis of magnesium plus nimodipine can further improve the results of DCI and CVS, Studies included in quantitative synthesis (meta-analysis) (n = 28) Figure 1: Flowchart of the study identification process.

Evidence-Based Complementary and Alternative Medicine
and it even has a trend of reducing the risk of death/PVS compared to nimodipine. A review showed that the lowdose statin therapy may have a beneficial effect in reducing the hemorrhagic transformation induced by thrombolysis [51]. erefore, whether the application of simvastatin on the basis of magnesium plus nimodipine inhibits hemorrhagic transformation after a residual secondary cerebral infarction to exert a neuroprotective effect and reduce the risk of death still needs to be confirmed by clinical studies.
In the GR results, an analysis based on a single study suggested that the intracranial application of magnesium with or without the antioxidative stress therapy may be able to improve the patient's neurological outcome. e characteristic of this study is that it significantly increases the level of magnesium in the brain but not in the peripheral circulation.
eoretically, it acts more directly on intracranial blood vessels and exerts neuroprotective effects. An intravenous magnesium injection has a limited effect on increasing its level in cerebrospinal fluid and may also affect other organs, causing bradycardia and hypotension. In addition, the study focused on Fisher grade 3-4 SAH patients and did not apply nimodipine [22]. erefore, the study suggests that for poorgrade SAH patients, increasing the intracranial magnesium concentration can help reduce DCI and CVS rates and improve neurological function recovery.
Observational studies suggest that hypomagnesemia is independently associated with hemorrhagic transformation, poor functional recovery, and DCI in SAH patients [10,17]. erefore, it is still reasonable to maintain magnesium levels in the normal range after SAH [17]. Magnesium showed improvement in neurological function in a phase 2 study [48] and was negative in a phase 3 study [30,34]     hydrogen antioxidant significantly promoted the GR rate of SAH patients immediately after surgery. e strategy of the simultaneous application of nimodipine and simvastatin with magnesium showed a trend of reducing the risk of mortality and PVS. ese two points may provide new information to modify the existing clinical application strategy of magnesium.
Based on the included clinical studies, magnesium supplementation has been shown to be beneficial in reducing the DCI risk. Modifying the magnesium clinical application strategy may further improve the effect on improving the GR rate and survival prognosis.
ese studies indicated that immediate intracranial magnesium therapy combined with hydrogen is beneficial to improve the GR rate for SAH patients, and the combination of magnesium plus nimodipine and simvastatin may have a tendency to improve survival outcome. erefore, two implications may impact future research. Firstly, the immediate intracranial application of magnesium after SAH. On the other hand, it is used in combination with other drugs, such as antioxidants, nimodipine, and simvastatin, to improve the effectiveness of interventions. ese directions deserve to be validated by further well-designed studies.
In conclusion, this study supports the beneficial effect of magnesium in reducing the risk of DCI. Based on a single RCT, immediate intracranial magnesium therapy with intravenous hydrogen after SAH can increase the GR rate. erefore, more high-quality RCTs are needed to confirm this finding.

Limitations.
ere were still several limitations in this analysis. First, this analysis is based on the study level but not on the individual level. Second, although this study analyzed the effect of magnesium dose on each outcome by metaregression, the dependent variable used was the total dose of magnesium application, and it is still not possible to analyze whether differences in magnesium levels in peripheral blood and in the cerebrospinal fluid have an effect on SAH treatment outcome.
ird, in pooling results, publication bias was detected. It indicates that there are still some potential negative results that have not been published. It will cause the results of this study to tend to be positive. Fourth, the definition of CVS is inconsistent in the included studies, and if assessor blinding is not performed at the same time, it may also make the result more positive.

Future Directions.
Hypomagnesemia occurs in more than 50% of patients with SAH and is independently associated with DCI, poor neurological prognosis, and hemorrhage severity. erefore, it is still necessary to maintain magnesium at a reasonable level. However, the results of phase 2 and phase 3 studies on magnesium in the treatment of SAH are incongruent, which may be because of the use of intravenous infusion and the difference in the combination of drugs.
e results of this study support the view that magnesium can reduce the risk of DCI. Based on a single study, the intracisternal infusion of magnesium immediately after SAH with intravenous hydrogen can increase the rate of GR, and the combination of nimodipine and simvastatin with magnesium has a tendency to improve survival/PVS prognosis. Future research can focus on the intracisternal infusion of magnesium immediately after SAH and the combination of magnesium and other drugs, such as antioxidants, nimodipine, and simvastatin, to further explore the application value of magnesium on SAH.