New-onset postoperative atrial fibrillation (POAF) is associated with increased risk of stroke, morbidity, mortality, respiratory failure and pneumonia, with additional treatment and thus increased postoperative cost.1–3
Clinical efforts to prevent and manage life-threatening complications after cardiac surgeries, such as POAF, have posed a significant challenge.4 Despite numerous trials investigating prophylactic and treatment approaches, POAF following cardiac surgery has persisted unchanged for several decades.5
Serum hypomagnesaemia is common after cardiac surgeries and has been linked to atrial fibrillation pathogenesis in experimental studies.6–8 Major cardiac effects of magnesium are prolongation of atrial and atrioventricular nodal refractory periods, which may facilitate rate and rhythm control in atrial fibrillation (AF).8
Clinical trials have examined the effectiveness of magnesium in reducing the occurrence of POAF, but there is a lack of updated systematic reviews to thoroughly summarize the findings. Previous studies have not assessed the certainty of evidence and did not consider the preoperative history of AF rhythm in the enrolled participants.7,9–13 We aimed to conduct a systematic review and meta-analysis of randomized control trials (RCTs) to assess the efficacy and safety of magnesium in preventing new-onset atrial fibrillation following cardiac surgery.
Materials and methods
Standard reporting
We registered our protocol with the Open Science Framework on 27 June 2022 and we followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) to report our findings.14–17 This is a sub-study from a larger network meta-analysis.3 In our protocol, we initially planned to analyse all available drug therapies. However, after assessing for incoherence and intransitivity, consulting with expert clinicians and considering magnesium’s clinical use, we prepared a report on interventions comparing magnesium with placebo or other control groups.
Search databases
We conducted a systematic search in MEDLINE, EMBASE, Web of Science Core Collection and Cochrane Library on 8 August 2022 and updated the search on 8 August 2023. The electronic search was developed and refined in MEDLINE. We did not apply any filters and language or publication date restrictions in our search to prevent missing evidence. Furthermore, we reviewed the reference lists of relevant guidelines, studies and reviews for additional potentially eligible studies.1,2,7,10–13,18–26 We gathered all search results in EndNote 20 and removed duplicate studies.27 Supplementary Table 1 presents additional details including the search terms for each database, date of searches and an exported strategy.
Eligibility criteria and study selection
We included RCTs that enrolled the entire population or a subset of it with adult participants (aged 18 years and older) without a history of previous AF or another supraventricular arrhythmia, undergoing cardiac surgery, including coronary artery bypass graft (CABG), valve surgery or other major cardiac surgeries, and were randomized to receive magnesium agents (such as magnesium sulfate, magnesium chloride and magnesium oxide) or placebo or standard care. Our main outcome of interest was the incidence of POAF. Supplementary Table 2 presents PICO and details on the eligibility criteria.
Based on the eligibility criteria, pairs of independent reviewers screened the titles, abstracts and full-text articles of potentially eligible studies. We resolved disagreements by discussion and, when necessary, by adjudication with a third reviewer. To improve consistency between reviewers and facilitate the screening of the literature, we used Rayyan online systematic review software.28
Data extraction
Pairs of reviewers extracted relevant data from full texts, tables and figures included in studies independently and in duplicate. We carried out calibration exercises to increase the consistency and accuracy of reviewer teams. We developed standardized data extraction forms to extract the following information: study characteristics (authors, publication year, country of recruitment and trial registrations), characteristics of participants (sample size, age, sex, underlying comorbidities and baseline medications), type of surgical procedure, details of interventions and comparators (formulation, description, dosage, route of delivery and time of intervention), risk of bias and measures of incidence of POAF, mortality and side effects.
Reviewers resolved discrepancies through discussion and consensus, with a third reviewer consulted if needed.
Risk-of-bias assessment
Pairs of reviewers independently assessed the quality of included trials using the Cochrane ROB 2.0 tool for randomized trials.29 We rated RCTs across the following five domains: randomization and allocation concealment process, blinding and deviation from intended intervention, loss to follow-up and missing outcome data, outcome measurement and selection of the reported results (deviations from the registered protocol).29,30 Reviewers discussed conflicts in risk of bias assessment to reach a consensus or consult with an arbitrator if necessary.
Data synthesis and statistical methods
For each outcome, we performed a random-effects pairwise meta-analysis of the direct comparisons using the restricted maximum likelihood estimator with the meta package in R (version 4.03, R Foundation for Statistical Computing).31,32
Heterogeneity in pairwise meta-analysis for all the direct comparisons was evaluated using the I2 statistic. I2 values from 0% to 40% may indicate potentially unimportant heterogeneity, 30% to 60% may indicate moderate heterogeneity, 50% to 90% may indicate substantial heterogeneity and 75% to 100% may indicate considerable heterogeneity.33 We assessed publication bias by visually inspecting funnel plots and conducting Egger’s statistical test for comparisons with 10 or more trials.
We summarized the effects of interventions using relative risks (RRs) and corresponding 95% confidence intervals (CIs) and absolute risk difference per 1,000 patients, with a baseline risk sourced from the median risk in the placebo and standard care arms across trials. For continuous outcomes, we reported mean differences (MDs) with associated 95% CI. All tests were two-sided and statistical significance was based on the 95% CIs excluding the null.
We performed subgroup analyses regardless of heterogeneity estimates to further investigate the sources of the heterogeneity. We investigated whether the timing of receiving interventions (treatment given preoperatively, postoperatively or intraoperatively) and risk of bias would affect the treatment effects. We conducted subgroup analyses only if each subgroup contained two or more studies and explored subgroup effects with a test of interaction using meta-regression analysis. We evaluated the credibility of the subgroup if the interaction p-value <0.05.
Certainty of evidence
We assessed the certainty of the evidence using the Grading of Recommendations, Assessment, Development and Evaluations (GRADE).34 We rated the certainty for each comparison and outcome as high, moderate, low or very low based on considerations of the risk of bias, inconsistency, indirectness, publication bias, intransitivity, incoherence and imprecision.
The results were reported using guidance from the GRADE working group, which involves using different adjectives based on the certainty of the evidence (drug x reduces mortality [high certainty], drug x probably reduces mortality [moderate certainty], drug x may reduce mortality [low certainty] and the effect of drug x on mortality is very uncertain [very low certainty]).4,34
Results
Search results
Our search identified 3,628 records and included 24 unique RCTs, including 3,373 participants. Supplementary Figure 1 and Supplementary Table 3 present the selection process and the list of included studies.35–58
Characteristics of included studies
The mean age across all included studies was 61.5 years. An estimated 76.3% of all participants were males. The average preoperative ejection fraction percentage for the entire population was 52.3% and the average cardiopulmonary bypass time was 92.5 minutes. The percentage of patients with hypertension, diabetes mellitus, myocardial infarction and chronic obstructive pulmonary disease was 53.1%, 24.1%, 45.4% and 5.5%, respectively. One of the included trials enrolled participants undergoing isolated valve surgery and one study enrolled participants undergoing mixed CABG and valve surgery.36,45 All other trials enrolled patients undergoing CABG.
Only two studies administered oral tablets of elemental magnesium and all other studies administered magnesium intravenously.46,54 The time of intervention administration varied across the trials. Three studies administered magnesium preoperatively, seven studies postoperatively, six studies both pre- and postoperatively and four studies intraoperatively. Four studies administered magnesium throughout surgery time and continued postoperatively.
Among the 24 studies, 15 did not provide a precise definition for POAF. In the remaining nine studies, five defined POAF as episodes lasting more than 10 minutes or requiring active intervention, three did not specify the duration criterion and one study defined POAF as AF lasting more than 30 s.
More details of patients’ and interventions’ characteristics are summarized in Supplementary Tables 3 and 4.
Risk of bias
Twelve studies (50%) were assessed as having some concerns and 11 studies (46%) were assessed as having a high risk of bias. Only one study was rated as having a low risk of bias. Twelve studies (50%) adequately generated their randomization process and allocation sequence concealment. We found a deviation from the intended intervention in three trials. Supplementary Figures 3 and 7 provide a summary of our risk of bias assessment.
Incidence of postoperative atrial fibrillation
Twenty-four trials, including 3,373 patients and 687 overall events, reported on the incidence of new-onset POAF. We found that magnesium may reduce the risk of new-onset POAF compared to the control group (RR: 0.55 [95% CI: 0.41, 0.74]; low certainty). We found substantial heterogeneity across the included 24 trials (I2=72%). We explored the source of heterogeneity and presented these results in the subgroup analyses section. Figure 1 presents the effect of magnesium on POAF and Table 1 presents the evidence profile and the summary of findings.35–58
Figure 1: Effect of magnesium on the incidence of postoperative atrial fibrillation35–58
CI = confidence interval; RR = relative risk.
Table 1: Evidence profile for the prophylactic effect of magnesium solutions versus control group
| Binary outcomes | Quality of evidence assessment | Relative effect RR† (95% CI) | Anticipated absolute effect | ||||||||
| N* | Risk of bias | Heterogeneity | Indirectness | Imprecision | Publication bias | Overall quality | Baseline risk (%) | Risk difference†† (95% CI) | |||
| New-onset POAF (all studies) | 24 | Serious limitation § | Not serious limitation §§ | Not serious limitation | Not serious limitation¶ | Serious limitation ¶¶ | Low | 0.55 (0.41, 0.74) | 26.2 | 117.9 fewer per 1,000 (95% CI: 68.1, 154.6 fewer) | |
| New-onset POAF ‡ (excluding high risk of bias studies) | 13 | Not serious limitation‡‡ | Not serious limitation** | Not serious limitation | Not serious limitation¶ | Not serious limitation††† | High | 0.70 (0.58, 0.84) | 25.5 | 76.5 fewer per 1,000 (95% CI: 40.8, 107.1 fewer) | |
| Postoperative all-cause mortality | 11 | Serious limitation § | Not serious limitation¶¶¶ | Not serious limitation | Serious limitation*** | Not serious limitation††† | Low | 1.00 (0.34, 2.90) | 0.8 | No change per 1,000 (95% CI: 5.3 fewer, 15.2 more) | |
| Continuous outcome |
|
|
|
|
|
|
| MD (95% CI)‡‡‡ | |||
| Duration of hospital stay (days) | 8 | Serious limitation § | Serious limitation §§§ | Not serious limitation | Serious limitation†††† | Serious limitation¶¶ | Very low | 0.34 (−0.94, +0.26) | |||
*Number of pooled studies in meta-analysis
†RR: Relative risk with corresponding 95% confidence interval
††Risk difference reported per 1,000 patients for each binary outcome
§Quality was rated down on the basis of risk of bias because most evidence came from high-risk-of-bias studies
§§Quality was not rated down on the basis of heterogeneity, as visual inspection was moderate and the I2 statistic result was in the substantial heterogeneity range [I2=74%]. (We considered heterogeneity ranging from 0% to 40% as potentially unimportant, 30% to 60% as moderate, 50% to 90% as substantial and 75% to 100% as considerable based on the Cochrane guides)
¶Quality was not rated down on the basis of imprecision, as the 95% CI for the pooled effect did not overlap a risk difference of 1 (no effect line)
¶¶ Quality was rated down on the basis of publication bias, as the funnel plot was asymmetric and Egger’s test was significant (Supplementary Figures 6 and 9)
‡‡ Quality was not rated down on the basis of risk of bias because all evidence came from non-high-risk-of-bias studies
**Quality was not rated down on the basis of heterogeneity because the I2 statistic result was in the potentially unimportant range [I2=4%]
†††Quality was not rated down on the basis of publication bias because the funnel plot was symmetric and Egger’s test was insignificant (Supplementary Figures 7 and 8)
¶¶¶ Quality was not rated down on the basis of heterogeneity because the I2 statistic result was in the potentially unimportant range [I2=0%]
***Quality was rated down on the basis of imprecision because the 95% CI for the pooled effect overlapped a risk difference of 1 (no effect line)
‡‡‡ MD: Mean difference with corresponding 95% CICI
§§§Quality was rated down on the basis of heterogeneity because visual inspection and the I2 statistic result was in the considerable heterogeneity range [I2=90%]
††††Quality was rated down on the basis of imprecision because the 95% CI for the pooled effect overlapped a mean difference of 0 (no effect line)
GRADE recommendations: The prophylactic effect of magnesium solutions versus the control group for the incidence of postoperative atrial fibrillation, all-cause mortality and duration of hospital stay after cardiac surgery
CI = confidence interval; POAF = postoperative atrial fibrillation; RR = relative risk.
All-cause mortality
Eleven studies, including 1,550 participants, reported the all-cause mortality, including 12 deaths.
We found that magnesium may have no significant effect on the prevention of all-cause mortality compared to the control group (RR: 1.00 [95% CI: 0.34, 2.90]; low certainty). We found unimportant heterogeneity (I2=0%). Figure 2 presents the effect of magnesium on all-cause mortality and Table 1 presents the summary of findings.36–38,40,43,44,50,51,54,56,58
Figure 2: Effect of magnesium on the incidence of all-cause mortality36–38,40,43,44,50,51,54,56,58
CI = confidence interval; RR = relative risk; SD = standard deviation.
Duration of hospitalization
Eight trials reported hospitalization duration, including 1,244 patients.
We found that magnesium may have no significant effect on the total days of hospital stay compared to the control group (MD: −0.34 [95% CI: −0.94, 0.26]; very low certainty). Figure 3 presents the corresponding forest plot and Supplementary Figure 9 shows the funnel plot.41,43–45,49,55–58
Figure 3: Effect of magnesium on duration of hospital stay41,43–45,49,55–58
CI = confidence interval; MD = mean difference; SD = standard deviation.
Serious adverse effects
Of 24 included RCTs, eight trials (including 1,372 patients) assessed the serious adverse events (SAEs) and reported that no SAEs occurred due to the magnesium intervention (Supplementary Table 6).
Heterogeneity assessment and subgroup analyses
Heterogeneity was substantial in the incidence of POAF, with an I² value of 72.0% (95% CI: 71.2, 84.8%). Upon conducting a subgroup analysis, studies rated as having ‘some concerns’ or ‘low’ risk of bias exhibited minimal heterogeneity (I²=4%), while those classified as having high-risk bias showed significant heterogeneity (I²=86%) (Figure 4). However, meta-regression analysis revealed that the difference in heterogeneity between these subgroups was not statistically significant (p-value of interaction=0.126). This finding is likely attributable to the fact that five studies reported RRs with wide CIs, whereas one study, which carried the most weight (5.5%), reported an RR greater than 1 (Figure 4).35–58 Consequently, there was no significant difference in the overall pooled effect between studies with ‘high’ risk of bias and those with ‘low’ or ‘some concerns’.
Figure 4: Subgroup analysis of risk of bias for the effect of magnesium on the incidence of postoperative atrial fibrillation35–58
CI = confidence interval; df = degrees of freedom; RR = relative risk.
It is important to highlight that most studies rated as ‘some concerns’ were downgraded due to reporting bias, specifically, the failure to register study protocols, rather than due to deficiencies in randomization, allocation concealment, blinding or follow-up (Supplementary Figure 3).
We also performed a separate meta-analysis and quality assessment on the 13 studies with ‘low’ or ‘some concerns’ risk of bias. This analysis produced more precise estimates and a higher quality of evidence, with an RR of 0.70 (95% CI: 0.58, 0.84), high certainty and low heterogeneity (I²=4%, Egger’s test=0.343) (Table 1).
Heterogeneity was unimportant for mortality (I2=0.0%; 95% CI: 0.0, 70.8) and considerable for duration of hospitalization (I2=90.0%; 95% CI: 82.0, 94.0). We investigated further sources of heterogeneity by running meta-regression and preidentified subgroup analyses. We did not find any statistically significant difference between subgroups based on the timing of the intervention (interaction p-value from meta-regression: 0.293; see Supplementary Figure 4 and Supplementary Table 5). Moreover, we did not find a significant difference between subgroups based on risk of bias (interaction p-value from meta-regression=0.266; see Supplementary Figure 5 and Supplementary Table 5). Supplementary Table 5 presents the summary of subgroup analyses and meta-regression for potential sources of heterogeneity.
Publication bias was assessed using Egger’s test or by visually inspecting funnel plots for less than 10 trials. An Egger’s value of <0.001 and an asymmetric funnel plot showed potential publication bias for the incidence of POAF (Supplementary Figure 7). Funnel plot visual symmetry indicated no publication bias for all-cause mortality (Supplementary Figure 8), whereas funnel plot visual asymmetry indicated publication bias for hospital stay (Supplementary Figure 9).
Discussion
Main findings
Our meta-analysis included 24 RCTs, covering over 3,300 patients, to assess the effectiveness and safety of magnesium in preventing new-onset POAF, a common complication following cardiac surgery. Our meta-analysis demonstrates that magnesium administration reduces the risk of new-onset AF after cardiac surgery without any serious adverse events. This finding is significant given the substantial burden POAF places on postoperative outcomes. AF is associated with a two-fold increased risk of 30-day and 6-month mortality, prolonged hospitalization (3.7 additional days on average) and higher rates of stroke, thromboembolism and cardiac arrest.59 However, this decrease in POAF did not translate to a significant impact on short-term mortality or hospital length of stay in our meta-analysis. It is important to note that the primary studies included in our analysis reported a low mortality event rate (0.87%) and had relatively short follow-up periods, with the longest being 30 days. These factors limit the statistical power to detect effects on mortality and may not adequately capture long-term complications associated with POAF. Regarding hospitalization duration, although statistically significant and points to a potential reduction in duration of hospital stay, it does not meet the minimal important difference of 1 day. Given these considerations, there is a need for large-scale RCTs with extended follow-up durations to more definitively assess the long-term effects of magnesium supplementation on mortality.
Management of POAF involves rate control (e.g. beta-blockers and calcium channel blockers) or rhythm control (e.g. amiodarone and electrical cardioversion), with both strategies showing comparable short-term outcomes. Haemodynamically unstable patients require immediate cardioversion.60 About 20–30% of patients may develop persistent AF. Anticoagulation is considered based on stroke risk and assessment of thromboembolic risk using tools such as the CHA₂DS₂-VASc score.60 Advanced interventions like catheter or surgical ablation may be considered for recurrent or refractory cases.61 Among the available options for preventing POAF, magnesium carries the lowest risk of side effects.13 It is widely available, well tolerated and cost-effective. Therefore, magnesium supplementation is a reasonable choice to prevent the future need for treatment.62
In relation to prior findings
The results of our meta-analysis suggest that magnesium may have a favourable effect on preventing new-onset POAF. This finding agrees with other previous large trials and meta-analysis studies. A meta-analysis conducted in 2004 enrolled 17 RCTs and concluded that magnesium reduced the risk of supraventricular arrhythmias (RR: 0.77 [95% CI: 0.63, 0.93]; p=0.002).11 In 2005, a meta-analysis of 20 RCTs found that magnesium administration reduced the risk of developing POAF (OR: 0.54 [95% CI: 0.38, 0.75]).7 Some other studies published the following year concluded similar results.10,25 One of these meta-analyses included seven RCTs that enrolled participants who had undergone CABG and did not have a prior history of AF. The pooled results showed that magnesium reduced the incidence of new-onset POAF (RR: 0.64 [95% CI: 0.50, 0.83]; p=0.001).25
In contrast, in some previous reports, magnesium did not prevent POAF.12,13 These disagreements could be justified. Some clinical trials investigating magnesium’s effectiveness in preventing POAF excluded participants with a history of supraventricular arrhythmia, while some others included them. This variation between the populations of these two study groups might explain the differences in the final results.3 Moreover, many previous systematic reviews did not exclude RCTs that enrolled patients with pre-existing AF, which could bias the final results. Some of these studies considered any type of POAF (pre-existing or new-onset) as their study outcome and reported the overall results (see Supplementary Table 7 for comparison and consistency of our results with some prior studies).
While our meta-analysis indicates that magnesium administration may reduce the incidence of POAF, the evidence pooled from the included studies does not allow for precise identification of patient subgroups who would benefit most. Existing literature suggests potential benefits in patients with hypomagnesaemia or elevated risk for POAF. A meta-analysis by Fanning et al. reported a reduction in POAF incidence with magnesium administration, especially in patients with low-baseline magnesium levels.40 Conversely, caution is advised when considering magnesium supplementation in patients with severe renal impairment or haemodynamic instability.63
Previous reports indicated that administering magnesium intravenously did not improve the cardioversion rates in patients with persistent AF.64,65 This suggests that magnesium may play a more significant role in preventing the onset of AF rather than aiding in its conversion. The TIGHT-K trial is a multicentre RCT that investigated the role of potassium supplementation in preventing POAF following cardiac surgery.66 The results indicated no significant effect of potassium on POAF incidence. Conversely, Mg supplementation has shown promise in reducing the incidence of POAF.
Strengths and limitations
The key strength of this research is its rigorous methodological approach, following PRISMA guidelines and employing advanced statistical analyses, meta-regression analyses and subgroup analyses to find the source of heterogeneity. As the most trustworthy evidence comes from RCTs, we only included RCTs. To prevent bias due to baseline comorbidity and reduce the heterogeneity, we only included studies that enrolled participants without a history of AF before surgery. This in-depth analysis provides a detailed understanding of the effectiveness of interventions. Furthermore, we provided a summary and evidence profile for our main results following the GRADE approach. This uses the study results for clinical applications.
However, our study had some limitations. We found a significant heterogeneity across the included trials, issues with risk of bias and a weak body of evidence, which pose notable limitations, potentially affecting the generalizability and reliability of our findings. The observed heterogeneity also highlights the complexity of POAF as a multifactorial condition, influenced by surgical techniques, patient comorbidities and perioperative management strategies. We performed pairwise meta-regressions and subgroup analyses to look for evidence of intransitivity and reassurance. We modified the heterogeneity effect by reporting the results from non-high risk of bias studies separately. This posed a high certainty and more precise results with minimal heterogeneity.
Future directions
The routine use of antiarrhythmic agents to prevent POAF has a drawback. Most patients undergoing cardiac surgery do not develop postoperative AF rhythm. However, they would still be exposed to possible side effects of these preventative treatments. Routine antiarrhythmic prophylaxis of POAF is estimated to expose more than 75% of patients to drugs that they may not need.7,67 Magnesium administration could be a suitable alternative to antiarrhythmic medications for preventing POAF. Hypermagnesaemia is a potential adverse event that may initially present with nausea, lethargy and loss of deep tendon reflexes with progression to an altered mental state, hypotension and bradycardia.68 However, none of the included RCTs mentioned any significant adverse events due to the administration of magnesium. Therefore, magnesium can be a noteworthy suggestion for the prophylaxis of POAF.
We identified one included RCT with a low risk of bias, while 12 RCTs raised some concerns and 11 RCTs were assessed as high risk of bias. We recommend conducting more high-quality research in this field in the future.
Conclusions
In conclusion, this meta-analysis presents the most up-to-date investigation on the effectiveness of magnesium agents in preventing the incidence of new-onset POAF after cardiac surgery. Our results showed that magnesium administration is effective in reducing the incidence of new-onset POAF. However, magnesium may not be beneficial in reducing all-cause mortality. The results of magnesium’s effect on reducing hospitalization are uncertain. Additional investments in new high-quality trials that address the limitations of the current evidence may improve the certainty of evidence and allow us to make stronger claims about the outcomes.
