Magnesium Sulfate to Prevent Perioperative Atrial Fibrillation in Cardiac Surgery: A Randomized Clinical Trial

Abstract

Objectives:

To determine whether perioperative IV magnesium sulfate infusion, targeting serum magnesium concentrations of 1.5–2.0 mmol/L, reduces the incidence of postoperative atrial fibrillation (POAF) in patients undergoing cardiac surgery.

Design:

Double-blind, randomized, placebo-controlled, single-center clinical trial with interim analysis for futility.

Setting:

HagaZiekenhuis, The Hague, The Netherlands (February 2022–November 2023).

Patients:

Adult patients undergoing coronary artery bypass grafting and/or valvular surgery without prior atrial arrhythmias or severe renal dysfunction.

Interventions:

Continuous IV infusion of magnesium sulfate (3 mmol/hr, with bolus if [baseline] magnesium < 1.0 mmol/L) or placebo (Ringer’s lactate), initiated after induction of anesthesia and continued until ICU discharge.

Measurements and Main Results:

A total of 265 patients underwent randomization before the trial was stopped at interim analysis for futility. Magnesium supplementation achieved clear separation in serum magnesium concentrations between groups. POAF occurred in 50 of 132 patients (37.9%) in the magnesium group and 38 of 133 patients (28.6%) in the placebo group (relative risk, 1.29; 95% CI, 0.92–1.80). No subgroup demonstrated benefit. Time-to-event and day-by-day analyses showed no early reduction in POAF with magnesium. Vasopressor use was more frequent in the magnesium group, although differences were not statistically significant. No safety signal was identified.

Conclusions:

In this randomized trial, perioperative magnesium infusion targeting serum concentrations of 1.5–2.0 mmol/L did not reduce POAF after cardiac surgery. These findings do not support routine prophylactic magnesium supplementation for prevention of POAF.

KEY POINTS.

Question: Does perioperative IV magnesium sulfate infusion targeting serum magnesium levels of 1.5–2.0 mmol/L reduce the incidence of postoperative atrial fibrillation (POAF) in patients undergoing cardiac surgery compared with placebo?

Findings: In the randomized, double-blind Perioperative Magnesium Infusion to Prevent Atrial FIbrillation Evaluated trial (n = 265), the incidence of POAF within 7 days after surgery was 37.9% in the magnesium group and 28.6% in the placebo group (relative risk 1.29; 95% CI, 0.92–1.83; p = 0.15). No benefit was observed across secondary outcomes, and the study was stopped early at interim analysis for futility. Safety concerns were not identified.

Meaning: Perioperative magnesium infusion to maintain serum magnesium levels between 1.5–2.0 mmol/L did not reduce POAF after cardiac surgery. Routine use of magnesium supplementation for this purpose is not supported.

Postoperative atrial fibrillation (POAF) is a common complication after cardiac surgery (1–3). Although typically intermittent and self-limiting, POAF is associated with increased morbidity and healthcare costs (4–7). POAF may necessitate postoperative anticoagulation, increasing the risk of hemorrhagic complications, including hemorrhagic stroke (8). POAF is also independently associated with prolonged ICU and hospital length of stay, thereby increasing overall healthcare utilization (9). Consequently, multiple strategies have been explored to reduce POAF incidence (10).

The pathophysiology of POAF remains incompletely understood (11–13). Several pharmacological interventions have been evaluated for POAF prevention, including beta-blockers, amiodarone, lidocaine, magnesium, and calcium channel blockers (14–16). Magnesium is of particular interest because hypomagnesemia is common and associated with an increased risk of arrhythmias across diverse populations, including cardiac surgery patients (17–19). Magnesium has a favorable safety profile (20) and is thought to exert antiarrhythmic effects through calcium channel blockade, reducing early after-depolarization–mediated triggered activity (5, 21).

Meta-analyses have suggested that higher magnesium levels may reduce POAF incidence, with several reporting significant benefit from perioperative magnesium administration (15, 22, 23). However, most studies were heterogeneous in dosing, timing, and study quality, and few evaluated targeted supplementation protocols. Importantly, no prior randomized trial applied a serum-adjusted magnesium strategy, limiting interpretability of existing evidence (24).

The Perioperative Magnesium Infusion to Prevent Atrial Fibrillation Evaluated (POMPAE) trial was therefore designed as a randomized, double-blind, placebo-controlled trial to determine whether perioperative IV magnesium infusion targeting total serum magnesium levels of 1.5–2.0 mmol/L reduces the incidence of POAF within 7 days after cardiac surgery.

METHODS

The complete POMPAE study protocol, including the algorithm for administration of the study medication and the methodological and statistical analysis plans, has been published previously (25).

The trial was approved by the medical ethics committee Leiden-Den Haag-Delft (ref. NL77956.058.21) on April 8, 2022 with the corresponding study title “POMPAE trial: Peri-Operative Magnesium infusion to Prevent Atrial fibrillation Evaluated.” The POMPAE trial was initially registered in the Dutch Trial Registry (NL9810) on October 7, 2021. However, following the final discontinuation of this registry in 2023 (link of the original registration is https://onderzoekmetmensen.nl/nl/trial/24637), the trial was retrospectively registered at ClinicalTrials.gov (NCT05669417) on December 21, 2022 (https://clinicaltrials.gov/ct2/show/NCT05669417) using the exact same protocol as for the registration in the Dutch Trial Registry. The first patient was enrolled on August 2, 2022, before the ClinicalTrials.gov registration. The research was conducted in accordance with the ethical standards of the institutional review board and with the principles of the Helsinki Declaration of 1975. The key elements of the study protocol are summarized below.

Trial Design

The POMPAE trial was a single-center, randomized, double-blind, placebo-controlled trial conducted at HagaZiekenhuis, The Hague, The Netherlands. The primary objective was to determine whether maintaining perioperative total serum magnesium concentrations between 1.5 and 2.0 mmol/L (3.6–4.9 mg/dL) reduced the incidence of POAF in patients undergoing cardiac surgery.

Eligibility

Adults (18 yr or older) undergoing elective or semi-elective coronary artery bypass grafting (CABG) and/or valvular surgery were eligible. A complete overview of the inclusion criteria is listed in Supplementary Table 1 (https://links.lww.com/CCM/H966). Patients with a history of atrial arrhythmias were excluded because of an increased risk of recurrent perioperative atrial fibrillation (26). Patients with severe renal dysfunction (creatinine clearance < 30 mL/min) were excluded due to reduced renal magnesium clearance and risk of hypermagnesemia (17, 27). Emergency procedures were excluded because informed consent could not be reliably obtained. Patients undergoing valvular surgery, with or without CABG, were included, with stratification by surgery type.

Anesthetic regime was not prospectively recorded for the trial patients. In routine practice within our institution, greater than 95% of patients received total IV anesthesia (propofol and remifentanil). Preoperatively, angiotensin-converting enzyme inhibitors, calcium antagonists and diuretics were generally withheld on the day of surgery with beta-blockers generally being continued.

All patients were managed postoperatively in a cardiac ICU with continuous electrocardiogram (ECG) monitoring. After ICU discharge, patients were monitored on the cardiac surgery ward with telemetry for a minimum of three additional days. Arrhythmia alerts were reviewed in real time and confirmed by ECG when indicated.

Randomization and Blinding

For more detailed information regarding this topic, we refer to the trial protocol as stated before. After written informed consent, patients were randomized using a secure web-based system (Castor electronic data capture) managed by the cardiac surgery department (28). Treatment allocation was communicated to the hospital pharmacy, which prepared the study medication.

Patients, treating clinicians (operating room, ICU, and ward), and investigators remained blinded throughout the trial. Magnesium sulfate and placebo (Ringer’s lactate) were indistinguishable clear solutions, ensuring allocation concealment.

Although magnesium levels were available as part of routine care and staff may have suspected allocation, atrial fibrillation was objectively assessed by ECG recordings minimizing the risk of bias. All other procedures (including rescue medications and electrocardioversion) were monitored.

Intervention

The study protocol is shown in Supplementary Figure 1 (https://links.lww.com/CCM/H966). In brief, the study medication was initiated immediately after induction of anesthesia and discontinued at ICU discharge or during any ICU readmission. Patients received a continuous infusion of magnesium sulfate (3 mmol/h) or placebo. If the preoperative serum magnesium concentration was less than1.0 mmol/L, a 10 mmol bolus of study medication was administered before initiation of the infusion.

Serum magnesium measurements were conducted directly after the bolus administration after induction of anesthesia and subsequently every 6 hours to guide study medication administration. Additional 10 mmol boluses were administered if levels fell below 1.5 mmol/L, whereas the infusion was temporarily stopped if levels reached greater than or equal to 2.0 mmol/L. This dosing strategy was based on prior pharmacokinetic studies in similar patient populations (27, 29, 30).

As part of normal clinical routine within the hospital, if magnesium values fell below 1.0 mmol/L, this justified the administration of unlabeled 10 mmol magnesium sulfate (and bolus of study medication) and was registered in the trial database. If a stopping rule was encountered (Supplementary Table 2, https://links.lww.com/CCM/H966), the study medication was halted but the patient remained part of the trial based on an intention-to-treat model.

Postoperative use of beta-blockers and other rhythm-modifying medications (apart from amiodaron use and/or electrocardioversion procedures) were not systematically collected. Overall, in ICU, beta-blockers are not (re)started and well over 95% of patients on the nursing ward (postintervention period), beta-blockers are (re)initiated with all perioperative medication use being summarized in Supplementary Table 3 (https://links.lww.com/CCM/H966).

Outcomes

The primary outcome was new-onset POAF lasting greater than or equal to 5 minutes, documented by ECG, occurring within 7 days after surgery or before hospital discharge or death, whichever occurred first. Secondary outcomes included POAF within 28 days, peak heart rate, ICU and hospital length of stay, duration of mechanical ventilation, duration of inotropic and/or vasopressor support, and a composite outcome of mortality, stroke, pulmonary embolism, delirium requiring antipsychotic medication, or infection requiring antibiotics within 28 days.

The trial was overseen by an independent Data Safety and Monitoring Board (DSMB). All serious adverse events are reported in Supplementary Table 4 (https://links.lww.com/CCM/H966).

Statistical Analysis

Sample size calculations were based on an expected baseline POAF incidence of 30% (2, 31). Lower rates are observed (around 20%) in isolated CABG surgery and higher rates (40–50% or more) in valvular surgery. With 530 patients, the study had 80% power to detect a 40% relative risk reduction with a two-sided alpha of 0.05. One interim analysis for futility was prespecified at approximately 50% enrollment. Futility was assessed using a non-binding boundary derived from the design assumption of a 40% relative risk reduction in 7-day POAF (25–15%): continuation required an interim treatment effect in the direction of benefit; a treatment effect less than or equal to 0% (no benefit or harm) supported stopping for futility. No formal harm-stopping boundary was prespecified; safety was monitored separately by the DSMB. The futility assessment did not incorporate external trials or meta-analytic priors and was assessed using a non-binding boundary based on stochastic curtailment (32).

Analyses were performed according to a modified intention-to-treat principle. Categorical variables are presented as numbers and percentages, and continuous variables as means (sd) or medians (interquartile range [IQR]), as appropriate. For approximately normal continuous outcomes we reported mean differences with Welch 95% CIs, for binary outcomes we reported risk differences using Newcombe’s method with Wilson score intervals in percentage points, and for skewed or ordinal outcomes we reported the Hodges–Lehmann median difference with percentile bootstrap 95% CIs. Missing data were minimal. No imputation was performed, and analyses used available data only. Denominators are reported for each outcome where applicable.

The primary outcome (POAF) and other binary secondary outcomes were analyzed using log-binomial models with a log link. Model adequacy was assessed by convergence and residual diagnostics. If log-binomial models failed to converge, Poisson regression with robust variance was used; if necessary, logistic regression was applied and odds ratios were reported. Sensitivity analyses were performed using logistic regression adjusted for prespecified baseline covariates associated with outcome (age, left ventricular function, type and duration of surgery, and use of vasopressors and inotropes), with results reported as odds ratios (95% CI).

Continuous duration outcomes were analyzed using quantile regression adjusted for surgery type, with results reported as adjusted medians (IQR) and median differences (95% CI). Longitudinal serum magnesium concentrations were analyzed using mixed-effects linear models with patient-level random effects and fixed effects for treatment, time, and their interaction.

All analyses were performed using SAS, Version 9.4 (SAS Institute, Cary, NC). A two-sided p value of less than 0.05 defined statistical significance for the primary outcome; secondary outcomes were considered exploratory without adjustment for multiple comparisons.

Exploratory post hoc subgroup analyses assessed treatment-effect heterogeneity across clinically relevant strata, including surgery type (CABG vs. valve), left ventricular function, aortic cross-clamp time, cardiopulmonary bypass time, age, sex, and baseline magnesium (median split, with missing as a separate category where applicable). Formal treatment–subgroup interaction tests were performed; results were interpreted as hypothesis-generating only.

RESULTS

Patients

The trial was stopped early after analysis following DSMB recommendation for futility, per the prespecified rule. Between February 2022 and November 2023, 663 patients were screened (Fig. 1). Of these, 207 met exclusion criteria, 171 eligible patients declined participation or were missed, and 20 were excluded for logistical reasons (e.g., study medication unavailable; Supplementary Table 5, https://links.lww.com/CCM/H966). Consequently, 265 patients were included in the analysis (132 intervention; 133 placebo). Baseline characteristics were well balanced between groups (Table 1 and Supplementary Table 6, https://links.lww.com/CCM/H966).

Figure 1.

Consort flow diagram Perioperative Magnesium Infusion to Prevent Atrial Fibrillation Evaluated (POMPAE) trial. Consort diagram for the POMPAE trial. In total 265 patients were analyzed as part of the interim analysis. AF = atrial fibrillation, eGFR = estimated glomerular filtration rate.

TABLE 1.

Baseline Characteristics of the Study Patients

Parameter	Magnesium (n = 132)	Placebo (n = 133)
Age, yra	66 (8.75)	66.6 (9.02)
Male, n (%)	105 (80)	116 (87)
Body mass index (n = 263), kg/m2a	28 (4.05)	28.3 (6.61)
euroSCORE 2 (n = 251)b	1.22 (0.8–2.05)	1.28 (0.85–2.11)
Previous medical history
Ischemic stroke, n (%)	6 (5)	11 (8)
Myocardial infarction, n (%)	44 (33)	48 (36)
Percutaneous coronary intervention, n (%)	25 (19)	26 (20)
Cardiac surgery, n (%)	2 (2)	2 (2)
Comorbidities
Hypertension, n (%)	70 (53)	55 (41)
Dyslipidemia, n (%)	27 (21)	31 (23)
Asthma/chronic obstructive pulmonary disease, n (%)	10 (8)	11 (8)
Chronic kidney disease, n (%)	6 (5)	5 (4)
Diabetes mellitus, n (%)	33 (25)	35 (26)
Arrhythmias other than AF, n (%)	0	1 (1)
Thyroid disease, n (%)	2 (2)	2 (2)
Obstructive sleep apnea syndrome, n (%)	13 (10)	7 (5)
None registered, n (%)	30 (30)	46 (35)
Smoking	107 (81)	89 (67)
Packyears (n = 100)a	26.6 (21.4)	27.2 (24.5)
Left ventricular ejection fraction
Normal (55–70), n (%)	91 (69)	95 (71)
Mild impaired (40–54), n (%)	27 (21)	28 (21)
Moderate (35–39), n (%)	7 (5)	6 (5)
Significantly impaired (< 35), n (%)	7 (5)	4 (3)
Type of surgery with valvular surgery present, n (%)	62 (24)	32 (24)

^aMean and sd.

^bMedian and interquartile range.

Intervention

Mean serum magnesium concentration after induction of anesthesia was 0.80 ± 0.12 mmol/L in the placebo group and 1.60 ± 0.31 mmol/L in the intervention group; preinduction levels were similar (0.82 ± 0.09 vs. 0.84 ± 0.07 mmol/L). Protocol adherence for bolus administration and infusion initiation was 93%, without between-group differences (Table 2). ECG monitoring and magnesium measurements were comparable between groups, with ICU dose adjustments performed in 89% of required instances. This resulted in sustained separation of serum magnesium concentrations throughout the observation period, with minimal intra-group variability (Fig. 2).

TABLE 2.

Compliance With Study Medication Protocol

Parameter	Magnesium (n = 132)	Placebo (n = 133)	Absolute Difference (95% CI), (Magnesium–Placebo)%	p
Magnesium preinduction, mmol/La	0.82 (0.09)	0.84 (0.07)	–0.01 (–0.03 to 0.01)	0.11
Magnesium postinduction, mmol/La	1.60 (0.31)	0.80 (0.12)	–0.81 (0.76 to 0.87)	< 0.0001
Bolus of study medication given, (n = 255), n (%)b	114 (91)	122 (94)	–2.6 % (–6.7 to 11.9)	0.42
Start of continuous infusion of study medication, (n = 255), n (%)c	115 (92)	123 (95)	–2.6 % (–6.3 to 11.5)	0.40
Time between induction and bolus (n = 235) (min)d	4.2 (0 to 7.8)	1.2 (0 to 7.8)	0.0 (0.0 to 3.0)	0.34
Time between induction and infusion (n = 236) (min)d	4.6 (4.2 to 16.8)	9.0 (1.2 to 19.2)	–4.2 (–7.8 to –0.0)	0.031
Compliance with required dose adjustments in ICU, (n, %)e	110 (83)	127 (96)	–12.2% (1.8 to 21.9)	0.001
Number of magnesium measurements in ICU per dayd,f	4 (3 to 4.5)	4 (3 to 4)	0 (0 to 0)	0.015
Number of electrocardiograms per patient patient (mean, sd)	8.1 (2.7)	8.1 (2.6)	0 (–0.5 to 0.8)	0.94

^aMean and sd.

^bA bolus of study medication given to a patient postinduction of anesthesia.

^cA continuous infusion of study medication initiated within 10 min postinduction of anesthesia.

^dMedian and interquartile range.

^eThe compliance as a ratio between the number of study medication dosing adjustments recorded divided by the number of dose adjustments required as per magnesium serum level.

^fThe median number of magnesium levels determined in the ICU period (protocol dictated every 4 hr, however, some patients were in ICU for < 24 hr in total.

^gThe number of electrocardiogram’s recorded per patients in the period from initiation of anesthesia until day 28 postoperatively.

Totals may not sum to 265 as the intervention was not documented for some participants.

Figure 2.

Magnesium levels in treatment and placebo group. Mean serum magnesium levels with 95% CIs over the first 48 hours after surgery in patients receiving the intervention (red) compared with control (blue). The intervention group achieved higher magnesium concentrations early and maintained them throughout the observation period, whereas levels in the control group remained lower and relatively stable. The number of patients contributing data at each time point is shown beneath the x-axis. The y-axis indicates the magnesium serum level in mmol/L.

Fourteen patients met predefined stopping criteria, most commonly due to hypotension requiring greater than 0.1 μg/kg/min noradrenaline for greater than 1 hour (base-dose), without significant differences between groups. Hemodynamic classification (vasoplegia vs. cardiogenic shock or a combination) was based on clinical assessment, echocardiography, and laboratory data (central venous oxygen saturation). Cardiogenic shock was managed with dobutamine with or without noradrenaline, whereas vasoplegia was primarily treated with noradrenaline. Protocol deviations are detailed in Supplementary Table 5 (https://links.lww.com/CCM/H966). No rescue therapies were used.

Outcomes

At interim analysis, POAF within 7 days occurred in 37.9% of patients in the intervention group and 28.6% in the placebo group (relative risk [RR] 1.29; 95% CI, 0.92–1.83; p = 0.15) (Table 3). No significant differences were observed for secondary outcomes. POAF within 28 days occurred in 39.4% vs. 29.3%, respectively (RR 1.31; 95% CI, 0.94–1.84; p = 0.11). The composite outcome (mortality, stroke, pulmonary embolism, delirium requiring antipsychotic medication, or infection requiring antibiotics within 28 d) at 28 days was also similar between groups (7.6% vs. 12.0%; RR 0.62; 95% CI, 0.29–1.32; p = 0.22).

TABLE 3.

Primary and Secondary Outcomes

Primary Outcome	Magnesium (n = 132)	Control (n = 133)	Adjusted Relative Risk (95% CI)a	Adjusted Risk Difference (95% CI)a	p
Atrial fibrillation within 7 d	50 (37.9%)	38 (28.6%)	1.29 (0.92 to 1.83)	10% (–1.2% to 21.2%)	0.15
Secondary outcomes
Atrial fibrillation within 28 d	52 (39.4%)	39 (29.3%)	1.31 (0.94 to 1.84)	10.7% (–0.6% to 22%)	0.11
28-d composite outcome (death, stroke, pulmonary embolism, delirium, infection)	10 (7.6%)	16 (12%)	0.62 (0.29 to 1.32)	–3.8% (–11.3% to 3.7%)	0.22

	Intervention	Control Median (Interquartile Range)		Difference (95% CI)
	Median (Interquartile Range)
Duration of vasopressor (min)	112.5 (77.0 to 176.5)	85.5 (43.0 to 137.5)		27 (–8 to 62)	0.13
Duration of ventilation (hr)	7.4 (6.3 to 9.7)	6.9 (6.0 to 8.9)		0.4 (–0.2 to 1)	0.15
Hospital length of stay (d)	6.5 (5.2 to 9.4)	6.7 (5.3 to 9.6)		–0.2 (–1.2 to 0.9)	0.77
ICU length of stay (hr)	20.6 (19.1 to 22.2)	20.5 (18.1 to 22.6)		0.1 (–0.9 to 1.2)	0.82
Duration of inotrope (min)	86 (68.5 to 110.5)	94 (36.5 to 163.5)		–8 (–99.7 to 83.7)	0.86
Time to atrial fibrillation (d)	4.1 (2.9 to 4.8)	4.1 (3.2 to 4.8)		0 (–0.5 to 0.4)	0.94
Highest heart rate	96.5 (87.0 to 122.5)	96.5 (85.0 to 121.5)		0 (–5.8 to 5.8)	1.00

^aAdjusted for strata (surgery type).

Because valvular surgery was balanced by stratification, POAF incidence varied by surgical subtype. Among patients undergoing isolated CABG, POAF occurred in 37.6% of the intervention group and 22.8% of the placebo group, whereas rates in valvular surgery were 38.7% and 46.9%, respectively (Supplementary Fig. 2, https://links.lww.com/CCM/H966). POAF incidence in control strata was consistent with assumptions used for sample size calculation.

Results were unchanged after sensitivity analyses adjusting for baseline imbalances and covariates associated with POAF (Supplementary Table 7, https://links.lww.com/CCM/H966). Exploratory post hoc analysis showed no difference in time to POAF between groups (Fig. 3 and Supplementary Fig. 3, https://links.lww.com/CCM/H966).

Figure 3.

Post hoc analysis of time to postoperative atrial fibrillation (POAF). Cumulative incidence of POAF as 1-incidence on the y-axis.

Use of inotropic therapy (dobutamine only) was similar between groups in frequency (6.8% vs. 5.3%; p = 0.6) and duration. Noradrenaline was used more frequently in the intervention group (40.2% vs. 28.6%; p = 0.05) and for a numerically longer duration (median [IQR] 112.5 [77.0–176.5] vs. 85.5 [43.0–137.5] min; p = 0.13), whereas maximum doses during the operating room and ICU periods were comparable (Supplementary Table 6, https://links.lww.com/CCM/H966).

Safety

Serious adverse events occurred in 24 patients (9.0%): 10 (7.6%) in the intervention group and 14 (10.5%) in the placebo group. All adverse events are summarized in Supplementary Table 4 (https://links.lww.com/CCM/H966) and were deemed unlikely related to study treatment by the DSMB. Two patients died beyond 28 days after surgery due to infection, with both opting for palliative care.

DISCUSSION

The POMPAE trial was a contemporary randomized controlled trial evaluating perioperative magnesium supplementation for the prevention of POAF after cardiac surgery. Although earlier and larger studies exist, POMPAE provides recent, high-quality evidence using protocolized, serum-adjusted magnesium dosing reflective of current surgical and ICU practice. Despite achieving clear separation in serum magnesium concentrations and high protocol adherence, magnesium supplementation did not reduce POAF incidence.

At interim analysis (n = 265), effect estimates favored placebo (RR 1.29), meeting the prespecified futility criterion, with very low conditional power to demonstrate benefit with continued enrollment. No signal of benefit was observed in any subgroup. Numerically higher rates of POAF and vasopressor use in the magnesium group raise the possibility of harm, although the study was not powered to detect differences in adverse outcomes. Taken together, these findings do not support the use of prophylactic magnesium for prevention of POAF.

Magnesium has long been considered a potential antiarrhythmic agent due to its effects on potassium and calcium channel function (33). Earlier evidence, including a 2013 Cochrane review, suggested a reduction in POAF, but was limited by heterogeneity in dosing, timing, and study quality (15). Subsequent studies have been small or methodologically heterogeneous. A randomized trial by Tohme et al (34) reported reduced POAF with preoperative oral magnesium loading; however, exclusion of patients requiring vasoactive support and absence of pharmacokinetic data limit generalizability.

The POMPAE trial uniquely targeted sustained total serum magnesium concentrations between 1.5 and 2.0 mmol/L throughout the perioperative and ICU periods, with frequent monitoring and dose adjustment. Although magnesium levels postinduction of anesthesia declined slightly by 0.04 mmol/L in the control group with consistent between-group separation (Fig. 2). Stratified randomization accounted for valvular surgery, a known POAF risk factor.

Exploratory analyses of achieved magnesium exposure (mean, minimum, and maximum levels over 48 hr) did not demonstrate a consistent association between higher magnesium levels and reduced POAF. These findings suggest that sustained elevation of serum magnesium does not mitigate POAF risk.

A modest increase in vasopressor use was observed in the magnesium group. Although not statistically significant, this finding raises a potential mechanistic hypothesis whereby magnesium-induced vasodilation may increase catecholamine exposure and contribute to arrhythmogenesis. However, causality cannot be inferred.

Importantly, additional analyses examining the timing of POAF did not demonstrate an early protective effect of magnesium. Both day-by-day cumulative incidence and time-to-event analyses showed no early separation in favor of magnesium, and the temporal pattern of events was consistent with the primary analysis. These findings suggest that the absence of benefit is unlikely to be explained by the relatively short duration of magnesium administration confined to the ICU period and post hoc subgroup analyses revealed no consistent benefit across surgical type, age, bypass duration, or ventricular function, with only weak signals in selected strata.

Strengths and Limitations

Key strengths include the randomized, double-blind design, protocolized intervention, and systematic outcome assessment. Serial magnesium measurements and high adherence ensured robust separation between groups, maximizing the ability to detect a true biological effect.

Several limitations merit consideration. First, only total serum magnesium was measured; ionized and intracellular magnesium, which may be more closely related to electrophysiologic effects, were not assessed. The relationship between total and biologically active magnesium is complex and influenced (among others) by albumin and intracellular distribution (35–37).

Second, POAF was recorded when documented on ECG (first 72 hr postsurgery, continuous telemetry was used besides minimal daily ECG recording), which may have underestimated brief or asymptomatic episodes, although observed incidence was consistent with prior literature.

Third, patients in the control group received magnesium as part of routine clinical care. This background exposure reflects contemporary practice but may have attenuated any treatment effect, limiting the ability to isolate the incremental benefit of the study intervention and representing an important constraint on interpretation.

Fourth, perioperative management variables, including duration of mechanical ventilation and timing of extubation, were not specifically analyzed. Although the short duration of ventilation and ICU length of stay suggest a cohort with largely uncomplicated postoperative recovery, unmeasured differences in perioperative care represent a potential source of residual confounding.

Fifth, postoperative beta-blocker use was not systematically recorded and represents an important source of potential unmeasured confounding. Although most patients were receiving chronic beta-blocker therapy preoperatively, variability in postoperative resumption or initiation may have influenced atrial fibrillation risk. Differential use of beta-blockers after ICU discharge represents a potential source of unmeasured confounding that would not be addressed by randomization and may have affected treatment effect estimates.

Sixth, magnesium administration was limited to the perioperative and early postoperative period, with a median duration of approximately 20 hours, whereas the peak incidence of POAF occurred several days later. Although additional analyses did not demonstrate an early treatment effect, a mismatch between duration of exposure and timing of outcome remains a potential limitation.

Seventh, other atrial arrhythmias and the management of POAF, including pharmacologic treatment and conversion to sinus rhythm, were not prospectively collected as part of the predefined study endpoints, as the trial focused on incidence rather than treatment of arrhythmia.

Eighth, the cohort was relatively low risk, with predominantly preserved ventricular function and short cardiopulmonary bypass and cross-clamp times. Although adjustment for these factors did not materially alter treatment effect estimates (Supplementary Table 7, https://links.lww.com/CCM/H966), generalizability may be limited to higher-risk populations with impaired ventricular function and greater surgical complexity, in whom atrial arrhythmias are more prevalent.

Finally, the trial was registered in a registry that is no longer actively maintained, and the current registry record does not accurately reflect study status. Although the study was conducted according to the originally registered protocol, this discrepancy represents a limitation in reporting transparency.

CONCLUSIONS

In this randomized trial, perioperative magnesium infusion targeting serum concentrations of 1.5–2.0 mmol/L did not reduce POAF after cardiac surgery. Despite achieving robust magnesium exposure, no clinical benefit was observed. These findings, together with the broader evidence base, do not support routine prophylactic magnesium for prevention of POAF. Future research should focus on alternative preventive strategies with stronger mechanistic rationale.

ACKNOWLEDGMENTS

The authors specifically acknowledge Prof. P.W.B. Nanayakkara (MD, PhD), J.A. Janson (MD), and H. Merten (PhD) for their participation within the DSMB. Nadine de Roode, A-M van den Heuvel, and E. Karijodikoro-Bogaards for their help in the logistics and management of the trial. J. Plischke for his active supervising role in the inclusion of patients from the ward and assistance in logistics. Furthermore, all nursing staff and physicians from the department’s cardiothoracic surgery, cardio-anesthesia, and ICU for the willingness to assist in the trial.

Supplementary Material

ccm-54-1635-s001.pdf ^{(1.1MB, pdf)}